University of Southern California Dissertations and Theses

Spatiotemporal visualization and analysis as a policy support tool: a case study of the economic geography of tobacco farming in the Philippines

(USC Thesis Other)

Spatiotemporal visualization and analysis as a policy support tool: a case study of the economic geography of tobacco farming in the Philippines

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

SPATIOTEMPORAL VISUALIZATION AND ANALYSIS AS A POLICY SUPPORT
TOOL:

A CASE STUDY OF THE ECONOMIC GEOGRAPHY OF TOBACCO FARMING IN
THE PHILIPPINES

by

Steven Louis Rubinyi

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

May 2014

Copyright 2014 Steven Louis Rubinyi
ii

DEDICATION
I dedicate this document to my mom and dad for always supporting me in my academic
endeavors and encouraging me to stay curious, and explore the world.

iii

ACKNOWLEDGMENTS
I will be forever grateful to my mentor, Professor Karen Kemp. It is safe to say that
without her patient guidance and meticulous eye for details, I would not have made it this
far. Thank you as well to the South East Asia Tobacco Control Alliance for helping me to
acquire necessary research documents and providing feedback on my initial research
idea.

iv

TABLE OF CONTENTS

Dedication ii
Acknowledgments iii
List of Tables vi
List of Figures vii
List of Abbreviations ix
Abstract x
Chapter One: Introduction 1
1.1 Project Objectives and Organization 5

Chapter Two: Background 7
2.1 Country Profile 7
2.2 Tobacco in the Philippines 10
2.3. Literature Review 15
2.3.1 Policy and Spatial Analysis 15
2.3.2 Tobacco Geography 19
2.3.3 Spatiotemporal Visualization and Analysis 21

Chapter Three: Data Sources and Preparation 26
3.1 Data Sources 26
3.1.1 Administrative Boundaries 27
3.1.2 Agricultural Data 28
3.1.3 Provincial Areas Data 30
3.1.4 Limitations 30
3.2 Methodology 33
3.2.1 Construction of the Geodatabase 33
3.2.2 Stationary Visualization for Spatiotemporal Analysis 37
3.2.3 Dynamic Visualization for Spatiotemporal Analysis 38

Chapter Four: Results of Spatiotemporal Analysis 40
4.1 Percent Total Provincial Area Devoted to Growing Tobacco 40
4.1.1 Stationary Visualization of Percent Area 41
4.2 Tobacco Volume of Production 44
4.2.1 Stationary Visualization of Volume of Production 44
v

4.2.2 Dynamic Visualization of Volume of Production 47
4.3 Tobacco Farm Gate Prices 55
4.3.1 Stationary Visualization of Farm Gate Prices 56
4.3.2 Comparative Analysis, La Union Province 59

Chapter Five: Discussion 64
5.1 Potential Causes 64
5.1 Key Takeaways 66

Chapter Six: Conclusions 68

References 71

Appendices
Appendix A: Percent of Provincial Area Devoted to Growing Tobacco 74
Appendix B: Total Volume of Tobacco Production, Metric Tons 82
Appendix C: Farm Gate Price of Tobacco, Pesos per Kilogram 92

vi

LIST OF TABLES
Table 1: Summary of Visualization-Based Techniques of Exploratory
Spatiotemporal Analysis (from Andrienko et al. 2003) 24

Table 2: NTA and BAS Statistics on Total Production in the
Philippines, 2007-2011 (Metric Tons) 31

Table 3: Average Yield in Metric Tons per Hectare 47

vii

LIST OF FIGURES
Figure 1: Philippines Regions and Provinces 8

Figure 2: Provinces Producing Tobacco at least 1 Year Since 1990 12

Figure 3: Provinces Producing Virginia and/or Native Type Tobacco
at least 1 Year Since 2000 13

Figure 4: Volume of Tobacco Production in the Philippines, by Type
(Metric Tons) 14

Figure 5: Statistics on Total Tobacco Production in the Philippines,
1990-2012 (Metric Tons) 32

Figure 6: Portion of Attribute Table for GADM Provincial
Administrative Boundaries 35

Figure 7: Portion of Joined Excel Files from CountrySTAT with
Added ID_1 Field 36

Figure 8: Percent of Total Provincial Area of Selected Provinces
Devoted to Growing Tobacco, All Types 42

Figure 9: Total Volume of Tobacco Production in Metric Tons, All
Types 45

Figure 10: The Philippines’ Mean Center and Focus Areas of the Mean
Centers of Each Type of Tobacco 48

Figure 11: Centers of Total Production by Year for Each Tobacco Type 49

Figure 12: Center of Total Tobacco Production by Year, All Types 51

Figure 13: Center of Total Tobacco Production by Year, Native Type 52

Figure 14: Total Volume of Production for Native Type Tobacco
(Metric Tons) 53

Figure 15: Center of Total Tobacco Production by Year, Virginia Type 54

viii

Figure 16: Total Volume of Production for Virginia Type Tobacco
(Metric Tons) 55

Figure 17: Native, Virginia, and Burley Type Farm Gate Prices for
Select Provinces (Pesos per Kilogram) 58

Figure 18: Farm Gate Prices for La Union Province (Pesos per
Kilogram) 60

Figure 19: Percent of La Union Province Total Area Devoted to
Tobacco Farming 61

Figure 20: Total Volume of Production for La Union Province (Metric
Tons) 62

Figure 21: All Tobacco Types (Appendix A) 74

Figure 22: Native Tobacco Type (Appendix A) 78

Figure 23: Virginia Tobacco Type (Appendix A) 80

Figure 24: All Tobacco Types (Appendix B) 82

Figure 25: Native Tobacco Type (Appendix B) 86

Figure 26: Virginia Tobacco Type (Appendix B) 89

Figure 27: Native Tobacco Type (Appendix C) 92

Figure 28: Virginia Tobacco Type (Appendix C) 96

Figure 29: Burley Tobacco Type (Appendix C) 100

ix

LIST OF ABBREVIATIONS
ASEAN Association of Southeast Asian Nations
FCTC 2003 WHO Framework Convention on Tobacco Control
GADM Database of Global Administrative Layers
GAUL Global Administrative Unit Layers
GDP Gross Domestic Product
GIS Geographic Information Systems
NRCCW National Resource Center for Information Technology in Child Welfare
NTA National Tobacco Administration
PSA Philippine Statistics Authority
SEATCA Southeast Asia Tobacco Control Alliance
SQL Structured Query Language
UNSALB United Nations Administrative Level Boundaries Dataset
USC University of Southern California
USD United States Dollar
VAT Value Added Tax
WHO World Health Organization

x

ABSTRACT
This study demonstrates the utility of visualization-based spatiotemporal analysis as a
policy support tool in the agricultural sector through a case study analyzing changes in
the spatial distribution of tobacco farming in the Philippines from 1990 through 2012.
Tobacco farming remains divisive in the Philippines; although often touted by tobacco
companies and supportive government agencies as integral to the Philippine economy
and an effective crop for poverty alleviation, recent studies dismiss these claims
altogether, suggesting that farmers would be better off diversifying or even switching
crops altogether (SEATCA 2008; Espino et al. 2009; World Health Organization 2012).
This study does not argue for or against tobacco farming; it simply illustrates how
spatiotemporal analysis can be successfully implemented to uncover deeper, more
nuanced insights that could be drawn upon to design efficient and effective tobacco
farming policies.
The analysis considers provincial level agricultural data from the Philippines
Bureau of Agricultural Statistics for tobacco area planted, volume of production and farm
gate pricing of three unique tobacco varieties: Native, Virginia, and Burley. Stationary
and dynamic techniques of spatiotemporal data visualization are used, and data are
analyzed for trends using outlined methods. The results holistically describe tobacco
farming in the Philippines and are drawn upon to determine which tobacco growing
provinces and types of tobacco are on the rise or decline, to investigate causation behind
spikes and dips in production, and to outline the future direction of the industry as a
whole. The spatiotemporal analysis provides empirical evidence for policy makers to
better understand regional and provincial trends in tobacco farming over time.
1

CHAPTER 1: INTRODUCTION
The negative health implications of tobacco use are well documented, with one in 10
adults killed by smoking worldwide according to the World Bank. Historically, the
smoking epidemic primarily affected rich countries, however by 2020 it is expected that
seven out of 10 smoking related deaths will be in low-income and middle-income
countries (World Bank 1999). Although there is no single panacea to reducing demand
for tobacco, increasing the price to consumers combined with innovative public health
education campaigns and government restrictions on marketing and packaging, have
recorded fair success at reducing consumption, particularly in high-income countries
(SEATCA 2008).
Still, as transnational tobacco companies continue to expand aggressively, both
production and consumption have steadily increased in the Association of Southeast
Asian Nations (ASEAN) region. The 2003 WHO Framework Convention on Tobacco
Control (FCTC), developed in response to the globalization of the tobacco epidemic,
asserts the importance of addressing both tobacco supply-side issues and demand
reduction in an attempt to structure global regulatory policy. It is within this framework
that governments in ASEAN have increasingly implemented policies and programs to
combat tobacco demand, primarily through strategy dependent on a combination of price
and tax measures with non-price measures.
As stated in the FCTC, one of the main tools for reducing consumption while
simultaneously raising funds for valuable government programs is the successful
implementation of a tobacco tax, often a combination of excise tax, value added tax
(VAT), import tariffs, and other inventive measures. Every country in ASEAN has some
2

