University of Southern California Dissertations and Theses

What leads to a happy life? Subjective well-being in Alaska, China, and Australia

(USC Thesis Other)

What leads to a happy life? Subjective well-being in Alaska, China, and Australia

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content WHAT LEADS TO A HAPPY LIFE? SUBJECTIVE WELL-BEING IN ALASKA, CHINA,
AND AUSTRALIA

by

Fengyu Wu

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirement for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)

August, 2018

Copyright 2018 Fengyu Wu
2

For my parents, in memory of my grandfather

3
ACKNOWLEGEMENTS

I am grateful to my parents for the selfless support they have given me in the past six years. I am
blessed to have Richard A. Easterlin as my advisor. He is one of the pioneers in the Economics of
Happiness and one of the most knowledgeable people I have ever met. He has taught me from how
to think to how to put things into words. When I was struggling with my research, he gave me not
only advice but also encouragement. I hope to continue learning from him. I also want to thank
Jeffrey B. Nugent, who patiently offered me advice on both academic and non-academic matters.
He gave me the opportunity to work with him on joint projects, an experience that I found valuable
and enjoyable. I want to thank John Strauss as well. I learned how to conduct rigorous quantitative
research from him in several classes. He has always been an incredible teacher and advisor.
I have learned from many other scholars in the past few years. My happiness classmates
Kelsey O’Connor and Robson Morgan have given me valuable feedback and friendship. Fellow
graduate students, including (but not limited to) Mingming Ma, Wenjing Chu, Rui He, Tushar
Bharati, and Urvashi Jain have offered me generous help with research and classes.
Last but not least, I would like to thank the administrative staff of Department of
Economics. Morgan Ponder, Young Miller, and Fatima Perez have been really patient and helpful
when handling administrative matters. I also want to thank the university for providing financial
support.

4
Table of Contents
DEDICATION............................................................................................................................... 2
ACKNOWLEGEMENTS ............................................................................................................ 3
LIST OF TABLES ........................................................................................................................ 6
LIST OF FIGURES ...................................................................................................................... 8
ABSTRACT ................................................................................................................................... 9
CHAPTER 1. INTRODUCTION .............................................................................................. 10
CHAPTER 2. LIFE SATISFACTION OF THE INDIGENOUS POPULATION OF
NORTHERN ALASKA .............................................................................................................. 15
2.1 Introduction ..................................................................................................................................... 15
2.2 The Study Population and Region ................................................................................................. 15
2.4 Data and Methods ........................................................................................................................... 22
2.4.1 Data ............................................................................................................................................ 22
2.4.1.1 Survey of Living Conditions in the Arctic (SLiCA) ......................................................................... 22
2.4.1.2 Other Sources of Data ........................................................................................................................ 23
2.4.2 Modelling Subjective Well-being .............................................................................................. 23
2.5 Results............................................................................................................................................... 25
2.5.1 Basic Models .............................................................................................................................. 25
2.5.2 Life Satisfaction and Connections with Culture and Communities ........................................... 32
2.5.3 Robustness Checks ..................................................................................................................... 34
2.6 Conclusion ........................................................................................................................................ 37
CHAPTER 3. THE EFFECT OF INTERGENERATIONAL SUPPORT ON THE LIFE
SATISFACTION OF OLDER PARENTS IN CHINA ........................................................... 38
3.1 Introduction ..................................................................................................................................... 38
3.2 Literature Review ............................................................................................................................ 43
3.3 Conceptual Framework .................................................................................................................. 45
3.4 Methodology..................................................................................................................................... 47
3.5 Data and Measurement ................................................................................................................... 49
3.5.1 Data ............................................................................................................................................ 49
3.5.2 Measurement .............................................................................................................................. 51
3.5.2.1 Measurement of Life Satisfaction ...................................................................................................... 51
3.5.2.2 Measurement of Structural and Functional Support .......................................................................... 51
3.5.2.3 Covariates .......................................................................................................................................... 56
3.6 Results............................................................................................................................................... 57
3.6.1 Rural Results .............................................................................................................................. 57
3.6.2 Urban Results ............................................................................................................................. 62
3.7 Further Analysis .............................................................................................................................. 66
3.7.1 Control for Individual Heterogeneity ......................................................................................... 66
3.7.2 A Less Selective Sample ............................................................................................................ 69
2.7.3 Floating Population in Urban Areas........................................................................................... 71
3.8 Conclusion and Discussion ............................................................................................................. 75

5
CHAPTER 4. AN EXAMINATION OF THE EFFECTS OF CONSUMPTION
EXPENDITURES ON LIFE SATISFACTION IN AUSTRALIA ......................................... 79
4.1 Introduction ..................................................................................................................................... 79
4.2 Literature Review ............................................................................................................................ 82
4.3 Data and Variables .......................................................................................................................... 86
4.3.1 Data ............................................................................................................................................ 86
4.3.2 Variables .................................................................................................................................... 87
4.3.2.1. Life Satisfaction ................................................................................................................................ 87
4.3.2.2 Variables on Material Living Conditions .......................................................................................... 88
4.3.2.3 Demographic and Socioeconomic Controls ...................................................................................... 96
4.4 Model Estimation ............................................................................................................................ 98
4.5 Results............................................................................................................................................. 102
4.5.1 Life Satisfaction and Income ................................................................................................... 102
4.5.2 Life Satisfaction, Total Consumption Expenditures, and Savings........................................... 105
4.5.3 Life Satisfaction, Conspicuous Consumption Expenditures, and Basic Consumption
Expenditures...................................................................................................................................... 106
4.5.3.1 Heterogeneity among Income Groups ............................................................................................. 111
4.6 Robustness Checks ........................................................................................................................ 113
4.6.1 Alternative Categorization of Conspicuous and Basic Consumption Expenditures ................ 113
4.6.2 Examining the Effects of 18 Components of Consumption Expenditures .............................. 116
4.6.3 Using Financial Satisfaction Instead of Life Satisfaction as Dependent Variable .................. 116
4.7 Conclusions .................................................................................................................................... 119
CHAPTER 5. SUMMARY AND CONCLUSIONS ............................................................... 121
REFERENCES .......................................................................................................................... 125
APPENDIX A Supplementary Material for Chapter 2......................................................... 138
APPENDIX B Supplementary Material for Chapter 3 ......................................................... 140
APPENDIX C Supplementary Material for Chapter 4......................................................... 142
Appendix C.1 Sample Attrition .......................................................................................................... 142
Appendix C.2 Measurement Error in Life Satisfaction ................................................................... 144

6

LIST OF TABLES

Table 2-1 Descriptive Statistics of Variables................................................................................ 26
Table 2-2 Life Satisfaction, Demographic & Socioeconomic Characteristics, Health, and Income:
Ordered Logit Regressions ................................................................................................... 27
Table 2-3 Understanding the Effects of Wage Income: Ordered Logit Regressions ................... 32
Table 2-4 Life Satisfaction and Connections with Culture and Community: Ordered Logit
Regressions ........................................................................................................................... 35
Table 2-5 Robustness Check (Controlling for Community Characteristics): Ordered Logit
Regressions ........................................................................................................................... 36
Table 3-1 Summary Statistics ....................................................................................................... 52
Table 3-2 Life Satisfaction and Structural Support (Rural Sample, OLS Estimation) ................. 58
Table 3-3 Life Satisfaction and Functional Support (Rural Sample, OLS Estimation) ................ 59
Table 3-4 Life Satisfaction, Structural Support, and Functional Support (Rural Sample, OLS
Estimation) ............................................................................................................................ 60
Table 3-5 Life Satisfaction and Structural Support (Urban Sample, OLS Estimation) ................ 63
Table 3-6 Life Satisfaction and Functional Support (Urban Sample, OLS Estimation) .............. 64
Table 3-7 Life Satisfaction, Structural Support, and Functional Support (Urban Sample, OLS
Estimation) ............................................................................................................................ 65
Table 3-8 . Life Satisfaction, Structural Support, and Functional Support (Rural Sample, FE
Estimation) ............................................................................................................................ 66
Table 3-9 Life Satisfaction, Structural Support, and Functional Support (Urban Sample, FE
Estimation) ............................................................................................................................ 67
Table 3-10 Life Satisfaction and Structural Support (Larger Sample) ......................................... 71
Table 3-11 Life Satisfaction, Structural Support, and Functional Support (Urban Sample with
Urban Hukou)........................................................................................................................ 73
Table 3-12 Life Satisfaction, Structural Support and Functional Support (Urban Sample with
Rural Hukou) ......................................................................................................................... 74
Table 4-1 Weekly Expenditures: Household Expenditure Survey (HES) (2009-10) and HILDA
(2010) .................................................................................................................................... 89
Table 4-2 Definitions of Conspicuous Goods vs. Basic Goods .................................................... 93
Table 4-3 Descriptive Statistics, Pooled over All Waves, HILDA 2006–2010 ............................ 97
Table 4-4 Life Satisfaction and Income ...................................................................................... 102
Table 4-5 Life Satisfaction, Total Household Consumption Expenditures, and Savings ........... 106
Table 4-6 Life Satisfaction, Conspicuous Consumption Expenditures, and Basic Consumption
Expenditures ....................................................................................................................... 107
Table 4-7 Life Satisfaction, Relative Conspicuous Expenditures, and Relative Basic
Expenditures ....................................................................................................................... 109
Table 4-8 Life Satisfaction and Expenditure on Durable Goods, Nondurable Goods, and Services
............................................................................................................................................. 110
Table 4-9 Life Satisfaction, Conspicuous Consumption Expenditures, and Basic Consumption
Expenditures: Heterogeneity among Income Groups ......................................................... 111
7
Table 4-10 Robustness Check: Alternative Definitions of Conspicuous and Basic Consumption
Expenditures ....................................................................................................................... 115
Table 4-11 Robustness Check: Life Satisfaction and Components of Consumption Expenditures
............................................................................................................................................. 117
Table 4-12 Robustness Check: Using Financial Satisfaction Instead of Life Satisfaction as a
Dependent Variable ............................................................................................................ 118

Appendix Table A-1 Definition of Key Variables ...................................................................... 138
Appendix Table A-2 The Distribution of Household Income Among the Five Income Sources by
Size of Income .................................................................................................................... 139
Appendix Table B-1 Summary Statistics (Larger Sample) ........................................................ 140
Appendix Table C-1 Test for Selective Attrition (Fixed Effects Regressions, HILDA 2006-2009)
............................................................................................................................................. 143
Appendix Table C-2 Consumption Categories ........................................................................... 143
Appendix Table C-3 Life Satisfaction, Income, Total Household Consumption Expenditures,
Savings, Conspicuous Consumption Expenditures, and Basic Consumption Expenditures
(Blundell and Bond GMM Estimation Results).................................................................. 144
Appendix Table C-4 Life Satisfaction and Total Consumption Expenditures, and Savings (OLS
and FE Results) ................................................................................................................... 146
Appendix Table C-5 Life Satisfaction, Conspicuous Consumption Expenditures and Basic
Consumption Expenditures (OLS and FE results) .............................................................. 146

8
LIST OF FIGURES
Figure 3-1 Conceptual Framework ............................................................................................... 46
Figure 4-1 Relationship between Conspicuous and Basic Consumption Expenditures ............... 95

9
ABSTRACT

This work presents new evidence on the determinants of subjective well-being, as measured by
life satisfaction, in three quite different populations. Chapter 2 studies the Inuit, an indigenous
hunter-gatherer population living on the barren northernmost fringes of Alaska. My analysis
indicates that the key factors that contribute to their life satisfaction are health, subsistence hunting
and fishing, and social support. In addition, the Inuit are more satisfied with their life if they have
both Christian religious beliefs and indigenous spiritual beliefs as part of their life. Surprisingly, a
higher level of wage income is associated with a lower level of life satisfaction, a finding that
challenges common preconceptions about the effects of modernization and points to the
importance of non-wage subsistence activities as a preferred substitute for wage employment for
this population. Chapter 3 examines the well-being of the elderly population in China, a country
characterized by a traditional three-generation family structure. In rural villages in China, life
satisfaction of older parents is still positively associated with traditional kinds of support such as
living in a three-generation household and the exchanges of financial and emotional support with
their children. However, in China's urban neighborhoods, this is not true. It seems that China’s
development, especially the development in urban areas, is breaking down the historical
association between intergenerational support and well-being. Chapter 4 aims to understand the
importance of social comparisons to people’s well-being. The study, which focuses on Australia,
investigates the effects of consumption expenditures on life satisfaction. It shows that conspicuous
(i.e. visible and positional) spending increases happiness while savings and spending on basic
goods and services, the less visible components of income, do not contribute to it. Moreover, this
research provides evidence for relationship heterogeneity across income groups.
10
CHAPTER 1. Introduction

This dissertation aims to understand the factors contributing to the determination of quality of life.
To achieve this, it investigates how material living conditions, family support, social comparison,
and various other factors affect well-being. Three original studies on three different populations
are presented.
Throughout this dissertation, quality of life or well-being is measured using subjective
well-being (SWB) measures. SWB is the general expression used to cover a variety of individual
self-reports of quality of life (Sachs et al. 2013). There are, in general, two types of measures of
SWB: cognitive life evaluations and emotional reports. Emotional reports, such as positive affect
(a range of positive emotions) and negative affect (a range of negative emotions), capture feelings
at one point in time. The focus of this dissertation is on life evaluations, represented by questions
asking how happy or satisfied people are with their lives as a whole. Such measures are considered
plausible measures of well-being since they are closely related to life circumstances, consistent
over a short period of time, and strongly correlated with both objective and subjective measures of
well-being.
1

There is a growing debate in economics on whether measures of material living conditions,
such as income, should be used exclusively when evaluating well-being in countries worldwide.
Some economists have argued that such measures, though providing information on material
aspects of life, fail to take into account non-material concerns such as family life, health, and work.
To better understand people’s well-being, researchers and policy makers have suggested collecting

1
Further discussions on the types of SWB questions and their reliability and validity can be found in Sachs et al. (2013) and OECD
(2013).

11
and analyzing data on SWB. For example, the recent Stiglitz-Sen-Fitoussi Report (2010), which
involved more than 20 leading economists, including several winners of the Nobel prize in
economics, recommends the use of SWB measures as a complement to measures like GDP per
capita in evaluating well-being in countries worldwide. Governments in more than ten countries
have now embarked on the collection of data on SWB as a component of official statistics. The
increasing availability of such data will offer policy makers a valuable new guide to decisions. The
present dissertation uses such data from three individual-level surveys conducted in three
countries/regions to understand, from different perspectives, the determinants of life satisfaction
in three very different populations. In the following paragraphs, I am going to provide a brief
description of the three studies.
Chapter 2 studies the life satisfaction of the indigenous population living in the Arctic
Alaska. The population, which is called the Inuit, was a hunter-gather population half a century
ago and still retains much of its hunter-gatherer life style. There is a common scholarly
preconception that modernization should make people better off. However, this study, using data
from the Alaska portion of the Survey of Living Conditions in the Arctic (SLiCA), finds that life
satisfaction of this indigenous population is surprisingly high, perhaps as high as that of the U.S.
population in general. In order to understand its high life satisfaction, the study estimates a
happiness equation, which includes not only traditional variables, such as gender, education,
income, and health, but also variables pertaining to connections with culture and community.
The regression analysis suggests that the factors contributing the Inuit’s life satisfaction
are health, subsistence hunting and fishing, social support, and having both Christian beliefs and
indigenous spiritual beliefs as part of their life. Surprisingly, wage income is found to be negatively
associated with the level of life satisfaction. One possible mechanism is that longer working time
12
spent on wage employment has restricted them from participating in hunting and fishing. The study
challenges the preconception that modernization increases happiness and indicates that the hunter-
gatherer life style, which the Inuit have been able to retain, is central to their life satisfaction.
Chapter 3 looks into the well-being of the elderly population in China. Traditionally, the
family, particularly adult children, serves as the primary source of the support for older parents.
However, economic growth, changes in public policies, the rise in internal migration as well as the
evolution of social attitudes have changed the pattern of living arrangements in China, and might
have altered older parents’ need for family support as well. Given the differences between rural
and urban China in terms of economic, policy, and social factors, the study investigates rural and
urban populations separately. The data from the first two waves of the China Health and
Retirement Longitudinal Study (CHARLS) are used.
Both the ordinary least squares and fixed-effects regressions suggest that the life
satisfaction of rural older parents, compared with that of their urban counterparts, depends more
on intergenerational support. Rural parents still enjoy the extended family experience. In
particular, they are happiest if living in a three-generation household, and living with grandchildren
plays an important role. I also find, for rural parents, that the exchanges of financial and emotional
support with their children have a positive and significant influence on life satisfaction. In contrast,
intergenerational support does not influence life satisfaction of urban parents as much in
contemporary China. For urban parents, living in a skip-generation household (i.e. living with
grandchildren only but without children) has a negative and significant influence on life
satisfaction while receiving help with self-care or household tasks from children is happiness
enhancing. It seems that China’s development, especially the development of urban China, is
changing the relationship between family support and well-being for older parents.
13
Chapter 4 tries to understand the importance of social comparison for well-being by
investigating the effects of consumption expenditures on life satisfaction in Australia. Previous
studies on social comparison mostly found a positive effect of relative income on life satisfaction.
My study looks at how consumption expenditures, especially spending on certain types of goods
and services, affect people’s life satisfaction. The study builds a dynamic model of life satisfaction
and estimates causal relationships by controlling for bias from not only unobserved individual
heterogeneity but also endogenous variables. The data come from 5 waves (wave 6 – 10) of
Household, Income and Labour Dynamics in Australia Survey.
Consistent with the existing literature, my study finds, causally, income has a positive and
moderate effect on life satisfaction. Specifically, I find that a doubling in household disposable
income leads to a 0.17 increase in the level of life satisfaction (on an 11-point scale), which is
about 1/3 of the within-person standard deviation of life satisfaction within my sample period.
When I look into the main components of income, or, in other words, how people spend their
income, I find that an important channel through which income influences an individual’s life
satisfaction is conspicuous spending, which is defined as spending on goods or services that are,
first, visible, and, second, positional (i.e. having value that depends on social comparisons). My
analysis also suggests that it is one’s conspicuous expenditures relative to those of others within
his or her reference group that really matter to individual’s life satisfaction. In contrast, basic
spending, such as spending on utilities and insurance, does not contribute to life satisfaction. In
addition, I find relationship heterogeneity across income groups: conspicuous expenditures have a
positive and significant influence on life satisfaction for individuals in all income groups.
However, basic expenditures have a negative and significant influence on life satisfaction for
14
individuals in the lowest income quartile only. My findings underscore the importance of social
comparison to people’s well-being and the necessity to model interdependence in utility functions.

15
CHAPTER 2. Life Satisfaction of the Indigenous Population of Northern Alaska
2.1 Introduction

The Inuit (Iñupiat, Yup’ik, and Siberian Yup’ik) is an indigenous population living in the Arctic
region. It is a hunter-gatherer population that still retains strong ties to the land. When compared
with the life satisfaction
2
of other indigenous populations living in the Arctic, that of the Alaskan
Inuit is surprisingly high, perhaps as high or higher than that of the U.S. population generally.
This paper, which builds on previous studies of subjective well-being of the Inuit (Poppel
& Kruse, 2009; Berman, 2009; Martin, 2012), explores the life satisfaction of the population
comprehensively. I include not only traditional variables (Blanchflower and Oswald, 2002), such
as gender, education, race, income, and health, but also variables measuring the connections with
culture and community. The main data source for this study is the Alaska portion of the Survey of
Living Conditions in the Arctic (SLiCA), which was the first to allow the comparison of living
conditions of indigenous people with similar cultures around the Arctic. The survey provides
opportunities to examine the importance of the availability of fishing and game, cash income,
family and social relations, and religious and spiritual beliefs for living in the Arctic.

2.2 The Study Population and Region

This study focuses on the Inuit (Iñupiat, Yup’ik, and Siberian Yup’ik) who live in the Alaskan
Arctic. Approximately 80 % of the people living in the region are indigenous, and the majority are
Iñupiat (Howe, 2009).

2
The term life satisfaction, happiness, and subjective well-being are used interchangeably in this paper, and refer to satisfaction
with life as a whole.
16
The study region contains the three northernmost census areas in the United States, the
North Slope Borough, the Northwest Arctic Borough, and the Bering Straits Census Area of
Alaska. There are 34 villages and three regional centers, which are connected primarily by air.
According to the 2000 Census, the average population size for a regional center is about 3500.
Villages range in size from about 100 to 750, with an average of 380 (Howe, 2009). The three
regional centers, Barrow, Kotzebue and Nome, are very different from villages in terms of labor
market opportunities and infrastructure. Regional centers have more opportunities for wage
employment than villages. According to the 2000 Census, 50 % of the working-age population
were unemployed or out of labor force in villages, compared to only 37% in regional centers.
Besides, the regional centers have more advanced infrastructure, such as piped sewer and water
systems and a maintained road system. There are also modern schools, hospitals, federal and state
offices in the regional centers (Howe, 2009).
The Alaska Arctic region has relatively poor economic performance in general. The
unemployment rate of the three areas are significantly higher than the state average with that of
Northwest Arctic Borough being the highest. For instance, in 2003, the unemployment rate in
Northwest Arctic Borough was close to 20 percent, while the state’s unemployment rate was only
about 8. The wage incomes are much lower in the Arctic (Huskey & Howe, 2010). However, the
mean and median household incomes are similar between the Alaskan Inuit population and the US
general population (SLiCA, 2002 & 2003; the US Census Bureau, 2003).
One reason for the relatively high household income of the Inuit population is the large
amount of transfer income from government and other organizations. According to SLiCA (2002,
2003), the median share of transfer income in total household income is one third for the Inuit,
compared with the that of the U.S. population as a whole of 8 percent. The percentage is even
17
higher for households with lower total income. For example, households with the lowest percentile
of income receive approximately 80% of their income from transfers. Transfers come to
individuals as Aid to Families with Dependent Children (ADFC), public assistance, retirement and
disability payments, the Alaska Permanent Fund Dividend (PFD), and dividends from regional
and village corporations. Regional and village corporations were established under the Alaska
Native Claims Settlement Act (ANCSA)
3
. Under this act, Alaskan natives who enrolled in the
native associations were made the shareholders in regional and village corporations (Ongtooguk,
2012).
The relatively low monetary economy in the region is also moderated by the region’s active
subsistence economy, which provides a significant source of real income to residents of the region.
According to the Alaska Division of Subsistence (Wolfe, 2000), most rural families in Alaska
depend on subsistence hunting and fishing. In surveyed communities, nearly 78% of households
fish and 63% hunt. The harvests, including whale, walrus, seals, caribou, moose, eggs, fish, and
berries, provide an important portion of dietary needs: residents receive over 300% of the
Recommended Dietary Allowance of protein and about half of their required calories are from
subsistence fish and game harvests (Huskey & Howe, 2010). For 52 percent of the households, no
less than half of the meat and fish consumed is harvested by their household members (SLiCA,
2002 & 2003). Subsistence harvesting is essential for the cultural continuity. According to the
Alaska Native Commission (1994), “it[subsistence] also involves cultural values and attitudes:
mutual respect, sharing, resourcefulness, and an understanding that is both conscious and mystical
of the intricate in interrelationships that link humans, animals, and the environment.” Besides the

3
The act was approved in 1971. Under this act, Alaska Natives received title to over 40 million acres of land and $ 962.5 million
to set up regional and village corporations. There were 12 regional corporations and over 200 village corporations at the time of
the Act. The 13
th
regional corporation was later created for Alaska Natives who no longer resided in Alaska. In Arctic Alaska, the
regional corporations are Arctic Slope Regional Corporation (ASRC), Northwest Arctic Native Association (NANA) and the
Bering Straits Native Corporation (BSNC).
18
food it produces, subsistence gives Alaska Natives the rather healthy perspective that they are
participating in the events of nature and providing the health and well-being of the community
(Martin, 2012). Subsistence activities seem be even more extensive in whaling communities. The
Alaska Eskimo Whaling Commission (AEWC) defines 9 whaling communities in Alaska that
participate in bowhead whale hunts. In a whaling community, nearly everyone is involved in whale
harvests, “as a member of a crew, helping to prepare for the hunt or in the butchering and
distribution of the harvest” (Martin, 2005). According to Huntington (1992), bowhead whale
harvests provide “life, meaning, and identity to the Eskimo whalers and their communities”.
Mixed cash- and harvest herding- based economies have become increasingly important
over time for communities living in the Arctic (Kruse, 1991; Kruse et al., 2008). Approximately
80 percent of our sample want to both work on wage jobs and harvest, herd or process their own
food (SLiCA, 2002 & 2003). Cash or employment become sources with which to acquire and
maintain modern inputs, such as snow machines, four wheelers, aluminum boats, and gasoline
motors, to do subsistence harvest (Howe 2009). As a result, households reporting higher wage
income also report more subsistence harvests (Kruse 1991; Kirkvliet & Nebesky, 1997). By
custom, they also share the subsistence harvests with others in the community (Martin, 2012).
Another interesting fact about the Inuit population is the transformation of the Inuit beliefs
in the 20th century (Laugrand & Oosten, 2010). In the 19th century, angakkuuniq (shamanism)
was the core of Inuit beliefs and practices. For Inuit, shamanism is embedded in a framework of
cosmological beliefs and practices, according to which not only human beings but also animals
are sentient beings that have to be respected. The Inuit consider the introduction of Christianity as
a break with the past: the missionaries taught that only human beings had souls. Today, many of
19
the Inuit follow Christianity but they continue to observe some rules of respect to non-human
beings: they observe shamanic features in some forms of Christianity.
Despite some general characteristics which the three regions share in common, there are
some differences between them. The most distinct difference is probably the relative abundance
of natural resources, and the wealth generated by the resources and distributed to the indigenous
individuals through the regional and village corporations. The communities in the North Slope
Borough are relatively wealthy (i.e. with higher GDP per capita) because the region contains a
major area of petroleum activities, including the large Prudhoe Bay oil fields. The oil tax revenue
to the local government provides the communities with employment opportunities (Berman, 2009).
The Red Dog zinc mine is located in the Northwest Arctic Borough. Even though, compared with
the other two regions, the Bering Straits region is relatively poor in natural resources, its
communities are connected to the world economy through Bering Sea fishery (Huskey and Howe,
2010).
It is the distinct features of Arctic Alaska that have attracted scientists and policy makers
to investigate how Alaska Natives with epoch-one (i.e. hunter-gatherer) lifestyle adapt to the job
opportunities and wage income in epoch three (i.e. modern industry economy). How the features
and the adaption affect the life satisfaction of the population is the focus of this study.

2.3 Literature Review

According to SLiCA, 55% of indigenous individuals in Alaska sample responded “5 very satisfied”
to the question “please tell me the number on this card that fits how satisfied you are with your life
as a whole”, whereas the other countries in the survey had lower percent of selecting “5 very
20
satisfied” on the same question (e.g. Canada: 50%; Greenland: 23%). In addition, the Inuit in
Arctic Alaska are as happy or happier than the U.S. population in general. This can be inferred
from the results of various surveys with questions on life satisfaction, even though the questions
and answers are phrased differently. According to the wave 4 of the World Values Surveys which
was conducted in the U.S. in 1999, only 35.2 percent of the U.S. population rank themselves as 9
or 10 on the 10-point-scale question for life satisfaction. And, the data from General Social Survey
(2000, 2002 and 2004) show that only 31.3 percent of individuals choose “very happy” when
responding to the question “taken all together, how would you say things are these days—would
you say you are very happy, pretty happy or not too happy?” The high life satisfaction of the
indigenous population in Northern Alaska actually coincides with the high life satisfaction of the
indigenous population in Australia. It has been found that, even though the indigenous population
of Australia are worse off than the non-indigenous population, they are more satisfied with their
lives (e.g. Ambrey & Fleming, 2014).
The research of life satisfaction or happiness is at the intersection of psychology, sociology
and economics. The study of happiness was the domain of psychology for a long time (Frey &
Stutzer, 2002). The groundbreaking contribution by Easterlin (1974) linked this psychological
research to economics, but there were few followers at that time. Since the late 1990s, the literature
on empirical analyses of the determinants of happiness in different regions and periods has started
to grown, for instance, the relationship between life satisfaction and age (Blanchflower & Oswald,
2004a; Blanchflower & Oswald, 2008; Hayo&Seifert, 2003; Helliwell, 2003), life satisfaction and
gender (Blanchflower, 2009; Helliwell, 2008; Senik, 2004; Stevenson & Wolfers, 2009), life
satisfaction and income ( Easterlin, 1995; Easterlin, 2001; Di Tella et al., 2001; Frey & Stutzer,
2000), life satisfaction and unemployment (Blanchflower & Oswald, 2004a; Clark & Oswald,
21
1994; Winkelmann & Winkelmann, 1998), life satisfaction and education(Frey & Stutzer, 2010;
Oreopoulus& Salvanes, 2011; Hartog & Oesterbeek, 1998 ), life satisfaction and marriage ( Senik,
2004; Knight et al., 2009; Stutzer & Frey, 2006), and life satisfaction and health (Oswald &
Powdthavee, 2008; Diener & Chan, 2011; Frey, 2011).
Earlier studies on well-being in Arctic Alaska have focused primarily on the relationship
between subsistence and well-being of the region (Langton, 1991; Huntington, 1992). However,
the quantitative analyses of this sort were limited to descriptive statistics (Wolfe and Walker, 1987;
Kruse, 1991). More recent research, however, has involved more substantial empirical analyses
using the data from SLiCA. Berman (2009) investigated the role of local subsistence as a factor
influencing individual’s decision to move, which can reflect household well-being from living in
a community. The results suggest that Inuit respondents in small Alaskan communities place a
high value on local subsistence opportunities, opportunities to earn wage income and quality-of-
life factors when deciding their place of residence. By including community characteristics in the
analysis, Martin (2012) found that what matters most for life satisfaction are family ties, social
support and opportunities to do things with other people. Her results indicate that the probability
of employment is associated with a lower level of satisfaction which may be explained by the fact
that people with jobs have less time for hunting and fishing and less time for activities with
extended family and friends. However, previous studies failed to examine important factors such
as income and spiritual beliefs of the population.