form of tobacco tax burden, ranging from 16 percent of retail price in Lao PDR to 70
percent of retail price in Thailand (SEATCA 2013). The Philippines sits relatively in the
middle of the pack, with a tax burden of 53 percent of retail price. A new excise tax
regime was approved by the Philippines government in 2012, leading Phillip Morris, with
a market share of 79.3%, to implement price increases of 30% to 75%.
A 2014 study by Citi Research shows just how effective a strong excise tax can
be. In 2013, after the excise tax was bolstered under Republic Act No. 10351, tobacco
sales volumes declined by 15.6% in the Philippines, a notable reversal from the 3.5%
compound annual growth rate experienced from 2007-2012. Although revenue from
tobacco taxes in ASEAN exceeded USD 13 billion in 2011, healthcare costs remain far
greater, up to 13.7 times the cumulative tax revenue in some instances. This conclusion
indicates that although the financial benefits of an excise tax remain significant, the more
important fight is to cut tobacco consumption, using tax revenue simply to mitigate
negative tobacco related externalities. It is also important for the countries in the ASEAN
region to work collaboratively to coordinate tobacco pricing, as illicit trade of tobacco
products is widespread and relatively difficult to prevent in total.
Surprisingly, although every country in ASEAN has some sort of policy to reduce
tobacco consumption, tobacco farming is still encouraged and supported by many
governments throughout the region. The Tobacco Regulation Act of 2003 represented the
first fully comprehensive national legislation on tobacco control in the Philippines. The
Act included five years of funding for the Tobacco Growers’ Assistance Program and the
Tobacco Growers’ Cooperative, with the intent of financially supporting farmers
3

displaced due to stricter tobacco policies and assisting tobacco farmers to develop
alternative farming systems and livelihoods.
Over the past century, tobacco production has increasingly shifted towards low-
income and middle-income countries (World Bank 1999). Tobacco cultivation not only
requires significant tracts of land that could be otherwise used for much-needed food
crops, but also is detrimental to the environment and notoriously linked to strenuous
working conditions, health hazards, and child labor (WHO 2012). It is an industry that is
dominated by a vertically integrated, globalized oligopoly, which is able to exert
immense power, often exploitatively, on tobacco farmers and national governments.
Regional advocacy groups such as the Southeast Asia Tobacco Control Alliance
(SEATCA), Save our Farmer, and Unfair Tobacco, have helped to shed light on many of
these issues, particularly the poverty cycle in tobacco farming perpetuated by the global
tobacco industry. Yet, even with much of this information well known, tobacco is still a
cash crop that remains an economically attractive option for many otherwise
impoverished farmers in suitable climates, particularly those already familiar with the
process and invested in necessary equipment and technology.
The positive economic ramifications of tobacco farming on a local economy are
often heralded by the tobacco industry as reason enough not to put pressure on tobacco
farming. However recent rebukes to this notion claim sustainable paths towards
alternative livelihoods are not only feasible, but would improve the economic standing of
local farmers by diversifying an economy that is currently undermined by corporate
practices (SEATCA 2008; Espino et al. 2009; WHO 2012).
4

The tobacco industry consists primarily of three sub-sectors in the value-added
production chain: farming, processing, and product manufacturing. Tobacco is a globally
traded commodity, with farming operations in over 120 countries, producing an estimated
global farm gate value of USD 8 billion (Geis et al. 2009). In the Philippines alone,
tobacco farming produced over USD 100 million in farm gate value in 2012 and provided
livelihood for 840,415 people, including those directly employed and their dependents
(National Tobacco Administration 2012).
Although inarguably significant, the total number of employed farmers is still
small when compared to overall national employment (SEATCA 2008). The relative
impact may be misleading however, as tobacco farming is heavily concentrated in only a
couple of provinces, playing a much more significant role in certain localized economies.
The cumulative lobby of the tobacco industry and those dependent on it for
livelihood, directly hinders efforts to increase tobacco control measures that would be
beneficial to public health. If sustainable alternative livelihoods can be realized for
tobacco growers, the tobacco industry will lose much of its political influence to
negotiate less stringent government policies towards tobacco consumption (SEATCA
2008). Ramifications could prove quite significant, particularly for the Philippines, which
is described by Alechnowicz and Chapman (2004) as having the strongest tobacco lobby
in Asia.
In this context, it is beneficial to analyze whether spatially nuanced trends in
tobacco farming emerge in the Philippines over time. The intent of this analysis is to
contribute novel research towards effective tobacco farming policy making. The
5

motivating research question is thus as follows: How does spatial variation in tobacco
farming change over time?

1.1 Project Objectives and Organization
This thesis is guided by the principle that detailed quantitative and qualitative research is
essential to building effective policy measures, particularly on controversial issues. The
main objective of this research is to demonstrate how spatiotemporal visualization and
analysis can contribute effectively towards informed policy development, exemplified in
this case by tobacco farming in the Philippines.
The project study uses available agricultural data to holistically investigate
tobacco farming through spatiotemporal visualization and analysis. Key questions were
developed to address challenges faced from the perspective of a national level policy
maker. The key questions are:
1. What provinces and regions are most dependent on tobacco farming and how has
this changed spatially over time?
2. What provinces and regions produce the highest total volume of tobacco
production and how has this spatially changed over time?
3. Is there any spatial variation in farm gate prices and, if so, how has this changed
spatially over time?
4. Considering the first three key questions, is there any spatial variation between
the three types of tobacco grown in the Philippines: Native, Virginia, and Burley?
If so, how has this changed over time?
5. If clear spatiotemporal trends are identified, can possible causes also be reasoned?
6

The investigation of the five key questions begins in chapter two, which describes
the country context behind tobacco farming in the Philippines and provides a literature
review of previous studies that have linked spatiotemporal visualization and analysis to
policy decision making. Chapter three follows with a detailed description of the data
sources and the methodology used to investigate each of the five key questions, followed
by chapter four, which contains the results. The first subsection of chapter four addresses
the relative dependence of each province on tobacco farming by normalizing the total
area of planted tobacco to the total area of each province. The second subsection
examines the total volume of production by province, including a section that uses
dynamic visualization to trace the center of tobacco production by year for each available
tobacco type. The third subsection examines spatiotemporal variation in farm gate prices,
including a section focused on a detailed comparison of data from La Union province.
Chapter five presents the discussion, which reasons possible causes from the observed
trends, discusses the broader significance of the results, validates methods used, and
presents suggestions for future work that could further strengthen the utility of
spatiotemporal analysis in support of tobacco farming policy development in the
Philippines. Lastly, Chapter six provides a summary of conclusions.
It is my working hypothesis that clear spatiotemporal trends will emerge from the
investigation of the first four key questions identified above and that it will be possible to
link these trends to potential causes. Combined, the results of this thesis will provide
novel information, useful to tobacco farming policy decision makers in the Philippines.
7

CHAPTER 2: BACKGROUND
Prior to outlining the methodology used in this study, it is critical to understand the
physical, economical, and historical context of the Philippines and to examine similar
work that has been completed on related topics. The country profile section aims to
establish a basic awareness of the local setting related to tobacco geography while
additionally serving as a springboard for deeper investigation into the causes behind
trends observed during the performed analysis. Following the Country Profile section, the
literature review discusses a range of techniques and insights drawn from previous
studies that provide a rationalization for the approach of this study.

2.1 Country Profile
The Philippines is a Southeast Asian archipelagic nation consisting of 7,107
unique islands in the western volcanic rim of the Pacific Ocean with a cumulative area of
nearly 300,000 square kilometers (Villaluz 2012). Approximately 95 percent of the land
area and a similar percentage of the population of 96.71 million Filipinos are associated
with one of the 11 largest islands in the archipelago (World Bank 2012). The islands are
split into three major island groups: the northern Luzon, 141,000 square kilometers; the
central Visayas, 57,000 square kilometers; and the southern Mindanao, 102,000 square
kilometers (Moog 2006).
The Philippines is politically divided into 81 provinces which are grouped into
one of 17 different regions. The National Capital Region, commonly referred to as Metro
Manila, is the largest urban area, with a population of nearly 12 million people. Each
province is subdivided into cities and municipalities, which are further subdivided into
8

barangays, the smallest local government unit. Figure 1 shows each of the 81 provinces
of the Philippines grouped into the 17 regions.

Figure 1 Philippines Regions and Provinces
Source: GADM Database of Global Administrative Areas
9

In 2012, the GDP of the Philippines totaled (current) USD 250.2 billion, with an
impressive growth rate of 6.8 percent (World Bank 2012). The past decade has seen
accelerated growth from the previous two decades, averaging about five percent per year
and raising gross national income per capita from (current) USD 1,000 in 2002 to
(current) USD 2,500 in 2012, as measured using the Atlas method (World Bank 2012).
Despite a decade of sustained growth, the economy of the Philippines is still
struggling to create more and better jobs (Chua et al. 2013). Policy distortions have
particularly hampered growth and productivity in the agriculture and manufacturing
sectors and much more needs to be done to accelerate inclusive growth and reduce
poverty. From 2003 to 2009, the percentage of the population living below the national
poverty line actually increased from 24.9 percent to 26.5 percent (World Bank 2012).
The Philippines is a warm and tropical climate, with a mean annual temperature
of 27 degrees Celsius that makes it ideal for tobacco farming (Moog 2006). The climate
is dominated by wet and dry seasons, more or less pronounced depending on localized
geography. Agricultural holdings account for approximately 33 percent of total land
utilization in the Philippines, encompassing nearly 100,000 square kilometers (Moog
2006).
According to the 2002 Agricultural Census, approximately 40,000 square
kilometers grow rice, 24,000 square kilometers grow maize, and 32,000 square
kilometers grow coconut, accounting for roughly 96 percent of total agricultural lands.
The 2002 Agricultural Census further reported a total of nearly five million agricultural
holdings, over 96 percent of which are less than seven hectares in area. In total, close to
10

six million Filipinos are engaged full-time in agricultural work, primarily on their own
small holdings.