22
2.4 Data and Methods
2.4.1 Data
2.4.1.1 Survey of Living Conditions in the Arctic (SLiCA)

The individual-level data of this study come from the Survey of Living Conditions in the Arctic
(SLiCA). SLiCA is an international partnership of indigenous people and researchers to measure,
compare and understand living conditions in Arctic Alaska, Canada, Greenland, Norway, Sweden,
Finland, and portions of Russia. It tries to measure living condition in a way that is relevant to
Arctic residents and improve the understanding of living conditions for the benefit of Arctic
residents. The survey design was based on previous studies on living conditions, social indicator
development and quality of life. SLiCA expanded the measurement of living conditions from
traditional measures such as income and unemployment to family relationship, social support, and
ethnic identity (Poppel et.al., 2007).
The Alaska portion of the survey consists of 663 respondents aged 16 and above with
information on approximately 3000 individuals in 20 communities in the North Slope, Northwest
Boroughs, and the Bering Straits region (i.e. Nome Census area). The sample used in this analysis
consists of 497 individuals with information on life satisfaction and other individual and household
characteristics. The around-90-minute face-to-face interviews took place in 2002 in the Northwest
Arctic region and in 2003 in the North Slope and Bering Straits regions. All the interviews in
Alaska were conducted in January and February because people are less likely to be away for
hunting and fishing and more likely to be at home during the winter (Martin, 2005). The response
rate was 84 percent and the maximum estimated sampling error was 4% (Kruse et. al., 2008). The
review and approval of the questionnaire, survey procedures, review of local communities, and
23
procedures for publication of results by other researchers were the responsibilities of Alaska Native
Management Board. This board consisted of members from each of the three regions and Alaska
Native Science Commission, and included international representation from the Inuit Circumpolar
Conference (Martin, 2012).

2.4.1.2 Other Sources of Data

Community level data come from the 2000 US Census and Alaska Eskimo Whaling Commission
(AEWC). The 2000 US Census provides community-level data on total employment and Alaska
Native share of the working age population, which can be used to construct employment per
working-age Native. It also provides information on the condition of the infrastructure in each
community, such as the percent of occupied housing unit that lack complete plumbing facilities.
The AEWC defines the communities that participate in bowhead whale hunts. These communities,
as described above, are more extensively involved in subsistence harvesting.

2.4.2 Modelling Subjective Well-being

The conceptual framework of this study is based on Blanchflower and Oswald (2002) and Frey
and Stutzer (2010). The following is an index model for a single latent
variable 𝑙𝑖𝑓𝑒 _𝑆𝑎𝑡𝑖𝑠𝑓𝑎𝑐𝑡𝑖𝑜𝑛 ∗, which is unobservable:
𝑙𝑖𝑓𝑒 _𝑆𝑎𝑡𝑖𝑠𝑓𝑎𝑐𝑡𝑖𝑜 𝑛 𝑖𝑗
∗= 𝑋 𝑖𝑗
′
𝛽 + 𝑍 𝑗 ′
𝛾 + 𝑢 𝑖𝑗
(1)

𝑙𝑖𝑓𝑒 _𝑆𝑎𝑡𝑖𝑠𝑓𝑎𝑐𝑡𝑖𝑜 𝑛 𝑖𝑗
= 𝑘 if 𝛼 𝑘 −1
< 𝑙𝑖𝑓𝑒 _𝑆𝑎𝑡𝑖𝑠𝑓𝑎𝑐𝑡𝑖𝑜 𝑛 𝑖𝑗
∗≤ 𝛼 𝑘 (2)
24

𝑃 𝑖𝑗𝑘
= 𝑃 (𝑙𝑖𝑓𝑒 _𝑆𝑎𝑡𝑖𝑠𝑓𝑎𝑐𝑡𝑖𝑜 𝑛 𝑖𝑗
= 𝑘 ) = 𝑃 (𝛼 𝑘 −1
< 𝑙𝑖𝑓𝑒 _𝑆𝑎𝑡𝑖𝑠𝑓𝑎𝑐𝑡𝑖𝑜 𝑛 𝑖𝑗
∗≤ 𝛼 𝑘 )
= 𝐹 (𝛼 𝑘 − 𝑋 𝑖𝑗
′
𝛽 − 𝑍 𝑗 ′
𝛾 ) − 𝐹 (𝛼 𝑘 −1
− 𝑋 𝑖𝑗
′
𝛽 − 𝑍 𝑗 ′
𝛾 ) (3)

𝑙𝑖𝑓𝑒 _𝑆𝑎𝑡𝑖𝑠𝑓𝑎𝑐𝑡𝑖𝑜 𝑛 𝑖𝑗
is the self-reported level of life satisfaction of individual i in
community j. X
ij
is a vector of individual characteristics, Z
j
is a vector of community
characteristics of community j, and u
ij
is the error term. The structure of the model is suitable to
be estimated using order logit or probit. In this study, I will present the results only from ordered
logit regressions, since the results from the ordered probit model appear to be very similar. For the
ordered logit model, F is the logistic cdf F(z) = e
z
/(1 + e
z
).
The main independent variables of interest include income measures, leisure and traditional
activities, religious or spiritual beliefs, quality of housing, public safety, family ties, and social
support. All household income measures, including total household income, household wage
income and household transfer income, used in this study are equivalized. Equivalized household
income is derived by calculating an equivalence factor according to an equivalence scale, and then
dividing income by the factor. The equivalence scale used is OECD equivalence scale, which
assigns a value of 1 to the first household member, of 0.7 to each additional adult and of 0.5 to
each child. Therefore, the equivalized household income can be viewed as an indicator of the
standardized economic resources available to each individual within a household, where the
standardization reflects the economies of scale relevant to the household.
Consistent with the existing literature, a wide variety of demographic, socio-economic and
health variables are included in this study: dummies for age groups, gender and race, employment
status, education, marital status and whether have any untreated medical problem. Community
25
characteristics, in my main analysis, are dummy variables (i.e. dummy for regional center, dummy
for the Bering Straits area, and dummy for the Northwest Arctic Borough) to control for the
geographic and economic differences between regional centers and villages as well as differences
between the three areas. As a robustness check, I control for three community level characteristics,
including measures on community level job opportunities and whether a community is involved
in whaling, which are proved to be important in the literature (Martin, 2012), and the percentage
of occupied housing units that lack complete plumbing facilities, which reflects the condition of
the infrastructure in a community. The descriptive statistics of the variables used can be found in
Table 2-1 and the definition of the key variables can be found in Appendix Table A-1.
Before presenting the findings, it is important to acknowledge that they do not demonstrate
causality; the relations are only associational.

2.5 Results

2.5.1 Basic Models

Following Blanchflower and Oswald (2002), I show, in Table 2-2 the ordered logit estimation
results for traditional happiness equations including demographic and socioeconomic
characteristics, health, income, and location dummies. Column (1) presents the specification
without an income measure. Individuals aged 65 and above are more satisfied with their life than
those in other age groups. In addition, living with a partner (i.e. husband, wife, partner, or

26
Table 2-1 Descriptive Statistics of Variables
Variable Mean Standard
Deviation
Dependent Variable
Life Satisfaction 4.374 0.826
Individual Characteristics
Age Group Dummies (the reference
group is age24 (i.e. respondents aged
below 25))
age2534 0.199 0.4
age3544 0.256 0.437
age4554 0.175 0.38
age5564 0.091 0.287
age65 0.095 0.293
female 0.579 0.494
Yupik 0.113 0.317
Employment Status Dummies (the
reference group is worked full or part
time for pay)
unemployed 0.046 0.21
retired 0.018 0.133
student 0.02 0.141
homemaker 0.074 0.263
other 0.052 0.223
partnered 0.547 0.498
Education Dummies (the reference group
is less than secondary education)
secondary 0.459 0.499
postsecondary 0.292 0.455
untreated medical problem 0.046 0.21
Income Measures Ln (household equivalized income) 10.507 0.701
Ln (household equivalized wage income) 9.284 2.913
Ln (household equivalized transfer
income)
8.753 0.993
number of leisure activities 8.457 2.955
Subsistence Activities number of traditional activities 6.034 3.934
subsistence harvests 0.423 0.26
Religious and Spiritual Beliefs Measures Christian beliefs 0.857 0.35
indigenous beliefs 0.759 0.428
Christian beliefs*indigenous beliefs 0.67 0.471
number of problem with your house 4.06 2.995
types of crimes committed 0.306 0.726
family ties index 12.103 1.936
social support index 29.014 5.4
Community Characteristics
Place Dummies regional center 0.557 0.497
Bering Straits Census Area 0.39 0.488
Northwest Arctic Borough 0.334 0.472
employment per native 0.675 0.284
whaling community 0.286 0.452
percentage of occupied housing units that lack complete plumbing facilities 24.274 29.888
Number of Observations: 497

companion) has a positive and significant influence while having untreated medical problem has
a negative and significant one. These three findings are consistent with the existing literature. The
coefficients of other variables are not significant, but the sign of some coefficients seem to be
inconsistent with the findings in the literature. For instance, female respondents appear to be less
happy, and the “other” group in the nonworking population, which includes those who are out for
hunting and fishing, are likely to be happier than the working population. Regarding the location
dummies, living in a regional center reduces life satisfaction, and so does living in the Bering
Straits or Northwest Arctic regions. These results seem to be intuitive because the employment
opportunities in the regional centers presumably restrict the Inuit from participating in traditional
activities and the Bering Straits area or the Northwest Arctic Borough are less wealthy than the
North Slope Borough.
Table 2-2 Life Satisfaction, Demographic & Socioeconomic Characteristics, Health, and
Income: Ordered Logit Regressions

Dependent Variable: Life Satisfaction
(1) (2) (3)

age2534 0.0996 0.108 0.0766

(0.322) (0.322) (0.323)
age3544 0.144 0.152 0.101

(0.345) (0.344) (0.346)
age4554 0.0183 0.0235 -0.0433

(0.324) (0.325) (0.326)
age5564 0.377 0.424 0.330

(0.370) (0.372) (0.369)
age65 0.883** 0.870** 0.423

(0.432) (0.429) (0.457)
female -0.192 -0.217 -0.211

(0.185) (0.186) (0.184)
Yupik -0.148 -0.162 -0.139

(0.278) (0.279) (0.279)
unemployed -0.195 -0.247 -0.237

(0.416) (0.419) (0.429)
retired -0.954 -0.941 -0.907*

(0.596) (0.611) (0.550)
student 0.0876 0.109 0.183

(0.726) (0.715) (0.720)
28
homemaker -0.128 -0.205 -0.142

(0.347) (0.357) (0.352)
other 0.328 0.290 0.336

(0.411) (0.412) (0.421)
partnered 0.413** 0.424** 0.463**

(0.192) (0.193) (0.191)
secondary 0.211 0.250 0.342

(0.260) (0.263) (0.268)
postsecondary 0.141 0.210 0.343

(0.280) (0.287) (0.297)
untreated medical problem -1.037*** -1.072*** -1.103***

(0.318) (0.319) (0.316)
Ln (household equivalized income)

-0.169

(0.148)

Ln (household equivalized wage
income)

-0.0724*

(0.0397)
Ln (household equivalized transfer
income)

0.139*

(0.0801)
regional center -0.396** -0.333* -0.365*

(0.189) (0.197) (0.189)
Bering Straits Census Area -1.179*** -1.198*** -1.189***

(0.256) (0.256) (0.256)
Northwest Arctic Borough -0.548** -0.551** -0.564**

(0.257) (0.256) (0.257)
Constant cut1 -5.547*** -7.271*** -4.971***

(0.603) (1.614) (0.971)
Constant cut2 -3.899*** -5.623*** -3.324***

(0.430) (1.592) (0.925)
Constant cut3 -2.695*** -4.419*** -2.113**

(0.367) (1.555) (0.915)
Constant cut4 -0.702** -2.421 -0.0949

(0.350) (1.545) (0.922)

Observations 497 497 497
Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

29
In column (2), I include the natural logarithm of equivalized household income, which
appears to have negative, though not significant, influence on life satisfaction. This result is
counterintuitive and inconsistent with what has been observed in the literature. To understand this
puzzle, I decompose the total household income by income sources, including income from sales
of crafts, self-employment, wage, government or other organization, and others. According to
Appendix Table A-2, wage and government or other organization are two main sources of income
for the Inuit in Arctic Alaska: wage is the primary source for households with income higher than
$28,000 (2/3 of the total households) while transfers from government or other organizations is
the most important source for households with income lower than that. In column (3), I include
natural logarithm of equivalized household wage income, the coefficient of which appears to be
negative and significant and natural logarithm of equivalized household transfer income, the
coefficient of which is positive and significant.
The negative effect of wage income seems unusual in the studies of happiness, but most of
the literature is on the epoch-three populations instead of populations having strong ties to epoch
one. For a population that lives on subsistence harvests, the opportunity cost of working in jobs on
the labor market consists of not only enjoying leisure activities but also participating in traditional
or subsistence activities, which both provide food and encourage the connections with nature,
culture, and communities. In order to understand the negative effect of wage income, I look at
several factors that may be correlated with wage income. First, I suspect that households with
higher wage income may purchase more equipment or tools for traditional or leisure activities.
There is indeed a positive and significant correlation (correlation coefficient = 0.100, p-
value=0.026) between wage income and the types of equipment/tools purchased in the past year.
Specifically, with a higher level of household wage income, the Inuit are significantly more likely
30
to purchase a truck, an outboard motor, or a computer. Second, I the types of equipment/tools
purchased are highly correlated with the types of leisure activities (e.g. participating in sports,
snowmobiling, and boating) (correlation coefficient=0.254, p-value= 0.000) as well as the types
of traditional activities (e.g. hunting, fishing, and trapping) (correlation coefficient=0.322, p-
value=0.000). These two points imply that higher wage income may lead to an increase in both the
types of leisure activities and the types of traditional activities through the purchasing of advanced
equipment and tools. I find the correlation between household wage income and types of leisure
activities is also positive and significant (correlation coefficient= 0.306, p-value=0.000) indeed.
Interestingly, those with higher wage income are more likely to participate in sports, boating or
kayaking, and snowmobiling or dog sledding. In addition, the correlation between household wage
income and types of traditional or subsistence activities to be positive and significant as well
(correlation coefficient= 0.150, p-value=0.000). This confirms the interdependence between
earnings and subsistence: higher wage income can make traditional activities more convenient and
efficient through purchasing of advanced equipment or machines.
So far, it seems that higher household wage income benefits the Inuit through an increase
in not only the types of leisure activities but also the types of traditional activities. Why do we
observe a negative effect of wage income? Obviously, an increase in the types of activities
involved does not necessarily mean an increase in spare time for these activities. Wage
employment can absolutely decrease the time for leisure and subsistence harvesting. Even though
the survey does not include any direct question on time allocation, there are some hints that shed
light on this. First, household wage income is negatively (though not significantly) associated with
the number of hours spent watching TV, a common leisure activity. Second, wage income is not
significantly related to the proportion of meat or fish consumed that is from a household’s own
31
harvest even if the number of traditional activities increases with wage income. These findings are
consistent with our suspicion that people working for paid jobs have less time for leisure activities
and less time for hunting and fishing.
To further examine the effect of income, leisure activities and traditional activities on life
satisfaction, I add the types of leisure activities, types of traditional activities, and proportion of
meat or fish consumed that is from a household’s own harvest (a measure on the level of
subsistence harvest)
4
, respectively, into the model (see Table 2-3). Column (1) of Table 2-3 is for
comparison (a copy of column (3) of Table 2-2). The number of leisure activities has a positive
but insignificant effect on life satisfaction (column (2)). In column (3) and (4), both the number of
traditional activities and the level of subsistence harvest have positive and significant influence.
We can observe that the estimated negative effect of wage income becomes larger when leisure
activities, traditional activities, or the level of subsistence harvests are controlled. This means that
the coefficient on wage income in column (1) contains, at least partially, the positive effect of
leisure and traditional activities. The negative effect of wage income after controlling for the
number of traditional activities (in column (3)) also suggest that wage employment might have
reduced the time the Inuit spent connecting with nature, culture and communities and thus create
dissatisfaction. My finding is consistent with the finding of Martin (2012) which shows that a
higher probability of employment is associated with a lower level of life satisfaction.

4
Proportion of meat or fish consumed that is from a household’s own harvest is strongly positively correlated with types of
traditional activities participated (correlation coefficient= 0.482, p-value=0.000)
32
Table 2-3 Understanding the Effects of Wage Income: Ordered Logit Regressions
Dependent Variable: Life Satisfaction
(1) (2) (3) (4)
Ln (household equivalized wage
income)
-0.0724* -0.0735* -0.0812** -0.0780**

(0.0397) (0.0403) (0.0394) (0.0395)
Ln (household equivalized transfer
income)
0.139* 0.138* 0.140* 0.135*

(0.0801) (0.0799) (0.0772) (0.0753)
number of leisure activities

0.00473

(0.0352)

number of traditional activities

0.0741***

(0.0277)

subsistence harvests

1.238***

(0.353)
Constant cut1 -4.971*** -4.946*** -4.587*** -4.500***

(0.971) (0.989) (0.948) (0.933)
Constant cut2 -3.324*** -3.299*** -2.936*** -2.847***

(0.925) (0.947) (0.902) (0.885)
Constant cut3 -2.113** -2.088** -1.717* -1.623*

(0.915) (0.934) (0.890) (0.873)
Constant cut4 -0.0949 -0.0697 0.322 0.426

(0.922) (0.940) (0.899) (0.879)

Observations 497 497 497 497
Notes:

Additional controls (all columns) include: age, gender, race, education, marital status,
employment status, health and place.
Robust standard errors in parentheses

Significance: *** p<0.01, ** p<0.05, * p<0.1

2.5.2 Life Satisfaction and Connections with Culture and Communities

In Table 2-4, additional explanatory variables are included based on the specification in column
(3) of Table 2-3. In columns (1) and (2), I focus on the effect of religious and spiritual beliefs.
Column (1) shows having Christian religious beliefs and having indigenous spiritual beliefs have
positive and insignificant effects. Column (2) adds an interaction term of the two beliefs variables.
Interestingly, the effects of having Christian religious beliefs and having indigenous spiritual
beliefs become negative though not significant by themselves but the effect of their interaction
appears to be positive. It seems that the two belief systems are complementary to each other for
33
the Inuit: it is the combination of the two that makes them happier. This implies the Inuit need to
keep their ancient beliefs and embrace new religious beliefs in order to sustain their happiness.
Columns (3) and (4) include the number of problems with respondent’s house and the
number of categories of crime victimized, respectively. The effects of both are negative but
insignificant. It seems the quality of housing and whether being a crime victim do not matter much
to their happiness.
Column (5) considers an index of family ties which is constructed based on the responses
to three questions on family connections in which higher values representing stronger family ties.
Similar to the family ties index, the social support index included in column (6) is derived from a
series of questions about how often the kinds of support are available when needed. Both indices
exert significant and positive influences on life satisfaction. In column (7), I include family ties
index and social support index in the same model. Because of the strong positive correlation
between two indexes, the positive effect of family ties index becomes insignificant after the control
of social support index. This suggests that the effect of family connections on life satisfaction is
mainly due to the support available. The positive effect of transfer income becomes insignificant
after either the family ties index or the social support index is added. In addition, the positive effect
of being partnered also becomes insignificant when the social support index is controlled for.
These findings from columns (5) to (7) suggest the importance of the availability of social support
to the life satisfaction of the Inuit.
Columns (8) is considered as a relatively complete specification, where all the variables in
the previous columns are included. The results are consistent with those in the previous columns,
except that the effect of having both Christian religious beliefs and indigenous spiritual beliefs
becomes more significant.
34

2.5.3 Robustness Checks

To make sure the finding on wage income or traditional activities isn’t just because of the
availability of the job opportunities, the extensiveness of the traditional activities, or the living
standard of a community, I run the same models but replacing the dummies for regional center,
the Bering Strait region and the Northwest Arctic Borough with community characteristics,
including employment opportunities per native aged 15 and older, whether a community
participates in whaling, and the percentage of occupied housing units that lack complete plumbing
facilities. To be concise, in Table 2-5, I only present the results for the specifications that are
comparable to column (3) of Table 2-3 and column (8) of Table 2-4. The effects of individual level
variables are generally robust to the control of community characteristics. More employment
opportunities per native is found to have a negative and significant effect on life satisfaction. Being
in a whaling community increases the life satisfaction of the Inuit. A higher percentage of occupied
housing units that lack complete plumbing facilities, which represents worse conditions with
respect to infrastructure, reduces their life satisfaction.

35
Table 2-4 Life Satisfaction and Connections with Culture and Community: Ordered Logit Regressions
Dependent Variable: Life Satisfaction
(1) (2) (3) (4) (5) (6) (7) (8)
Ln (household equivalized wage income) -0.0782** -0.0764* -0.0797** -0.0825** -0.0851** -0.105** -0.106** -0.101**

(0.0399) (0.0411) (0.0396) (0.0398) (0.0394) (0.0419) (0.0419) (0.0449)
Ln (household equivalized transfer income) 0.144* 0.143* 0.136* 0.137* 0.126 0.130 0.128 0.128

(0.0780) (0.0796) (0.0783) (0.0775) (0.0811) (0.0811) (0.0810) (0.0853)
number of traditional activities 0.0682** 0.0653** 0.0729*** 0.0746*** 0.0647** 0.0598** 0.0586** 0.0515*

(0.0280) (0.0283) (0.0277) (0.0276) (0.0278) (0.0281) (0.0282) (0.0289)
Christian beliefs 0.290 -0.226

-0.531

(0.261) (0.443)

(0.454)
indigenous beliefs 0.0944 -0.552

-0.853

(0.220) (0.502)

(0.529)
Christian beliefs*indigenous beliefs

0.804

1.048*

(0.555)

(0.580)
number of problem with your house

-0.0296

-0.0231

(0.0309)

(0.0311)
types of crimes committed

-0.0579

0.0242

(0.139)

(0.163)
family ties index

0.101**

0.0174 0.0222

(0.0481)

(0.0518) (0.0525)
social support index

0.0904*** 0.0879*** 0.0883***

(0.0207) (0.0220) (0.0226)
Constant cut1 -4.314*** -4.707*** -4.737*** -4.638*** -3.750*** -2.675** -2.584** -3.029**

(0.996) (1.039) (0.974) (0.938) (1.067) (1.117) (1.148) (1.263)
Constant cut2 -2.662*** -3.052*** -3.085*** -2.987*** -2.098** -1.016 -0.924 -1.363

(0.948) (1.009) (0.940) (0.910) (1.037) (1.077) (1.118) (1.268)
Constant cut3 -1.441 -1.827* -1.865** -1.768** -0.879 0.216 0.307 -0.125

(0.936) (1.003) (0.927) (0.899) (1.031) (1.066) (1.111) (1.266)
Constant cut4 0.604 0.224 0.178 0.272 1.170 2.321** 2.412** 1.993

(0.946) (1.012) (0.936) (0.909) (1.050) (1.088) (1.137) (1.292)

Observations 497 497 497 497 497 497 497 497
Additional controls (all columns) include: age, gender, race, education, marital status, employment status, and place.
Robust standard errors in parentheses

Significance: *** p<0.01, ** p<0.05, * p<0.1

Table 2-5 Robustness Check (Controlling for Community Characteristics): Ordered Logit
Regressions
Dependent Variable: Life Satisfaction
(1) (2)
Ln (household equivalized wage income)
-0.0650* -0.0818*

(0.0384) (0.0438)
Ln (household equivalized transfer income) 0.138* 0.122

(0.0738) (0.0803)
number of traditional activities 0.0623** 0.0472*

(0.0271) -0.028
Christian beliefs

-0.595

(0.452)
indigenous beliefs

-0.942*

(0.530)
Christian beliefs*indigenous beliefs

1.136**

(0.579)
number of problem with your house

-0.0331

(0.0313)
types of crimes committed

0.0231

(0.157)
family ties index

0.0226

(0.0528)
social support index

0.0823***

(0.0230)
employment per native -1.421*** -1.645***

(0.407) (0.409)
whaling community 0.803*** 0.867***

(0.227) (0.229)
percentage of occupied housing units that lack complete
plumbing facilities -0.00908** -0.00867**

(0.00407) (0.00416)
Constant cut1 -4.877*** -3.518***

(1.003) (1.309)
Constant cut2 -3.227*** -1.856

(0.964) (1.315)
Constant cut3 -2.012** -0.628

(0.947) (1.304)
Constant cut4 0.0105 1.467

(0.958) (1.331)

Observations 497 497
Additional controls (all columns) include: age, gender, race, education, marital status,
employment status.
Robust standard errors in parentheses

Significance: *** p<0.01, ** p<0.05, * p<0.1

37
2.6 Conclusion

This study investigates what individual and household characteristics explain life satisfaction
among the Inuit population in the Arctic Alaska. Both a baseline analysis and a robustness check
suggest the factors that are positively associated with life satisfaction of the Inuit are health,
participating in traditional activities, social support and having both Christian religious beliefs and
indigenous spiritual beliefs as part of life. One interesting and novel finding is the negative effect
of the wage income. This implies that the time and effort spent working might have prevented the
Inuit from participating in traditional harvest and maintaining connections with nature, culture and
community. The implication can be validated by the positive and significant effect of the number
of traditional activities in which they have participated and the negative and significant effect of
employment opportunities per native adult.
The negative effect of wage income supports the policy recommendations of others (Berkes
et al., 1995; Martin, 2012). The continued essentiality of subsistence harvest or traditional
activities and the discomfort from employment mean that the policy makers should empower the
indigenous people to discuss and plan the future of their regions. The importance of having both
Christian religious beliefs and indigenous spiritual beliefs as part of life suggests the Inuit need to
keep their ancient spiritual beliefs along with the Christian beliefs in order to sustain their
happiness. Future studies may investigate the effects of state or local policies and the abundance
of natural resources on life satisfaction of the population. In addition, since men and women
specialize in different types of traditional activities as well as different types of employment, future
analysis could usefully be conducted for males and females separately to explore gender
heterogeneity.

38

CHAPTER 3. The Effect of Intergenerational Support on the Life Satisfaction of Older
Parents in China
3.1 Introduction

In 2010, the percentage of people aged 60 or over in China was 12.3%, and it is expected to rise
to 30.5% by the early 2040s. Because the population in China is aging so rapidly, scholars have
been increasingly interested not only in the systems by which older parents are being supported,
but in their sense of well-being within those systems.
Traditionally, the family, particularly adult children, serves as the primary source of
support. Family responsibility for taking care of the elderly is not only a right protected by law but
also a tenet of Confucianism (Davis-Friedmann, 1991). The absence of a well-established pension
system further necessitates that adult children take care of their parents.
However, the degree to which the well-being of older parents relies on their children might
differ according to whether the parents are rural or urban
5
, because they have faced quite different
economic, policy and social constraints in the past few decades. Economic reforms since 1978,
which have included the opening up of the country to foreign investment and the introduction of
free market principles, have revitalized the economy and created job opportunities in cities. The
reforms, however, have widened the income gap between urban and rural areas. According to most

5
Because of the household registration system, there are actually two ways to distinguish rural from urban population in China:
the actual place of residence and the hukou registered place. Hukou registration system officially identifies a person as a
resident of an area. Following the two definitions on the categorization of rural and urban populations, there should be four
subpopulations in total: urban natives (urban residents with urban hukou), rural natives (rural residents with rural hukou), rural-
to-urban migrants (urban residents with rural hukou), and urban-to-rural migrants (urban residents with rural hukou). The social
environment and income levels largely depend on the place of residence while the implementation of public policies, such as
pension schemes and family planning policies, mostly depend on one’s hukou registration status. In the baseline analysis, the
study will distinguish “urban” from “rural” according to whether the respondents are living in urban communities or rural
villages (the place of residence).
39
estimates, mean per capita income of urban residents is approximately three times as much as that
of rural residents (Knight & Gunatilaka, 2010).
The opening up of the country not only attracted foreign investment but also stimulated the
spread of western values. Since the influence of western culture, which places more emphasis on
independence and privacy, was greater in large cities than in remote areas, the traditional value of
filial piety has been better preserved in rural areas.
In addition to the income disparity and differences in values between rural and urban
residents, the current pension scheme for urban employees is far more generous than that for rural
residents. Since the founding of the PRC in 1949, old-age support in rural areas has relied on the
elderly’s own labor income and family support. It was not until 2009 that the government
introduced a new pension scheme to formally address the needs of this population
6
. But the current
amount
7
is still too low to guarantee a basic standard of living. By contrast, urban employees can
count on approximately 40-50% of their previous income for support when they retire.
Another difference between rural and urban population
8
pertains to family planning
policies. The implementation of family planning policies can be divided into three periods. The
implementation began in the early 1960s
9
, after which the policies became progressively more
restrictive (Wang, 2014)
10
. Because manpower is crucial in the agricultural sector, a larger number
of children (especially males) is preferred. Hence, these family planning policies were always more
restrictive for urban than for rural people (Wang, 2014). For example, the one-child policy allowed

6
National rural pension pilot (New Rural Social Pension Scheme) was announced in 2009 and started in late 2009 with an aim to
achieve full geographic coverage no later than 2013 (State Council 2009b).
7
The basic pension level is 55 yuan (less than $10) per month, which is below the rural poverty line.
8
Here, “rural” and “urban” refer to the household registration status rather the place of residence.
9
The three periods: the period with mild and narrowly implemented policies, the period with strong and widely enforced
policies and the period with the harshest one-child policy.
10
Family planning policies in China are primarily designed for Han families and non-Han families are covered in more relaxed
forms.
40
a married urban couple to have only one child, whereas their rural counterparts, under certain
conditions, were allowed to have another. Therefore, rural parents generally had more children to
rely on.
Despite the many differences between rural and urban populations, the trend of living
separately from children has been observed in both areas. Traditionally, older Chinese parents live
with a child, usually the oldest son, who continues to live in the parental household even after
marriage and the birth of children. Although this kind of stem-family household is still the most
common family structure in China, the number of non-traditional households has grown
substantially (Zeng & Wang, 2003). These include several new types of arrangements: empty nest
(grandparents living by themselves), skip-generation (grandparents living with grandchildren) and
network households (children and older parents living separately but nearby) (Chen & Silverstein,
2000). Even though it has become more likely for older parents in both rural and urban areas to
live separately from their children, the dynamics guiding the living arrangements are quite
different. In rural areas, young adults have moved to cities to pursue jobs that have been created
by the economic reforms (Goldstein et al., 1997). This has resulted not only in an increased
geographic separation between adult children and their parents but also in an increased number of
grandchildren left behind with their grandparents. On the contrary, in urban areas, the increasing
availability of pensions and housing as well as the influence of western attitudes has encouraged
older parents to live separately from children (Palm & Deng, 2008; Meng & Luo, 2008; Sheng,
2005; Zeng & Wang, 2003).
It is still uncertain whether this trend of living away from their children has adversely
affected the well-being of the older parents. Most relevant studies have not been optimistic about
it, and suggesting that living away from children may limit the support and care provided for older
41
parents (Zimmer & Kwong, 2003; Sun, 2002). Studies have also found co-residence with adult
children may directly protect the psychological well-being of the older parents, especially for the
widowed. (Wang et al., 2013; Chou et al., 2006; Silverstein et al. 2006). In addition, functional
support, which includes financial, emotional and instrumental support, has proved to be a channel
through which co-residence may benefit older parents indirectly (Chen & Silverstein, 2000;
Silverstein et al. 2006; Krause & Liang, 1993).
Another viewpoint less discussed in the literature is that living alone and getting support
from the family are not mutually exclusive. This implies that living away from children does not
necessarily mean not getting help from them. For example, Strauss et al. (2011), using data from
China Health and Retirement Longitudinal Study (CHARLS), showed that children living nearby
visited their parents more often while children living far away provided a larger amount of net
transfers. From this point of view, living alone, which induces a large amount of functional support
from children, may not be detrimental to the well-being of older parents. In addition, older parents
may prefer living separately from their children because they value privacy (Doty, 1986; Martin,
1989; Kotlikoff & Morris, 1990; Mutchier & Burr, 1991). A few studies on China found that
elderly people living without their children enjoyed higher levels of psychological well-being due
to their independence and lack of intergenerational conflicts (Yang & Chandler, 1992; Zhou &
Qian, 2008).
Most of the studies on the effects of intergenerational support in China have mainly focused
on psychological well-being of older parents. The purpose of this study is to examine the effects
of various sources of intergenerational support, including family size, gender composition of
children, living arrangements, and functional (i.e. financial, instrumental, and emotional) support,
on the life satisfaction of older parents in China.
42
Traditionally, older parents with more children are thought to be more content with their
life. This is not only because of the functional benefits their children provide but also because of
the cultural norms. As a popular Chinese proverb says, “More children bring greater happiness.”
With respect to their children’s gender, the presence of a son is particularly important for old
parents because of the cultural traditions and economic benefits. However, recent studies on China
have shown that daughters are valued more than before (Gu et al. 1995; Yang, 1996; Zeng et al.,
2016). I expect that, without any family planning, the life satisfaction of rural parents should be
more sensitive to family size and gender composition than urban parents since the traditional
values are more prevalent and manpower is more important in rural areas. However, with stronger
family planning policy in cities, it is possible that family size and gender composition have stronger
influences on the life satisfaction of urban parents. In addition, because rural parents tend to adhere
more to cultural norms, it’s more likely that they will be happier with traditional household
structures and support systems. This is reinforced by the weaknesses of the pension system in rural
areas. However, in cities, the parents may not have as strong a preference for traditional households
and may depend less on the support of their offspring because family size is smaller, income is
higher, pensions are more comprehensive and the influence of western values is stronger. My
empirical analysis will clear the doubts raised in this paragraph.
There are, in general, three problems, with the existing studies on the effects of
intergenerational support on life satisfaction in China: First, to my knowledge, only one study (i.e.
Silverstein et al., 2006) in this literature looked into the proximity to children if older parents live
without children. With the rising trend in living alone accompanied with a rise in living closer to
children, it will be meaningful to investigate whether the proximity to children plays a role on
subjective well-being. Second, studies on a national sample haven’t looked at urban and rural
43
populations separately. The preferences on intergenerational support can be different between rural
and urban residents because of the differences in economics, public policies, and social attitudes
as discussed in the previous section. Therefore, studying China’s population as a whole, which is
a common practice in this literature, blurs an investigation. Last but not least, all of the existing
studies have used cross-sectional data, the analysis on which some time-invariant individual
characteristics, which can be correlated with both intergenerational support and one’s life
satisfaction, are unable to be controlled. With these three problems in mind, I try to provide a more
comprehensive analysis in this research.
Specifically, the study makes two contributions to the existing literature. First, I examine
two aspects of living arrangements: co-residence and proximity to children. Second, this is one of
the first studies to investigate the relationship between life satisfaction and family support in rural
and urban China separately to compare the results for the two populations.