2.2 Tobacco in the Philippines
Tobacco was first introduced to the Philippines in the late 16
th
century by the
Spanish; however, full commercialization did not begin for another 200 years (De Jesus
1980). In the late 18
th
century, the colonial Spanish government began searching for ways
to make the Philippines a profitable and self-financing colony. Although the Manila
galleon trade had proven quite successful for those directly involved, the Philippines
remained largely financially dependent on subsidies from Mexico (Crouch 1985). One of
the first acts to raise revenue was the creation of a government supported tobacco
monopoly, established on March 1, 1782 by Governor-General Jose Basco y Vargas (De
Jesus 1980).
Although initially unsuccessful, the business quickly took off and made the
colony self-sustaining once it became open to foreign shipping and businessmen (Crouch
1985). Over the subsequent 100 years, the Philippines emerged as the largest tobacco
producer in Asia, realizing profits of USD 3 million in 1881, the year before the
monopoly’s eventual dissolution (De Jesus 1980).The tobacco monopoly focused on
cultivation solely in Luzon, particularly the Cagayan Valley, the Illocos provinces, Nueva
Ecija and Marinduque. The legacy of the tobacco monopoly remains strong, and tobacco
farming is still most prevalent in the Luzon provinces.
Three types of tobacco are currently grown in the Philippines: Native, Virginia,
and Burley. In total, the National Tobacco Administration (NTA), reports that tobacco
11

farming in the Philippines only encompasses an area slightly larger than 30 square
kilometers as of 2012. Different types of tobacco are traditionally grown in different
provinces. The Native tobacco type constitutes 24 percent of total area planted and is
grown to some extent in nearly all tobacco growing provinces. The Virginia tobacco type,
introduced to the Philippines in 1927, constitutes 58 percent of all tobacco area planted,
and is concentrated in Ilocos Sur, Abra and La Union (NTA 2014).
The Burley tobacco type was the last to be cultivated in the Philippines. Today,
Burley tobacco constitutes 18 percent of total area planted and is primarily grown in
Pangasinan, Tarlac, Nueva Ecija and Occidental Mindoro. Burley tobacco also produces
the highest average yield per hectare (2,200 kilograms) followed by Virginia (2,000) and
Native (1,653). Seventy-five percent of Virginia and Burley tobacco is distributed
domestically, compared to only 30 percent for Native tobacco. The remaining tobacco is
distributed to foreign markets all over the world.
Figures 2 and 3 give a visual overview of tobacco producing provinces in the
Philippines using data from the Bureau of Agricultural Statistics (BAS). Figure 2 shows
all provinces that reported tobacco production for at least one year since 1990.
Of the 52 provinces where tobacco production was reported, 24 produced tobacco
for the entire 23 year period and 45 out of 52 provinces reported tobacco production for
at least 10 years out of 23. No province that reported tobacco production reported less
than five total years.
12

Figure 2 Provinces Producing Tobacco at least 1 Year Since 1990
Source: Data from the Bureau of Agricultural Statistics

Figure 3 shows all of the provinces that produced Virginia and/or Native type tobacco
over the duration of this study. Note that Burley type is excluded from this map. The
BAS does not have any segregated data on Burley type production, yet it is inferred to be
included in all cumulative tobacco statistics.

13

Figure 3 Provinces Producing Virginia and/or Native Type Tobacco at least 1 Year
Since 2000
Source: Data from the Bureau of Agricultural Statistics

The NTA indicates that Burley type tobacco accounts for approximately a quarter
of total production, which the “unspecified” type in Figure 4 is inferred to be from 2000
onward, when segregated statistics of Virginia and Native type tobacco were first
recorded by the BAS. Even so, because this was not explicitly stated, it is not included in
the analysis.
14

Figure 4 Volume of Tobacco Production in the Philippines, by Type (Metric Tons)
Source: Data from the Bureau of Agricultural Statistics

Also note that although Virginia type tobacco production accounts for roughly half of
cumulative production, BAS data show that it is limited to only nine provinces, of which,
only Benguet does not also grow Native type tobacco. From tabular analysis, it is
possible to see that only four provinces reported growing Virginia type tobacco for the
full 13 years statistics are available: Abra, Ilocos Norte, Ilocos Sur, and La Union. Each
of these four provinces is located in the northwestern section of the island of Luzon.
In contrast, 41 provinces reported growing Native type tobacco at some point during
the 13 years and of these provinces, 26 produced Native type tobacco continuously
throughout the entire period. Although this initial gross level of data analysis can provide
a general level understanding of the distribution, the significance of these varying
continuity records becomes apparent only with the application of spatiotemporal
visualization.
0
20000
40000
60000
80000
100000
120000
140000
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
Volume of Tobacco Production
Native Virginia Unspecified
15

2.3 Literature Review
The literature review focuses on three topics; policy and spatial analysis, tobacco
geography, and spatiotemporal visualization and analysis. Each of the topics was chosen
after considering their relative importance to the study. The first section focuses on the
relationship between policy analysis and spatial analysis. The focus on policy is integral
to understand the utility of spatial analysis as a support tool for policy decision making
and can be seen as the overarching theme of this thesis. The next section focuses on
tobacco geography and is particularly important for drawing parallels to research
previously completed in this area, where it has been done, and to determine the
comparative novelty of this thesis in the context of tobacco farming and the Philippines.
The final section focuses on spatiotemporal visualization and analysis. The
literature sheds light into the future of spatiotemporal analysis and provides a template
for examining spatiotemporal data using readily available visualization techniques. Taken
as a whole, the literature review provides important context for the present study, shows
how the study extends the research base in the area of spatiotemporal visualization, and
illustrates how policy analysts can make use of some emerging capabilities for analyzing
spatiotemporal data.

2.3.1 Policy and Spatial Analysis
Geographic Information Systems (GIS) technology has increasingly played an
important role as a decision support tool to policy makers in a wide variety of sectors.
Although occasionally criticized as gratuitous, GIS offers a powerful range of operational
functions, from simple spatial pattern identification to complex statistical analysis, which
16

allow for raw data to be transformed to digestible information useful in policy making
(Dyke et al. 1996). Map presentation of data can often communicate large amounts of
spatially nuanced information more effectively than other methods which may only
provide a tabular display of results.
Many studies have come out in recent years illustrating cases and outlining
practice for successful GIS implementation as a support tool for policy decision making.
The National Resource Center for Information Technology in Child Welfare (NRCCW)
recognized the potential of GIS for child welfare policy and planning for child welfare
quite early, publishing a GIS tips, tools, and trends document in 2002 (NRCCW 2002).
The document highlights the utility of GIS to help support effective management
decisions, illustrate the flow of clients to and from community services, and predict
future needs. NRCCW also present a brief case study of how GIS has helped New York
City better monitor resource allocations, performance evaluations, and quality
improvements after first integrating GIS into their service networks beginning in 1999.
Desai et al. (2009) illustrate the increasing volume of applications for GIS as a
tool for policy decision making through a case study of Medicaid expenditures in Ohio.
The article describes the myriad of analysis tools available, and emphasizes the
responsibility of the analyst to educate decision makers on source data and underlying
assumptions.
Despite taking a critical look into the pitfalls of data distortion and
misrepresentation, Desai et al. (2009) make it quite clear that spatial visualization
remains a powerful tool to present and understand policy implications and guide decision
17

making. Yet, full utility remains unrealized in many instances, as advancements in spatial
analysis and modeling often outpace the rate of adoption.
Almost two decades ago, Worrall and Bond (1997) illustrated some of the
struggles of implementing GIS into successful spatial decision support systems,
particularly at the local government level. The study analyzed the early days of GIS from
a uniquely British perspective, dating back to the inception of the 1987 Department of
Environment Chorley Report, which explored the potential significance of GIS in
operational and applied policy settings. The Chorley Report also delved into issues
surrounding the development and integration of spatial data systems into public sector
organizations.
Suggesting that hype had played a role in unrealistically heightened expectations,
Worrall and Bond viewed the state of GIS in the British public sector at the time of
publication as a considerable failure. The article outlined many of the early challenges of
linking spatial analysis effectively in a policy making context, discussed root causes for
these setbacks, and itemized tensions that must be resolved. One of the main problem
areas discussed is the inability to keep up with the dynamic and progressive nature of GIS
technology in a fully functional manner, particularly in the development of spatial
decision support systems in public sector organization.
Even today, many of the problems mentioned by Worrall and Bond are magnified
in the developing world, where limited budgets and technical expertise often delay
adoption and restrict utility of GIS in government policy analysis and operations. Berisso
and de Vries (2010) illustrate these impediments through a case study of the utility sector
in Ethiopia. The study suggests that contextual determinants related to economic,
18

technological, legal and financial infrastructure present major impediments to technology
access, and thus to technology adoption in developing countries. These impediments also
lead to inherent internal weaknesses in government-created and government-maintained
authoritative data, which severely limits the utility of complicated methods of spatial
analysis.
The World Bank recognizes the increasing importance of making data driven
decisions for development, pushing governments around the world to escalate
investments in open, countrywide authoritative data (World Bank 2013). Hillygus et al.
(2006) discuss the political and social challenges of census mobilization, including
inherent difficulties accounting for displaced individuals in hard to reach areas. These
difficulties tend to be magnified in countries where capacity is underdeveloped,
documentation is less stringent, and corruption is more prolific.
GIS adoption in the Philippines has progressed continually and important policy
studies utilizing spatial analysis have come out in recent years, yet full application of GIS
analysis remains unrealized in some instances. Launio et al. (2008) use data from
household surveys of Philippine rice growers to analyze the spatial diversity of modern
rice varieties. The article focuses on indices of spatial diversity for each province in the
Philippines, providing strong analytical interpretation but without much spatial
visualization. The analytic conclusions of Launio et al. are meant to encourage policy
makers to strengthen the public rice breeding program. However, the authors miss out on
the opportunity to strengthen the argument through visual representation.
An example of a more effective use of spatial analysis is demonstrated in the
study by Dahly et al. (2013) which analyzes the spatial distribution of obese Filipinos in
19

Metro Cebu. The article successfully identifies obese clusters and relates them to urban
areas at the neighborhood level. The authors include multiple maps showing where these
clusters exist, and suggest that these results may facilitate research aimed at combatting
the increasing prevalence of obesity in the Philippines.
Over the past twenty years, spatial analysis has evolved from a relative novelty
for policy makers to an essential tool in their decision making process. Large increases in
the scope and volume of spatial applications have been made, however much of this
additional utility is dependent on the availability and fidelity of authoritative data, and a
capacity to manage and analyze the data effectively. The National Statistics Office in the
Philippines has made strides in recent decades to bolster local authoritative data and
encourage GIS use, including an initiative to widely distribute CDs containing census
data preloaded into a program that allows for basic GIS analysis (Philippine Statistics
Authority 2010). Quick adoption and availability of high-quality data have become
increasingly important for policy decision making in countries such as the Philippines,
which still face steep development challenges.

2.3.2 Tobacco Geography
The geography of tobacco farming has yet to be directly explored in the
Philippines. Although a SEATCA survey of tobacco growing regions by Espino et al.
(2009) provides considerable detailed insight into overarching trends in tobacco farming,
it neither probes into local variation nor provides spatial visualization of the data
collected. Studies on tobacco geography done in other countries and regions are helpful
in providing relevant insight.
20

In the pre-GIS era, Coppock (1965) examined tobacco geography in Nigeria,
producing multiple maps showing the different tobacco growing regions and the different
types of tobacco grown. Using rudimentary observations made from hand-drafted maps,
the author analyzed the geography of where different types were grown. The study uses
these findings to predict where production of different types might increase based on
factors such as availability of machinery, access to infrastructure, and climate.
Although not much has been done since to look into spatial trends of tobacco
farming geography in the developing world, many studies exist detailing the geography
of tobacco farming in the United States. Both the United States and the Philippines have
concentrated belts of tobacco production, and parallels can be drawn for determining the
driving forces behind historical geographic change.
Grise (1970) provided insight into the optimum geographic location of Burley
type tobacco production in the United States. The study created a model that detailed the
intricate geographic and economic drivers behind the tobacco farming industry that
enable certain competitive advantages. Considering all crop alternatives, inputs to the
model included: available resource supplies, crop yields, prices for inputs and outputs,
and prices for land and labor. The results help to understand the driving economic force
behind changes in tobacco geography.
The National Agricultural Statistics Service (NASS) under the United States
Department of Agriculture (USDA) provides detailed statistics of tobacco farming back
to 1934 and basic statistics on acreage and production by state back to 1866.
Visualization of the data facilitates analysis of tobacco farming trends. Birdsall (2001)
uses historical tobacco statistics to look into the implications of the federal Tobacco
21

Stabilization Program, which was aimed at buffering the region’s small-farm tobacco
landscapes from change. The author argues that the increasing geographic fragmentation
of tobacco growing communities drives the introduction of alternative crops.
Lessons from previous studies on the geography of tobacco farming such as
Birdsall’s conclusions regarding geographic fragmentation are important for determining
the factors of analysis for this thesis, and for drawing parallels with policy measures
taken by other countries. The United States is particularly of interest because of the
similarities between the Virginia, Kentucky, Tennessee, South Carolina and North
Carolina tobacco belt and the Luzon tobacco region of the Philippines; both have
hundreds of years of historical roots, and both represent highly concentrated, regional
tobacco farming landscapes.

2.3.3 Spatiotemporal Visualization and Analysis
Integration of time and space is useful for visualizing real time tracking,
transactional changes, and temporal attributes. Generally, spatiotemporal visualizations
can be grouped into four types: dynamic, discrete, stationary, and change. A dynamic
feature follows a path or track, a discrete feature displays separate events in distinctive
areas, a stationary feature changes value but not location, and a change feature may
change both value and location (Esri 2012).
Visualization techniques remain a challenge, particularly when presented in a
format that restricts or is not compatible with animations or videos. It is possible to adapt
to static images by displaying time in a series of maps, possibly using trackers to show
22

movement for dynamic visualization, or supplementing discrete and stationary maps with
graphs.
In this thesis, the locations for each of the Philippines provinces do not change
over time, but the values do; making stationary spatiotemporal visualization the most
straightforward approach. This approach is analyzed by Pickle (2009), who evaluates the
effectiveness of American mortality atlases in exploratory spatial analysis.
Although occasionally criticized as ineffective (Tukey 1979), Pickle proves that
despite their inherent limitations, mortality atlases served the purpose of describing
patterns in United States mortality data. Each mortality atlas reviewed had between 99
and 148 citations, contributing to novel etiologic findings such as the discovered links
between asbestos and arsenical air pollution to lung cancer.
Much of the inspiration to include an additional focus on dynamic visualization in
this thesis comes from a McKinsey global Institute study that tracks the world’s
economic center of gravity across time (McKinsey 2012). The study, based off of work
done by Angus Maddison at the University of Groningen, took each nation’s geographic
center of gravity and weighted the locations by GDP in three dimensions, which was then
projected to the nearest point on the earth’s surface.
A line connects points beginning with historical estimates from AD 1 and ending
in projections for 2025. A clear trend northwestward trend emerges from the year 1000 to
1950, at which point the trend sharply transitions back eastward, where it is projected to
continue through 2025. While historical estimates and future projections can be inexact,
the method quite effectively visualizes large-scale trends that transcend possible
inaccuracies within each data set.
23

Recently, algorithm-based methods of spatiotemporal analysis and modeling have
advanced substantially. Sahu and Mardia (2005) cite the desire to predict time evolution
of select response variables over a certain domain as the primary reason for this
development. As computing advances, so does the desire to create more intricate
statistical models that can better analyze and predict spatial patterns.
Deng et al. (2013) present an example of recent advancements. The authors
consolidate spatiotemporal clustering analysis methods into three types: space-time
scanning methods, density-based methods, and distance-based methods, and propose a
novel algorithm that addresses autocorrelation and heterogeneities in space and time.
The techniques presented by Deng et al. are incredibly useful for fields with many
data points such as climate change, epidemiology, earthquake, and crime analysis. The
cumulative yearly data in this thesis however, are not sufficient to effectively implement
higher powered statistical analysis, leaving visualization techniques as the primary source
for spatial analysis.
Andrienko et al. (2003) consider the confines and potential of visualization-based
techniques for exploratory analysis of spatiotemporal data. The study provides a useful
catalog of existing techniques, recognizing elementary and general methods for
identifying and comparing visualizations supporting tasks of different types. The
identified techniques for representation and exploration of spatiotemporal data are
grouped into four categories and summarized in Table 1.

24

Table 1: Summary of Visualization-Based Techniques of Exploratory
Spatiotemporal Analysis (from Andrienko et al. 2003)

Category Technique
Universal techniques Querying
Map animation
Map iteration
Techniques suitable for data
about existential changes
Time labels
Representation of age by color
Aggregation of data about events
Space-time cubes
Techniques for studying
thematic images
Change map
Time-series graphs
Aggregation of attribute values
Techniques applicable to data
about moving objects
Trajectory lines
Arrows
Tracing
Time labels
Space-time cube
Animation modes, i.e. snapshot in time, movement
history, and time window

Each technique identified by Andrienko et al. (2003) is described in detail in
relation to the analysis scenario. For example, for map iteration of stationary data, change
can be detected through visually scanning each map and comparing fragments. The
analysis can be supplemented by overlay or fading techniques, however for the stationary
data in this thesis, scanning is the most reasonable option due to the required format
lacking the capacity to display short animations.
Unfortunately, as is explained in the following chapter, the intrinsic limitations of
tobacco data in the Philippines restrict the potential for advanced spatiotemporal
modeling and statistical analysis. This thesis, therefore, instead relies on the
visualization-based techniques outlined by Andrienko et al. (2003) to temporally analyze
available stationary data and derived dynamic data. Although such analysis is not as
detailed as it would be using statistical models, Pickle (2009) and McKinsey (2012)
25

support the notion that visualization-based techniques are still quite valuable for
recognizing general trends and hypothesis generation.

26

CHAPTER 3: DATA SOURCES AND PREPARATION
This section contains a description of the data, their sources and characteristics, used in
the analysis and a step by step outline of the process used to incorporate that data into the
project geodatabase and integrate it into the spatiotemporal analysis. The intent of this
section is to illustrate the process that a policy analyst might be able to follow with
limited data and GIS tools.
The study joins administrative boundaries from the database of global
administrative areas (GADM) to provincial level tobacco farming data acquired through
CountrySTAT Philippines. The tobacco farming data are then prepared and
spatiotemporally analyzed using stationary and dynamic visualization techniques for each
tobacco type. The analysis is focused on determining spatiotemporal trends in tobacco
farming for relative provincial dependence, volume of production, and farm gate prices.

3.1 Data Sources
The major source of tobacco statistics used in this study is the CountrySTAT
Philippines database. As explained below, although data limitations exist, CountrySTAT
represents a landmark achievement in the collection and retention of basic agricultural
statistics in the developing world. This study presents an opportunity to highlight the
potential of spatial analysis derived from data available through the Food and Agriculture
Organization of the United Nations (FAO)’s global FAOSTAT databases.
Additionally, public and participatory GIS efforts such as the GADM database
used to acquire data for the Philippines administrative areas illustrate how, in countries
that lack resources to develop a strong infrastructure for spatial information, a more
27

inclusive approach can be used to increase integration of spatial analysis in public policy
decision making.

3.1.1 Administrative Boundaries from GADM
Country boundary and provincial boundary layers were acquired from GADM
Version 2.0 (www.gadm.org). GADM is a high-resolution database of global
administrative areas, which allows for free and open use towards academic pursuits. The
user can download the entire global dataset or select by country. Data are available in
multiple file formats including: shapefile, Esri personal geodatabase, Esri file
geodatabase, Google Earth .kmz file, or Rdata file.
The database is hosted by Prof. Robert Hijmans’ lab at the University of
California Davis. Regional, municipal, and barangay layers are also available on GADM,
however none are used in this project. The coordinate reference system for all GADM
data is latitude/longitude and the WGS84 datum.
Gleditsch and Weidmann (2012) refer to GADM as one of the most
comprehensive databases for national and subnational political borders and the World
Bank (2013) highlights its flexibility for quickly responding to public feedback and
resolving problems. The database is constantly updated, however it does not track
changes over time.
GADM Version 2.0 included Shariff Kabunsuan as a province, even though it
only existed from 2006 to 2008. Before becoming a province, Shariff Kabunsuan made
up the northern half of Maguindanao. After the law establishing Shariff Kabunsuan was
ruled void by the Supreme Court of the Philippines, the land returned to Maguindanao. I
28

chose to keep Shariff Kabunsuan in this study to reflect its brief existence over the study
period. Tobacco farming in Maguindanao exists but is marginal when compared to the
Luzon provinces.
Two other two sources of administrative data considered for use in this project
were the Global Administrative Unit Layers (GAUL) and the United Nations Second
Administrative Level Boundaries Dataset (UNSALB). Although each has advantages and
drawbacks, GADM was ultimately chosen due to the dynamic nature of Philippines
administrative units over the past decade, and the personal desire to support open and
community-driven data.