3.2 Literature Review

The studies on the relationship between intergenerational support and life satisfaction in China
have reached different conclusions especially because they look at different types of samples and
use different measures of living arrangements.
With respect to family size, Chen and Short (2008), using a national probability sample of
individuals aged 80 or older in China, finds the number of children has no obvious effect on
subjective well-being. Silverstein et al. (2006), studying parents aged 60 and older living in rural
Anhui Province, finds that the number of children has a positive influence on life satisfaction.
44
Regarding the gender composition of children, Zhang and Liu (2007) finds that, for Chinese aged
65 and above, it has no significant influence on life satisfaction.
Different studies have used different definitions on measures of living arrangements as
well. What has been ascertained is that living alone is associated with a lower level of life
satisfaction for elderly Chinese parents (Chen & Short, 2008; Zhang, 2015; Silverstein et al., 2006;
Ren & Treiman, 2015). Many studies have compared co-residence with other types of living
arrangements using various waves of Chinese Longitudinal Healthy Longevity Survey. For
example, Zhang and Liu (2007) shows that, for Chinese aged 65 and above, co-residing with
household members is associated with a lower level of life satisfaction than living in a nursing
home but that living alone is associated with the lowest level of life satisfaction; Chen and Short
(2008) finds that living with spouse or children is associated with higher subjective well-being
than living alone for the oldest old (aged 80 years or above); and Wang et al. (2014), studying the
same age group, finds that co-residence with children improves life satisfaction of widowed but
this does not hold for married people. However, little is known on the consequences for living in
a skip-generation household. Silverstein et al. (2006) shows that, for the rural residents, living in
a skip-generation household is as life satisfying as living in a three-generation household.
With respect to the receipt of intergenerational functional support, receiving financial
support and emotional support has been found to be beneficial to life satisfaction of older parents
(Chen & Short, 2008; Silverstein et al. 2006). However, less is known about the effects of
providing support for children. On one hand, providing support could also be a burden for a support
giver, but on the other hand, it could be a source of self-efficacy and social approval. Schwarz
(2010) finds that, for elderly mothers, the effects of providing help for adult daughters vary
according to cultural contexts. Even though, traditionally, adult children are responsible for caring
45
older parents in China instead of older parents supporting adult children, the country has undergone
rapid social changes with a profound effect on parent-child relationship. For instance, according
to Chen and Silverstein (2000), the parent-child relationship has become less hierarchical. This
evolving parent-child relationship makes it interesting to examine the effect of providing
functional support on life satisfaction of older parents in contemporary China.

3.3 Conceptual Framework

The model of this study is an extension of conceptual framework of Blanchflower and Oswald
(2004) and Frey and Stutzer (2010).
self-reported life satisfaction = h(u(demographic characteristics, socioeconomic
characteristics, health, support, natural and social environment, personality, and genes, t)) + e
(4)

where u(…) is defined as a person’s true well-being or utility, h(…) is a continuous non-
differentiable function relating actual to reported well-being, t is the time period, e is an error term.
It is assumed that u(…) is a function that is only observable to the individual. The error term, e,
incorporates among other factors the inability of individuals to convey their satisfaction level.
According to this expression, self-reported life satisfaction can be influenced by demographic
characteristics (e.g. age, gender, and race), socioeconomic characteristics (e.g. income, education,
marital status, and employment status), health, support from and to others (e.g. family and social
support), natural and social environment, personality and genes, and reporting errors. Figure 1
illustrates my model in detail.
46
Figure 3-1 Conceptual Framework

Even though there are studies that look at the direct effect of natural or social environment
on life satisfaction, in general, this study models them as the factors that can affect it only indirectly
by shaping the influence of individual characteristics on life satisfaction. For instance, an
individual’s happiness will be less influenced by family support if he or she lives in a place with
more comprehensive welfare policies. In this case, it is reasonable to examine rural and urban
population separately given the sharp differences in economic development, public policies and
social attitudes between them.
47
As highlighted in Figure 1, this study focuses on the relationship between life satisfaction
and intergenerational support, which is one dimension of family support. There are two general
types of such support: structural and functional support. Structural support contains the
composition of the social network and the availability of people in the network who may help the
individual (Chen & Silverstein, 2000). When applied to older parents and their children, structural
support includes the number of children, their gender, and intergenerational living arrangements.
Intergenerational living arrangements, including co-residence with children and the proximity to
children, reflect traditional values of a society and influence the exchanges of actual or functional
support. Functional aspects of support are divided into instrumental, financial, and emotional
forms of support. This study will examine both the reception and provision of functional support.
Instrumental support includes personal help such as help with personal care and household chores,
and taking care of grandchildren. Financial support includes money and in-kind support.
Emotional support reflects intimacy, trust, and confidences with others. In my model, structural
support influences life satisfaction not only indirectly through functional support but also directly
(see Figure 1 for graphical illustration). In other words, structural support can influence life
satisfaction beyond functional benefits and costs: this relationship contains the elements of culture,
social values, and attitudes.

3.4 Methodology

The general specification used is as follows:

𝐿 𝑆 𝑖𝑗𝑘𝑡 = 𝛼 ′𝑆 𝑗𝑡
+ 𝛽 ′𝐹 𝑖𝑡
+ 𝛾 ′𝑋 𝑖𝑡
+ 𝛿 ′𝑍 𝑖 + 𝜇𝑊 2
𝑡 + 𝐶𝑜𝑚𝑚𝑢𝑛𝑖𝑡𝑦 𝑘 + 𝘀 𝑖𝑗𝑘𝑡 (5)
48

𝐿 𝑆 𝑖𝑗𝑘𝑡 represents life satisfaction of individual i of household j in community k at time t.
𝑆 𝑗𝑡
is a vector of structural support variables (i.e. family size, gender composition of children, and
living arrangements) for household j at time t, 𝐹 𝑖𝑡
is a vector of functional support variables for
individual i at time t, 𝑋 𝑖𝑡
is a vector of time-varying individual characteristics, 𝑍 𝑖 is a set of time-
invariant individual characteristics, 𝑊 2
𝑡 is a dummy variable indicating wave 2. 𝐶𝑜𝑚𝑚𝑢𝑛𝑖𝑡𝑦 𝑘
is an unobserved random variable which includes all the time-invariant factors of community k
that may influence life satisfaction. And, 𝘀 𝑖𝑗𝑘𝑡 is the error term.
I estimate (1) using ordinary least squares (OLS) estimation. This is similar to the common
practice in the existing literature. However, estimators from OLS estimation may be biased if it
fails to take into account the potential correlation between unobserved individual characteristics
and observables in the model. So, I rewrite equation (1) as

𝐿 𝑆 𝑖𝑗𝑘𝑡 = 𝛼 ′𝑆 𝑗𝑡
+ 𝛽 ′𝐹 𝑖𝑡
+ 𝛾 ′𝑋 𝑖𝑡
+ 𝛿 ′𝑍 𝑖 + 𝜇𝑊 2
𝑡 + 𝜃 𝑖 + 𝑢 𝑖𝑗𝑘𝑡 (6)
11

𝜃 𝑖 is an unobserved random variable including all the time-invariant factors of individual i
that can influence life satisfaction but are not observable. In this case, it includes not only
individual characteristics such as personality and genes but also time-invariant unobserved
household and community level characteristics because all individuals remain in the same
household and community in two waves. 𝜃 𝑖 is potentially correlated with observed regressors. I
use fixed-effects estimation method to estimate equation (2). Fixed-effects estimators is also called

11
Because no individual move across communities between the two waves, the community fixed effects 𝐶𝑜𝑚𝑚𝑢𝑛𝑖𝑡𝑦 𝑘 are
included in the individual fixed effects 𝜃 𝑖 .
49
within estimator or variance estimator because it only makes use of the difference within
individuals.
Given the advantage of the fixed-effects estimation, however, I still use OLS estimation
for the main analysis and fixed-effects estimation as a robustness check. The primary reason is that
most of the intergenerational support variables do not vary much within 2 years for an individual.
Even though the fixed-effects estimator is able to reduce endogeneity bias, I may get a lot of
insignificant estimates simply due to the lack of variations between the 2 waves of data. Another
limitation of fixed-effects estimation is that the coefficients of observable time-invariant individual
characteristics (i.e. 𝛿 ) cannot be estimated.

3.5 Data and Measurement
3.5.1 Data

The data for this study come from the first two waves of the China Health and Retirement
Longitudinal Study (CHARLS). It is part of a set of longitudinal aging surveys that include surveys
in the United States, England, nineteen countries in continental Europe, Korea, Japan, and India,
focusing on the mid-aged and elderly aged 45 or over and their spouse in China
12
. The sample in
the dataset is nationally representative
13
and is chosen through multi-stage probability sampling.
It covers 450 villages/urban communities in 150 counties/districts of 28 provinces across the
country. The first wave, CHARLS national baseline, was fielded from June 2011 to March 2012.

12
Institutionalized mid-aged and elderly are not sampled, but wave 1 respondents who later enter into an institution will be
followed.
13
Tibet is excluded
50
It contains 17, 708 individuals in 10, 257 households. The second wave was fielded in 2013. It
contains 18, 648 individuals, among which 15, 684 are respondents from the 2011 baseline sample.
Since I focus on the effect of intergenerational support on life satisfaction of older parents
in China, my sample consists of individuals with data on life satisfaction, number and gender
composition of their children, living arrangements, functional support, and other individual
characteristics available. This leaves me with 5,497 respondents, who were aged 40 to 91 at the
baseline survey and were interviewed in both waves. They all have biological or adopted children
alive, non-coresident children
14
, and grandchildren under 16
15
. About 90 percent of my sample
were below 30 years old when the family programs were established in the 1970s. Besides, more
than half of the sample were below 30 years old when the strict one-child policy was implemented
in the 1980s. Therefore, the sample should be strongly influenced by the family planning policies.
In the baseline analysis, the study will distinguish “urban” from “rural” according to
whether the respondents are living in urban communities or rural villages (the place of residence).
I will call them “urban residents” or “rural residents”, respectively. Urban-to-rural migrants (i.e.
those who have urban hukou but live in rural villages) are not common, so I don’t examine them
separately from other rural residents. Rural-to-urban migrants (i.e. those who have rural hukou but
live in urban communities), however, do account for 1/3 of the urban residents and, in general, do
not share the same sets of public policies with urban natives. They are not investigated separately
from urban natives in the baseline analysis because the size of this population is not large in my
sample. I will study urban natives and rural-to-urban migrants separately as a robustness check.

14
Only the respondents with non-coresident children were asked on financial transfers and frequency of seeing children.
15
Only the respondents with grandchildren under 16 were asked about whether taking care of grandchildren or not.
51
3.5.2 Measurement
3.5.2.1 Measurement of Life Satisfaction

The CHARLS survey measures life satisfaction from the following question:

Please think about your life-as-a-whole. How satisfied are you with it? Are you completely
satisfied, very satisfied, somewhat satisfied, not very satisfied, or not at all satisfied?

The variable takes value 5 if an individual chooses “completely satisfied” and takes value 1 if an
individual chooses “not at all satisfied”.
In 2011, the average life satisfaction for the whole sample was 3.071 with about 23 percent
of the respondents choosing very satisfied or completely satisfied. In 2013, the average life
satisfaction increased a little to 3.134 with about 26 percent of the respondents choosing the top
two responses. Life satisfaction of urban and rural residents were quite similar on average with
urban residents enjoying a slightly higher level in both waves. The summary statistics of both
variables of interest and control variables by wave and place of residence can be found in Table 3-
1.

3.5.2.2 Measurement of Structural and Functional Support

The explanatory variables that on which I focus are intergenerational support measures, one aspect
of which is structural. Structural support measures include the number of children (biological or

52
Table 3-1 Summary Statistics
Panel A Whole Rural Urban

W1: life satisfaction 3.071 3.061 3.123

W2: life satisfaction 3.134 3.127 3.175

W1: number of observations 5497 4629 868

W2: number of observations 5497 4629 868
Panel B Whole Rural Urban
Structural Support Measures

W1: live with children and grandchildren (reference
group)
0.313 0.311 0.326

W2: live with children and grandchildren (reference
group)
0.271 0.268 0.282

W1: live with children only 0.194 0.201 0.152

W2: live with children only 0.155 0.157 0.139

W1: live without children and without
grandchildren*closest child lives in same county/city
0.333 0.326 0.372

W2: live without children and without
grandchildren*closest child lives in same county/city
0.368 0.363 0.396

W1: live without children and without
grandchildren*closest child lives outside same county/city
0.05 0.051 0.046

W2: live without children and without
grandchildren*closest child lives outside same county/city
0.062 0.064 0.053

W1: live with grandchildren only*closest child lives in
same county/city
0.067 0.064 0.084

W2: live with grandchildren only*closest child lives in
same county/city
0.1 0.099 0.104

W1: live with grandchildren only*closest child lives
outside same county/city
0.043 0.047 0.02

W2: live with grandchildren only*closest child lives
outside same county/city
0.045 0.048 0.025

W1: living with parents 0.046 0.049 0.029

W2: living with parents 0.01 0.009 0.016

W1: number of children 3.011 3.078 2.653

W2: number of children 3.06 3.121 2.737

W1: have one child (reference group) 0.052 0.042 0.106

W2: have one child (reference group) 0.046 0.036 0.1

W1: have two children 0.351 0.334 0.445

W2: have two children 0.346 0.329 0.437

W1: have three children and above 0.597 0.624 0.449

W2: have three children and above 0.608 0.635 0.463

W1: have daughters only (reference group) 0.112 0.102 0.168

W2: have daughters only (reference group) 0.098 0.087 0.152

W1: have sons only 0.164 0.155 0.214

W2: have sons only 0.148 0.139 0.195

W1: have sons and daughters 0.724 0.744 0.618

W2: have sons and daughters 0.755 0.774 0.653
Functional Support Measures

W1: don’t need help with ADLs/IADLs (reference group) 0.867 0.857 0.924

W2: don’t need help with ADLs/IADLs (reference group) 0.848 0.841 0.883

W1: need help but no one helps 0.026 0.027 0.02

W2: need help but no one helps 0.029 0.029 0.032

W1: need help and children help 0.017 0.018 0.008

W2: need help and children help 0.036 0.038 0.029

W1: need help and others help 0.09 0.098 0.048

W2: need help and others help 0.087 0.092 0.056

W1: whether receive financial transfers from non-
coresident children
0.448 0.466 0.354

W2: whether receive financial transfers from non-
coresident children
0.834 0.847 0.765
53

adopted) dummies, gender composition of children (i.e. whether having sons only, daughters only,
or both sons and daughters), and living arrangements. The average number of children is lower in
urban areas (2.65 at baseline) than in rural areas (3.08 at baseline). Specifically, around 55 percent
of urban parents have no more than 2 children compared with 38 percent for rural parents. Besides,

W1: whether provide financial transfers for non-
coresident children
0.086 0.079 0.122

W2: whether provide financial transfers for non-
coresident children
0.272 0.264 0.317

W1: ln(max financial transfer received) 3.444 3.552 2.871

W2: ln(max financial transfer received) 6.543 6.576 6.365

W1: ln(max financial transfer provided) 0.726 0.657 1.099

W2: ln(max financial transfer provided) 2.094 1.975 2.724

W1: whether see non-coresident children more than once
a month
0.701 0.68 0.809

W2: whether see non-coresident children more than once
a month
0.733 0.714 0.836

W1: whether take care of grandchildren under 16 0.409 0.401 0.449

W2: whether take care of grandchildren under 17 0.514 0.498 0.603
Demographic & Socioeconomic Characteristics

W1: age49 0.113 0.119 0.085

W2: age49 0.057 0.06 0.038

W1: age5059 0.426 0.436 0.376

W2: age5059 0.37 0.38 0.318

W1: age6069 0.368 0.359 0.414

W2: age6069 0.435 0.428 0.47

W1: age7079 0.087 0.081 0.12

W2: age7079 0.127 0.121 0.159

W1: age80 0.005 0.005 0.006

W2: age80 0.011 0.011 0.015

W1: male 0.459 0.462 0.444

W2: male 0.459 0.462 0.444

W1: Han nationality 0.928 0.931 0.909

W2: Han nationality 0.928 0.931 0.909

W1: married 0.908 0.91 0.899

W2: married 0.889 0.891 0.879

W1: illiterate (reference group) 0.293 0.323 0.134

W2: illiterate (reference group) 0.278 0.308 0.125

W1: some primary education 0.453 0.467 0.38

W2: some primary education 0.457 0.471 0.38

W1: some secondary education 0.254 0.21 0.486

W2: some secondary education 0.265 0.221 0.495

W1: ln(per capita expenditure) 9.112 9.026 9.571

W2: ln(per capita expenditure) 9.112 9.026 9.571

W1: whether have any ADL or IADL difficulty 0.133 0.144 0.076

W2: whether have any ADL or IADL difficulty 0.152 0.159 0.118

W1: working 0.724 0.802 0.306

W2: working 0.687 0.752 0.34

W1: whether participate in or receive any pension 0.802 0.801 0.803

W2: whether participate in or receive any pension 0.802 0.801 0.803

W1: age of respondents 59.166 58.938 60.382

W2: age of respondents 61.166 60.938 62.382

W1: mean(age of repondent's children) 32.384 32.187 33.434

W2: mean(age of repondent's children) 34.825 34.631 35.848
54
a higher proportion of those who live in urban areas have only daughters or sons though those
living in rural areas are more likely to have both sons and daughters.
With detailed information on household composition and proximity to children in
CHARLS, I construct two sets of living arrangements measures: co-residence (i.e.
intergenerational structure of a household) and proximity to children. There are four types of co-
residence: (a) lives with children and grandchildren; (b) lives with children only; (c) lives with
grandchildren but not with children; and (d) lives with neither children nor grandchildren. There
are three types of proximity to children: (a) lives with children; (b) lives without children, but at
least one child lives in the same county/city; and (c) lives without children, and all children live
beyond the same county/city. Since the interactions between these two sets of variables may also
influence the exchanges of functional support as well as life satisfaction, I interact the two sets of
variables to create six categories of mutually exclusive living arrangements, and the traditional
three-generation household serves as the reference group in the regression analysis. I control for
whether a respondent is living with parents or parents-in-law but this is not the focus of the study
16
.
For the country as a whole, in wave 1, a majority (38 percent) of the older parents live
without children and grandchildren, 31 percent of the older parents live with both children and
grandchildren, 19.4 percent live with children only, and 11 percent live with grandchildren only.
Two years later, the percentage of one-generation households and skip-generation households
increase to 43 percent and 14.5 percent, respectively, but the percent of three-generation or two-
generation households decrease. In general, I observe that both rural and urban residents become
more likely to live in non-traditional households as they age.

16
For the whole sample, no more than 5 percent of the older parents lived with their parents or parents-in-law in wave 1. The
percentage declines to 1 percent in wave 2.
55
Comparing between the two populations, I also find some differences: The proportion of
rural residents living in a skip-generation household is higher. In addition, around 42.3 percent of
rural skip-generation households have the closest child beyond a county or city while it is only
29.2 percent for their urban counterparts. These imply that rural migrants are more likely to leave
their children behind with grandparents, especially when they have long-distance migration. On
the contrary, there is a higher percentage of single-generation households in urban areas due to a
larger proportion of network households (i.e. children and parents live separately but in the same
county or city). Furthermore, as expected, traditional households (i.e. two- or three-generation
households) are more prevalent in rural areas.
In addition to structural support, functional support, which includes financial, instrumental,
and emotional support, is examined (the summary statistics can be found in Table 3-1 as well). I
consider older parents as both receivers and providers of such support. Financial exchange
measures are whether the parents received or provided any money and in-kind support (from or
for non-coresident children) in the past year and the natural logarithm of the maximum amount
among the support (to or from all non-coresident children). In my sample, rural older parents were
more likely to receive financial support from non-coresident children (also with a larger transfer
amount) than urban older parents. The situation is the opposite with respect to providing financial
support for non-coresident children. Instrumental support measures include whether children or
others provided help with activities of daily living (ADL
17
) or instrumental activities of daily
living (IADL
18
) when needed and whether the respondent took care of grandchildren under 16 in
the past year. For help with ADL or IADL, four dummies are created: no difficulty with ADL or

17
ADL include dressing, bathing or showering, eating, getting into or out of bed, using the toilet, or controlling urination
and defecation.
18
IADL include doing household chores, preparing hot meals, shopping for groceries, managing your money or taking
medications.
56
IADL (reference group), have difficulty with ADL or IADL and children help, have difficulty with
ADL or IADL and others help, and have difficulty with ADL or IADL and no one helps. For both
rural and urban samples, the major source of instrumental help when needed is others not children.
The percentage of urban residents taking care of grandchildren under 16 is higher than that of rural
residents in both waves. Since no direct measure on intimacy and trust with others is available in
the survey questionnaire, the frequency of seeing children is considered as the exchanges of
emotional support between older parents and their children in this study. The highest frequency
among all non-coresident children is recorded as the frequency of seeing children for each
respondent. The variable is constructed as a dummy on whether seeing children at least once a
month. I find urban older parents see their non-coresident children more frequently than their rural
counterparts.

3.5.2.3 Covariates

I also include a set of individual and household characteristics as control variables, including age,
gender, nationality, marital status, per capita household expenditure, education, whether having
any ADL/IADL difficulty, work status, and whether participating in or receiving any pension (The
summary statistics of the covariates can also be found in Table 3-1.). Age group dummies are used
to better capture the non-linear effect of age. Nationality is measured by a dummy on whether the
respondent is of Han Nationality. Marital status indicator is dummy which takes value 1 if married
or cohabited, 0 otherwise. Per capita household expenditure rather than income is used as a
measure of living standard since it can be measured with less noise and reflects the expenditure
planning on the expected lifetime income. Three education groups are constructed: illiterate
57
(reference group), have some primary education, and have some secondary education. Work status
is a dummy indicating whether the respondent engaged in agricultural work for more than 10 days
in the past year, or worked for wage, own business or unpaid family business for at least one hour
last week.

3.6 Results
3.6.1 Rural Results

Around 84 percent of our whole sample are rural residents. First, I examine various types of
intergenerational support in separate regression models before putting them together in a single
model. Tables 3-2, 3-3, and 3-4 present the OLS results for the rural sample.
Table 3-2 presents the results with structural support measures, including a set of
demographic and socioeconomic characteristics and the health measure (i.e. whether have any
ADL/IADL difficulty) as controls for all specifications. Surprisingly, neither more children nor
the presence of sons have significant effects on the life satisfaction of rural older parents. In column
(2), I focus on co-residence aspect of living arrangements. With the reference group being living
with both children and grandchildren, living without children and without grandchildren or living
with children only has a negative and significant effect while the effect of living with grandchildren
only is not significant. Living with parents also has no obvious effect. Column (3), on the other
hand, investigates the proximity to children aspect (living with children is the reference group). I
find that not co-residing with children do not have a significant negative influence and, to my
surprise, having all children living beyond the county/city has a positive (though not significant)
influence. In column (4), I study the effects of six intergenerational living arrangements from the
58
interactions between co-residence dummies and proximity-to-children dummies (living in a three-
generation household is the reference category). I still find that living with children only (i.e. in a
two-generation household without grandchildren) has a negative and significant influence. Living
in a single-generation or a skip-generation household is significantly harmful to life satisfaction
only if the closest child lives within the same county or city. In other words, having children not
living together but nearby seems to be detrimental to life satisfaction of rural older parents but the
significance of the effect disappears if children live farther away.
Table 3-2 Life Satisfaction and Structural Support (Rural Sample, OLS Estimation)
Dependent Variable: Life Satisfaction
(1) (2) (3) (4)
OLS OLS OLS OLS

have two children -0.0116 -0.0108 -0.0114 -0.00791

(0.0501) (0.0506) (0.0505) (0.0505)
have three children 0.00784 0.0165 0.0109 0.0225

(0.0530) (0.0538) (0.0539) (0.0539)
have sons only 0.000753 -0.00127 -0.00294 -0.00537

(0.0371) (0.0369) (0.0372) (0.0370)
have sons and daughters 0.00984 0.00722 0.00901 0.00635

(0.0325) (0.0324) (0.0325) (0.0324)
live without children and without
grandchildren

-0.0690***

(0.0236)

live with grandchildren only

-0.0437

(0.0300)

live with children only

-0.140***

-0.140***

(0.0273)

(0.0273)
live with parents

-0.0227

-0.0242

(0.0496)

(0.0495)
closest child lives in same county/city

-0.0158

(0.0196)

closest child lives outside same county/city

0.0291

(0.0336)

live without children and without
grandchildren*closest child lives in same
county/city

-0.0740***

(0.0238)
live without children and without
grandchildren*closest child lives outside
same county/city

-0.0382

(0.0460)
live with grandchildren only*closest child
lives in same county/city

-0.0619*

(0.0347)
live with grandchildren only*closest child
lives outside same county/city

-0.0031

(0.0425)
Observations 9,258 9,258 9,258 9,258
R-squared 0.106 0.11 0.107 0.11
59

Next, I study the set of regressions with functional support in Table 3-3. In column (1), I
examine the exchanges of instrumental help from OLS estimations. Receiving children’s help with
ADL or IADL difficulty is significantly harmful than having no ADL or IADL difficulty, and the
magnitude of its negative effect seems smaller than receiving others’ help with the difficulty or
receiving no help. However, the three coefficients are actually not statistically significant from
each other. This means children’s help in self-care or household tasks is not significantly more
beneficial than others’ help or no help when an individual has ADL or IADL difficulty. Taking
care of grandchildren under 16, which I consider as providing instrumental help, has a positive and
significant effect. In column (2) and (3), I find both receiving and providing financial support have
a positive and significant effect on life satisfaction of the rural population both qualitatively and
quantitatively. Noticeably, the coefficients of receiving financial help almost doubles those of
providing financial help. In column (4), seeing non-coresident children frequently (i.e. no less than
once a month) also appears to have a positive and significant effect.
Table 3-3 Life Satisfaction and Functional Support (Rural Sample, OLS Estimation)
Dependent Variable: Life Satisfaction

(1) (2) (3) (4)

OLS OLS OLS OLS

need help but no one helps -0.207***

(0.0513)

need help and children help -0.160***

(0.0533)

need help and others help -0.223***

(0.0309)

whether take care of grandchildren under
16
0.0454**

(0.0179)

whether receive financial transfers from
non-coresident children

0.0937***

(0.0194)

Controls for all the specifications: age, gender, nationality, marital status, education, natural logarithm of per capita
expenditure, whether have any difficulty with ADL or IADL, work status, whether participate in a pension system,
wave 2 dummy and community fixed effects. Robust standard errors clustered at the individual level in parentheses.
*** p<0.01, ** p<0.05, * p<0.1

60
whether provide financial transfers for
non-coresident children

0.0475**

(0.0225)

ln(max financial transfer received)

0.0159***

(0.0025)

ln(max financial transfer provided)

0.00650**

(0.0028)

whether see non-coresident children more
than once a month

0.0487**

(0.0210)
Observations 9,258 9,258 9,258 9,258
R-squared 0.107 0.109 0.112 0.107
Controls for all the specifications: age, gender, nationality, marital status, education, natural logarithm of per
capita expenditure, whether have any difficulty with ADL or IADL, work status, whether participate in a pension
system, wave 2 dummy and community fixed effects. Robust standard errors clustered at the individual level in
parentheses.
*** p<0.01, ** p<0.05, * p<0.1

I then look at the models with both structural and financial help in Table 2.3. Overall, the
signs and significances of all support variables are consistent with those in Table 2.1 and 2.2 except
that the positive effect of taking care of grandchildren under 16 becomes insignificant.
Table 3-4 Life Satisfaction, Structural Support, and Functional Support (Rural Sample,
OLS Estimation)
Dependent Variable: Life Satisfaction
(1) (2)
OLS OLS

have two children -0.00854 -0.0102

(0.0503) (0.0501)
have three children 0.00936 0.0053

(0.0538) (0.0536)
have sons only -0.000755 -0.00173

(0.0371) (0.0370)
have sons and daughters 0.00736 0.00512

(0.0324) (0.0324)
live with children only -0.130*** -0.130***

(0.0278) (0.0278)
live without children and without
grandchildren*closest child lives in same
county/city
-0.0790*** -0.0839***

(0.0244) (0.0244)
live without children and without
grandchildren*closest child lives outside same
county/city
-0.00878 -0.0185

(0.0463) (0.0462)
live with grandchildren only*closest child lives
in same county/city
-0.0746** -0.0831**

(0.0348) (0.0348)
live with grandchildren only*closest child lives
outside same county/city
0.00802 -0.00655

(0.0432) (0.0434)
live with parents -0.0245 -0.026

(0.0497) (0.0498)
need help but no one helps -0.203*** -0.202***

(0.0511) (0.0510)
61
need help and children help -0.170*** -0.172***

(0.0528) (0.0526)
need help and others help -0.222*** -0.221***

(0.0307) (0.0306)
whether take care of grandchildren under 16 0.019 0.0153

(0.0189) (0.0189)
whether receive financial transfers from non-
coresident children
0.0895***

(0.0196)

whether provide financial transfers for non-
coresident children
0.0473**

(0.0226)

ln(max financial transfer received)

0.0155***

(0.0025)
ln(max financial transfer provided)

0.00641**

(0.0028)
whether see non-coresident children more than
once a month
0.0647*** 0.0637***

(0.0219) (0.0218)
Observations 9,258 9,258
R-squared 0.114 0.117
Controls for all the specifications: age, gender, nationality, marital status, education, natural logarithm of per
capita expenditure, whether have any difficulty with ADL or IADL, work status, whether participate in a
pension system, wave 2 dummy and community fixed effects. Robust standard errors clustered at the
individual level in parentheses.
*** p<0.01, ** p<0.05, * p<0.1

In summary, interestingly, number and gender composition of children have no significant
effect on life satisfaction of rural older parents. Living in a traditional three-generation household
is associated with the highest level of life satisfaction, confirming the findings in the existing
literature. Having older parents live by themselves or live with children only has a negative and
significant influence while the negative effect of living with grandchildren only is not that obvious.
Besides, after taking into account proximity to children, I find the negative influence of living in
a single-generation or skip-generation household is only significant when the closest child lives in
the same county or city. It seems that, if not co-residing with children, rural older parents are
happier with children living far away from them rather than with children living close to them.
One tentative explanation for this finding is that longer-distance migration tends to be more job-
oriented (e.g. job assignment, migrant worker/business, and study/training)
19
. This suggests that

19
China Census, 1990
62
those moving out of a county are likely to be more educated and wealthier than their counterparts
who stay within the county. Therefore, older parents in rural villages may feel satisfied even if
their children all live far away. Regarding various types of functional support, the exchanges of
financial and emotional support have positive influences on life satisfaction while the exchanges
of instrumental support have no obvious effects.