3.1.2 Agricultural Data from CountrySTAT
CountrySTAT is an initiative by the Food and Agriculture Organization of the
United Nations (FAO), aimed at providing reliable statistical data on food and agriculture
for the many countries that lack the internal capacity (www.countrystat.org). Developing
countries may have limited staff, a lack of adequate tools, insufficient budgets, or a lack
of analytic capacity to effectively monitor national trends. CountrySTAT provides a web-
based information system, based on the open-source FENIX platform currently used in
the global FAOSTAT, to improve access to food and agriculture statistics at a national
and sub-national level. The data are intended to support data analysis and evidence-based
decision making and facilitate policy making.
There are currently 25 member countries, each with their own CountrySTAT team
responsible for collecting and maintaining data, and making sure that all 80 uniform
agricultural indicators conform to international standards. The network is expanding
29

rapidly, with the latest launch in Sierra Leone on 13 February 2014. A further 16
countries remain interested in launching a CountrySTAT system in the next couple years.
The Philippines CountrySTAT team was founded in 2006 and the database now
consists of national core and sub-national statistics for 9 categories: production, trade,
food consumption, prices, fertilizer and pesticides, land use, labor and employment, costs
and returns, and others. Each category is broken down into sub-categories from which the
user can extract information through a basic query interface. The data can then be
downloaded by the user into an XLS, XML, TXT, or CSV file. This thesis focuses on
provincial statistics because they are the most detailed and most comprehensive available
on this system. Each of the data sets used are described below:
 Tobacco Area Planted
o Availability: Data for total annual tobacco area planted by province are
available from 1990 – 2012 (Burley, Native, and Virginia types). Also,
segregated data for Native and Virginia tobacco types total annual area
planted by province are available from 2002 – 2012.
o Units: Hectares
o Source: Bureau of Agricultural Statistics
 Tobacco Volume of Production
o Availability: Data for total annual tobacco volume of production by
province are available from 1990 – 2012 (Burley, Native, and Virginia
types). Also, segregated data for Native and Virginia tobacco types total
annual volume of production by province are available from 2000 – 2012.
o Units: Metric tons
30

o Source: Bureau of Agricultural Statistics
 Tobacco Farm Gate Price
o Availability: Data for yearly average tobacco farm gate price by province
are available for Native, Virginia, and Burley tobacco types from 1990 –
2012.
o Units: Pesos per kilogram
o Source: Bureau of Agricultural Statistics

3.1.3 Provincial Areas Data from the Philippine Statistics Authority
Data for provincial areas in hectares were not included in the GADM database,
and were thus acquired for all provinces from the Philippine Statistics Authority (PSA).
The PSA manages the Philippine Statistical System, which provides both the government
and the general public with basic data in support of planning and decision making.

3.1.4 Limitations
One of the major limitations in this study is that data collection only began in
1990, with distinction between unique tobacco types not being made until 2002 and 2000
respectively for tobacco area planted and tobacco volume of production. Additionally, no
statistics have been kept for area planted and volume of production for Burley type
tobacco. Farm gate prices are available from 1990 – 2012 however a comparative
analysis of farm gate prices and percent area planted or volume of production can only
be done for Native and Virginia tobacco types from 2002 or 2000 to 2012. Another
limitation is that there are only 81 provinces, of which, the National Tobacco
31

Administration (NTA) reports only 23 grew tobacco in 2012. Such a lack of basic data, as
is common in the developing world, makes this collection largely insufficient for the
most comprehensive spatial analysis techniques. Despite this limitation, as this study
shows, it is still useful and productive to base a policy analysis study such as this on
visualization-based techniques.
Another major limitation is the lack of consensus on tobacco production statistics
in the Philippines, particularly between the Bureau of Agricultural Statistics (BAS) and
the NTA, both surprisingly housed within the Department of Agriculture. According to
the NTA, tobacco is currently grown in 23 out of the 81 provinces however the BAS
reports that 37 provinces grew tobacco in 2012. Additionally, Euromonitor International,
a privately owned market intelligence firm, received significantly different production
statistics from the NTA for their 2013 industry profile on tobacco in the Philippines
compared to CountrySTAT data from the BAS, as shown in Table 2.

Table 2 NTA and BAS Statistics on Total Tobacco Production in the Philippines,
2007-2011 (Metric Tons)
Source: NTA Data extracted from Euromonitor (2013)

2007 2008 2009 2010 2011
NTA 41,951 42,779 58,572 73,767 79,092
BAS 34,289 32,466 36,383 40,530 44,944

NTA statistics indicate tobacco production nearly doubled from 2007 to 2011
while BAS statistics only show an increase of approximately 34 percent. Figure 5 shows
the nationwide downward trend in tobacco production from 1990 to 2012 as reported by
32

BAS. The data indicate a slight increase in tobacco production beginning in 2008 that
could possibly signal a slight trend in the opposite direction.
If NTA statistics prove true, tobacco production in only 4 years trends from a
recorded low to the highest level in nearly 20 years. This example serves as a reminder
that statistics, particularly in the developing world, are to be approached cautiously.
Unfortunately, the NTA data are not available at the provincial level, so it is not possible
to compare the statistics spatially over time against data from the BAS.

Figure 5 Statistics on Total Tobacco Production in the Philippines, 1990-2012
(Metric Tons)

For this study, full data sets are only available from the CountrySTAT data
supplied by BAS. Yet, even with these limitations, it is still possible to find significance
in spatial trends. Although data are often incomplete and unreliable in the developing
world, it can still be used as an important tool in policy analysis as long as the limitations
are communicated and understood by policymakers.
0
20,000
40,000
60,000
80,000
100,000
120,000
140,000
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
Volume of Production
BAS
NTA
33

3.2 Methodology
The methodology for this thesis can be broken down into two parts: the steps
taken to construct the geodatabase, and the techniques used for visualization-based
spatiotemporal analysis. Geodatabase construction consisted of downloading and
formatting CountrySTAT data, merging the data to the GADM provincial administrative
boundaries, and producing the necessary output layers to be used in the visualization-
based spatiotemporal analysis. The analysis focused on the created layers of both
stationary and dynamic time-enabled data, using techniques guided by the framework
outlined in Andrienko et al. (2003).

3.2.1 Construction of the Geodatabase
An Esri file geodatabase was created to hold the following 15 layers necessary for
visualization-based spatiotemporal analysis:
 Base Layers
1. Country level administrative boundary
2. Provincial level administrative boundaries
3. Provincial mean centers
4. Country mean center
 Stationary Visualization
5. Percent of total provincial area devoted to growing tobacco, all types
6. Percent of total provincial area devoted to growing tobacco, Native
type
34

7. Percent of total provincial area devoted to growing tobacco, Virginia
type
8. Total volume of tobacco production by province, all types
9. Total volume of tobacco production by province, Native type
10. Total volume of tobacco production by province, Virginia type
11. Farm gate pricing by province, Native type
12. Farm gate pricing by province, Virginia type
13. Farm gate pricing by province, Burley type
 Dynamic Visualization
14. Mean center by year for volume of tobacco production, all types
15. Mean center by year for volume of tobacco production, Native type
16. Mean center by year for volume of tobacco production, Virginia type

All layers used for stationary visualization required provincial level
administrative boundaries to be joined with CountrySTAT data. All layers used for
dynamic visualization required the mean center of each province to be joined with
CountrySTAT data. This required the mean center of each province first to be calculated
and exported from the provincial level administrative boundaries layer prior to the join.
Unjoined base administrative boundaries were also included to provide background
spatial reference for spatiotemporal analysis of dynamic data.
To begin, it was necessary to download GADM layers and CountrySTAT data.
GADM layers were downloaded from the website into a file geodatabase. The database
attributes can be seen in Figure 6. Importantly, ID_1 served as the key field for joining
35

GADM provincial administrative boundaries to CountrySTAT data. Each province name
corresponds to a number in the ID_1 field. It should also be noted that the Shape_Area
field downloaded from the GADM database is not in hectares. This was manually
replaced with provincial data from the Philippine Statistics Authority (PSA) prior to
export into the project file geodatabase.

Figure 6 Portion of Attribute Table for GADM Provincial Administrative
Boundaries

From the downloaded GADM file geodatabase, the country administrative
boundary (layer 1) and provincial administrative boundaries (layer 2) were exported to
the project file geodatabase. The mean center of each province was then calculated, and
the resulting provincial mean centers layer was exported to the project file geodatabase
36

(layer 3). Lastly, the geographic mean center of all the provinces was calculated and
exported to the project geodatabase as well (layer 4).
CountrySTAT data were downloaded from the website into three separate
Microsoft Excel files; one for area planted, one for volume of production, and one for
farm gate prices. These three files were then joined into a single Excel file and the ID_1
field was added to align province names with the ID_1 field used in GADM provincial
administrative boundaries, as seen in Figure 7. The Date_1 field was also slightly
reformatted to ensure that it would import into ArcGIS as a time field. Once complete,
the data were imported into ArcGIS as a table within the project file geodatabase.

Figure 7 Portion of Joined Excel Files from CountrySTAT with Added ID_1 Field

To create the necessary stationary and dynamic visualization layers, this study
used ArcGIS’s Make Query Table tool. A sequence of SQL queries were applied to the
37

imported table of tobacco data, each time joining the selected data to the provincial
administrative boundaries layer (layers five through 13) or the provincial mean centers
layer (layers 14 through 16). Each run of the Make Query Table tool produced a single,
ephemeral output layer, which then had to be imported into the file geodatabase.
Layers five through seven were additionally normalized to the total area of each
province. The idea behind this was that normalization would help account for the large
variation in geographic area between many of the provinces, address the key question
regarding relative provincial dependence on tobacco farming, and provide a metric more
distinct from volume of production. Normalization of these layers required a couple of
extra steps; first a new field was created and labeled for percent area, then the field
calculator was used to divide the total area of tobacco planted by the total provincial area.
Layers 14 through 16 also required an extra step for preparation; the weighted
mean center of the joined provincial mean centers was calculated for each year, allowing
change over time to be visualized as a single, dynamic point. Mean center for layer 14
was weighted using volume of tobacco production for all types, layer 15 for Native type,
and layer 16 for Virginia type.
In total, the project geodatabase contained 16 layers; nine of which were analyzed
by stationary visualization techniques, three by dynamic visualization techniques, and
four for reference.