3.6.2 Urban Results

The examination of the urban sample follows the same procedure as used to the rural one. Table
3-5 contains specifications with structural support. From OLS estimations, having more children
has a positive though insignificant effect while having sons only significantly improve the life
satisfaction of older parents in urban areas. In columns (2) and (3), I examine the effect of co-
residence and proximity to children, respectively. Regarding types of co-residence, living in a skip-
generation household has a negative and significant effect while the effects of other types of co-
residence are not obvious. Regarding the proximity to children, both types of non-coresidence have
negative but insignificant influences. When interacting these two dimensions of living
arrangements (column (4)), I find, no matter how far non-coresident children live, living in a skip-
generation household is life satisfaction deteriorating.
Next, I look at various types of functional support in Table 3-6. Among the exchanges of
all types of functional support, only children’s help with self-care and household tasks appears to
be significantly beneficial while the exchanges of financial and emotional support mostly have
63
positive and insignificant effects
20
. If I examine structural and functional support measures
together in one model, the results from previous specifications still hold (See Table 3-7).
Table 3-5 Life Satisfaction and Structural Support (Urban Sample, OLS Estimation)
Dependent Variable: Life Satisfaction
(1) (2) (3) (4)

OLS OLS OLS OLS

have two children 0.0137 -0.0093 -0.00888 -0.00123

(0.0775) (0.0771) (0.0774) (0.0750)
have three children 0.0723 0.0453 0.0534 0.0537

(0.0898) (0.0883) (0.0883) (0.0860)
have sons only 0.149** 0.149** 0.143** 0.148**

(0.0713) (0.0708) (0.0709) (0.0701)
have sons and daughters 0.0627 0.0575 0.0583 0.0618

(0.0655) (0.0659) (0.0653) (0.0657)
live without children and without
grandchildren

-0.0506

(0.0556)

live with grandchildren only

-0.224***

(0.0706)

live with children only

-0.0865

-0.0846

(0.0731)

(0.0732)
live with parents

0.0474

0.0547

(0.1050)

(0.1030)
closest child lives in same county/city

-0.0668

(0.0472)

closest child lives outside same county/city

-0.0293

(0.0918)

live without children and without
grandchildren*closest child lives in same
county/city

-0.06

(0.0561)
live without children and without
grandchildren*closest child lives outside
same county/city

0.0482

(0.1110)
live with grandchildren only*closest child
lives in same county/city

-0.203***

(0.0733)
live with grandchildren only*closest child
lives outside same county/city

-0.277*

(0.1420)
Observations 1,736 1,736 1,736 1,736
R-squared 0.154 0.161 0.155 0.162
Number of individuals
Controls for all the specifications: age, gender, nationality, marital status, education, natural logarithm of per capita
expenditure, whether have any difficulty with ADL or IADL, work status, whether participate in a pension system,
wave 2 dummy and community fixed effects. Robust standard errors clustered at the individual level in
parentheses.
*** p<0.01, ** p<0.05, * p<0.1

20
While whether receiving financial help from non-coresident children has an insignificant influence, the maximum quantity
received does have a significant and positive influence.
64

In conclusion, for urban older parents, having sons only positively influences life
satisfaction while living in a skip-generation household negatively influences it. In addition,
receiving help with self-care and household tasks from children is life satisfaction improving
(compared with receiving no help or others’ help).

Table 3-6 Life Satisfaction and Functional Support (Urban Sample, OLS Estimation)

Dependent Variable: Life Satisfaction

(1) (2) (3) (4)

OLS OLS OLS OLS

need help but no one helps -0.261**

(0.1150)

need help and children help 0.17

(0.1630)

need help and others help -0.190**

(0.0895)

whether take care of grandchildren under 16 0.0112

(0.0420)

whether receive financial transfers from non-
coresident children

0.0535

(0.0440)

whether provide financial transfers for non-
coresident children

0.0656

(0.0450)

ln(max financial transfer received)

0.0101*

(0.0052)

ln(max financial transfer provided)

0.00712

(0.0052)

whether see non-coresident children more than
once a month

0.0296

(0.0550)
Observations 1,736 1,736 1,736 1,736
R-squared 0.154 0.152 0.154 0.15
Number of individuals
Controls for all the specifications: age, gender, nationality, marital status, education, natural logarithm of per capita
expenditure, whether have any difficulty with ADL or IADL, work status, whether participate in a pension system,
wave 2 dummy and community fixed effects. Robust standard errors clustered at the individual level in parentheses.
*** p<0.01, ** p<0.05, * p<0.1

65

Table 3-7 Life Satisfaction, Structural Support, and Functional Support (Urban Sample,
OLS Estimation)
Dependent Variable: Life Satisfaction

(1) (2)

OLS OLS

have two children -0.00654 -0.0103

(0.0737) (0.0733)
have three children 0.0414 0.036

(0.0844) (0.0838)
have sons only 0.144** 0.143**

(0.0699) (0.0692)
have sons and daughters 0.0617 0.0601

(0.0651) (0.0649)
live with children only -0.0821 -0.0798

(0.0733) (0.0729)
live without children and without
grandchildren*closest child lives in same
county/city
-0.0656 -0.0664

(0.0580) (0.0576)
live without children and without
grandchildren*closest child lives outside same
county/city
0.0763 0.0711

(0.1120) (0.1110)
live with grandchildren only*closest child lives
in same county/city
-0.206*** -0.211***

(0.0745) (0.0745)
live with grandchildren only*closest child lives
outside same county/city
-0.252* -0.256*

(0.1410) (0.1400)
live with parents 0.0637 0.0657

(0.1050) (0.1040)
need help but no one helps -0.252** -0.249**

(0.1130) (0.1130)
need help and children help 0.143 0.146

(0.1640) (0.1630)
need help and others help -0.194** -0.195**

(0.0904) (0.0904)
whether take care of grandchildren under 16 0.0167 0.0154

(0.0444) (0.0443)
whether receive financial transfers from non-
coresident children
0.0504

(0.0441)

whether provide financial transfers for non-
coresident children
0.0638

(0.0449)

ln(max financial transfer received)

0.0100*

(0.0052)
ln(max financial transfer provided)

0.00715

(0.0052)
whether see non-coresident children more than
once a month
0.0479 0.0494

(0.0539) (0.0537)
Observations 1,736 1,736
R-squared 0.168 0.17
Number of individuals
Controls for all the specifications: age, gender, nationality, marital status, education, natural logarithm of per
capita expenditure, whether have any difficulty with ADL or IADL, work status, whether participate in a pension
66
system, wave 2 dummy and community fixed effects. Robust standard errors clustered at the individual level in
parentheses.
*** p<0.01, ** p<0.05, * p<0.1

3.7 Further Analysis
In this section, I test the robustness of my main findings in three dimensions. First, I show that my
findings are robust to the control of individual heterogeneity generally. Second, I test whether the
findings on structural support are robust when I use a less selective sample of older parents. Third,
I show whether there are any differences between urban natives and rural-to-urban migrants.

3.7.1 Control for Individual Heterogeneity

As I discussed earlier, there may be some time-invariant individual characteristics that are
potentially correlated with both life satisfaction and intergeneration support, and I am able to
control for them using fixed-effects estimation. The fixed-effects results for rural and urban
residents are presented in Table 4.1 and 4.2, respectively.
Table 3-8 . Life Satisfaction, Structural Support, and Functional Support (Rural Sample,
FE Estimation)
Dependent Variable: Life Satisfaction
(1) (2)

FE FE

have two children -0.0634 -0.0649

(0.1110) (0.1110)
have three children -0.159 -0.16

(0.1260) (0.1260)
have sons only -0.136 -0.136

(0.1220) (0.1220)
have sons and daughters -0.0379 -0.0389

(0.0927) (0.0926)
live with children only -0.0799* -0.0799*

(0.0431) (0.0431)
live without children and without
grandchildren*closest child lives in same
county/city
-0.0779* -0.0793*

(0.0412) (0.0412)
live without children and without
grandchildren*closest child lives outside
same county/city
-0.0687 -0.0701
67

(0.0622) (0.0624)
live with grandchildren only*closest child
lives in same county/city
-0.0656 -0.0678

(0.0447) (0.0447)
live with grandchildren only*closest child
lives outside same county/city
0.00128 -0.00167

(0.0617) (0.0617)
live with parents -0.0388 -0.0396

(0.0554) (0.0556)
need help but no one helps -0.103* -0.103*

(0.0614) (0.0613)
need help and children help -0.077 -0.0777

(0.0654) (0.0653)
need help and others help -0.126*** -0.126***

(0.0362) (0.0361)
whether take care of grandchildren under 16 0.00651 0.00543

(0.0239) (0.0239)
whether receive financial transfers from non-
coresident children
0.0245

(0.0245)

whether provide financial transfers for non-
coresident children
0.0165

(0.0271)

ln(max financial transfer received)

0.00478

(0.0032)
ln(max financial transfer provided)

0.00226

(0.0034)
whether see non-coresident children more
than once a month
0.035 0.0343

(0.0289) (0.0288)
Observations 9,258 9,258
R-squared 0.015 0.015
Number of individuals 4,629 4,629
Controls for all the specifications: age, gender, nationality, marital status,
education, natural logarithm of per capita expenditure, whether have any
difficulty with ADL or IADL, work status, whether participate in a pension
system, and wave 2 dummy. Robust standard errors clustered at the individual
level in parentheses.
*** p<0.01, ** p<0.05, * p<0.1

Table 3-9 Life Satisfaction, Structural Support, and Functional Support (Urban Sample,
FE Estimation)
Dependent Variable: Life Satisfaction

(1) (2)

FE FE

have two children 0.448 0.45

(0.2950) (0.2960)
have three children 0.686** 0.688**

(0.3050) (0.3060)
have sons only 0.590** 0.584**

(0.2520) (0.2520)
have sons and daughters 0.388** 0.384**

(0.1630) (0.1620)
live with children only 0.0793 0.0807

(0.0972) (0.0977)
live without children and without
grandchildren*closest child lives in same
county/city
0.109 0.11
68

(0.0863) (0.0864)
live without children and without
grandchildren*closest child lives outside
same county/city
0.0911 0.0909

(0.1540) (0.1540)
live with grandchildren only*closest child
lives in same county/city
0.0282 0.0283

(0.0925) (0.0931)
live with grandchildren only*closest child
lives outside same county/city
-0.0291 -0.0319

(0.1890) (0.1900)
live with parents 0.125 0.125

(0.1240) (0.1240)
need help but no one helps -0.172 -0.174

(0.1150) (0.1140)
need help and children help -0.042 -0.0431

(0.1910) (0.1910)
need help and others help -0.0607 -0.0612

(0.1030) (0.1030)
whether take care of grandchildren under
16
-0.00824 -0.00898

(0.0576) (0.0576)
whether receive financial transfers from
non-coresident children
-0.0465

(0.0534)

whether provide financial transfers for non-
coresident children
0.0306

(0.0584)

ln(max financial transfer received)

-0.00484

(0.0063)
ln(max financial transfer provided)

0.00225

(0.0067)
whether see non-coresident children more
than once a month
-0.0704 -0.0705

(0.0687) (0.0688)
Observations 1,736 1,736
R-squared 0.028 0.028
Number of individuals 868 868
Controls for all the specifications: age, gender, nationality, marital status, education, natural
logarithm of per capita expenditure, whether have any difficulty with ADL or IADL, work
status, whether participate in a pension system, and wave 2 dummy. Robust standard errors
clustered at the individual level in parentheses.
*** p<0.01, ** p<0.05, * p<0.1

As expected, the results from fixed-effects estimation are consistent with those from OLS
estimation, but the coefficients are smaller and less significant in general. This could be explained
by the few changes on the intergenerational support measures between the two years. There is one
exception: the effects of the family size and gender composition variables on the life satisfaction
of urban older parents appear to be more influential from fixed-effects estimation than those from
OLS estimation. One tentative explanation for this is that the effects of family size and gender
69
composition are highly correlated with time-invariant characteristics, such as personality, in urban
areas. Detailed analysis is needed to understand the puzzle.

3.7.2 A Less Selective Sample

It is possible that the findings above may not be representative of older parents in China in general
because, in order to exam functional support, those who don’t have non-coresident children or
grandchildren under 16 are not included in the main analysis. In this subsection, I am going to
investigate a larger sample, which includes these older parents. By construction of CHARLS, this
larger sample should be more representative of older parents than my baseline sample since it will
not exclude older parents with certain characteristics. I want to check whether the findings on
structural support for this larger sample are consistent with those for the smaller sample of my
main analysis.
First, I compare the summary statistics of the two samples to understand how different they
are in terms of various characteristics (The summary statistics of the larger sample can be found
in Appendix Table B-1.). The sample size of my main analysis is about half of that of the larger
sample. The larger sample is 2 years younger on average with around 12 percent higher in the age
range “aged 49 or younger”
21
. Therefore, I expect that the larger sample should be more influenced
by the family planning policies and have smaller family sizes. Comparing the relevant statistics
between the two samples confirms our expectation: the average number of children for the larger
sample is smaller especially for the urban sample. Specifically, the proportion of urban older
parents having a single child is around 0.2 higher, and the percentage of rural parents having one

21
There is also a slightly higher percent of those over 70 as expected for the larger sample. (Those who are over 70 are less
likely to have grandchildren under 16 and individuals without grandchildren under 16 are excluded from the main sample of
analysis.)
70
or two children is also higher. With a smaller average family size, I also find a higher percentage
of sons or daughters only families in the larger sample. Regarding the living arrangements, I find
the proportion of two-generation households (i.e. living with children only) is around 0.1 higher
for the larger sample. This should be mainly due to the fact that the larger sample has a larger
proportion of middle-aged parents who do not have any grandchild. The differences in the
percentage of other types of living arrangements are not that obvious. The demographic and
socioeconomic characteristics are similar between the two samples in general except that the larger
sample is slightly wealthier, healthier, and more educated. To summarize, the two samples are
different in terms of some characteristics mainly because the larger sample is younger on average
than the sample of main analysis.
What about the relationship between life satisfaction and structural support? In Table 3-10,
I present the OLS results for the larger sample that are comparable to those for the sample of main
analysis in column (4) of Table 3-2 and 3-5. For rural residents (column (1)), I find the results for
the larger sample are quite consistent to the results for my main analysis except that the negative
effect of living in a single-generation household with the closet child living outside the same
county becomes significant. For urban residents (column (2)), I find having a larger family size
has a significant and positive effect while having sons only no longer has a significant influence.
These may be potentially explained by the fact that a larger proportion of older parents in the larger
sample are highly influenced by the family planning policies. Regarding living arrangements, the
negative influence of living with children only becomes significant in the larger urban sample.
In conclusion, the findings on the relationship between life satisfaction and structural
support, especially the living arrangements, for the large representative sample are quite similar to
71
those for the sample of my main analysis. By and large, the findings from my main analysis can
be considered representative for older parents in China.

Table 3-10 Life Satisfaction and Structural Support (Larger Sample)
Dependent Variable: Life Satisfaction

(1) (2)

Rural Urban

OLS OLS

have two children -0.0076 0.0601*

(0.0246) (0.0358)
have three children 0.0195 0.0955**

(0.0279) (0.0465)
have sons only -0.017 -0.00496

(0.0241) (0.0330)
have sons and daughters -0.00459 -0.0357

(0.0231) (0.0388)
live with children only -0.123*** -0.0891**

(0.0180) (0.0363)
live with parents -0.0177 0.034

(0.0340) (0.0538)
live without children and without
grandchildren*closest child lives in same
county/city
-0.0709*** -0.00841

(0.0173) (0.0346)
live without children and without
grandchildren*closest child lives outside same
county/city
-0.0570* -0.0147

(0.0293) (0.0538)
live with grandchildren only*closest child lives
in same county/city
-0.0493* -0.128**

(0.0281) (0.0523)
live with grandchildren only*closest child lives
outside same county/city
-0.00916 -0.202*

(0.0370) (0.1140)
Observations 17,346 4,686
R-squared 0.09 0.1
Controls for all the specifications: age, gender, nationality, marital status, education, natural
logarithm of per capita expenditure, whether have any difficulty with ADL or IADL, work
status, whether participate in a pension system, wave 2 dummy and community fixed effects.
Robust standard errors clustered at the individual level in parentheses.
*** p<0.01, ** p<0.05, * p<0.1

2.7.3 Floating Population in Urban Areas

Economic expansion has lured a large number of rural people to work in cities, but the household
registration system prevents them from urban rights and benefits because they don’t have urban
72
hukou. Although the policies for urban social security and old-age insurance are designed to cover
all workers in urban employment including migrants, the system has a number of characteristics
in operation that act as barriers to the participation of migrant workers. From the sample of analysis,
about 29 percent of the urban residents have rural hukou (we call them rural-to-urban migrants),
and only 7 percent of them participate in or receive urban workers’ basic pension compared with
65 percent for urban residents with urban hukou (we call them urban natives). Over half of the
rural-to-urban migrants still participate in or receive rural pension. Besides, rural-to-urban
migrants are not restricted by the stricter family planning policies (e.g. one-child policy since
1980s) designed for residents with urban hukou. They are only restricted by more relaxed family
planning policies designed for those with rural hukou. In my sample, about 14 percent of urban
natives have only one child while less than 3 percent of rural-to-urban have one child. Since the
population of rural migrants may not receive adequate social support in cities and are not affected
by stricter family planning policies for urban hukou, they may have different preferences on
support from urban natives. Therefore, in the following analysis, I am going to look at urban
natives and rural migrants separately.
Since over 70 percent of the urban residents have urban hukou, the results for this
population (in Table 3-11) are generally consistent with the results for the whole sample of urban
residents. If comparing the regression results for urban natives (Table 3-11) and rural-to-urban
migrants (Table 6.2), we can find the relationships between various types of support and life
satisfaction are similar between these two samples in general. There are only two exceptions: first,
rural migrants do not enjoy children’s instrumental help as formal urban residents. Specifically,
rural migrants prefer others’ help with ADL or IADL rather than children’s help; and second, the
effect of taking care of grandchildren under 16 appears to be positive and significant for rural
73
migrants but not urban natives. These two points imply that rural migrants do not feel happier with
receiving instrumental help from children and they are happier with providing instrumental help
instead. One possible explanation is that rural migrants consider the opportunity costs of their
children taking care of them and their grandchildren as so high that they do not want children to
spend time taking care of them and are willing to help children take care of young grandchildren.

Table 3-11 Life Satisfaction, Structural Support, and Functional Support (Urban Sample
with Urban Hukou)
Dependent Variable: Life Satisfaction

(1) (2)

OLS OLS

have two children 0.00789 0.00427

(0.0820) (0.0814)
have three children 0.109 0.104

(0.0942) (0.0937)
have sons only 0.118 0.117

(0.0858) (0.0850)
have sons and daughters 0.0646 0.0625

(0.0764) (0.0763)
live with children only -0.0417 -0.0395

(0.0927) (0.0921)
live without children and without
grandchildren*closest child lives
in same county/city
-0.0248 -0.0267

(0.0655) (0.0653)
live without children and without
grandchildren*closest child lives
outside same county/city
0.16 0.155

(0.1360) (0.1350)
live with grandchildren
only*closest child lives in same
county/city
-0.149* -0.154*

(0.0798) (0.0799)
live with grandchildren
only*closest child lives outside
same county/city
-0.15 -0.164

(0.1990) (0.1980)
live with parents 0.0646 0.063

(0.1330) (0.1340)
need help but no one helps -0.19 -0.189

(0.1470) (0.1470)
need help and children help 0.328** 0.329**

(0.1620) (0.1620)
need help and others help -0.157 -0.157

(0.0997) (0.0996)
whether take care of grandchildren
under 16
-0.0325 -0.0342

(0.0515) (0.0513)
whether receive financial transfers
from non-coresident children
0.0426

74

(0.0502)

whether provide financial transfers
for non-coresident children
0.0423

(0.0528)

ln(max financial transfer received)

0.00919

(0.0058)
ln(max financial transfer provided)

0.00543

(0.0061)
whether see non-coresident
children more than once a month
0.0168 0.0194

(0.0667) (0.0667)
Observations 1,232 1,232
R-squared 0.196 0.198
Number of individuals
Controls for all the specifications: age, gender, nationality, marital status,
education, natural logarithm of per capita expenditure, whether have any
difficulty with ADL or IADL, work status, whether participate in a pension
system, wave 2 dummy and community fixed effects. Robust standard errors
clustered at the individual level in parentheses.
*** p<0.01, ** p<0.05, * p<0.1

Table 3-12 Life Satisfaction, Structural Support and Functional Support (Urban Sample
with Rural Hukou)
Dependent Variable: Life Satisfaction

(1) (2)

OLS OLS

have two children -0.189 -0.199

(0.3810) (0.3850)
have three children -0.223 -0.235

(0.4100) (0.4130)
have sons only 0.0465 0.0481

(0.1720) (0.1720)
have sons and daughters 0.00292 0.00399

(0.1510) (0.1510)
live with children only -0.102 -0.098

(0.1220) (0.1220)
live without children and without
grandchildren*closest child lives in
same county/city
-0.0678 -0.062

(0.1190) (0.1180)
live without children and without
grandchildren*closest child lives outside
same county/city
-0.138 -0.142

(0.2160) (0.2150)
live with grandchildren only*closest
child lives in same county/city
-0.344** -0.343**

(0.1680) (0.1690)
live with grandchildren only*closest
child lives outside same county/city
-0.299 -0.3

(0.2490) (0.2470)
live with parents 0.00284 -0.000803

(0.1970) (0.1970)
need help but no one helps -0.255* -0.254*

(0.1480) (0.1470)
need help and children help -0.651* -0.648*

(0.3760) (0.3740)
need help and others help -0.245 -0.246

(0.2020) (0.2020)
75
whether take care of grandchildren under
16
0.205** 0.201**

(0.0908) (0.0913)
whether receive financial transfers from
non-coresident children
0.0127

(0.1040)

whether provide financial transfers for
non-coresident children
0.152

(0.1040)

ln(max financial transfer received)

0.00331

(0.0123)
ln(max financial transfer provided)

0.0146

(0.0117)
whether see non-coresident children
more than once a month
0.0204 0.0182

(0.1100) (0.1100)
Observations 503 503
R-squared 0.298 0.297
Controls for all the specifications: age, gender, nationality, marital status,
education, natural logarithm of per capita expenditure, whether have any
difficulty with ADL or IADL, work status, whether participate in a pension
system, wave 2 dummy and community fixed effects. Robust standard errors
clustered at the individual level in parentheses.
*** p<0.01, ** p<0.05, * p<0.1

By and large, the separate examinations on urban natives and rural migrants show that,
with few exceptions, the findings for rural migrants are similar to those for urban natives. However,
the results from the sample of rural migrants may not be generalizable because of the potential
small-cell-size problem caused by the small sample size.

3.8 Conclusion and Discussion

Economic growth, development in public policies, rise in internal migration and evolution of social
attitudes have changed the pattern of living arrangements in China, and possibly have altered the
tradition of older parents relying on children for support. This study investigated the effect of
intergenerational support on the life satisfaction of older parents in China, examining urban and
rural populations separately. Structural component of support, which includes family size, gender
76
composition of children and intergenerational living arrangements, and functional component of
support, which includes exchanges of instrumental, financial and emotional helps, are studied.
Regarding family size, I expected that more children should make older parents happier
not only because more children means more support available but also because having more
children is a cultural norm. However, it is not that obvious to predict whether the presence of sons
promotes older parents’ happier in contemporary China. Besides, it is ambiguous to judge which
population should be more affected by the number and gender composition of children: cultural
tradition and family planning policies gave us different expectations. With respect to living
arrangements and functional support, I suspected that rural older parents were happier with
traditional living arrangements than with nontraditional ones and depended on children for
functional support. On the contrary, due to more generous pension policies, smaller family size,
and influence of western culture, life satisfaction of urban older parents may not be negatively
influenced by living in a non-traditional household and may depend less on functional helps from
children.
The empirical analysis helps understand the actual situation. I find happiness of rural
parents is not significantly influenced by the number and gender composition of children at all.
But, they do enjoy the extended family integration. Rural older parents living with both children
and grandchildren experience the highest level of life satisfaction. Living with neither children nor
grandchildren or living with children only has a significant and negative influence while living
with grandchildren only has a negative but insignificant influence. This implies that the co-
residence with grandchildren plays an important role to happiness of rural parents. After taking
distance to children into consideration (in the case of non-coresidence), I find rural parents are
happier with children living far away than with children living close to them. Possibly, children’s
77
improved standard of living due to a long-distance job-oriented migration pleases their parents.
Regarding various types of functional support, the exchanges of financial and emotional support
turn out to be happiness improving for rural parents. That all the findings on living arrangements
are robust to the control of intergenerational transfers implies that they influence well-being of
individuals beyond functional costs and benefits.
The results from my rural sample is most comparable to the results of Silverstein et al.
(2006) which studied a small rural sample in Anhui Province. The main difference, besides my
sample is more nationally representative, is that I have two waves of data and thus are able to
employ fixed-effects estimation method. My findings from OLS estimations on living
arrangements are consistent with their findings in general while the findings on functional support
are not. Specifically, I find providing financial support an important factor while they don’t.
The findings for the urban sample, as expected, are very different from those for the rural
one. Having sons only has a positive and significant influence. Living in a skip-generation
household is detrimental to happiness while receiving help with self-care or household tasks from
children is happiness improving.
It seems that life satisfaction of the rural older parents is more positively associated with
living in a traditional three-generation household. Three tentative explanations are income gap,
disparity in pension system and differences in social attitudes. Lower level of income, inferior
public pension scheme and ideals of filial piety make it favorable for rural older parents to co-
reside with their children and grandchildren. Happiness of urban older parents, however, is not
negatively influenced by living on their own. Even though, compared with rural parents, urban
parents may place more importance on privacy because of the influence by western values, privacy
78
is a normal good for all. With considerably higher wage and pension income, urban parents may
afford living separately from children.
Exchanges of various types of functional support are found to be not as important as what
we have learnt from the literature. In particular, life satisfaction of urban older parents is only
significantly influenced by instrumental helps from children.
It seems that China’s development is breaking down the historical association between
intergenerational support and well-being. While the life satisfaction of China’s rural population
still depends on traditional kinds of support such as living in a three-generation household and
receiving financial transfers, that of China’s urban population does not.
The Chinese population as a whole has become more homogeneous due to rural-urban
migration trend and the establishments of more unified policies. The tremendous rural-urban
migration flows have sent not only income but also social attitudes back from urban to rural areas.
The pension system has become more established and unified in the past 10 years. The new rural
pension scheme introduced in 2009, though still less generous than pension for urban employees,
has better coverage over rural population and involves government subsidies when compared with
the old rural pension scheme. Changing filial norms and rising standard of living have created new
form of interdependence between generations in China. If the social norms and the pension system
of rural China continue to converge to those of urban China, older parents in rural China will be
less dependent on family as a source of support and be happier with living independently in the
future.

79
CHAPTER 4. An Examination of the Effects of Consumption Expenditures on Life
Satisfaction in Australia
4.1 Introduction

Social comparison, which refers to the tendency of people to compare their own situations
with those of others, has long been considered important to subjective well-being (SWB).
22
Studies
focusing on the link between SWB and social comparisons have mostly investigated the effect of
relative income on SWB (e.g., Clark and Oswald 1996; McBride 2001; and Ferrer-i-Carbonell
2005). The effects of social comparisons, or “demonstration effects” if we look at individuals’
consumption behavior, however, naturally apply more strongly to some types of goods than to
others (Veblen 1899; Frank 1985a, 1985b). For example, it is very likely that we know the makes
and models of the cars our friends drive and the brands of clothes they wear but we are less likely
to know how much they save or how much they spend on insurance or health care. Some early
studies have formally defined consumption types according to their demonstration effects, or
impacts on social comparisons. Veblen (1899) coined the term “conspicuous consumption” to refer
to the consumption that is intentionally used as a signal for status. Frank (1985b) used the term
“positional goods” to mean “those things whose value depends relatively strongly on how they
compare with things owned by others”; in contrast, he called goods that depend relatively less
strongly on such comparisons “nonpositional.”
This paper aims to understand social comparisons by looking at the effects of consumption
expenditures on life satisfaction among the general population of Australia. Specifically, I

22
The terms subjective well-being, life satisfaction, and happiness are used interchangeably in this paper and refer to
satisfaction with life as a whole.

80
distinguish between their spending on conspicuous goods and services and spending on basics.
Following the existing literature, I define conspicuous consumption expenditures as any spending
on goods or services that are first, literally “visible” to outsiders, and second, positional
(Winkelmann 2012). Basic consumption expenditures, on the other hand, refers to any spending
on goods or services that are only consumed for their utilitarian value. Therefore, conspicuous
purchases reflect one’s material aspirations and social status, and are more subject to social
comparisons than basic purchases, which are more likely to relate to one’s basic needs.
At a general level, this study contributes to the little empirical literature on the influence of
consumption expenditures on SWB. The main contributions of the present study in relation to
previous works are the following: First, the study examines not only the effect of total consumption
expenditures on SWB but also the effects of types of consumption expenditures. The results show
that a channel through which income influences an individual’s SWB is the conspicuous
component of income. Household spending on conspicuous goods and services has a positive and
significant influence on individual’s life satisfaction, and this effect is close to the effect of income
on life satisfaction: a 1 percent increase in conspicuous consumption expenditures increases life
satisfaction by 0.0013 on an 11-point scale. My analysis also suggests that it is one’s household’s
ranking of conspicuous consumption expenditures within his or her reference group that really
matters, a conclusion that supports the importance of social comparisons to one’s SWB.
Expenditures on basic goods and services, however, do not contribute to happiness, a finding that
agrees with Scitovsky’s argument that consumption satisfying our basic needs (or comfort) does
not contribute much to SWB (Scitovsky 1976; Friedman and McCabe 1996). Another “invisible”
component of income, savings, also has no significant effect, probably due indeed to its invisibility.
81
Another contribution of the study is to apply the Arellano-Bond generalized method of
moments (GMM) to estimate causal relationships. Most of the previous findings on consumption
expenditures and SWB have been cross-sectional and, therefore, can only be interpreted as
associational or relational. There are various potential sources of bias associated with micro data
estimation of the effect of consumption expenditures on life satisfaction in studies of this sort.
First, while most of the existing studies do not account for unobserved individual heterogeneity, it
may, in fact, be correlated with both consumption expenditures and life satisfaction. This will result
in what psychologists called a “personality bias” on obtained estimates. For example, it is possible
that people who are extroverted are more likely to be happy as well as spend more, which will
result in an upward bias in the estimates. Second, there may be reverse causality from happiness
to consumption expenditures. If it is the case that happier people have more control over their
spending, the estimated effect of expenditures will be biased downwards. Third, the measurement
error in consumption expenditures that is independent of the level of expenditures can bias the
estimated effect towards zero (i.e., attenuation bias). Since there are different sources of bias and
these biases are not in the same direction, it is difficult to understand the direction of the overall
bias of the estimator. The use of the Arellano-Bond GMM estimation helps to eliminate each source
of bias and provide a consistent estimation of the model.
The final contribution of the study is that it provides evidence of relationship heterogeneity
across income groups. It seems that conspicuous consumption expenditures contribute positively
to life satisfaction for individuals in all income groups. By contrast, basic consumption
expenditures influence life satisfaction negatively for those whose household income belongs to
the lowest income quartile.
82
4.2 Literature Review

Of the numerous studies examining the relationship between SWB and material living conditions,
most use income as the sole measure of material living conditions. One of the earliest and most
important works in this literature has been Easterlin’s seminal article (Easterlin 1974), which finds
self-reported happiness levels do not rise over time as a country’s real income rises even though
rich people are happier than poor people in the same country. There has been considerable
discussion surrounding this paradox on either the explanations for it (e.g., Easterlin 1995; Easterlin
2001; and Clark et al. 2008) or whether it actually exists (i.e., whether the happiness and income
rise jointly over time) (e.g., Easterlin 1995; and Stevenson and Wolfers 2008).
Even though country-level studies have not reached a consensus on how important income
is to happiness, within-country studies often agree that income has a positive, albeit moderate,
effect on happiness (e.g., Blanchflower and Oswald, 2004a, 2004b; Graham and Pettinato 2004;
Winkelmann and Winkelmann 1998; Luttmer 2005; and Kingdon and Knight 2007). In general,
the effects found in panel studies are smaller than those found in cross-sectional studies. In
addition, the effects found in causal analyses, using either exogenous variations in income or
instrumental variables, are larger than those found in associational analyses. Studies also have
found that it is one’s reference group income or one’s income relative to that of others within one’s
reference group that matters (e.g., Clark and Oswald 1996; McBride 2001; and Ferrer-i-Carbonell
2005), a finding that suggests the importance of social comparisons to well-being.
There are quite a few reasons, however, why current income is not a perfect measure of
material living conditions and numerous scholars have argued that consumption expenditures may
be superior (e.g., Cutler and Katz 1991, 39; and Meyer and Sullivan 2006, 2). When it comes to
83
explaining SWB, consumption expenditures may turn out to be a better measure for three particular
reasons (Noll and Weick 2015). First, they provide information on the content of the consumer
basket. We can investigate whether or not the acquisition of specific types of goods or services
makes one happy. Second, consumption expenditures provide a more accurate measure of long-
term resources than does income (Poterba 1989; Cutler and Katz 1992). Third, they can be
measured with less noise than income (Strauss and Thomas 1998).
Given the relevance and advantage of studying consumption expenditures and their
composition, the literature is still relatively silent about how individuals’ spending choices affect
SWB. There seem to be two main reasons for this situation (Stanca and Veenhoven 2015;
Zimmermann 2014). First, many researchers still equate consumption and happiness: a utility
maximizing agent is fully informed about his or her preferences and constraints, and he or she is
supposed to distribute income efficiently so that no other allocation of income can increase his or
her utility. However, they often fail to consider the differences between expected utility and
experienced utility, the latter being the proxy for happiness (Kahneman and Thaler 2006). Second,
appropriate data on consumption expenditures are not as frequently and widely collected in surveys
as those on income.
The relatively little literature on consumption expenditures and happiness mainly focuses
on total household expenditures. It has been found that total household spending has a positive and
significant, though sometimes moderate, influence on life satisfaction (Headey et al. 2004;
Guillen-Royo 2008; Lewis 2011; Zimmermann 2014; Noll and Weick 2015; Wang et al. 2015).
However, almost all of these studies only examine expenditures on non-durable goods
23
even if
they claim them to be “total household expenditures.” In addition, only three empirical studies

23
The only durable expenditure included is the housing expenditure.
84
have used panel data and studied within-person relationships. Headey et al. (2004) and
Zimmermann (2014), the former using household panel data from Britain and the Netherlands and
the latter using a representative panel sample of US individuals, show that total consumption
expenditures are at least as important to happiness as income. Wang et al. (2015), using two waves
of data from the China Family Panel Studies (CFPS), finds that consumption expenditures have a
significant and positive influence on happiness and that relative consumption matters.
Nevertheless, none of the studies tries to estimate causal relationships.
There is a growing body of literature on conspicuous consumption, an important focus of
which is what should be classified as conspicuous or visible. Charles et al. (2009) defines certain
goods as visible based on a survey of business students of the University of Chicago. Friehe and
Mechtel (2014), based on Charles et al. (2009), includes more items as conspicuous. Heffetz
(2011) measures the visibility of thirty-one consumption categories using a US national telephone
survey.
Another part of the related literature has focused on the effects of conspicuous and basic
spending on life satisfaction. As far as I know, there are only two relevant studies. Perez-Truglia
(2013) shows that an individual’s life satisfaction increases with his or her ranking of clothing
expenditures (highly observable), but it is not affected by his or her ranking of food expenditures
(highly unobservable). Zimmermann (2014) finds that consumption is associated with SWB
mainly through conspicuous rather than basic expenses. However, all these estimations cannot be
understood as demonstrating causality.
In addition, there are a few studies investigating how specific components of consumption
expenditures are associated with happiness, but almost all of these studies have conducted cross-
sectional analysis. DeLeire and Kalil (2010) finds that, for the elderly population in the United
85
States, only leisure expenditures are positively associated with life satisfaction. For the US
population in general, Zimmermann (2014) finds that expenditures on food at home, dining out,
clothing, child education, sport, and leisure have positive and significant influences while
expenditures on health care have a negative and significant effect. Noll and Weick (2015) also
shows, for the population of Germany, that expenditures on clothing and leisure are the drivers of
life satisfaction while expenditures on housing and mobility decrease it. Gokdemir (2015)
demonstrates that, for the Turkish population as a whole, expenditures on durable goods and
savings have positive and significant influences on life satisfaction. Dumludag (2015) indicates
that expenditures on clothing and durable goods have positive and significant influences on life
satisfaction at different levels of economic development even though the relationship between life
satisfaction and other consumption categories differs at these different levels generally. It appears
that the components that contribute to life satisfaction are more visible than those that do not.
The relationship between consumption expenditures and life satisfaction is not necessarily
one way. It is also possible that well-being affects consumption expenditures or income. Guven
(2012), instrumenting happiness with unexpected sunshine, finds that happier people save more
and have a lower marginal propensity to consume. De Neve and Oswald (2012) shows that
adolescents and young adults who report higher life satisfaction earn significantly higher levels of
income later in life.
The aforementioned relatively few investigations on consumption expenditures compared
with the vast literature on income motivate us to examine the causal effects of different types of
consumption expenditures on life satisfaction.

86
4.3 Data and Variables
4.3.1 Data

The empirical analysis will be based on data drawn from the Household, Income and Labour
Dynamics in Australia (HILDA) survey, which began in 2001 and from which 15 waves of data
are now available. The HILDA survey is financed by the Australian government and is conducted
by the Melbourne Institute of Applied Economic and Social Research. It is a nationwide panel
survey that contains a wide range of social, demographic, and socioeconomic information. Its
design owes much to other household panel studies such as the German Socio-Economic Panel
(GSOEP) and the British Household Panel Survey (BHPS) (Haisken-DeNew 2001; Frick et al.
2007; Watson and Wooden 2012).
An advantage of HILDA over GSOEP or BHPS is that it contains more detailed
information on household spending, which makes this study possible. Household expenditures on
a wide range of nondurable goods and services were first collected in the wave 5 Self-Completion
Questionnaire (SCQ). The list of items collected was expanded to include consumer durables from
wave 6 although some were dropped in wave 11. While the person in the household responsible
for the household bills was asked to complete the household-level expenditure (weekly, monthly
or yearly) questions in the SCQ, sometimes more than one person in a household provided answers.
Therefore, the responses across all individuals who provided a response (excluding the responses
from dependent students) are averaged to provide an estimate. Expenditures on housing
24
and child
care are collected regularly in more detail in the Household Questionnaire. This study makes use
of waves 6–10 (2006–2010) of the HILDA since it contains the most comprehensive list of

24
Following the practice of Household Expenditure Survey, no imputed rent has been taken into account in calculating the
housing expenditures of home owners.
87
consumption items. Excluding those without life satisfaction (and lagged life satisfaction), income,
household expenditures, and other individual characteristics used in this study, I am left with
14,107 individuals and 47,405 observations. Since I want to build a dynamic framework and use
the Arellano-Bond GMM estimation method, the sample used in my main analysis consists only
of individuals who have been observed in at least three consecutive waves.
All longitudinal surveys have to confront the problem of sample attrition. If attrition is
selective (i.e., individuals with certain characteristics are more likely to drop out of a study than
others), the relationship to be estimated will be biased. This problem can be an issue in the HILDA
since only less than 72% percent of the original wave 1 (2001) respondents remain after wave 5
(2005), and the average wave-on-wave sample attrition of my sample period (waves 6–10) is about
4%. Because, as I will show below, my estimation method involves taking a first difference,
attrition due to time-invariant individual characteristics will not cause any problem. My test for
selective attrition suggests that the attrition is, at least, random conditional on current individual
observables. This test will be discussed in Appendix Table C.1.

4.3.2 Variables
4.3.2.1. Life Satisfaction

The dependent variable of the regression analysis will be overall life satisfaction. It is based on the
question “All things considered, how satisfied are you with your life?”. This is measured on an 11-
point scale, with a higher value representing a higher level of life satisfaction. The mean level of
overall life satisfaction is 7.90 and the median is 8. The distribution of life satisfaction is left-
skewed.

88
4.3.2.2 Variables on Material Living Conditions

As mentioned above, waves 6–10 of the HILDA contain information on household consumption
expenditures. The dataset contains 28 consumption categories. Based on the findings in the
existing literature (e.g., Charles et al. 2009; Heffetz 2011; and Noll and Weick 2015), I choose to
sort the existing consumption categories into 18 broader consumption categories: housing,
groceries, health care (health insurance included) and child care, meals eaten out, vehicle
purchases, holidays, motor vehicle fuels and engine oil, clothing and footwear, phone rent and
calls and internet charges, other insurance, home utilities, alcohol, education, furniture and
household appliances, recreational devices and equipment, motor vehicle repairs and maintenance,
tobacco, and public transportation. Detailed definition of the categories can be found in Appendix
Table C-2.
The HILDA survey asks retrospective questions on various expenditures in the past
financial year. This is comparatively simple compared with what most official surveys do.
Household Expenditure Survey (HES), which is conducted by Australian Bureau of Statistics
every six years, collects information on over 600 categories of consumption expenditures by
making use of a diary method over a period of two weeks.
25
A comparison between the HILDA
2010 and HES 2009–10 on weekly household expenditures in Table 4-1 shows considerable
consistency in the relative importance of each type of consumption between these two sources.
The consistency suggests that the consumption expenditure data from the HILDA can be of good
quality.

25
Less frequent and often large expenditures were collected on a “recall” basis.
89
Table 4-1 Weekly Expenditures: Household Expenditure Survey (HES) (2009-10) and
HILDA (2010)
In % of total expenditure
HES HILDA
Expenditures on
Housing 29.40 28.84
Groceries 16.44 17.93
Health care (health insurance included) and child care 6.57 5.48
Meals out 5.56 4.75
Vehicle purchases 5.32 7.02
Holidays 5.19 5.61
Motor vehicle fuels and engine oil 4.50 4.26
Clothing and footwear 3.82 3.48
Phone rent and calls and internet charges 3.48 3.76
Other insurance 3.13 2.48
Home utilities 2.89 2.96
Alcohol 2.86 2.63
Education 2.70 2.53
Furniture and household appliances 2.66 2.00
Recreational devices and equipment 2.07 1.89
Motor vehicle repairs and maintenance 1.44 1.68
Tobacco 1.11 1.72
Public transportation 0.87 0.97
Total Expenditures 100.00 100.00
Notes:
Database: Household Expenditure Survey 2009–10; Household, Income and Labour Dynamics in
Australia (HILDA) 2010
The HILDA calculations are based on responses to various retrospective questions on expenditures in the
last financial year. (13,482 respondents)
The HES calculations are based on detailed expenditure tables (2009–10) published by Australian Bureau
of Statistics (9,774 households)

The total household consumption expenditures are the sum of all the components of
household expenditures available in the survey. More importantly, I want to distinguish
conspicuous consumption expenditures from other expenditures—namely, basic consumption
expenditures. Conspicuous consumption expenditures, as mentioned earlier, can be understood as
spending on goods or services that are (1) readily observable and (2) positional (i.e., having value
that depends on social comparisons (Frank, 1985b)). Based on an online survey and common
sense, Charles et al. (2009) considers visible or conspicuous consumption expenditures to consist
90
of expenditures on apparel, personal care
26
, and vehicles because these items are readily observable
across anonymous social interactions. Friehe and Mechtel (2014) uses a wider definition of
conspicuous consumption expenditures. Because there is evidence showing that status concerns
are mainly determined by the respect and admiration we earn in face-to-face interactions with those
we are close to (Senik 2009; Clark and Senik 2010), its definition of conspicuous goods or services
also includes items that can be regarded as observable and positional only vis-a-vis colleagues,
friends, and family. These items are furniture and household appliances, recreational devices and
equipment, meals eaten out, and holidays. My definition adds one more item—alcohol—to the
definition of Friehe and Mechtel (2014) (See Table 4-2). Alcohol can be considered conspicuous
because it is not a primary necessity of life, and it has been widely advertised that an iconic and
desirable lifestyle can be achieved by investing in a “luxury” alcoholic beverage. On the other
hand, my definition of basic consumption expenditures includes expenditures on education,
tobacco, groceries, housing (i.e., mortgage, rent, repairs, renovations, and maintenance), public
transportation, motor vehicle repairs and maintenance, motor vehicle fuels and engine oil, health
care (health insurance included) and child care, phone rent and calls and internet charges, home
utilities, and other insurance. These items have little or no visibility and/or limited status effects.
I add the consumption expenditures of all the categories in each type (i.e., conspicuous or basic)
and get the conspicuous consumption expenditures and basic consumption expenditures,
respectively. I will use the definitions proposed by Charles et al. (2009) and Friehe and Mechtel
(2014) as robust checks for my baseline analysis. In the case of Charles et al. (2009), since its
definition on conspicuous consumption expenditures is the most restrictive, I define “intermediate
consumption expenditures” as spending on the goods and services that are only observable and

26
HILDA does not have a separate spending category on personal care. Personal care products are included in “groceries”
category.
91
positional among colleagues, friends, and family (i.e., furniture and household appliances,
recreational devices and equipment, meals eaten out, holidays and alcohol), and basic consumption
expenditures as the spending on the rest of the items.
The baseline definition of conspicuous and basic consumption expenditures seems to be
plausible according to a visibility index by Heffetz (2011),
27
which identifies visibility or
conspicuousness as the speed with which members of a society notice a household’s expenditures
on different commodities. The value of the index arranges from 0 to 1, with a higher value
corresponding to a higher level of conspicuousness. Heffetz (2011) finds a strong and positive
correlation between visibility and expenditure elasticity. This implies that those items with higher
values of visibility index are more likely to be luxury and, therefore, are less likely to be related to
basic needs.
28
The last column of Table 4-2 shows the visibility index. We can observe that, in
general, those purchases I define as conspicuous are ranked higher in terms of the visibility index
than those I define as basic. However, expenditures on tobacco products, which are not considered
conspicuous, are the most visible. Following Friehe and Mechtel (2014), I do not consider tobacco
products to be conspicuous because I believe that they do not fulfill the second requirement of my
definition (i.e., the goods should be positional).
29
Another item that seems to be relatively visible
is spending on education. For example, school uniforms, which are compulsory in all Catholic
schools as well as in most private and public schools in Australia, are signals for the types of
schools and the levels of tuition fees. As robustness checks, I will include expenditures on tobacco

27
Heffetz (2011) conducted a survey with the main survey question being “Imagine that you meet a new person who lives in a
household similar to yours. Imagine that their household is not different from other similar households, except that they like to, and
do, spend more than average on[X]. Would you notice this about them, and if so, for how long would you have to have known
them, to notice it? Would you notice it almost immediately upon meeting them for the first time, a short while after, a while after,
only a long while after, or never?” for 31 different expenditure category titles. Based on the results of the survey, he develops an
index of expenditure visibility.
28
Expenditure or income elasticity, though more discussed by traditional economist, is not directly related to visibility of
conspicuousness. For example, insurance can be highly elastic but it is not visible at all.
29
In addition, the empirical analysis by Heffetz (2011) suggests that tobacco products seem to be inferior goods for a large interval
of the income scale.
92
products and expenditures on education as components of conspicuous consumption expenditures,
respectively.
In addition, categorizing expenditures on housing as basic seems to be controversial.
30

However, since my measure of housing expenditures does not include the imputed rent for home
owners without any loans, it may not well represent the conspicuous aspect of housing. Therefore,
the housing expenditures in this study might be understood as financial burdens instead. On the
other hand, because housing expenditures seem to be reasonably observable and positional and
account for approximately 30 percent of one’s total household consumption expenditures, I will
separate housing expenditures from other basic consumption expenditures as a robustness check.
Because most of the related literature uses income as a measure of material living
conditions, I, too, examine the effect of household income at least for comparison purpose. The
level of household income is defined as the household’s gross income from all sources minus
estimated taxes. In addition to income, I also construct household savings, which, in this study, is
defined as the difference between total household income and total household consumption
expenditures.
31

It is widely acknowledged that financial variables, such as income and consumption
expenditures, are hard to measure in surveys and are potentially under-reported or reported with
error. In order to reduce the potential bias relating to these measures, I use imputed derived
variables contained in the HILDA survey. The details of the imputation method can be found in
Hayes and Watson (2009) and Sun (2010). To minimize the influence of outliers, I exclude
observations with household expenditures or income in the top and bottom 0.1% of the respective

30
According to Marx (1942) and Frank (1999), housing can be positional.
31
A negative value of savings means that a household spent more than its income in that period.
Table 4-2 Definitions of Conspicuous Goods vs. Basic Goods
Category Baseline Charles et al.
(2009)
Friehe & Mechtel
(2014)
Heffetz (2011)'s Visibility Index*
Vehicle purchases C C C 0.73
Clothing and footwear C C C 0.71
Furniture and household appliances C I C 0.68
Recreational devices and equipment C I C 0.66
Meals out C I C 0.62
Alcohol C I B 0.60
Holidays C I C 0.58
Education B B B 0.56
Tobacco B B B 0.76
Groceries B B B 0.50
Housing B B B 0.50
Public transportation B B B 0.45
Motor vehicle repairs and maintenance B B B 0.42
Motor vehicle fuels and engine oil B B B 0.39
Phone rent and calls, and internet charges B B B 0.38
Health care (health insurance included) and child care B B B 0.36
Home utilities B B B 0.31
Other Insurance B B B 0.21
Notes:
C: Conspicuous; B: Basic; I: Intermediate
*For four items—alcohol, groceries, phone rent and calls, and other insurance—each is a combination of multiple items in Heffetz (2011). So, I compute a budget-share weighted average of
the visibility score for each of these four broader categories. The budget shares are derived from Household Expenditure Survey, 2009–10. Consumption on child care is not categorized in
Heffetz (2011), and I assign it a visibility index the same as health care.
distribution. This treatment leaves a few observations with negative income values, and,
consequently, these observations (about 0.20% of observations) are omitted from the analysis.
32

The variables of consumption expenditures, income, and savings
33
are deflated into
constant price of 2006 using CPI information from Australian Bureau of Statistics
34
. The average
individual in my sample has an annual household disposable income of $75,703.26 and spends
$13,641.00 on conspicuous goods or services and $34,969.80 on basic goods or services. Figure 1
illustrates the distribution of conspicuous and basic consumption expenditures. For the vast
majority of individuals, the share of household conspicuous consumption expenditures increases
with household total consumption expenditures—there are some fluctuations only among the top
2.5 percent of the individuals. This pattern is largely consistent with the Eagle curves found for
highly observable goods, such as cars and clothing, reported in Heffetz (2011). All monetary
measures are treated as logarithmized
35
variables due to the expected nonlinear relationships and
in order to normalize the skewed income and expenditure distributions.
To directly understand the social comparison effects, I also construct each individual’s
ranking of income, savings, and total, conspicuous, and basic consumption expenditures within his
or her reference group. Ranking is defined as the share of individuals that have less or equal
income, savings, or expenditures (in the corresponding consumption category) in one’s reference
group.

32
The results of my analysis turn out to be similar without eliminating extreme values of income and consumption expenditure
as well as omitting negative values of income.
33
All financial variables constructed so far were household level measures instead of per-capita measures. In order to explore
whether this is problematic, I also constructed equivalized measures (i.e., adjusting the level of income, savings, and
expenditure variables by dividing over the OECD-modified equivalence scale). The regression results are virtually the same.
34
I also tried price unadjusted measures, and results didn’t change much. This is not surprising because there is no severe price
change within the five years.
35
Since there are also a small number of observations (around 0.2 percent of the sample) have zero income, I approximate
logarithm with inverse hyperbolic sine. In the case of negative savings, I take the absolute value of savings before taking
logarithm and add a negative sign after that.
95

Figure 4-1 Relationship between Conspicuous and Basic Consumption Expenditures

Notes: The darker line (left axis) corresponds to a local polynomial regression of share of conspicuous consumption expenditures
on total consumption expenditures, with the shaded area denoting the corresponding 95% confidence interval. The lighter line (right
axis) corresponds to a Kernel estimate of the density distribution of total consumption expenditures. Data is from the period 2006
to 2010 of the HILDA. All expenditures are expressed in constant price of 2006.

The reference group of each individual needs to be defined. Every individual can have his
or her own definition of reference group (e.g., friends, colleagues, and family members), so,
ideally, we want to have individuals identify their own reference groups. Such information,
however, is usually not asked in surveys. In the existing literature, various other approaches have
been taken to define an individual’s reference group in the context of income. Some researchers
choose to define a reference group in a broader manner. For example, McBride (2001) defines
reference groups by age while Stutzer (2004) assumes that individuals compare themselves with
others in the same region. Other researchers define more specific reference groups. For example,
Van de Stadt et al. (1985) defines reference groups according to education level, age, and
96
employment status. Ferrer-i-Carbonell (2005) uses age, education, and region. Paul and Guilbert
(2013) defines reference groups by age and education. In some studies, gender is also considered
as a relevant factor.
In this study, the definition of a reference group is based on a variety of characteristics
including respondent’s age, gender, education level, and geographic location. Specifically, the
reference group is defined as individuals 5 years younger and 5 years older than the respondent.
Gender is separated into males and females, and education is divided into four categories according
to the highest level of education obtained: below high school; high school; vocational degree; and
bachelor’s degree or above. In addition, the geographical region is based on 13 major statistical
regions. Each of my ranking variables takes a value from 0 (corresponding to the lowest level of
expenditure in the reference group) to 1 (the highest). The absolute levels of these monetary
measures are strongly correlated with their relative levels, but the relationships are not perfect: the
correlation coefficients are approximately 0.70. Moreover, because individual fixed effects will be
included in the regression models, it is important to verify that there is significant within-individual
variation in each of the main independent variables. Indeed, according to the one-way analysis of
variance (ANOVA), the within-individual variability in the ranking variables is comparable to the
corresponding between individual variability.

4.3.2.3 Demographic and Socioeconomic Controls

Consistent with the existing literature, a wide variety of demographic and socioeconomic variables
are included in the econometric analysis, including age categories—25–34, 35–44, 45–54, 55–64,
65–74, and 75 or above, with below 25 omitted; highest level of educational—high school,
97
vocational degree, and bachelor’s degree or above, with below high school omitted; marital
status—separated, divorced, or widowed, and never married, with married or de facto omitted;
labor market status—part-time employed, unemployed, and not in the labor force, with full-time
employment omitted; and finally, health status, represented by a dummy on whether the respondent
has any long-term health condition. In order to take the demand-side information into account, I
also control for natural logarithm of number of adults and natural logarithm of number of
dependent children in the household. In addition, year, month of interview, and region dummies
are controlled to take year, seasonal, and regional effects into account. Table 4-3 presents the
summary statistics of all variables used in the empirical analysis.
Table 4-3 Descriptive Statistics, Pooled over All Waves, HILDA 2006–2010
Mean Standard
Deviation
Life satisfaction 7.89 1.45
Household disposable income 75,703.26 62,897.63
Total consumption expenditures 48,610.79 33,488.58
Savings 27,092.47 55,848.01
Conspicuous consumption expenditures 13,641.00 15,330.07
Basic consumption expenditures 34,969.80 24,821.27
Rank (income) 0.51 0.29
Rank (total consumption expenditures) 0.51 0.29
Rank (savings) 0.51 0.29
Rank (conspicuous consumption expenditures) 0.51 0.29
Rank (basic consumption expenditures) 0.51 0.29
Number of adults at home 2.28 1.18
Number of dependent children at home 0.61 1.04
Female 0.53 0.50
Reference group: age 15–24
Age 25–34 0.16 0.37
Age 35–44 0.18 0.38
Age 45–54 0.18 0.38
Age 55–64 0.13 0.34
Age 65–74 0.09 0.29
Age 75 0.07 0.26
Reference group: below high school
High school 0.16 0.36
Vocational degree 0.28 0.45
Bachelor’s degree or above 0.21 0.41
Reference group: married or de facto
Separated, divorced, or widowed 0.14 0.35
Never married 0.25 0.43
Reference group: full-time employed
98
Part-time employed 0.21 0.41
Unemployed 0.04 0.18
Not in the labour force 0.32 0.47
Whether have any long-term health condition 0.27 0.44
Number of observations 47,405
Variables not reported: year, month of interview, and region dummies.
A reference person is a male, aged 24 or below, married or de facto, has below-high-
school education, works full-time, and has no long-term health condition.