3.2.2 Stationary Visualization for Spatiotemporal Analysis
Stationary spatiotemporal visualization requires a static feature to change value
over time (ESRI 2012). In this case, individual, stationary provinces change field values
38

yearly for layers five through 13. Change over time is displayed in a series of choropleth
maps that when strung together, can be viewed as an animation or displayed as a time
series. In this document, since each time series produced a multi-page figure, the maps
were placed in the appendices.
While in most instances, it is difficult to distinguish substantial changes year by
year, this type of analysis is able to highlight important trends that emerge in the tobacco
geography, particularly when the different types of tobacco are considered against each
other for each of the three indicators: percent of provincial area planted, volume of
production, and farm gate prices.
For each layer, it was first necessary to enable time, using the Date_1 field to
define the temporal intervals. Data were then visually analyzed to spot regional trends
and changes over time, guided by techniques outlined in Andrienko et al. (2003) for
exploratory analysis using visualization. Although animations were produced to guide the
analysis of each layer, the confines of static presentation limit the presentation of results
to individualized choropleth maps for each year in the time series.
In regions or provinces where trends were spotted, supplemental graphs were used
to look at comparative relationships at a more nuanced level. Without first understanding
the larger scale spatiotemporal trends however, it would be difficult to pick out the
important bits of information simply from tabular data.

3.2.3 Dynamic Visualization for Spatiotemporal Analysis
A dynamic feature changes location over time (Esri 2012). For layers 14 through
16, the dynamic feature is the weighted mean center for volume of tobacco production of
39

all tobacco types, Native tobacco type, and Virginia tobacco type. While it is possible to
visualize dynamic temporal features using a time series approach, dynamic features can
also be displayed on a single output map, connecting the yearly locations by line.
Andrienko et al. (2003) refer to this method as a form of overlay visualization.
Layers 14 through 16 each contained a single data point for each year data were
available. The points were connected from one year to the next, and the length of these
lines was calculated to determine the magnitude of change year by year. The direction of
the change was also considered in the analysis to spot major events and trends over time.
The findings from dynamic visualization were considered with the findings from
stationary visualization for volume of tobacco production, and were additionally
supplemented by graphs to analyze spotted trends in further detail. Unfortunately, it
would not be effective to produce dynamic visualization for area planted data because
they were normalized to provincial area, and it would also not be effective for farm gate
data, as they were already mean statistics for each province.

40

CHAPTER 4: RESULTS OF SPATIOTEMPORAL ANALYSIS
The results of the spatiotemporal analysis are split into three subchapters: percent total
provincial area devoted to growing tobacco, tobacco volume of production, and tobacco
farm gate prices. Each subchapter provides a different perspective on the economic
geography of tobacco farming in the Philippines.
The subchapters are further split into second-level subchapters based on the
technique used for visualization-based spatiotemporal analysis. Each subchapter has a
second-level subchapter for stationary visualization. The tobacco volume of production
subchapter has an additional second-level subchapter for dynamic visualization, and the
tobacco farm gate prices subchapter has an additional second-level subchapter for an in-
depth comparative analysis of farm gate prices in La Union province.

4.1 Percent Total Provincial Area Devoted to Growing Tobacco
Three long map series which illustrate the percent of total provincial area devoted
to growing tobacco by year for all types (Native, Virginia, and Burley), and segregated
for Native and Virginia types are included in Appendix A (Figures 21 through 23). The
CountrySTAT database contains statistics on the total hectares of tobacco farming by
province. For this analysis, the total area of tobacco farming by province was normalized
to the total area of each province. The idea behind this is to distinguish area planted data
from volume of production in a way that provides another perspective on tobacco
farming. By normalizing to the area of each province, Appendix A makes it possible to
compare provincial dependence on tobacco farming by year. The next sections discuss
the results of a visual analysis of these map series.
41

4.1.1 Stationary Visualization of Percent Area
Figure 21 in Appendix A shows the percent of total provincial area planted for all
tobacco types. Data for Figure 1 are available over the full period from 1990 to 2012, so
change can be examined through a succession of 23 maps showing percentage by year. A
quick glance at 1990 shows a small concentration of provinces with a higher percentage
of area planted in the Ilocos Region of Luzon. Lower percentages are scattered
throughout the rest of the country, with a small gap of no tobacco growing provinces
between the northern and southern growing regions.
As the time series progresses, the provincial dependence on tobacco farming in
the Ilocos Region declines substantially. Additionally, there is an increasing geographic
separation between the northern and southern tobacco growing provinces. Figure 8 details
the three provinces with the highest percent area planted of all types; La Union, Ilocos
Sur, and Pangasinan, as well the province that recorded the highest recorded percent area
planted of all types outside of Luzon; Misamis Oriental.

42

Figure 8 Percent of Total Provincial Area of Select Provinces Devoted to Growing
Tobacco, All Types
Source: Data from the Bureau of Agricultural Statistics

The single highest recorded percentage of total provincial area devoted to
growing tobacco was 9.8 percent in La Union in 1992. It is very interesting to see the
quick decline in tobacco farming in La Union. Each of the three Luzon provinces in
Figure 8 experiences a sharp downturn from 1992 to 1994 and again from 2003 to 2006.
Interestingly, the later downturn corresponds with the passage of the Tobacco Regulation
Act of 2003. Ilocos Sur remains at approximately five percent in 2012, about three
percent higher than the next highest province, La Union. This appears to be an anomaly
in the northwestern Luzon tobacco growing region, yet remains much lower than the
highs experienced in the early 1990s.
Percent area planted was further broken down in Figures 22 and 23 of Appendix
A to look at the distinction between Native and Virginia tobacco types. Although data on
Native and Virginia tobacco are only available from 2002 to 2012, there are still several

0
2
4
6
8
10
12
Percent Area Planted
Ilocos Sur La Union
Pangasinan Misamis Oriental
43

distinct differences in spatial distribution and temporal trends between the two types.
Native type tobacco, as discussed earlier and shown in Figure 4, is grown in many more
provinces than Virginia type tobacco.
Figures 22 and 23 further illustrate precisely how much more dispersed Native
type tobacco farming is than Virginia type tobacco farming. During the 13 years of
record, total Native type tobacco area planted never exceeded one percent of total
provincial area. In contrast, Virginia type tobacco farming accounted for higher
percentages dispersed between fewer provinces, even managing to exceed five percent in
Ilocos Sur in 2002 and 2003. Interestingly, despite the higher rate of dispersion, the
provinces with the highest concentration of Native tobacco farming are still La Union and
Ilocos Sur, with the southern province of Davao del Sur coming in a distant third.
One of the more important temporal observations for the Native tobacco type is
the rise observed within Ilocos Sur. No Native type tobacco farming was reported in
Ilocos Sur until 2004, when a marginal .13 percent was first reported. This number rose
to a high of .66 percent by 2008, making it the highest of any province that year, before
dropping down to .33 percent in 2012. This trend was first spotted through visual analysis
of Figure 22.
Percent area for Virginia type tobacco is first dominated by Ilocos Sur and La
Union before a decline of La Union that mirrors the observed decline in La Union for all
types in Figure 8. Negros Oriental is the only province that consistently grows Virginia
type tobacco outside of Luzon, although total area planted is marginal when normalized
by provincial area.

44

4.2 Tobacco Volume of Production
Unlike percent total area planted, this spatiotemporal analysis for volume of
production is not normalized by area. The total volume of production metric shows the
total output of the tobacco farming industry in each province. Although normalizing by
area provides unique insights into dependence of a province on tobacco farming, it is also
important to assess the total size of the industry within a province’s economy.

4.2.1 Stationary Visualization of Volume of Production
Appendix B contains Figures 24 through 26, which collectively illustrate the total
volume of tobacco production in each province by year and type. General trends over
time are consistent with the patterns observed in Appendix A, however the dominance of
the northern Luzon provinces appears even more demarcated from southern counterparts
when larger provinces that grow high volumes of tobacco, such as Isabela, Cagayan, and
Pangasinan, are not normalized by area. Figure 9 details the four provinces with the
largest tobacco producing years by volume and additionally includes Davao del Sur, the
largest tobacco producing province by volume in the southern provinces.

45

Figure 9 Total Volume of Tobacco Production in Metric Tons, All Types
Source: Data from the Bureau of Agricultural Statistics

The most visible trend in both Figure 24 and Figure 9 is the sharp decline in
production from 1992 to 1994 by the three provinces in the Ilocos Region: Ilocos Sur,
Isabela and La Union. All three provinces set record highs in 1992 or 1993, followed by
an incredible collapse in production in 1994. Pangasinan experienced the most significant
decrease in tobacco production, dropping from over 30,000 metric tons in 1992 to
approximately 8,000 metric tons in 1994.
Each of the three provinces of Figure 9 located in the Ilocos Region follows a
similar trend until 2008, when Ilocos Sur increases production by approximately 5,000
metric tons over the subsequent four years while production in La Union and Pangasinan
stagnates. A similar rise in production begins in 2008 for Isabela, which is located in the
Cagayan Valley Region, still on the island of Luzon. Davao del Sur represents the largest
tobacco growing province by volume in the southern provinces, yet it remains largely
trivial in comparison to volumes produced in Luzon.