4.4 Model Estimation

This study employs a dynamic model of life satisfaction, which includes a lagged dependent
variable as an independent variable. The inclusion of the lagged life satisfaction is theoretically
plausible since past happiness is related to a “hedonic capital”
36
(i.e., a collection of stock-like
variables that affect an individual’s well-being) or investment for future happiness (Graham and
Oswald 2010). Besides, it is also empirically reasonable because past happiness is potentially
related to various current individual socioeconomic characteristics as well as current happiness.
Therefore, the inclusion of the lagged life satisfaction mitigates the problem of reverse causality.
Consider the following dynamic model with unobserved individual fixed effects.
𝐿𝑆
𝑖𝑡
= 𝜆 𝐿𝑆
𝑖𝑡 −1
+ 𝛽 ′𝑀 𝑖𝑡
+ 𝛾 ′
𝑋 1,𝑖𝑡
+ 𝛿 ′𝑋 2,𝑖𝑡
+ 𝜃 ′𝑍 𝑖 + 𝛼 𝑖 + 𝘀 𝑖𝑡
(7)
where 𝐿 𝑆 𝑖𝑡
is life satisfaction of individual i at time t. 𝑀 𝑖𝑡
denotes a vector of measures of material
living conditions, such as income and consumption expenditures. 𝑋 1,𝑖𝑡
is a vector of time-varying
exogenous controls, such as age, month, year, and region
37
dummies. 𝑋 2,𝑖𝑡
is a vector of
endogenous characteristics, including education attainment, marital status, employment status,
health, and household structure. 𝑍 𝑖 is a set of time-invariant individual characteristics (i.e., gender

36
Examples of components of hedonic capital are marital status and health.
37
Region dummies can be endogenous potentially. However, the control of individual fixed effects (𝛼 𝑖 ) largely reduce the
endogeneity of region dummies since over 90% of individuals do not change region of residence during the sample period. I try
to treat region dummies as endogenous as well, and the estimation results appear to be similar.
99
in this study). 𝛼 𝑖 accounts for unobserved individual characteristics that are considered invariant
or almost invariant over time, such as genetics. 𝘀 𝑖𝑡
is the error term.
There is some difficulty in estimating equation (7) because of the existence of both the
lagged dependent variable and the individual fixed effects. A serious bias will arise because a
typical demeaning process in fixed effects estimation results in a correlation between the demeaned
lagged dependent variable and the demeaned error term (Nickell 1981). The bias may be quite
sizable in the “small T” context. Since the bias is due to the construction of the model, one way to
avoid this is to exclude either the lagged dependent variable or the individual fixed effects.
Therefore, I can use the fixed effects estimation when excluding the lagged life satisfaction and
the OLS estimation when excluding the fixed effects.
To estimate equation (7) comprehensively (i.e., include both the lagged life satisfaction
and the individual fixed effects), I use the Arellano-Bond GMM estimator (Arellano and Bond
1991). It not only solves the bias caused by the inclusion of the lagged dependent variable but also
deals with the possible endogeneity of other independent variables. The Arellano-Bond estimation
starts by transforming all regressors, usually by differencing, and uses a set of instrumental
variables to solve endogeneity problems.
Taking the first difference of (7), I have the unobserved individual fixed effects eliminated
from the model:
𝐿𝑆
𝑖𝑡
− 𝐿𝑆
𝑖𝑡 −1
= 𝜆 (𝐿𝑆
𝑖𝑡 −1
− 𝐿𝑆
𝑖𝑡 −2
) + 𝛽 ′(𝑀 𝑖𝑡
− 𝑀 𝑖𝑡 −1
) + 𝛾 ′(𝑋 1,𝑖𝑡
− 𝑋 1,𝑖𝑡 −1
) + 𝛿 ′(𝑋 2,𝑖𝑡
−
𝑋 2,𝑖𝑡 −1
) + (𝘀 𝑖𝑡
− 𝘀 𝑖𝑡 −1
) (8)
The use of instruments is required to deal with two issues. First, some of the explanatory variables,
consumption expenditure or income variables in particular, are endogenous. This is reflected in
the potential correlations between these variables (i.e., 𝑀 and 𝑋 2
) and the error term. Second, by
100
construction, there is correlation between (𝑌 𝑖𝑡 −1
− 𝑌 𝑖𝑡 −2
)and (𝘀 𝑖𝑡
− 𝘀 𝑖𝑡 −1
). Assuming, for the
moment, that (a) the error term 𝘀 𝑖𝑡
is not serially correlated, and (b) the explanatory variables are
weakly exogenous—that is, they may be affected by past and current realizations of the dependent
variable but not by its future innovations—the following moment conditions apply:
𝐸 [𝐿𝑆
𝑖𝑡 −𝑠 (𝘀 𝑖𝑡
− 𝘀 𝑖𝑡 −1
)] = 0 for 𝑠 ≥ 2, t = 3,…,T, (9)
𝐸 [𝑀 𝑖𝑡 −𝑠 (𝘀 𝑖𝑡
− 𝘀 𝑖𝑡 −1
)] = 0 for 𝑠 ≥ 2, t = 3,…,T, (10)
𝐸 [𝑋 2,𝑖𝑡 −𝑠 (𝘀 𝑖𝑡
− 𝘀 𝑖𝑡 −1
)] = 0 for 𝑠 ≥ 2, t = 3,…,T, (11)
However, as suggested by Powdthavee (2009), it is reasonable to assume systematic
measurement errors in life satisfaction (i.e., reporting errors). The details are discussed in
Appendix C.2. With a time-varying measurement error in the lagged dependent variable, the
regression error will follow moving average of order 1. Therefore, following Arellano (2003), I
use higher-order lagged dependent variables as instruments (i.e., 𝑠 ≥ 3 in [3]). Using the moment
conditions (3)–(5) for the difference equation (2), I employ the two-step Arellano-Bond GMM
estimation with Windmeijer (2005) corrected robust standard errors to generate consistent and
efficient estimates.
A crucial assumption for the validity of my estimation is that the instruments are
exogenous. Two specification tests are used to check whether the lagged values of the explanatory
variables are valid instruments. The first is the Hansen test of over-identification (Hansen, 1982),
which tests the overall validity of the instruments by analyzing the sample analogue of the moment
conditions used in the estimation. Failure to reject the null hypothesis gives support to the model.
I rely on the Hansen instead of the Sargen test since the former is robust to heteroskedasticity in
errors while the latter is not. The second test examines the hypothesis that the error term 𝘀 𝑖𝑡
is not
serially correlated. This is done by investigating whether the differenced error is first- or second-
101
order serially correlated. First-order serial correlation of the differenced error term is expected by
construction (i.e., the original error term [in levels] is uncorrelated). However, if the second-order
serial correlation of the differenced error term is found, the original error term is serially correlated
and the instruments are misspecified. Otherwise, we can accept the null hypothesis of no second-
order serial correlation and conclude that the moment conditions are well specified.
The results from the Arellano-Bond GMM method can be understood as causal since it
allows not only for unobserved individual heterogeneity but also for endogeneity of regressors.
The estimates are considered reliable if the two specification tests pass. Nevertheless, there are
two general concerns about the Arellano-Bond estimator. First, as T rises, the instrument count
can easily grow large relative to the sample size, making some asymptotic results on the estimators
and related specification tests misleading. According to Hsiao and Zhou (2017), however, the
Arellano-Bond GMM estimator is consistent if
𝑇 𝑁 → 0 as (𝑁 , 𝑇 ) → ∞ and asymptotically unbiased
if
𝑇 3
𝑁 → 0. Since T is small (T=5) and the panel is wide (number of individuals > 10,000) in this
study, the problem of too many instruments may not be an issue
38
. Second, the instruments in the
Arellano-Bond GMM may be weak, and there is no standard test for weak instruments in this
setting. Nevertheless, the problem may not be an issue in this case as well because I only have a
few lags as instruments. Some studies have used the Blundell-Bond GMM methodology to avoid
the weak-instrument problem. This method augments a level equation to the difference equation
and obtains additional instruments. The initial condition assumptions in the Blundell-Bond GMM,
however, can be violated. I try this method in my analysis, and the results appear to be similar to
the Arellano-Bond GMM results. The problem is that the second-order serial correlation tests do

38
To be conservative, I follow Roodman’s (2009) suggestion to reduce the number of instruments by combining instruments
through their addition into sets. Calderón et al. (2002), Beck and Levine (2004), Carkovic and Levine (2005), and Powdthavee
(2009) have used this technique. The results turn out to be similar.
102
not pass for all of the specifications (the results are presented in Appendix Table C-3). Therefore,
I stick to the Arellano-Bond GMM methodology in my study.

4.5 Results

4.5.1 Life Satisfaction and Income
I first look at what has been widely discussed: the relationship between income and life satisfaction
(see Table 4). My finding on this nationally representative Australian population is quite consistent
with what has been discovered among other populations or in a broader context.
Table 4-4 Life Satisfaction and Income
Dependent Variable: Life Satisfaction

(1) (2) (3) (4) (5) (6) (7) (8)
OLS OLS FE FE AB-GMM
a
AB-GMM
a

AB-
GMM
b

AB-
GMM
b

L life satisfaction 0.558*** 0.535***

0.101*** 0.0990*** 0.220** 0.228**

(0.00625) (0.00633)

(0.0134) (0.0135) (0.112) (0.0943)
Ln (income) 0.0900*** 0.0398*** 0.0433*** 0.0261*** 0.0989 0.172** 0.174** 0.0226

(0.00794) (0.00869) (0.00855) (0.00884) (0.117) (0.0871) (0.0877) (0.139)
Rank (income)

0.128**

(0.0634)
Socioeconomic controls NO YES NO YES NO YES YES YES
Adjusted R-squared 0.342 0.353 0.00214 0.00964

Observations 47,405 47,405 47,405 47,405 33,045 33,045 33,045 33,045
Number of individuals

14,107 14,107 12,277 12,277 12,277 12,277
Number of instruments

52 106 103 109
Specification tests (p-value)

(a) Hansen test of over-identification

0.002 0.493 0.610 0.599
(b) Serial correlation

First-order

0.000 0.000 0.000 0.000
Second-order

0.019 0.015 0.152 0.128
Notes:

Additional controls (all columns) include: age, month of interview, year, and region. Socioeconomic controls include education, marital status,
employment status, health, natural logarithm of number of dependent children at home, and natural logarithm of number of adults at home.
In Arellano-Bond GMM (AB-GMM) estimations, all explanatory variables whenever they appear in the model—except age, year, month, and
region dummies—are treated as endogenous variables; that is, their lagged values are used as the instruments in the difference equation. Age,
year, month, and region dummies are treated as exogenous variables. Their differences are the instruments for themselves in the difference
equation.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1
a: instrumenting from the second-lag life satisfaction

b: instrumenting from the third-lag life satisfaction

103
Columns 1 and 2 show the OLS results, which provide cross-sectional interpretations.
Column 1 only has a set of exogenous variables, including age, year, month of interview, and
region dummies, as controls. Column 2 includes a set of socioeconomic controls, which are
considered potentially endogenous. The lagged life satisfaction has a positive and significant
influence on current life satisfaction. This is consistent with the hedonic capital theory (i.e., our
happiness today depends on how happy we were in the past [Graham and Oswald, 2010]). The
effect of income on life satisfaction is positive and significant though the magnitude is small.
Columns 3 and 4 show the FE results, which provide with-in person interpretations. The lagged
life satisfaction is dropped because its coefficient cannot be estimated consistently by construction
of the fixed effects estimation. In both column 3 and 4, the coefficient on income also turns out to
be positive and significant. The estimated relationships in the first four columns are not necessarily
causal, especially because of the potential endogeneity of some of the independent variables.
Columns 5–7 intend to make causal inference using the Arellano-Bond GMM estimation.
In column 5, the most recent lags (i.e., from the second-lag) of life satisfaction are used as a part
of the instruments, and the controls are the same as what we have in columns 1 and 3. Here, the
effect of the lagged life satisfaction is also positive and significant as expected, but the effect of
household income appears to be positive and insignificant. Furthermore, this specification is
rejected by the second-order serial correlation test as well as the Hansen test of over-identification.
In other words, it may have (1) omitted variables that are correlated with both life satisfaction and
income and/or (2) serial correlations in the error caused by the presence of time-varying
measurement error.
In column 6, I control for a number of endogenous individual characteristics as in columns
2 and 4. This specification passes the Hansen test of over-identification. It indicates the importance
104
of controlling for the individual socioeconomic characteristics that have been omitted in the
previous specification. In this case, the coefficient of household income becomes positive and
significant, and it is approximately four times as large as that in column 2 and almost seven times
as large as that in column 4. This is actually consistent with what has been found in the literature:
Knight et al. (2009), Luttmer (2005), and Powdthavee (2010), all of which attempt to estimate the
causal effect of income on happiness, find that the estimated IV income coefficient is considerably
larger than the estimated OLS or FE income coefficient. However, the second-order serial
correlation test still does not pass. This can be expected because I have not considered the time-
varying measurement error in the lagged life satisfaction.
To solve the problem of the time-varying measurement error bias, in column 7, I drop the
most recent lag of life satisfaction from the instrument list and use only the third lag and beyond
as instruments in the GMM estimation. This specification is now supported by the serial correlation
tests: we do not reject the null hypothesis of no second-order serial correlation with the use of
higher order lags of life satisfaction as instruments. The coefficient on the income variable is
similar to that in column 6, and the coefficient on the lagged life satisfaction becomes larger and
closer to those from the OLS estimations in columns 1 and 2. Specifically, I find that a 1 percent
increase in household disposable income leads to a 0.0017 increase in the level of life satisfaction.
This influence cannot be considered large since the within-person standard deviation of life
satisfaction is around 0.73. In column 8, I add one’s ranking of household income within one’s
reference group into the specification. Consistent with the literature on relative income and well-
being (e.g., Clark et al. 2008), the income ranking has a positive and significant influence on life
satisfaction, and the effect of the level of income disappears when income ranking is controlled.
105
In summary, the Arellano-Bond GMM estimation, which addresses both unobserved
heterogeneity and omitted time-varying variables bias, provides an estimate of the causal effect of
income which is larger than the estimated effect from the OLS or FE estimations. This causal
effect, however, is still considered moderate, and the effect can be mainly explained by one’s
improvement in his or her relative income position. The results are consistent with what has been
found in the past empirical works. This consistency provides some support for the Arellano-Bond
GMM estimation as a reliable method to analyze how material living conditions influence life
satisfaction. Therefore, in order to make my analysis concise, I am only going to present results
from the Arellano-Bond GMM estimation in the following sections, and interested readers can
find the results from the OLS and FE estimation in Appendix Table C-4 and C-5.

4.5.2 Life Satisfaction, Total Consumption Expenditures, and Savings
In order to understand how income influences life satisfaction, I look into components of income
or, in other words, how people spend their income. Specifically, the effects of total consumption
expenditures and savings will be examined in Table 4-5.
The results in columns 1 and 4 show that total household consumption expenditures and
savings have no obvious effects on life satisfaction. In columns 2 and 5, the ranking of
consumption expenditures and the ranking of savings are added, respectively. They also turn out
to have insignificant influences. The results in columns 3 and 6 confirm that those in columns 1
and 4 hold even when income is included among the controls. The finding that savings has no
significant influence on life satisfaction is not surprising. Savings, even though offering financial
security for the future, is largely unobservable. The result that total consumption expenditures have
an insignificant influence motivates us to look into their components.
106
Table 4-5 Life Satisfaction, Total Household Consumption Expenditures, and Savings
Dependent Variable: Life Satisfaction
(1) (2) (3) (4) (5) (6)
AB-GMM AB-GMM AB-GMM AB-GMM AB-GMM AB-GMM
L life satisfaction 0.210* 0.208* 0.184* 0.110* 0.101* 0.193*
(0.124) (0.113) (0.108) (0.063) (0.055) (0.109)
Ln (total consumption expenditures) 0.0445 0.0329 0.0483
(0.176) (0.164)
Rank (total consumption expenditures) 0.144
(0.290)
Ln (savings) 0.00868 0.00634 0.00768
(0.010) (0.011) (0.015)
Rank (savings) 0.0357
(0.289)
Ln (income) 0.151** 0.131*
(0.075) (0.078)
Socioeconomic controls YES YES YES YES YES YES
Observations 33,045 33,045 33,045 33,045 33,045 33,045
Number of individuals 12,277 12,277 12,277 12,277 12,277 12,277
Number of instruments 103 109 109 103 109 109
Specification tests (p-value)
(a) Hansen test of over-identification 0.613 0.605 0.556 0.402 0.432 0.472
(b) Serial correlation
First-order 0 0 0 0 0 0
Second-order 0.172 0.184 0.232 0.759 0.732 0.203
Notes:
Additional controls (all columns) include: age, month of interview, year, and region. Socioeconomic controls include education, marital
status, employment status, health, natural logarithm of number of dependent children at home, and natural logarithm of number of adults
at home.
In Arellano-Bond GMM (AB-GMM) estimations, all explanatory variables whenever they appear in the model—except age, year,
month, and region dummies—are treated as endogenous variables; that is, their lagged values are used as the instruments in the
difference equation (instrumenting from the third-lag life satisfaction). Age, year, month, and region dummies are treated as exogenous
variables. Their differences are the instruments for themselves in the difference equation.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1

4.5.3 Life Satisfaction, Conspicuous Consumption Expenditures, and Basic Consumption
Expenditures
As suggested above, conspicuous consumption expenditures may reflect one’s social status or
aspirations and be more subject to social comparisons. I hypothesize that spending on conspicuous
goods and services should have a stronger positive effect on life satisfaction than spending on basic
ones. I test this hypothesis in Table 4-6.

107
Table 4-6 Life Satisfaction, Conspicuous Consumption Expenditures, and Basic
Consumption Expenditures
Dependent Variable: Life Satisfaction

(1) (2) (3) (4)
AB-GMM AB-GMM AB-GMM AB-GMM
L life satisfaction 0.220** 0.248** 0.239** 0.242**

(0.112) (0.122) (0.120) (0.112)
Ln (conspicuous consumption expenditures)

0.129** 0.112**

(0.0626) (0.0484)

Ln (basic consumption expenditures)

-0.0749 -0.0952

(0.168) (0.222)

Proportion of conspicuous consumption expenditures

0.638

(0.503)
Ln (income) 0.174**

0.0522

(0.0877)

(0.0958)

Socioeconomic controls YES YES YES YES
Observations 33,045 33,045 33,045 33,045
Number of individuals 12,277 12,277 12,277 12,277
Number of instruments 103 109 115 103
Specification tests (p-value)

(a) Hansen test of over-identification 0.610 0.628 0.606 0.548
(b) Serial correlation

First-order 0.000 0.000 0.000 0.000
Second-order 0.152 0.170 0.122 0.595
Notes:

Additional controls (all columns) include: age, month of interview, year, and region. Socioeconomic controls include
education, marital status, employment status, health, natural logarithm of number of dependent children at home, and
natural logarithm of number of adults at home.
In Arellano-Bond GMM (AB-GMM) estimations, all explanatory variables whenever they appear in the model—except
age, year, month, and region dummies—are treated as endogenous variables; that is, their lagged values are used as the
instruments in the difference equation (instrumenting from the third-lag life satisfaction). Age, year, month, and region
dummies are treated as exogenous variables. Their differences are the instruments for themselves in the difference
equation.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1

Column 1 on income is the same as column 7 of Table 4-4, and it is included here for
comparison. Column 2 shows that conspicuous consumption expenditures have a positive and
significant effect, while basic consumption expenditures have a negative and insignificant effect
on life satisfaction. Specifically, a 1 percent increase in conspicuous consumption expenditures
leads to about a 0.0013 increase in the level of life satisfaction. This effect is moderate and similar
to that of income in column 1 (0.0017). When I add household income to the specification (see
column 3), the influence of the conspicuous consumption expenditures remains positive and
108
significant, while that of income becomes no longer important. This implies that spending on
conspicuous goods and services can be an important mechanism through which income influences
life satisfaction.
Another way to look at an individual’s consumption basket is to look at consumption ratios
rather than consumption levels. Consumption ratio is defined as the expenditures on a particular
consumption category divided by total consumption expenditures. In column 4 of Table 4-6, I
include the proportion of conspicuous consumption, which turns out to have a positive though not
significant influence. This suggests that it is the level of conspicuous consumption expenditures
not their share in total consumption expenditures that matters to one’s happiness.
In Table 4-7, I investigate the effect of social comparisons directly by including one’s
relative conspicuous or basic consumption expenditure position, which is measured as one’s
ranking of certain types of consumption expenditures in his or her reference group. Column 1 on
income is the same as column 8 of Table 4-4, and it is included here for comparison. In column 2,
the effect of one’s ranking of conspicuous consumption expenditures within his or her reference
group is positive and statistically significant, but that of one’s ranking of basic consumption
expenditures is not. A 0.1 increase in the ranking of conspicuous consumption expenditures (i.e.,
a one decile increase) translates into a happiness gain of 0.05 (i.e., about 7% of within-person
standard deviation of life satisfaction). In column 3, I add one’s income ranking, which turns out
to have a positive and insignificant influence. In column 4, I add the level of conspicuous and basic
consumption expenditures to the specification in column 2. The effect of the ranking of
conspicuous spending remains positive and significant, but the effect of the level of conspicuous
spending becomes no longer significant. This indicates that the effect of conspicuous consumption
expenditures on life satisfaction comes mainly through social comparisons. In column 5, I add
109
both household income level and one’s ranking of household income within his or her reference
group. Compared with the result in column 1, one’s income ranking becomes no longer significant
when expenditure variables are controlled. The results in Table 4-7, in general, indicate that one’s
relative conspicuous consumption expenditures or social comparisons play an important role in
one’s life satisfaction.
Table 4-7 Life Satisfaction, Relative Conspicuous Expenditures, and Relative Basic
Expenditures
Dependent Variable: Life Satisfaction

(1) (2) (3) (4) (5)
AB-GMM AB-GMM AB-GMM AB-GMM AB-GMM
L life satisfaction 0.228** 0.202** 0.206** 0.252** 0.224**

(0.0943) (0.0896) (0.0865) (0.102) (0.0988)
Ln (conspicuous consumption expenditures)

0.0640 0.0946

(0.0962) (0.0729)
Ln (basic consumption expenditures)

-0.0398 -0.120

(0.0682) (0.348)
Rank (conspicuous consumption expenditures)

0.535** 0.505** 0.512** 0.321**

(0.272) (0.225) (0.202) (0.141)
Rank (basic consumption expenditures)

0.0131 -0.0171 0.00270 0.00692

(0.255) (0.290) (0.492) (0.123)
Ln (income) 0.0226

0.0233

(0.139)

(0.117)
Rank (income) 0.128**

0.113

0.101

(0.0634)

(0.287)

(0.354)
Socioeconomic controls YES YES YES YES YES
Observations 33,045 33,045 33,045 33,045 33,045
Number of individuals 12,277 12,277 12,277 12,277 12,277
Number of instruments 109 109 115 121 133
Specification tests (p-value)

(a) Hansen test of over-identification 0.599 0.776 0.731 0.689 0.614
(b) Serial correlation

First-order 0.000 0.000 0.000 0.000 0.000
Second-order 0.128 0.510 0.564 0.358 0.528
Notes:

Additional controls (all columns) include age, month of interview, year, and region. Socioeconomic controls include
education, marital status, employment status, health, natural logarithm of number of dependent children at home, and
natural logarithm of number of adults at home.
In Arellano-Bond GMM (AB-GMM) estimations, all explanatory variables whenever they appear in the model—except
age, year, month, and region dummies—are treated as endogenous variables; that is, their lagged values are used as the
instruments in the difference equation (instrumenting from the third-lag life satisfaction). Age, year, month, and region
dummies are treated as exogenous variables. Their differences are the instruments for themselves in the difference equation.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1

110
Readers might also be interested in knowing what the result will be if I classify
expenditures according to whether they are on nondurable goods, durable goods, and services.
Table 4-8 shows the result. It seems that expenditures on nondurable goods, durable goods, and
services all have insignificant influences. This implies that this classic classification by economists
cannot really tell us what kinds of consumption expenditures are most associated with life
satisfaction.
Table 4-8 Life Satisfaction and Expenditure on Durable Goods, Nondurable Goods, and
Services
Dependent Variable: Life Satisfaction

(1)
AB-GMM
L life satisfaction 0.299**

(0.129)
Ln (expenditures on durable goods) 0.0445

(0.176)
Ln (expenditures on nondurable goods) 0.0079

(0.045)
Ln (expenditures on services) -0.0326

(0.170)
Socioeconomic controls YES
Observations 33,045
Number of individuals 12,277
Number of instruments 115
Specification tests (p-value)

(a) Hansen test of over-identification 0.386
(b) Serial correlation

First-order 0.000
Second-order 0.291
Notes:

Additional controls (all columns) include age, month of interview, year, and region. Socioeconomic
controls include education, marital status, employment status, health, natural logarithm of number of
dependent children at home, and natural logarithm of number of adults at home.
In Arellano-Bond GMM (AB-GMM) estimations, all explanatory variables whenever they appear in the
model—except age, year, month, and region dummies—are treated as endogenous variables; that is, their
lagged values are used as the instruments in the difference equation (instrumenting from the third-lag life
satisfaction). Age, year, month, and region dummies are treated as exogenous variables. Their differences
are the instruments for themselves in the difference equation.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1

111
4.5.3.1 Heterogeneity among Income Groups

Even though, for the population as a whole, conspicuous consumption expenditures have a positive
and significant influence on life satisfaction and basic consumption expenditures have no obvious
influence, each may have different effects across different income groups. I am going to explore
whether the effects vary across different income quartiles.
To that end, in Table 4-9, I interact the Ln (conspicuous consumption expenditures) and
Ln (basic consumption expenditures) with the income quartile dummies. For conspicuous
consumption expenditures, Ln (conspicuous consumption expenditures) by itself has a positive
and significant influence, and the three relevant interaction terms all have positive but not
significant effects. This means conspicuous consumption expenditures have a positive and
significant influence on life satisfaction for all income quartiles, and the effects for the higher
income quartiles are not statistically different from the effect for the lowest income quartile.
Table 4-9 Life Satisfaction, Conspicuous Consumption Expenditures, and Basic
Consumption Expenditures: Heterogeneity among Income Groups
Dependent Variable: Life Satisfaction

(1)
AB-GMM
L life satisfaction 0.158*

(0.0958)
Ln (conspicuous consumption expenditures) 0.108*

(0.0613)
Ln (conspicuous consumption expenditures) × 2nd income quartile 0.0788

(0.129)
Ln (conspicuous consumption expenditures) × 3rd income quartile 0.0223

(0.220)
Ln (conspicuous consumption expenditures) × 4th income quartile 0.00651

(0.129)
Ln (basic consumption expenditures) -0.528**

(0.257)
Ln (basic consumption expenditures) × 2nd income quartile 0.514

(0.325)
Ln (basic consumption expenditures) × 3rd income quartile 0.209

(0.325)
Ln (basic consumption expenditures) × 4th income quartile 0.632**

(0.313)
112
Socioeconomic controls YES
Tests (p-value)

Ln (basic consumption expenditures) + Ln (basic consumption expenditures) × 2nd
income quartile=0
0.9521
Ln (basic consumption expenditures) + Ln (basic consumption expenditures) × 3rd
income quartile=0
0.1723
Ln (basic consumption expenditures) + Ln (basic consumption expenditures) × 4th
income quartile=0
0.5711
Observations 33,045
Number of individuals 12,277
Number of instruments 163
Specification tests (p-value)

(a) Hansen test of over-identification 0.855
(b) Serial correlation

First-order 0.000
Second-order 0.479
Notes:
Additional controls (all columns) include age, month of interview, year, and region. Socioeconomic controls
include education, marital status, employment status, health, natural logarithm of number of dependent children at
home, natural logarithm of number of adults at home, and income quartiles.
In Arellano-Bond GMM (AB-GMM) estimations, all explanatory variables whenever they appear in the model—
except age, year, month, and region dummies—are treated as endogenous variables; that is, their lagged values are
used as the instruments in the difference equation (instrumenting from the third-lag life satisfaction). Age, wave,
month, and region dummies are treated as exogenous variables. Their differences are the instruments for
themselves in the difference equation.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1

For basic consumption expenditures, Ln (basic consumption expenditures) by itself has a
negative and significant influence, but all the relevant interaction terms turn out to have positive
coefficients with the coefficient of the interaction with the fourth income quartile being significant.
In order to figure out whether the positive effects of the interaction terms offset the negative effect
of Ln (basic consumption expenditures) by itself. I test whether the summation between the
coefficient of Ln (basic consumption expenditures) and that of its each interaction term is
statistically different from zero. The tests show that all three summations are not statistically
different from zero. This means basic consumption expenditures have a negative influence on life
satisfaction only for those in the lowest income quartile.
In summary, conspicuous consumption expenditures contribute to life satisfaction for
individuals in all income groups, while basic consumption expenditures negatively influence life
113
satisfaction for individuals in the lowest income group only. Therefore, it seems that the happiness
of all individuals, no matter which income group they belong to, depends primarily on social
comparisons.

4.6 Robustness Checks

In this section, I test the robustness of my main findings in three dimensions. First, I show that the
findings are robust to varying the definition of conspicuous and basic consumption expenditures.
Second, I examine the effects of the 18 components of consumption expenditures. Third, I test
whether the findings are robust to using financial satisfaction, a domain satisfaction that is more
focused on material sources of happiness, as a dependent variable.

4.6.1 Alternative Categorization of Conspicuous and Basic Consumption Expenditures

As discussed earlier, various definitions of conspicuous consumption expenditures have been
employed in the literature. In this section, I consider whether the identified effect of conspicuous
consumption expenditures is robust to various other classifications. Table 4-10 presents the results.
Column 1 is a copy of baseline result (the same as column 2 of Table 4-6). In column 2, I
include tobacco expenditures, which are highly observable but not necessarily positional, as a part
of conspicuous consumption expenditures. The coefficient on the conspicuous consumption
variable remains positive but it is less significant (at 10% level) and the magnitude is smaller than
the baseline coefficient. In column 3, I consider education expenditures as conspicuous and get a
similar result as in column 1.
114
In column 4, I follow the definition of conspicuous consumption expenditures in Friehe
and Mechtel (2014), which excludes alcohol expenditures as a part of conspicuous consumption
expenditures. The coefficient of conspicuous consumption expenditures is still positive and as
significant as the baseline coefficient but smaller in magnitude.
In column 5, I follow the definition in Charles et al. (2009), which, in my case, only
considers spending on vehicle purchases and clothing and footwear to be conspicuous. As
discussed earlier, this definition is very restrictive, and including spending on furniture and
household appliances, recreational devices and equipment, meals eaten out, alcohol products, and
holidays as parts of basic spending is also not reasonable. Therefore, a new category, namely
“intermediate consumption expenditures,” has been created to include the spending on these items,
which are observable and positional only among colleagues, friends, and family. The results in
column 5 show that both conspicuous consumption expenditures and intermediate consumption
expenditures have positive and significant influences. The effect of the intermediate consumption
expenditures indeed turns out to be slightly larger and more significant than the effect of
conspicuous consumption expenditures.
In column 6, I treat spending on housing as a separate category. The result shows that the
effect of housing expenditures is negative and close to zero. This finding confirms my suspicion
that the housing expenditure measure in this study relates more to financial burdens than to
conspicuousness. My result is consistent with the finding of Plagnol (2011), which shows that
mortgage payments have a negative and significant influence on financial satisfaction. Excluding
spending on housing from basic spending, I still find the coefficient on basic spending to be
negative and insignificant and that on conspicuous spending to be positive and significant.

Table 4-10 Robustness Check: Alternative Definitions of Conspicuous and Basic Consumption Expenditures

Dependent Variable: Life Satisfaction

(1) (2) (3) (4) (5) (6)

Baseline
Tobacco
expenditures as
conspicuous
Education
expenditures as
conspicuous
Friehe & Mechtel
(2014) definition
Charles et al.
(2009) definition
Housing as a
separate
category
AB-GMM AB-GMM AB-GMM AB-GMM AB-GMM AB-GMM
L life satisfaction 0.248** 0.246** 0.256** 0.289** 0.254** 0.258**

(0.122) (0.122) (0.124) (0.123) (0.103) (0.118)
Ln (conspicuous consumption expenditures) 0.129** 0.0996* 0.127** 0.0912** 0.0980* 0.116**

(0.0626) (0.0594) (0.0536) (0.0409) (0.0540) (0.0486)
Ln (intermediate consumption expenditures)

0.116**

(0.0577)

Ln (basic consumption expenditures) -0.0749 -0.0646 -0.0581 -0.0786 -0.0544 -0.0655

(0.168) (0.167) (0.159) (0.162) (0.181) (0.0664)
Ln (housing expenditures)

-0.00113

(0.0290)
Socioeconomic controls YES YES YES YES YES YES
Observations 33,045 33,045 33,045 33,045 33,045 33,045
Number of individuals 12,277 12,277 12,277 12,277 12,277 12,277
Number of instruments 109 109 109 109 115 115
Specification tests (p-value)

(a) Hansen test of over-identification 0.628 0.579 0.651 0.723 0.873 0.454
(b) Serial correlation

First-order 0.000 0.000 0.000 0.000 0.000 0.000
Second-order 0.170 0.112 0.156 0.341 0.121 0.226
Notes:

Additional controls (all columns) include age, month of interview, year, and region. Socioeconomic controls include education, marital status, employment status,
health, natural logarithm of number of dependent children at home, and natural logarithm of number of adults at home.
In Arellano-Bond GMM (AB-GMM) estimations, all explanatory variables whenever they appear in the model—except age, year, month, and region dummies—
are treated as endogenous variables; that is, their lagged values are used as the instruments in the difference equation (instrumenting from the third-lag life
satisfaction). Age, year, month, and region dummies are treated as exogenous variables. Their differences are the instruments for themselves in the difference
equation.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1
4.6.2 Examining the Effects of 18 Components of Consumption Expenditures

The result on conspicuous consumption expenditures seems not surprising if we look into the effect
of each of the 18 components of household consumption expenditures. As can be seen in Table 4-
11, the components that are categorized as conspicuous in the baseline definition mostly have
positive and significant influences on life satisfaction in either the OLS regression or the FE
regression, but those categorized as basic mostly have negative or unobvious influences.

4.6.3 Using Financial Satisfaction Instead of Life Satisfaction as Dependent Variable

Table 4-12 reproduces the main specifications, but instead of using life satisfaction as a dependent
variable, it uses a question about how satisfied you are with your financial situation, the answer to
which is also on an 11-point scale. This measure of subjective well-being seems to be more closely
related to one’s material living conditions. The correlation between life satisfaction and financial
satisfaction is 0.44. Since life satisfaction and financial satisfaction variables have the same scale,
the magnitudes of the coefficients are directly comparable across regressions with these two
measures of subjective well-being. The coefficients from Table 4-12 indicate that the results are
qualitatively similar when using financial satisfaction instead of life satisfaction as a dependent
variable. I also find, quantitatively, the coefficients from Table 4-12 are generally larger than their
counterparts in Table 4-7. Specifically, the effect of income level remains statistically significant
when income ranking is included (column 2), and the effect of the income ranking remains positive
and significant when expenditure measures are included (column 7). The finding seems to be
117
intuitive since financial satisfaction, compared with life satisfaction, is more associated with
objective financial situations.
Table 4-11 Robustness Check: Life Satisfaction and Components of Consumption Expenditures
Dependent Variable: Life Satisfaction

(1) (2)
OLS FE
L life satisfaction 0.531***

(0.00637)

Ln (vehicle purchases) 0.000297 0.00171

(0.00126) (0.00105)
Ln (clothing and footwear) 0.00663** 0.00430

(0.00283) (0.00303)
Ln (furniture and household appliances) 0.00457*** 0.00419***

(0.00147) (0.00131)
Ln (recreational devices and equipment) 0.00279 0.00430***

(0.00171) (0.00155)
Ln (meals out) -0.000700 0.00538**

(0.00216) (0.00239)
Ln (alcohol) 0.00374** 0.00165

(0.00164) (0.00235)
Ln (holidays) 0.0107*** 0.00479***

(0.00168) (0.00180)
Ln (tobacco) -0.00904*** -0.00697***

(0.00164) (0.00253)
Ln (education) -4.80e-05 0.00252

(0.00171) (0.00200)
Ln (groceries) 0.00857 0.00359

(0.00603) (0.00619)
Ln (housing) -0.00648*** -0.00307

(0.00202) (0.00243)
Ln (public transportation) -0.00333** -0.00217

(0.00170) (0.00207)
Ln (motor vehicle repairs and maintenance) -0.00455 -0.00262

(0.00298) (0.00296)
Ln (motor vehicle fuels and engine oil) 0.000577 0.00246

(0.00343) (0.00370)
Ln (phone rent and calls, and internet charges) -0.00766* -0.00231

(0.00465) (0.00478)
Ln (health care (health insurance included) and child care) -0.0151*** -0.00670*

(0.00352) (0.00397)
Ln (home utilities) -0.00342 -0.00439

(0.00457) (0.00473)
Ln (other Insurance) 0.0105*** 0.00498

(0.00316) (0.00343)
Socioeconomic controls YES YES
Observations 47,405 47,405
Number of individuals

14,107
Adjusted R-squared 0.355 0.0108
Notes:

Additional controls (all columns) include: age, gender, month of interview, year, and region. Socioeconomic controls include
education, marital status, employment status, health, natural logarithm of number of dependent children at home, and natural
logarithm of number of adults at home.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1
Table 4-12 Robustness Check: Using Financial Satisfaction Instead of Life Satisfaction as a Dependent Variable

Dependent Variable: Financial Satisfaction

(1) (2) (3) (4) (5) (6) (7)

AB-GMM AB-GMM AB-GMM AB-GMM AB-GMM AB-GMM AB-GMM
L financial satisfaction 0.257*** 0.286*** 0.250** 0.380*** 0.223* 0.321*** 0.346***

(0.0925) (0.0950) (0.109) (0.104) (0.116) (0.103) (0.0847)
Ln (income) 0.456*** 0.299**

0.239

0.129

(0.119) (0.144)

(0.151)

(0.158)
Rank (income)

0.844**

0.247*

(0.400)

(0.133)
Ln (conspicuous consumption expenditures)

0.234*** 0.342***

0.151 0.269

(0.0710) (0.112)

(0.132) (0.763)
Ln (basic consumption expenditures)

0.0442 0.153

0.0502 0.204

(0.222) (0.304)

(0.0916) (0.129)
Rank (conspicuous consumption expenditures)

1.633*** 1.511** 1.044**

(0.465) (0.725) (0.425)
Rank (basic consumption expenditures)

0.331 0.871 0.399

(0.395) (1.289) (0.495)
Socioeconomic controls YES YES YES YES YES YES YES
Observations 33,047 33,047 33,047 33,047 33,047 33,047 33,047
Number of individuals 12,273 12,273 12,273 12,273 12,273 12,273 12,273
Number of instruments 103 109 109 115 109 121 133
Specification tests (p-value)

(a) Hansen test of over-identification 0.508 0.325 0.720 0.851 0.500 0.497 0.743
(b) Serial correlation

First-order 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Second-order 0.121 0.137 0.543 0.307 0.267 0.348 0.222
Notes:

Additional controls (all columns) include: age, month of interview, year, and region. Socioeconomic controls include education, marital status, employment status,
health, natural logarithm of number of dependent children at home, and natural logarithm of number of adults at home.
In Arellano-Bond GMM (AB-GMM) estimations, all explanatory variables whenever they appear in the model —except age, year, month, and region dummies—are
treated as endogenous variables; that is, their lagged values are used as the instruments in the difference equation (instrumenting from the third-lag financial
satisfaction). Age, year, month, and region dummies are treated as exogenous variables. Their differences are the instruments for themselves in the difference equation.
Robust standard errors in parentheses (clustered at individual level)

Significance: *** p<0.01, ** p<0.05, * p<0.1

4.7 Conclusions

The relationship between material living conditions and SWB has been widely explored in the
literature, but most of the studies have focused on income. Recent research has begun to investigate
the effect of consumption expenditures on SWB mainly because expenditure data have become
more widely collected in household surveys.
This study contributes to the relatively scarce literature on consumption expenditures and
life satisfaction. It suggests that one plausible mechanism through which income influences
happiness is spending on conspicuous goods and services. A 1 percent increase in the level of
conspicuous expenditures leads to about a 0.0013 increase in the level of life satisfaction. This
magnitude is moderate and close to the effect of income. On the other hand, spending on basic
goods and services does not influence life satisfaction. I also test heterogeneity among income
groups and find that conspicuous consumption expenditures contribute to life satisfaction for
individuals of all income groups but that basic consumption expenditures only negatively influence
the life satisfaction for individuals in the lowest income quartile.
Moreover, the study finds that it is indeed one’s household spending on conspicuous
consumption relative to that of others in one’s reference group that contributes to life satisfaction.
This finding underscores the importance of social comparisons to people’s wellbeing. Social
comparison or interdependence, however, has not been taken into account in utility functions by
traditional economists. Therefore, my study, together with previous studies on social comparisons
of income, suggests that interdependence should be modeled in utility functions.
The analysis presented here is subject to certain limitations. First, the current study focuses
on Australia, a developed country. It would be interesting and important to test the relationship
120
between conspicuous or basic consumption expenditures and life satisfaction in other developed
countries and in developing countries. I suspect that the outcomes may well be based on the social
structure and the wealth of different societies. Second, there might be heterogeneity among
genders. Males and females often have different preferences and face different constraints, so their
SWB can be influenced differently by different factors. There might also be life-cycle
heterogeneity. Family structures and financial burdens change over the life course, and these
changes can influence the effect of conspicuous or basic consumption expenditures on life
satisfaction. In order to understand the resulting heterogeneity, detailed life-cycle analysis is
needed.

121

CHAPTER 5. Summary and Conclusions

This dissertation provides new evidence on the association between various life circumstances and
people’s quality of life measured by self-reported life satisfaction. In the past few decades,
economists have been increasingly interested in studying subjective well-being measures, such as
life satisfaction, and policy makers have also started to use these measures to design and evaluate
policies. The three studies presented in this dissertation provide analysis mostly on how material
living conditions, family support, and social comparison are each associated with life satisfaction.
The results are based on individual-level data from three very different populations. The
dissertation provides new knowledge that could be used to design policies aiming at promoting the
overall societal well-being.
In Chapter 2, I study the life satisfaction of the Inuit, an indigenous population living on
the barren northernmost fringes of Alaska. The population is very distinct from the U.S. general
population since it was a hunter-gatherer population half a century ago and still retains much of its
hunter-gatherer life style nowadays. My analysis suggests that the factors that are positively
associated with the Inuit’s life satisfaction are health, subsistence hunting and fishing, social
support, and having both Christian beliefs and indigenous beliefs as part of their life. However,
wage income is found to be negatively associated with their life satisfaction. One possible
explanation is that the Inuit’s active involvement in wage employment makes them feel
disconnected from nature, culture, and community, thus creating dissatisfaction. My analysis
suggests that it is essential to design policies that are suitable for the region.

122
Future studies may further investigate the reason for the negative association between wage
income and life satisfaction. In particular, it will be interesting to look into the types of occupations
and how they relate to wage income as well as life satisfaction. My preliminary analysis (not
presented here) suggests that males and females specialize in different types of traditional activities
and different types of occupations. Therefore, it is reasonable to conduct analysis for males and
females separately to explore relationship heterogeneity between genders.
Chapter 3 studies the relationship between intergenerational support and the life
satisfaction of older parents in China. My regression analysis suggests that, in general, the life
satisfaction of the rural parents is more influenced by intergenerational support than that of their
urban counterparts. Living in a three-generational household is associated with the highest level
of life satisfaction for rural parents but this is not true for urban parents. Rural parents’ life
satisfaction is positively influenced by the exchanges of both financial and emotional support. In
contrast, urban parents’ life satisfaction is only positively influenced by receiving help with self-
care or household tasks from children. This study implies that China’s development, especially the
development in urban China, is changing the older parents’ need for support.
Future studies can look into the relationship between family support and well-being in
China from different perspectives. First, the reasons for the relatively weak relationship between
family support and life satisfaction in urban China can be explored. Is it mainly because of the
more established pension system for urban parents? Or, is it also because of the stricter family
planning policy in urban China in the past few decades? Second, there was a reform on the family
planning policies in 2012, and the current policy allows urban parents to have two children like
rural parents. It will be interesting and inspiring to investigate the casual effect of an increase in
family size on life satisfaction of urban parents using this exogenous policy change.
123
Chapter 4 investigates the importance of social comparison to people’s well-being by
studying the effects of consumption expenditures on life satisfaction in Australia. Using data from
a nationally representative household panel survey in Australia, the study finds that one important
channel through which income influences life satisfaction is conspicuous spending. My analysis
also suggests that it is one’s relative conspicuous expenditures within his or her reference group
that really matter to an individual’s life satisfaction. The result also shows that conspicuous
expenditures positively influence the life satisfaction of individuals in all income groups, a finding
further pointing out the importance of social comparison to people’s well-being.
Because there is relatively little literature on how individuals’ spending choices affect their
SWB, there are several possible directions for future research. First, this relationship has generally
not yet been explored in many countries and regions. While my research on Australia finds that it
is conspicuous spending that contributes to life satisfaction, I suspect that the findings will be
different for other societies because the relationships seem to depend on the social structure and
the wealth of different societies. Second, it would be interesting and meaningful to investigate
gender heterogeneity. Males and females often have different preferences and face different
constraints, so their SWB might be influenced differently by different factors. Besides, there might
be life-cycle heterogeneity as well. For example, family structures and financial burdens change
over the life course, and these changes can influence the effect of conspicuous or basic spending
on life satisfaction. In order to understand the resulting heterogeneity, detailed life-cycle analysis
is needed.
While the economics of happiness is still relatively young, beginning in 1974 with Richard
Easterlin’s famous work (Easterlin, 1974), research in this field is becoming increasingly
prominent. With a growing interest among economists in the study of SWB measures and an
124
increasing availability of data, the field of happiness economics will continue to grow. As far as I
am concerned, there are two general topics that researchers in this field can look more into in the
future. First, it will be important and meaningful to evaluate the causal relationship between
environmental damage and SWB. A number of policy goals, such as Sustainable Development
Goals by the United Nations, highlight the importance of the environment. However, it is still
unclear how much people value good environment. Second, more attention should be drawn to the
relationship between the use of social media and happiness. The increasing availability of social
network sites has altered the ways we interact with our colleagues, friends and family. We can
study, for example, whether making more friends in social network sites makes us happy and
whether the interactions with our friends through these sites improve our happiness more than the
interactions with our friends in the real life. The increasing use of social media has not only
changed the ways we can get support but also given us more opportunities to get information. For
example, we can know, through the social network sites, whether our friends recently got promoted
to a new position or whether they just returned from an oversea travel. The large amount of online
information makes us easier to compare with our peers. The relationship between the frequency of
social network sites use and happiness can be explored to understand social comparison in the
contemporary world.

References
Ambrey, C. L., & Fleming, C. M. (2014). Life satisfaction in Australia: Evidence from ten years
of the HILDA survey. Social Indicators Research, 115(2), 691-714.
Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence
and an application to employment equations. The Review of Economic Studies, 58(2), 277–
297.
Arellano, M. (2003). Panel data econometrics. Oxford University Press.
Beck, T., & Levine, R. (2004). Stock markets, banks, and growth: Panel evidence. Journal of
Banking & Finance, 28(3), 423–442.
Berman, M. (2009). Moving or staying for the best part of life: theory and evidence for the role of
subsistence in migration and well-being of Arctic Inupiat residents. Polar Geography, 32(1-
2), 3-16.
Blanchflower, D. G. (2009). International evidence on well-being. In Measuring the subjective
well-being of nations: National accounts of time use and well-being (pp. 155-226). University
of Chicago Press.
Blanchflower, D. G., & Oswald, A. J. (2004a). Well-being over time in Britain and the
USA. Journal of Public Economics, 88(7), 1359–1386.
Blanchflower, D. G., & Oswald, A. J. (2004b). Money, sex and happiness: An empirical study. The
Scandinavian Journal of Economics, 106(3), 393–415.
Blanchflower, D. G., & Oswald, A. J. (2008). Is well-being U-shaped over the life cycle? Social
Science & Medicine, 66(8), 1733-1749.
126
Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data
models. Journal of Econometrics, 87(1), 115–143.
Brickman, P., Coates, D., & Janoff-Bulman, R. (1978). Lottery winners and accident victims: Is
happiness relative? Journal of Personality and Social Psychology, 36(8), 917.
Carkovic, M., & Levine, R. (2005). “Does foreign direct investment accelerate economic growth.”
In The impact of foreign direct investment on development: New measurements, new
outcomes, new policy approaches.
Charles, K. K., Hurst, E., & Roussanov, N. (2009). Conspicuous consumption and race. The
Quarterly Journal of Economics, 124(2), 425–467.
Chen, F., & Short, S. E. (2008). Household context and subjective well-being among the oldest
old in China. Journal of family issues, 29(10), 1379-1403.
Chen, X., & Silverstein, M. (2000). Intergenerational social support and the psychological well-
being of older parents in China. Research on aging, 22(1), 43-65.
Chou, K. L., Ho, A. H. Y., & Chi, I. (2006). Living alone and depression in Chinese older
adults. Aging and Mental Health, 10(6), 583-591.
Clark, A. E., & Oswald, A. J. (1994). Unhappiness and unemployment. The Economic Journal,
648-659.
Clark, A. E., & Oswald, A. J. (1996). Satisfaction and comparison income. Journal of Public
Economics, 61(3), 359–381.
Clark, A. E., & Senik, C. (2010). Who compares to whom? The anatomy of income comparisons
in Europe. The Economic Journal, 120(544), 573–594.
127
Clark, A. E., Frijters, P., & Shields, M. A. (2008). Relative income, happiness, and utility: An
explanation for the Easterlin paradox and other puzzles. Journal of Economic
Literature, 46(1), 95–144.
Costa, D. L. (1998). Displacing the family. In The Evolution of Retirement: An American
Economic History, 1880-1990 (pp. 106-132). University of Chicago Press.
Cutler, D. M., & Katz, L. F. (1992). Rising inequality? changes in the distribution of income and
consumption in the 1980's. The American Economic Review, 82, 546–551.
De Neve, J. E., & Oswald, A. J. (2012). Estimating the influence of life satisfaction and positive
affect on later income using sibling fixed effects. Proceedings of the National Academy of
Sciences, 109(49), 19953–19958.
DeLeire, T., & Kalil, A. (2010). Does consumption buy happiness? Evidence from the United
States. International Review of Economics, 57(2), 163–176.
Di Tella, R., MacCulloch, R. J., & Oswald, A. J. (2001). Preferences over inflation and
unemployment: Evidence from surveys of happiness. American Economic Review, 335-341.
Diener, E., & Chan, M. Y. (2011). Happy people live longer: Subjective well-being contributes to
health and longevity. Applied Psychology: Health and Well‐Being, 3(1), 1-43.
Doty, P. (1986). Family care of the elderly: The role of public policy. The Milbank Quarterly, 34-
75.
Dumludag, D. (2015). Consumption and life satisfaction at different levels of economic
development. International Review of Economics, 62(2), 163–182.
Easterlin, R. A. (1974). Does economic growth improve the human lot? Some empirical
evidence. Nations and households in economic growth, 89, 89-125.
128
Easterlin, R. A. (1995). Will raising the incomes of all increase the happiness of all? Journal of
Economic Behavior & Organization, 27(1), 35–47.
Easterlin, R. A. (2001). Income and happiness: Towards a unified theory. Economic Journal, 465-
484.
Easterlin, R. A. (2010). Happiness, growth, and the life cycle. OUP Catalogue.
Ferrer-i-Carbonell, A. (2005). Income and well-being: An empirical analysis of the comparison
income effect. Journal of Public Economics, 89(5), 997–1019.
Frank, R. H. (1985a). Choosing the right pond: Human behavior and the quest for status. Oxford
University Press.
Frank, R. H. (1985b). The demand for unobservable and other nonpositional goods. The American
Economic Review, 75(1), 101–116.
Frank, R. H. (1999). Luxury fever: Money and happiness in an era of excess. Princeton University
Press.
Frey, B. S. (2011). Happy people live longer. Science, 331(6017), 542-543.
Frey, B. S., & Stutzer, A. (2000). Happiness, economy and institutions. The Economic
Journal, 110(466), 918-938.
Frey, B. S., & Stutzer, A. (2002). What can economists learn from happiness research? Journal of
Economic literature, 402-435.
Frey, B. S., & Stutzer, A. (2010). Happiness and economics: How the economy and institutions
affect human well-being. Princeton University Press.
Frick, J. R., Jenkins, S. P., Lillard, D. R., Lipps, O., & Wooden, M. (2007). The Cross-National
Equivalent File (CNEF) and its member country household panel studies. Schmollers
Jahrbuch: Journal of Applied Social Science Studies, 127(4), 627–654.
129
Friedman, J., & McCabe, A. (1996). Preferences or happiness? Tibor Scitovsky's psychology of
human needs.
Friehe, T., & Mechtel, M. (2014). Conspicuous consumption and political regimes: Evidence from
East and West Germany. European Economic Review, 67, 62–81.
Gokdemir, O. (2015). Consumption, savings and life satisfaction: the Turkish case. International
Review of Economics, 62(2), 183–196.
Graham, C. L., & Pettinato, S. (2004). Happiness and hardship: Opportunity and insecurity in new
market economies. Brookings Institution Press.
Graham, L., & Oswald, A. J. (2010). Hedonic capital, adaptation and resilience. Journal of
Economic Behavior & Organization, 76(2), 372–384.
Gu, S., Zhu, N., Chen, X., & Liang, J. (1995). Old-age support system and policy reform in
China. Korea journal of population and development, 24(2), 245-273.
Guillen-Royo, M. (2008). Consumption and subjective wellbeing: Exploring basic needs, social
comparison, social integration and hedonism in Peru. Social Indicators Research, 89(3), 535–
555.
Guven, C. (2012). Reversing the question: Does happiness affect consumption and savings
behavior? Journal of Economic Psychology, 33(4), 701–717.
Haisken-DeNew, J. P. (2001). A hitchhiker's guide to the world's household panel data
sets. Australian Economic Review, 34(3), 356–366.
Hansen, L. P. (1982). Large sample properties of generalized method of moments
estimators. Econometrica: Journal of the Econometric Society, 1029–1054.
Hartog, J., & Oosterbeek, H. (1998). Health, wealth and happiness: why pursue a higher education?
Economics of Education Review, 17(3), 245-256.
130
Hayes, C., & Watson, N. (2009). HILDA imputation methods.
Hayo, B., & Seifert, W. (2003). Subjective economic well-being in Eastern Europe. Journal of
Economic Psychology, 24(3), 329-348.
Headey, B., Muffels, R., & Wooden, M. (2004). Money doesn't buy happiness . . . Or does it? A
reconsideration based on the combined effects of wealth, income and consumption.
Heffetz, O. (2011). A test of conspicuous consumption: Visibility and income elasticities. Review
of Economics and Statistics, 93(4), 1101–1117.
Helliwell, J. F. (2003). How's life? Combining individual and national variables to explain
subjective well-being. Economic Modelling, 20(2), 331-360.
Helliwell, J. F. (2008). Life satisfaction and quality of development (No. w14507). National
Bureau of Economic Research.
Helliwell, J. F., Layard, R., & Sachs, J. (2013). World happiness report.
Howe, E. L. (2009). Patterns of migration in Arctic Alaska. Polar Geography, 32(1-2), 69-89.
Hsiao, C., & Zhou, Q. (2017). JIVE for Panel Dynamic Simultaneous Equations Models.
Huntington, H. P. (1992). Wildlife management and subsistence hunting in Alaska. Belhaven
Press.
Kahneman, D., & Thaler, R. H. (2006). Anomalies: Utility maximization and experienced
utility. The Journal of Economic Perspectives, 20(1), 221–234.
Kingdon, G. G., & Knight, J. (2007). Community, comparisons and subjective well-being in a
divided society. Journal of Economic Behavior & Organization, 64(1), 69–90.
Knight, J., & Gunatilaka, R. (2010). The rural–urban divide in China: income but not
happiness?. The Journal of Development Studies, 46(3), 506-534.
131
Knight, J., Lina, S. O. N. G., & Gunatilaka, R. (2009). Subjective well-being and its determinants
in rural China. China Economic Review, 20(4), 635-649.
Knight, J., Lina, S. O. N. G., & Gunatilaka, R. (2009). Subjective well-being and its determinants
in rural China. China Economic Review, 20(4), 635–649.
Kotlikoff, L. J., & Morris, J. N. (1990). Why don't the elderly live with their children? A new look.
In Issues in the Economics of Aging (pp. 149-172). University of Chicago Press, 1990.
Krause, N., & Liang, J. (1993). Stress, social support, and psychological distress among the
Chinese elderly. Journal of Gerontology, 48(6), P282-P291.
Kruse, J. (1991). Alaska Inupiat subsistence and wage employment patterns: understanding
individual choice. Human Organization, 50(4), 317-326.
Kwan, C. H. (2013). Pension Reform in China: Meeting the Challenges Posed by Societal
Aging. Nomura Journal of Capital Markets, 4(3).
Langdon, S. J. (1991). The integration of cash and subsistence in Southwest Alaskan Yup'ik
Eskimo communities. Senri Ethnological Studies.
Laugrand, F. B., & Oosten, J. G. (2010). Inuit shamanism and Christianity: Transitions and
transformations in the twentieth century (Vol. 58). McGill-Queen's Press-MQUP.
Lei, X., Strauss, J., Tian, M., & Zhao, Y. (2011). Living Arrangements of the Elderly in China.
Lewis, J. (2011). Income, expenditure and personal well-being.
Luttmer, E. F. (2005). Neighbors as negatives: Relative earnings and well-being. The Quarterly
Journal of Economics, 120(3), 963–1002.
Martin, L. G. (1989). Living arrangements of the elderly in Fiji, Korea, Malaysia, and the
Philippines. Demography, 26(4), 627-643.
132
Martin, S. L. (2005). Determinants of well-being in Inupiat and Yupiit Eskimos: Do communities
matter? (Doctoral dissertation, University of Texas at Dallas).
Martin, S. (2012). Cultural continuity and communities and well-being.
Marx, K., & Engels, F. (1942). Wage-labour and Capital. Workers’ Literature Bureau, 208–209.
McBride, M. (2001). Relative-income effects on subjective well-being in the cross-
section. Journal of Economic Behavior & Organization, 45(3), 251–278.
McGarry, K., & Schoeni, R. F. (2000). Social security, economic growth, and the rise in elderly
widows’ independence in the twentieth century.Demography, 37(2), 221-236
Meng, X., & Luo, C. (2008). What determines living arrangements of the elderly in urban
China. Pp267 –286 in Inequality and Public Policy in China, Edited by BA Gustafsson, S. Li.
Mutchier, J. E., & Burr, J. A. (1991). A longitudinal analysis of household and nonhousehold
living arrangements in later life. Demography, 28(3), 375-390.
Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica: Journal of the
Econometric Society, 1417–1426.
Nijman, T., & Verbeek, M. (1992). Nonresponse in panel data: The impact on estimates of a life
cycle consumption function. Journal of Applied Econometrics, 7(3), 243–257.
Noll, H. H., & Weick, S. (2015). Consumption expenditures and subjective well-being: empirical
evidence from Germany. International Review of Economics, 62(2), 101–119.
OECD Better Life Initiative. (2013). OECD guidelines on measuring subjective well-being.
Oreopoulos, P., & Salvanes, K. G. (2011). Priceless: The nonpecuniary benefits of schooling. The
Journal of Economic Perspectives, 159-184.
Oswald, A. J., & Powdthavee, N. (2008). Does happiness adapt? A longitudinal study of disability
with implications for economists and judges. Journal of Public Economics, 92(5), 1061-1077.
133
Palmer, E., & Deng, Q. (2008). What has economic transition meant for the well-being of the
elderly in China. Pp182 –203 in Inequality and Public Policy in China, Edited by BA
Gustafsson, S. Li.
Paul, S., & Guilbert, D. (2013). Income–happiness paradox in Australia: Testing the theories of
adaptation and social comparison. Economic Modelling, 30, 900–910.
Perez-Truglia, R. (2013). A test of the conspicuous–consumption model using subjective well-
being data. The Journal of Socio-Economics, 45, 146–154.
Plagnol, A. C. (2011). Financial satisfaction over the life course: The influence of assets and
liabilities. Journal of Economic Psychology, 32(1), 45–64.
Poppel, B., & Kruse, J. (2009). The Importance of a Mixed Cash-and Harvest Herding Based
Economy to Living in the Arctic–An Analysis on the Survey of Living Conditions in the
Arctic (SLiCA). In Quality of Life and the Millennium Challenge (pp. 27-42). Springer
Netherlands.
Poterba, J. M. (1989). Lifetime incidence and the distributional burden of excise taxes. The
American Economic Review, 79, 325–330.
Powdthavee, N. (2009). I can’t smile without you: Spousal correlation in life satisfaction. Journal
of Economic Psychology, 30(4), 675–689.
Powdthavee, N. (2010). How much does money really matter? Estimating the causal effects of
income on happiness. Empirical Economics, 39(1), 77–92.
Ren, Q., & Treiman, D. J. (2015). Living arrangements of the elderly in China and consequences
for their emotional well-being. Chinese Sociological Review, 47(3), 255-286.
134
Rojas, M. (2007). The complexity of well-being: A life-satisfaction conception and a domains-of-
life approach. Researching well-being in developing countries: From theory to research, 259-
280.
Sachs, J. D., Helliwell, J. F., & Layard, R. (2013). World Happiness Report 2012. New York:
Earth Institute, Columbia University.
Schwarz, B., Albert, I., Trommsdorff, G., Zheng, G., Shi, S., & Nelwan, P. R. (2010).
Intergenerational support and life satisfaction: A comparison of Chinese, Indonesian, and
German elderly mothers. Journal of Cross-Cultural Psychology.
Scitovsky, T. (1976). The joyless economy: An inquiry into human satisfaction and consumer
dissatisfaction.
Senik, C. (2004). When information dominates comparison: Learning from Russian subjective
panel data. Journal of Public Economics, 88(9), 2099-2123.
Senik, C. (2009). Direct evidence on income comparisons and their welfare effects. Journal of
Economic Behavior & Organization, 72(1), 408–424.
Sicular, T., Ximing, Y., Gustafsson, B., & Shi, L. (2007). The urban–rural income gap and
inequality in China. Review of Income and Wealth, 53(1), 93-126.
Silverstein, M., Cong, Z., & Li, S. (2006). Intergenerational transfers and living arrangements of
older people in rural China: Consequences for psychological well-being. The Journals of
Gerontology Series B: Psychological Sciences and Social Sciences, 61(5), S256-S266.
Stanca, L., & Veenhoven, R. (2015). Consumption and happiness: an introduction. International
Review of Economics, 62(2), 91.
Stevenson, B., & Wolfers, J. (2008). Economic growth and subjective well-being: Reassessing the
Easterlin paradox (No. w14282). National Bureau of Economic Research.
135
Stevenson, B., & Wolfers, J. (2009). The paradox of declining female happiness (No. w14969).
National Bureau of Economic Research.
Strauss, J., & Thomas, D. (1998). Health, nutrition, and economic development. Journal of
Economic Literature, 36(2), 766–817.
Stutzer, A. (2004). The role of income aspirations in individual happiness. Journal of Economic
Behavior & Organization, 54(1), 89–109.
Stutzer, A., & Frey, B. S. (2006). Does marriage make people happy, or do happy people get
married? The Journal of Socio-Economics, 35(2), 326-347.
Sun, C. (2010). HILDA expenditure imputation. HILDA Project Technical Paper Series, (1/10).
Sun, R. (2002). Old age support in contemporary urban China from both parents’ and children’s
perspectives. Research on Aging, 24(3), 337-359.
Van de Stadt, H., Kapteyn, A., & Van de Geer, S. (1985). The relativity of utility: Evidence from
panel data. The Review of Economics and Statistics, 179–187.
Van Praag, B. M., & Ferrer-i-Carbonell, A. (2004). Happiness quantified: A satisfaction calculus
approach. Oxford University Press, Chapter 4.
Veblen, T. (1925 [1899]). The theory of the leisure class; An economic study of institutions.
Macmillan.
Wang, H., Cheng, Z., & Smyth, R. (2015). Does consuming more make you happier? Evidence
from Chinese panel data.
Wang, J., Chen, T., & Han, B. (2014). Does co-residence with adult children associate with better
psychological well-being among the oldest old in China?. Aging & mental health, 18(2), 232-
239.
136
Watson, N., & Wooden, M. P. (2012). The HILDA Survey: a case study in the design and
development of a successful household panel survey. Longitudinal and Life Course
Studies, 3(3), 369–381.
Windmeijer, F. (2005). A finite sample correction for the variance of linear efficient two-step
GMM estimators. Journal of Econometrics, 126(1), 25–51.
Winkelmann, L., & Winkelmann, R. (1998). Why are the unemployed so unhappy? Evidence from
panel data. Economica, 65(257), 1-15.
Winkelmann, R. (2012). Conspicuous consumption and satisfaction. Journal of Economic
Psychology, 33(1), 183–191.
Wolfe, R. J., & Walker, R. J. (1987). Subsistence economies in Alaska: productivity, geography,
and development impacts. Arctic Anthropology, 56-81.
Wooldridge, J. M. (2010). Econometric analysis of cross section and panel data. MIT Press.
Yang, H. (1996). The distributive norm of monetary support to older parents: A look at a township
in China. Journal of Marriage and the Family, 404-415.
Yang, H., CHHANDLER, D., & Chandler, D. (1992). Intergenerational relations: Grievances of
the elderly in rural China. Journal of Comparative Family Studies, 431-453.
Yi, Z., & Wang, Z. (2003). Dynamics of family and elderly living arrangements in China: New
lessons learned from the 2000 census. China Review, 95-119.
Zeng, Y., Brasher, M. S., Gu, D., & Vaupel, J. W. (2016). Older Parents Benefit More in Health
Outcome From Daughters’ Than Sons’ Emotional Care in China. Journal of aging and health,
0898264315620591.
Zhang, L. (2015). Living Arrangements and Subjective Well-Being among the Chinese
Elderly. Open Journal of Social Sciences, 3(03), 150.
137
Zhou, M., & Qian, Z. (2008). Social Support and Self-Reported Quality of Life China’s Oldest
Old. In Healthy longevity in China (pp. 357-376). Springer Netherlands
Zimmer, Z., & Kwong, J. (2003). Family size and support of older adults in urban and rural China:
Current effects and future implications. Demography,40(1), 23-44.
Zimmermann, S. (2014). The Pursuit of subjective well-being through specific consumption
choice.