0
5000
10000
15000
20000
25000
30000
35000
1990
1992
1994
1996
1998
2000
2002
2004
2006
2008
2010
2012
Volume of Production
Ilocos Sur Isabela Pangasinan
La Union Davao Del Sur
46

Figure 25 in Appendix B shows the spatial distribution of total production for
Native type tobacco from 2000 to 2012. The northern provinces of Luzon dwarf the
volume of production in the southern provinces continuously from 2000 to 2012. Isabella
is visibly the largest producer of Native type tobacco in the Philippines, reaching a high
of nearly 6,000 metric tons in 2005. Ilocos Sur reached a record high for Native type
tobacco production in 2008, recording over 2,700 metric tons. La Union also recorded
high levels of Native type tobacco production in the mid-2000s and Cagayan, located just
north of Isabela, has produced significant volumes as well.
Figure 26 in Appendix B shows the spatial distribution of total production for
Virginia type tobacco from 2000 to 2012. The map reinforces the point that the Ilocos
Region, with the exception of Pangasinan, contains a vast majority of total Virginia type
tobacco production, and thus a majority of total tobacco production in the Philippines.
The only other province with significant Virginia type tobacco production is the adjacent
Abra province, officially placed in the Cordillera Administrative Region. Lastly, although
small in volume, it is important to note that the provinces of Isabela and Davao del Sur
both reported Virginia type tobacco production for the first time in 2012.
Overall, there does not appear to be a large increase in the tobacco yield per
hectare over the last 23 years, at least in the Ilocos Region. Table 3 looks at the
productivity in tons per hectare of three provinces, Ilocos Sur and La Union from the
Ilocos Region, and Isabela from the Cagayan Valley Region. It is important to note that
changing ratios of tobacco types grown in each province may influence average yield per
hectare in addition to poor crop years, new technologies, or variations in productivity.

47

Table 3 Average Yield in Metric Tons per Hectare
Source: Data from the Bureau of Agricultural Statistics

1990 1995 2000 2005 2010 2012
Ilocos Sur
Total Production 24107 19542 14174 11629 13833 17527
Total Area Planted 17199 15488 13515 7989 12171 13758
Tons per Hectare 1.40 1.26 1.05 1.46 1.14 1.27

Isabela
Total Production 5369 8296 6448 7944 8273 10183
Total Area Planted 3850 10271 6521 3911 4873 6235
Tons per Hectare 1.39 0.81 0.99 2.03 1.70 1.63

La Union
Total Production 23273 16117 9814 8830 5464 5487
Total Area Planted 11976 8327 6364 3987 2841 2911
Tons per Hectare 1.94 1.94 1.54 2.21 1.92 1.88

4.2.2 Dynamic Visualization of Volume of Production
Building on the example from McKinsey (2012) discussed in Chapter 2 in which
they tracked the economic center of gravity, a similar form of dynamic visualization was
used here as a means of capturing and tracking the movement over time of the center of
tobacco production in the Philippines on a single map.
By calculating the mean center (also known as the centroid) of each province and
attributing volume of production data to those points, it is possible to perform a weighted
mean center calculation for each type of tobacco for each year, using all tobacco growing
provinces of the selected type for the input features. The output of this calculation shows
the center of tobacco production, and the temporal change each year can provide valuable
insights into tobacco production trends. Because the Philippines is an archipelago, it is
possible for the center of production to be in the ocean.
48

Figure 10 shows the geographic mean center of the Philippines based on the
centroids of all provinces. The mean center is located in the Visayas Islands, near the
provinces of Capiz, Aklan, and Masbate. Figure 10 also shows the bounding rectangles
that surround the set of points for each of the tobacco type weighted mean centers by
year, all of which are located in the northern section of Luzon.

Figure 10 The Philippines’ Geographic Mean Center and Focus Areas of the Mean
Centers of Each Type of Tobacco
49

Figure 11 looks closer into the individual focus areas, and shows individual points
indicating mean center calculations by year for all tobacco types, Native tobacco type,
and Virginia tobacco type. It is quite apparent that the spatial volatility is highest year by
year for the calculated mean center of Native tobacco. This is likely attributed to a greater
rate of spatial distribution for Native tobacco in comparison to Virginia tobacco.

Figure 11 Centers of Total Production by Year for Each Tobacco Type
50

Virginia tobacco production is largely confined to four provinces, and all of the
yearly mean center calculations fall right in the middle, in the province of Ilocos Sur.
Although Native tobacco is heavily weighted towards the Ilocos region as well, it is also
grown in large quantities in the Cagayan Valley Region and in smaller quantities
throughout select southern provinces. The mean center for all types of tobacco falls
between the two other map areas, a bit closer to the mean center calculations for Virginia
tobacco because yearly Virginia tobacco production is two times higher than yearly
native tobacco production.
Figure 12 looks at the center of total tobacco production by year for all types. A
line connects dots representing each year from 1990 to 2012. The center of production
begins near the center of Benguet province then drifts eastward from 1991 to 1997. From
1998 to 2006, the center of production floats slightly north or south along the border
between Bunguet province and Nueva Vizcaya province before experiencing a massive
shift northward of 52.37 kilometers in 2007.
This shift was likely caused by the increase in production in Ilocos Sur and
Isabella, both located north of where the center of total tobacco production had been
located in previous years, coupled with the stagnation in production in Pangasinan and
the decline in production in La Union. By 2012, the center of tobacco production had
shifted nearly 50 kilometers northeast into Ifugao province.
51

Figure 12 Center of Total Tobacco Production by Year, All Types

The center of production for Native type tobacco, visible in Figure 13, was off the
coast of Aurora province in the year 2000, taking a big jump of 53.53 kilometers
northward from 2003 to 2004, another big jump of 67.78 kilometers southward from
2005 to 2006, and a final big jump of 66.04 kilometers northward from 2006 to 2007.
52

From 2007 to 2012, the center of production jumps around Quirino province, ending up
nearly 80 kilometers northwest of where it began in 2000 as shown in Figure 13.

Figure 13 Center of Total Tobacco Production by Year, Native Type

Figure 14 shows the changes in volume of production from 2000 to 2012 for
Native type tobacco. Both the Isabela and Cagayan provinces are in the Cagayan Valley
53

Region, while the La Union and Ilocos Sur provinces are in the Ilocos region. Most of
“All other” provinces represent the large distribution of southern provinces with very low
levels of production.

Figure 14 Total Volume of Production for Native Type Tobacco (Metric Tons)

There is a substantial dip from 2005 to 2006, particularly for Isabela and Cagayan
that likely explains the shift southwestward experienced by center of production at that
time; note that in 2006, roughly one third of Native type tobacco came from “All other”
provinces. The general trend northwestward is likely explained by the increased
production on Ilocos Sur in the latter half of the timeframe.
As visible in Figure 15, the center of production for Virginia type tobacco never
leaves Ilocos Sur. It begins by slowly drifting southward until 2003 when the trend
reverses, shifting slightly northward instead each year until 2012. The changes by year all
follow the same line, bearing roughly five degrees east of north. The largest change by
0
2000
4000
6000
8000
10000
12000
14000
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Volume of Production
Native Type Tobacco
All Other
La Union
Cagayan
Isabela
Ilocos Sur
54

year occurred from 2010 to 2011 when the center of production moved 6.46 kilometers.
The final position for center of production is less than 15 kilometers north of the initial
position in the year 2000.

Figure 15 Center of Total Tobacco Production by Year, Virginia Type

Figure 16 illustrates precisely how dominant Ilocos Sur is for total Virginia
production. Including statistics for all tobacco types, Ilocos Sur accounted for over 30
55

percent of the total volume of all tobacco produced in the Philippines from 2000 to 2012.
The northward drift that begins after 2003 for the center of tobacco production is likely
due to the proportional decrease in tobacco produced in La Union province, the
southernmost of the four dominant Virginia tobacco producing provinces. The largest
shift for the Virginia type center of production occurred between 2010 and 2011, when
Ilocos Sur and Ilocos Norte both experienced their largest increases in total production
over the timeframe.

Figure 16 Total Volume of Production for Virginia Type Tobacco (Metric Tons)
Source: Data from the Bureau of Agricultural Statistics

4.3 Tobacco Farm Gate Prices
Farm gate prices are available for Native, Virginia, and Burley type tobacco from
1990 to 2012 for each province in the Philippines. These prices are shown spatially over
time in Appendix C for each of the tobacco types in Figures 27 through 29. Although
unable to fully determine causality, changing patterns in farm gate prices may explain
0
5000
10000
15000
20000
25000
30000
20002001 20022003 200420052006 20072008 200920102011 2012
Volume of Production
Virginia Type Tobacco
All other
La Union
Ilocos Sur
Ilocos Norte
Abra
56

some of the increases and decreases in tobacco production over time. Factors such as
provincial price volatility, rate of price change, and spatial distribution and variation of
price each have implications that provide insight into industry shifts. These are explored
in detail in a subsection focusing on La Union province.

4.3.1 Stationary Visualization of Farm Gate Prices
Farm gate prices for the native tobacco type are shown in Figure 27. Native
tobacco has the largest volatility of all tobacco types, with farm gate prices ranging from
just over 10 pesos per kilogram in Davao del Sur in 1994 to nearly 450 pesos per
kilogram for the same province in 2012. Map 7 also shows the geographic variability of
farm gate prices for native type tobacco. In 2012, when farm gate prices in Davao del Sur
reached the record high of 450 pesos per kilogram, farm gate prices in Abra remained
under 50 pesos per kilogram. Interestingly, prices in Luzon remained much less volatile,
as indicated by Isabela and La Union provinces in Figure 18.
The geographic variability of farm gate pricing for Native type tobacco is in direct
contrast to Virginia type tobacco, where the range in prices remains similar for each
province by year, with only slight variations. In 2012, the highest farm gate price for
Virginia type tobacco was in Ilocos Sur, at 68.36 pesos per kilogram and the lowest was
in Ilocos Norte, at 53.34 pesos per kilogram. Of course, there were only four total
provinces reporting farm gate prices for Virginia type tobacco in 2012, all of which
located in the same geographic area. The farm gate price of Virginia type tobacco tends
to increase gradually over time, with a slight decrease from 2002 to 2004 before
recovering again in 2005.
57

Burley type tobacco farm gate pricing also remains similar in each of the
provinces by year, with slightly higher prices consistently found in Isabela and Cagayan
provinces. Interestingly, the farm gate price of Burley has increased at a faster rate than
that of Virginia type tobacco, inferring that production should also increase in relation.
Unfortunately, we do not have production information for Burley type tobacco however
we can conclude that more and more tobacco farmers may be tempted to switch from
Virginia type tobacco to Burley type tobacco if this trend continues.
Figure 17 shows farm gate prices for select provinces over time. Trend lines were
added on each graph for La Union province because La Union province has the highest
percentage of its total area devoted to tobacco farming and it has consistently reported
farm gate prices for the entire 23 year period for all three types of tobacco. For the
regression, y is the farm gate price, and x is the year. The multiplier before x is the slope
for each of the trend lines, indicating the rate of change for farm gate prices each year.
The R squared value describes how well the data fit the trend; the closer to 1, the better
the fit. Also note that the graph for Native type in Figure 17 has a different scale than for
Burley type or Virginia type. This is to accommodate for the large temporally variability
in farm gate price in Davao Del Sur.
The slope and R squared values of the created trend lines confirm much of what
was observed by visually analyzing the maps in Appendix C. The slope for both Native
and Burley tobacco types far exceeds that of Virginia type tobacco and the lowest R
squared value is associated with Native type tobacco.

58

Figure 17 Native, Virginia, and Burley Type Farm Gate Prices for Select Provinces
(Pesos per Kilogram)

y = 2.4703x + 11.047
R² = 0.8164
0
100
200
300
400
500
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
Farm Gate Price
Native Type
Davao Del Sur
Isabela
La Union
Linear (La Union)
y = 1.5586x + 21.306
R² = 0.8849
0
20
40
60
80
100
120
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
Farm Gate Price
Virginia Type
Ilocos Sur
La Union
Linear (La Union)
y = 2.3896x + 5.7953
R² = 0.8764
0
20
40
60
80
100
120
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
Farm Gate Price
Burley Type
Isabela
La Union
Pangasinan
Linear (La Union)
59

4.3.2 Comparative Analysis, La Union Province
From 2002 to 2012, it is possible to compare percent area planted data for Native
and Virginia tobacco types to corresponding farm gate prices. It is also possible to
compare volume of production data to corresponding farm gate prices from 2000 to 2012.
La Union is the only province in the Philippines which produced both types of tobacco
and reported corresponding farm gate prices for the duration. It is also the only province
to report farm gate prices for all three types of tobacco each year since 1990, and was
thus used in Figure 18 to show trend lines, slope, and R squared values. This section
builds off of Figure 18, and digs deeper into the relatedness of farm gate prices to percent
area planted and volume of production.
Figure 18 begins by investigating farm gate prices for each of the three tobacco
types. Burley type tobacco is included in Figure 18 even though no data are available for
percent area planted and volume of production because Burley tobacco may help to
explain the gap between the cumulative tobacco statistics and the sum of Native and
Virginia tobacco statistics.
Figure 18 clearly shows that from 2000 to 2012, the farm gate price for Native
and Burley tobacco has increased substantially in comparison to Virginia tobacco. The
farm gate price of Native tobacco surpassed that of Virginia tobacco around 2005 and the
farm gate price of Burley tobacco surpassed that of Virginia tobacco around 2008.
Although Virginia tobacco produces approximately a 20 percent higher yield per hectare
than Native tobacco, in 2011 and 2012 the farm gate price of Native tobacco was high
enough to match this difference entirely. Burley tobacco produces the highest yield per
60

hectare, and the gross income per hectare of Burley tobacco has likely exceeded that of
Virginia tobacco for every year since 2006.

Figure 18 Farm Gate Prices for La Union Province (Pesos per Kilogram)
20
30
40
50
60
70
80
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Farm Gate Price
La Union
Native
Virginia
Burley
Native
y = 2.7377x + 33.441
R² = 0.7065
Virginia
y = 1.1534x + 40.234
R² = 0.5071
Burley
y = 3.6487x + 19.988
R² = 0.8875
20
30
40
50
60
70
80
Farm Gate Price
La Union
Linear (Native)
Linear (Virginia)
Linear (Burley)
61

From a purely economic perspective, tobacco farmers in La Union province
should feel natural pressures to increasingly shift from Virginia tobacco to Native or
Burley tobacco. It is quite unfortunate that there is no further data available for Burley
tobacco, as the combination of high yield per hectare and high farm gate price make it the
most economically attractive type going forward.
Figure 19 shows a sharp decrease in percent area farmed after 2003 before
evening out again at a lower level in 2007. From 2002 to 2012, Virginia type tobacco
went from nearly three percent of total provincial area to just above one percent. Native
type tobacco also experienced a decrease in area farmed after 2003; however the decrease
was not nearly as sharp. Percent area planted for Native tobacco has been increasing
slightly since 2008 and by 2012, recovered to nearly 80 percent of 2003 levels.

Figure 19 Percent of La Union province total area devoted to tobacco farming
Source: Data from the Bureau of Agricultural Statistics
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Percent Area Planted
La Union
Other
Virginia
Native
62

Figure 20 shows a similar graph to Figure 19; replacing percent total area with
volume of production. Although most of the same trends are visible in this graph, the
years 2000 and 2001 are additionally visible. Production for both Native and Virginia
tobacco appears relatively steady from 2000 to 2003, decreases substantially from 2003
to 2007, and becomes steady again until 2012. The decrease largely mirrors rates in
Figure 19, indicating that no substantial increases in yield per hectare offset the decrease
in tobacco farming area.

Figure 20 Total Volume of Production for La Union Province (Metric Tons)
Source: Data from the Bureau of Agricultural Statistics

Although the trend is clear that tobacco farming in La Union province is
decreasing, Native tobacco has endured better than Virginia tobacco, perhaps partially
due to observations regarding farm gate price noted above. Although Burley data would
0
2000
4000
6000
8000
10000
12000
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Volume of Production
La Union
Other
Virginia
Native
63

be an insightful addition, there is no official indication that the “Other” type tobacco
listed in Figures 20 and 21 represents Burley tobacco. The “Other” type data were simply
calculated by subtracting Virginia and Native data from the available cumulative figures.

64

CHAPTER 5: DISCUSSION
Although tobacco area planted and volume of production decreased dramatically in the
Philippines from 1990 to 2012, many nuanced and localized trends emerged from
analysis performed using both stationary and dynamic visualization techniques that
provide much deeper, more contextualized insight into the industry. Tobacco farming
remains an incredibly pertinent and controversial issue, particularly in the ASEAN region
where tobacco regulation is being continuously updated and expanded.

5.1 Potential Causes
One of the first observations noted while analyzing the percent area planted and
volume of production data was the two large downturns that occurred from 1992 to 1994
and from 2003 to 2007. Although not insinuating causation, it appears as if two very
separate occurrences may have played roles in each of these downturns.
The Mount Pinatubo eruption began in the summer of 1991 and continued
through 1992, causing nearly USD 500 million in damage and completely disrupting the
economy of Luzon. Much of the damage locally came from lahars, rain induced torrents
of volcanic debris, in the months following the eruptions. Mount Pinatubo is located on
the border between Zambales, Tarlac, and Pampanga provinces on Luzon Island. This
would explain why tobacco growing provinces closer to Mount Pinatubo, such as
Pangasinan and La Union, experienced the sharpest drop in production while provinces
further away, such as Ilocos Sur and Isabela, experienced less severe drops.
Prior to the eruption, nearly 10 percent of the total area of La Union province was
devoted to tobacco farming. Within two years, the number dropped to under five percent.
65

In Pangasinan, production dropped from a high of 31,168 metric tons in 1992, to 8,122
metric tons in 1994. As further confirmation, the southern province of Davao Del Sur,
over 1,500 kilometers away, actually increased tobacco production from 654 metric tons
in 1993 to 1,639 metric tons in 1994.
The downturn in production from 2003 to 2007 appears to be less of a dramatic,
geographically variable shock, and more of an all-encompassing, increasing rate of
decline. The sustained decline in tobacco farming more or less corresponds to the passage
of the Tobacco Regulation Act of 2003, and the ensuing five year program aimed at
transitioning tobacco farmers into alternative farming systems and livelihoods.
The Act established the Tobacco Growers’ Assistance Program and the Tobacco
Growers’ Cooperative. The Tobacco Growers’ Assistance Program provided financial
support both to farmers displaced due to the implementation of the Act, and those who
voluntarily gave up tobacco farming. The Tobacco Growers’ Cooperative assisted
tobacco farmers in developing alternative farming systems and encouraged tobacco
farmers to plant alternative crops.
Because the trend is less sharp and less geographically distinct, it is not possible
to say with complete confidence that the two are related, however it remains likely. It
should also be noted that there is an increase in tobacco production between 2008 and
2012 directly following the expiry of the five year program implemented under the
Tobacco Regulation Act of 2003. It would be an interesting study to determine how much
of the increase is from former tobacco farmers switching back again as a result of the
expiry of funded programs under the Act.
66

Further localized disrupters such as typhoons or droughts, along with natural
economic pressure to shift production to more productive crops, may contribute towards
national trends and regional anomalies, leaving many questions unanswered and in need
of further research. Why does Native tobacco production reduce by nearly 50 percent
from 2005 to 2006? Why did La Union not experience the same increase in production
from 2008 to 2012 that its northern neighbor, Ilocos Sur, experienced?
Although this study exposes trends evident in available spatial data, it does not
provide all of the necessary answers regarding causation. To more confidently correlate
causation and pick up on smaller trends, it would be helpful to have a longer track record
of tobacco farming data, inclusive of all three types: Native, Virginia, and Burley. It
would also be helpful to obtain more detailed tobacco farming data at the barangay level
to open up the possibility of spatiotemporal analysis methods beyond visualization.
Lastly, a comparative investigation to other crops, particularly those which can be grown
in lieu of tobacco, could provide valuable information for policy decision-makers.

5.2 Key Takeaways
Although farm gate prices for all tobacco types have risen at a relatively linear
rate since 1990, the rate of increase in relative farm gate price for Native and Burley type
tobacco has readily outpaced that of Virginia type tobacco, yet Virginia tobacco still
accounts for over half of all tobacco produced in the Philippines. This would seem to
indicate that more tobacco farmers will increasingly feel pressure to change the type of
tobacco grown, particularly in the four largest Virginia tobacco producing provinces.
Additionally, 2008 to 2012 marked the first time in the entire 23 year study period with
67

four years of consecutive growth in total tobacco production, and could be a sign that
tobacco farming is making a comeback.
In general, Native tobacco farming was more widespread geographically, while
Virginia tobacco was largely limited to only four provinces in northwestern Luzon. All of
the major tobacco growing provinces in northern Luzon share