138
APPENDIX A Supplementary Material for Chapter 2

Appendix Table A-1 Definition of Key Variables
Variable Definition
Life Satisfaction
Response to the question "please tell me the number on this card that fits how satisfied
you are with your life as a whole". 1=very dissatisfied, 2=somewhat dissatisfied,
3=neither satisfied nor dissatisfied, 4=somewhat satisfied, 5=very satisfied
untreated medical problem
Constructed from the question "do you have an untreated medical problem?" 1=Yes,
0=No or Don't know
number of leisure activities
Constructed from the question "Which of these activities did you do in the past 12
months? Please just tell me the letters of the activities you did: a. play bingo, b. take part
in a native festival, c. read books or magazines, d. listen to the radio or stereo, e. visit
neighbors, friends, and family, f. listen to or tell a native story, g. go to sports events, h.
participate in sports, i. take part in a native dance, j. take part in native traditional games,
k. go snowmobiling or dog sledding, l. hike, run, jog, or walk, m. boat or kayak, n. be out
in the country,". number of leisure activities=the number of activities that a respondent
mentioned.
number of traditional
activities
Constructed from the question "please tell me the letters of any of the activities you did in
te last 12 months: a. member of a whaling crew, b. skinned and butchered a caribou, c.
help whaling crews by cooking, giving money or supplies, cutting meat, d. sew skins,
make parkas and kamiks, make sleds or boats, e. make sleds or boats, f. hunt caribou,
moose, or sheep, g. hunt seal or ugruk, h. hunt walrus, i. hunt waterfowl, j. gather eggs, k.
fish, l. gather greens, roots, or other plants, m. preserve meat or fish, n. trap, o. pick
berries, p. make native handicrafts". number of traditional activities=the number of
activities that a respondent mentioned.
subsistence harvests
Constructed from the question "thinking about all the meat and fish your household ate in
the past 12 months, how much did members of your household harvest: none, less than
half, about half, or more than half". 0=none, 0.25=less than half, 0.5=about half,
0.75=more than half.
Christian beliefs
Christian as religious beliefs part of your life: constructed from the question" are
Christian religious beliefs part of your life?" 0=yes, 1=no.
indigenous beliefs
Constructed from the question" are Inupiat or Yupik spiritual beliefs part of your life?"
0=yes, 1=no.
number of problem with
your house
Constructed from the question "Does your house have any of these problems? a. too little,
b. dampness, c. mold or mildew, d. water leaking from the ceiling from condensation or
melting, e. frost on the windows, f. draft from the doors or windows, g. drafts from places
other than doors & windows, h. cold floors, i. generally cold, j. stale air-inadequate
ventilation, k. shifting of house from active permafrost, l. water that is not safe to drink, at
least at some times of the year. " The value of the variable is the number of the problems
mentioned.
types of crimes committed
Constructed from the question "within the past 12 months, have you been a victim of 1.
theft? sexual assault? another type of assault other offenses? "
family ties index
A count of responses to three questions about family: (1) “How strong are the links
among family members not living with you?” (2) “During the last month, how often were
you in touch with members of your family not living with you by phone or mail?” (3)
“During the last month, how often were you in contact with family members not living
with you by visiting or being visited?” For the first question, response categories are:
1=very weak, 2=weak, 3=neither weak nor strong, 4=strong and 5=very strong. For the
other two questions, response categories are: 1=never, 2=once, 3=a few times, 4=more
than a few times, 5=every day.
social support index
A count of responses to a series of questions about the kinds of support available to
people when they need it: (1) someone you can count on to listen to you when you need
to talk; (2) someone you can count on when you need advice; (3) someone who shows
you love and affection; (4) someone to have a good time with; (5) someone to confide in
139
or talk about yourself and your problems; (6) someone to get together with for relaxation;
(7) someone to do something enjoyable with. Response categories are: 1=not at all,
2=very seldom, 3=some of the time, 4=most of the time and 5=all the time.

Appendix Table A-2 The Distribution of Household Income Among the Five Income
Sources by Size of Income
Sales of crafts,
and etc. (%)
Self-
employment
(%)
Wages (%) Government
and other
organizations
(%)
Other sources
(%)
N of
Respondents

Total 0.02 0.03 0.61 0.33 0.01 643
$1-1500 1 0 0 0 0 1
$1501-5000 0.19 0 0.12 0.69 0 12
$5001-8000 0.02 0.01 0.05 0.93 0 14
$8001-12000 0.08 0.01 0.16 0.73 0.02 21
$12001-16000 0.02 0 0.28 0.7 0 36
$16001-23000 0.01 0.03 0.46 0.46 0.03 58
$23001-28000 0.01 0.03 0.46 0.49 0.02 37
$28001-37000 0.01 0.01 0.65 0.33 0 78
$37001-50000 0.02 0.02 0.62 0.33 0.01 101
$50001-70000 0.01 0.02 0.73 0.22 0.02 111
$70001-90000 0.01 0.02 0.81 0.16 0 64
$90001-
110000
0.01 0.07 0.79 0.11 0.01 46
$110001-
150000
0.03 0.07 0.79 0.11 0.01 38
$150001-
200000
0 0.1 0.74 0.13 0.03 20
Above 200000 0.07 0.18 0.5 0.08 0.18 6

140
APPENDIX B Supplementary Material for Chapter 3

Appendix Table B-1 Summary Statistics (Larger Sample)
Appendix Table 1. Summary Statistics (Larger Sample)
Panel A Whole Rural Urban

W1: life satisfaction 3.065 3.051 3.117

W2: life satisfaction 3.124 3.113 3.166

W1: number of observations 11016 8673 2343

W2: number of observations 11016 8673 2343
Panel B Whole Rural Urban
Structural Support Measures

W1: live with children and grandchildren (reference group) 0.291 0.302 0.248

W2: live with children and grandchildren (reference group) 0.254 0.262 0.224

W1: live with children only 0.327 0.313 0.38

W2: live with children only 0.25 0.238 0.295

W1: live without children and without grandchildren*closest
child lives in same county/city 0.256 0.253 0.265

W2: live without children and without grandchildren*closest
child lives in same county/city 0.305 0.305 0.305

W1: live without children and without grandchildren*closest
child lives outside same county/city 0.056 0.056 0.056

W2: live without children and without grandchildren*closest
child lives outside same county/city 0.088 0.084 0.101

W1: live with grandchildren only*closest child lives in same
county/city 0.046 0.047 0.043

W2: live with grandchildren only*closest child lives in same
county/city 0.073 0.076 0.06

W1: live with grandchildren only*closest child lives outside same
county/city 0.025 0.029 0.008

W2: live with grandchildren only*closest child lives outside same
county/city 0.031 0.034 0.016

W1: living with parents 0.056 0.058 0.048

W2: living with parents 0.011 0.011 0.012

W1: number of children 2.684 2.819 2.184

W2: number of children 2.768 2.905 2.258

W1: have one child (reference group) 0.166 0.121 0.332

W2: have one child (reference group) 0.148 0.102 0.319

W1: have two children 0.363 0.363 0.362

W2: have two children 0.358 0.359 0.355

W1: have three children and above 0.471 0.516 0.306

W2: have three children and above 0.494 0.539 0.326

W1: have daughters only (reference group) 0.135 0.109 0.229

W2: have daughters only (reference group) 0.12 0.096 0.209

W1: have sons only 0.254 0.237 0.318

W2: have sons only 0.224 0.205 0.295

W1: have sons and daughters 0.611 0.654 0.453

W2: have sons and daughters 0.656 0.699 0.496
Demographic & Socioeconomic Characteristics

W1: age49 0.232 0.231 0.233

W2: age49 0.149 0.149 0.148

W1: age5059 0.365 0.368 0.354

W2: age5059 0.363 0.364 0.358

W1: age6069 0.283 0.286 0.271

W2: age6069 0.328 0.332 0.31

W1: age7079 0.104 0.099 0.124

W2: age7079 0.134 0.128 0.156

W1: age80 0.016 0.016 0.018

W2: age80 0.026 0.025 0.028

W1: male 0.468 0.475 0.443

W2: male 0.468 0.475 0.443
141

W1: Han_nationality 0.926 0.927 0.922

W2: Han_nationality 0.926 0.927 0.922

W1: married 0.898 0.902 0.885

W2: married 0.882 0.886 0.869

W1: illiterate (reference group) 0.251 0.29 0.107

W2: illiterate (reference group) 0.237 0.275 0.101

W1: some primary education 0.42 0.451 0.303

W2: some primary education 0.423 0.456 0.3

W1: some secondary education 0.329 0.259 0.59

W2: some secondary education 0.34 0.269 0.599

W1: ln(per capita expenditure) 9.192 9.069 9.647

W2: ln(per capita expenditure) 9.192 9.069 9.647

W1: whether have any ADL or IADL difficulty 0.119 0.134 0.067

W2: whether have any ADL or IADL difficulty 0.137 0.147 0.102

W1: working 0.716 0.801 0.399

W2: working 0.685 0.761 0.404

W1: whether participate in or receive any pension 0.776 0.772 0.79

W2: whether participate in or receive any pension 0.776 0.772 0.79

W1: age of respondents 0.858 0.858 0.858

W2: age of respondents 0.84 0.84 0.841

W1: age of respondents 57.866 57.786 58.16

W2: age of respondents 59.866 59.786 60.16

W1: mean(age of repondent's children) 30.915 30.837 31.204

W2: mean(age of repondent's children) 33.345 33.314 33.459

142
APPENDIX C Supplementary Material for Chapter 4

Appendix C.1 Sample Attrition
To test whether there is any selection bias due to attrition, I apply a test suggested by Wooldridge
(2010). It shows that sample selection due to attrition in the fixed effects context is only a problem
when the selection is related to the idiosyncratic errors. I write the fixed effects model as
𝐿𝑆
𝑖𝑡
= 𝛽 ′
𝑋 𝑖𝑡
+ 𝛼 𝑖 + 𝘀 𝑖𝑡
(12)
where 𝑋 𝑖𝑡
is a vector of all time variant individual or regional characteristics in my setting. Let 𝑠 𝑖𝑡

be a selection indicator, where 𝑠 𝑖𝑡
= 1 if (𝐿𝑆
𝑖𝑡
, 𝑋 𝑖𝑡
) is observed. In order to have no selection bias
due to attrition, we need to have strict exogeneity: 𝐸 (𝘀 𝑖𝑡
|𝑿 𝒊 , 𝒔 𝒊 , 𝛼 𝑖 ) = 0, 𝑡 = 1,2, … , 𝑇 .
In the case where the selection is entirely random, 𝒔 𝒊 is independent of (𝜺 𝒊 , 𝑿 𝒊 , 𝛼 𝑖 ) , so the
selection will not result in any inconsistency under standard fixed effects assumptions. Besides,
under much weaker conditions, for example, 𝘀 𝑖𝑡
is mean independent of 𝑠 𝑖𝑡
given (𝑿 𝒊 , 𝛼 𝑖 ) for all
𝑡 , FE on the unbalanced panel is also consistent.
There is a simple test suggested by Nijman and Verbeek (1992). We can include a lead of
the selection indicator 𝑠 𝑖𝑡 +1
in equation (6), estimate the new equation using fixed effects, and do
a t test for the significance of 𝑠 𝑖𝑡 +1
. Under the null hypothesis, 𝘀 𝑖𝑡
should be uncorrelated with 𝑠 𝑖𝑟

for all 𝑟 , so the selection in the future period should be insignificant in the equation at time 𝑡 .
I apply this simple test to my fixed effects models on income (shown in Appendix Table
C-1). Being present in the sample one period ahead, “Selection Lead”, is not correlated with life
satisfaction conditional on the covariates and the individual fixed effects. The results suggest that
the attrition in my sample is, at least, random conditional on current individual observables.
Applying the test to fixed effects models of total consumption expenditures or conspicuous
consumption expenditures gives similar results (not shown).
143

Appendix Table C-1 Test for Selective Attrition (Fixed Effects Regressions, HILDA 2006-
2009)
Appendix Table 5. Test for Selective Attrition (Fixed Effects Regressions, HILDA 2006-2009)
Dependent Variable: Life Satisfaction
(1) (2)
FE FE
Selection Lead 0.0423 0.0459
(0.0289) (0.0287)
Ln (income) 0.0509*** 0.0340***
(0.0106) (0.0109)
Socioeconomic controls NO YES
Observations 51414 51414
Number of individuals 15585 15585
Adjusted R-squared 0.00205 0.00925
Notes:
Additional controls (all columns) include age, gender, month of interview, year, and region.
Socioeconomic controls include education, marital status, employment status, health, natural
logarithm of number of dependent children at home, and natural logarithm of number of adults at
home.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1

Appendix Table C-2 Consumption Categories
Category Definition
Housing Rent, mortgage and repairs, and renovations and maintenance to your home
Groceries Food, cleaning products, pet food, and personal care products, excluding alcohol or tobacco
Health care and child care Health care spending includes fees paid to doctors, dentists, opticians, physiotherapists, chiropractors
and any other health practitioner, medicines, prescriptions and pharmaceuticals (include alternative
medicines.), and private health insurance; child care spending includes child care costs for children
while parents work, during school holidays, and non-employment related childcare, for school-aged and
not yet at school children. Assumes 12 weeks school holidays a year for school-aged children.
Meals eaten out Restaurants, take-away food, and bought lunches and snacks. Do not include alcohol
Vehicle purchase Buying brand new or used motor vehicles, motorbikes, or other vehicles (include boats, planes,
caravans, trailers and jet skis)
Holidays Holidays and holiday travel costs (include short & long holidays.)
Motor vehicle fuels and engine oil Motor vehicle fuel (petrol, diesel, LPG) and engine oil
Clothing and footwear Men’s, women’s and children’s clothing and footwear
Phone rent and calls and internet
charges
Telephone rent and calls (include rent and charges on mobile phones), and internet charges
Other insurance Home and contents, and motor vehicle insurance
Home utilities Electricity bills, gas bills, and other heating fuel (such as firewood and heating oil)
Alcohol Alcohol consumed at home or with meals eaten out
Education Education fees paid to schools, universities, and other education providers (include private tuition fees)
Furniture and household appliances Any bedroom and outdoor furniture (do not include floor coverings), household appliances, such as
ovens, fridges, washing machines, and air conditioners
Recreational devices and equipment Computers and related devices (such as printers, digital cameras, iPods, MP3 players, electronic
organizers and game consoles), televisions, home entertainment systems, and other audio visual
equipment (such as DVD players and video cameras)
Motor vehicle repairs and
maintenance
Motor vehicle repairs and maintenance (include regular servicing)
Tobacco Cigarettes and other tobacco products
Public transportation Public transport and taxis

144
Appendix C.2 Measurement Error in Life Satisfaction

Consider the following dynamic model with unobserved individual fixed effects.
𝐿𝑆
𝑖𝑡
∗
= 𝛽 0
𝐿𝑆
𝑖𝑡 −1
∗
+ 𝛽 1
′
𝑀 𝑖𝑡
+ 𝛽 2
′
𝑋 1,𝑖𝑡
+ 𝛽 3
′
𝑋 2,𝑖𝑡
+ 𝛽 4
′
𝑍 𝑖 + 𝛼 𝑖 + 𝘀 𝑖𝑡
(13)
𝐿 𝑆 𝑖𝑡
∗
is true self-rated life satisfaction. However, we cannot observe 𝐿 𝑆 𝑖𝑡
∗
. What we can observe is
a proxy 𝐿 𝑆 𝑖𝑡
, which is reported by an individual. Individuals may have a systematic tendency to
misreport their true life satisfaction scores. For example, more extroverted individuals may always
overrate their life satisfaction when asked. Following Powdthavee (2009), I assume that there is a
systematic error to how life satisfaction is measured and that this measurement error is driven by
a fixed effect and a time-varying component:
𝐿 𝑆 𝑖𝑡
= 𝐿 𝑆 𝑖𝑡
∗
+ 𝜃 𝑖 + 𝜇 𝑖𝑡
(14)
Assume the error 𝜇 𝑖𝑡
is not serially correlated and is uncorrelated with 𝐿 𝑆 𝑖𝑡
∗
as well as 𝑀 𝑖𝑡
, 𝑋 1,𝑖𝑡
,
and 𝑋 2,𝑖 𝑡 . Substituting (8) into (7)
𝐿𝑆
𝑖𝑡
= 𝛽 0
𝐿𝑆
𝑖𝑡 −1
+ 𝛽 1
′
𝑀 𝑖𝑡
+ 𝛽 2
′
𝑋 1,𝑖𝑡
+ 𝛽 3
′
𝑋 2,𝑖𝑡
+ 𝛽 4
′
𝑍 𝑖 + [𝛼 𝑖 + (1 − 𝛽 0
)𝜃 𝑖 ] + (𝘀 𝑖𝑡
+ 𝜇 𝑖𝑡
−
𝛽 0
𝜇 𝑖𝑡 −1
) (15)
The presence of time-varying measurement error would make the regression error follow
a first order moving average process. Therefore, the error is serially correlated, and I cannot use
the second lag of life satisfaction as an instrument in the first-difference equation. What I can do
is to use higher order lags.

145
Appendix Table C-1 Life Satisfaction, Income, Total Household Consumption
Expenditures, Savings, Conspicuous Consumption Expenditures, and Basic Consumption
Expenditures (Blundell and Bond GMM Estimation Results)

Dependent Variable: Life Satisfaction

(1) (2) (3) (4)
BB-GMM BB-GMM BB-GMM BB-GMM
L life satisfaction 0.440*** 0.435*** 0.437*** 0.382***

(0.0855) (0.0847) (0.0812) (0.0660)
Ln (income) 0.189**

(0.0896)

Ln (total consumption expenditures)

0.0357

(0.0962)

Ln (savings)

0.0153

(0.0109)

Ln (conspicuous consumption expenditures)

0.120***

(0.0323)
Ln (basic consumption expenditures)

-0.106

(0.126)
Socioeconomic controls YES YES YES YES
Observations 33,045 33,045 33,045 33,045
Number of individuals 12,277 12,277 12,277 12,277
Number of instruments 143 143 143 152
Specification tests (p-value)

(a) Hansen test of over-identification 0.282 0.236 0.205 0.350
(b) Serial correlation

First-order 0.000 0.000 0.000 0.000
Second-order 0.000 0.000 0.000 0.000
Notes:

Additional controls (all columns) include age, gender, month of interview, year, and region. Socioeconomic controls include
education, marital status, employment status, health, natural logarithm of number of dependent children at home, and natural
logarithm of number of adults at home.
In Blundell-Bond GMM (BB-GMM) estimations, all explanatory variables whenever they appear in the model—except age,
year, month, and region dummies—are treated as endogenous variables; that is, their lagged values are used as the instruments
in the difference equation (instrumenting from the third-lag life satisfaction), and their lagged differences are used as the
instruments in the level equation (instrumenting from the second-lag differences in life satisfaction). Age, year, month, and
region dummies are treated as exogenous variables. Their differences are the instruments for themselves in the difference
equation, and their levels are the instruments for themselves in the level equation.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1

146
Appendix Table C-2 Life Satisfaction and Total Consumption Expenditures, and Savings
(OLS and FE Results)
Dependent Variable: Life Satisfaction
(1) (2) (3) (4)
OLS FE OLS FE

L life satisfaction 0.536*** 0.536***
(0.00633) (0.00631)
Ln (total consumption expenditures) 0.0363*** 0.0329**
(0.0106) (0.0135)
Ln (savings) 0.00274*** 0.000478
(0.000678) (0.000658)
Socioeconomic controls YES YES YES YES
Observations 47,405 47,405 47,405 47,405
Number of individuals 14,107 14,107
Adjusted R-squared 0.353 0.00971 0.353 0.00959
Notes:
Additional controls (all columns) include age, gender, month of interview, year, and region. Socioeconomic controls
include education, marital status, employment status, health, natural logarithm of number of dependent children at
home, and natural logarithm of number of adults at home.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1

Appendix Table C-3 Life Satisfaction, Conspicuous Consumption Expenditures and Basic
Consumption Expenditures (OLS and FE results)
Dependent Variable: Life Satisfaction
(1) (2) (3) (4) (5) (6)
OLS FE OLS FE OLS FE

L life satisfaction 0.535*** 0.544*** 0.544***
(0.00633) (0.00738) (0.00739)
Ln (conspicuous consumption expenditures) 0.0250*** 0.0175*** 0.0134 0.0115
(0.00492) (0.00549) (0.00917) (0.00807)
Rank (conspicuous consumption
expenditures)
0.139*** 0.0859*** 0.0898** 0.0457**
(0.0240) (0.0217) (0.0392) (0.0222)
Ln (basic consumption expenditures) -0.0189* -0.00475 -0.0292 -0.0155
(0.0113) (0.0141) (0.0314) (0.0319)
Rank (basic consumption expenditures) -0.0765*** -0.00701 -0.0274 0.0178
(0.0240) (0.0253) (0.0573) (0.0573)
Socioeconomic controls YES YES YES YES YES YES
Observations 47,405 47,405 47,405 47,405 47,405 47,405
Number of individuals 14,107 14,107 14,107
Adjusted R-squared 0.353 0.00995 0.353 0.00979 0.353 0.00972
Notes:
Additional controls (all columns) include age, gender, month of interview, year, and region. Socioeconomic controls include education, marital status,
employment status, health, natural logarithm of number of dependent children at home, and natural logarithm of number of adults at home.
Robust standard errors in parentheses (clustered at individual level)
Significance: *** p<0.01, ** p<0.05, * p<0.1

Abstract (if available)

Abstract This work presents new evidence on the determinants of subjective well-being, as measured by life satisfaction, in three quite different populations. Chapter 2 studies the Inuit, an indigenous hunter-gatherer population living on the barren northernmost fringes of Alaska. My analysis indicates that the key factors that contribute to their life satisfaction are health, subsistence hunting and fishing, and social support. In addition, the Inuit are more satisfied with their life if they have both Christian religious beliefs and indigenous spiritual beliefs as part of their life. Surprisingly, a higher level of wage income is associated with a lower level of life satisfaction, a finding that challenges common preconceptions about the effects of modernization and points to the importance of non-wage subsistence activities as a preferred substitute for wage employment for this population. Chapter 3 examines the well-being of the elderly population in China, a country characterized by a traditional three-generation family structure. In rural villages in China, life satisfaction of older parents is still positively associated with traditional kinds of support such as living in a three-generation household and the exchanges of financial and emotional support with their children. However, in China's urban neighborhoods, this is not true. It seems that China’s development, especially the development in urban areas, is breaking down the historical association between intergenerational support and well-being. Chapter 4 aims to understand the importance of social comparisons to people’s well-being. The study, which focuses on Australia, investigates the effects of consumption expenditures on life satisfaction. It shows that conspicuous (i.e. visible and positional) spending increases happiness while savings and spending on basic goods and services, the less visible components of income, do not contribute to it. Moreover, this research provides evidence for relationship heterogeneity across income groups.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Four essays on how policy, the labor market, and age relate to subjective well-being

PDF

Explaining well-being: essays on the socio-economic factors accompanying life satisfaction

PDF

Three essays on social policy: institutional development, and subjective well-being as a cause and consequence of labor market outcomes

PDF

Essays on health and well-being

PDF

Adaptation, assets, and aspiration. Three essays on the economics of subjective well-being

PDF

Beyond revealed preferences: how gender, relative socioeconomic status, and social norms drive happiness and behavior

PDF

Essays on the economics of subjective well-being in transition countries

PDF

The dynamics of well-being in daily life: a multilevel perspective

PDF

Essays on behavioral decision making and perceptions: the role of information, aspirations and reference groups on persuasion, risk seeking and life satisfaction

PDF

A neuropsychological exploration of low-SES adolescents’ life goals and their motives

PDF

Essays on health and aging with focus on the spillover of human capital

PDF

Lifting up ourselves by lifting up others: examining cognitive-affective states linking generativity to well-being

PDF

Essays on macroeconomics and income distribution

PDF

Selection and impacts of early life events on later life outcomes

PDF

Later life success of former college student-athletes as a function of retirement from sport and participant characteristics

PDF

Three essays in health economics

PDF

Intergenerational transfers & human capital investments in children in the era of aging

PDF

Perspectives on state capacity and the political geography of conflict

PDF

Essays on development and health economics: social media and education policy

PDF

Three essays on cooperation, social interactions, and religion

Asset Metadata

Creator Wu, Fengyu (author)

Core Title What leads to a happy life? Subjective well-being in Alaska, China, and Australia

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Publication Date 07/13/2018

Defense Date 04/20/2018

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag conspicuous consumption,Happiness,life satisfaction,OAI-PMH Harvest,social comparison,subjective well-being

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Easterlin, Richard (committee chair), Myers, Dowell (committee member), Nugent, Jeffrey (committee member)

Creator Email fengyuwu@usc.edu,fywu1989@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-16734

Unique identifier UC11669026

Identifier etd-WuFengyu-6398.pdf (filename),usctheses-c89-16734 (legacy record id)

Legacy Identifier etd-WuFengyu-6398.pdf

Dmrecord 16734

Document Type Dissertation

Format application/pdf (imt)

Rights Wu, Fengyu

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA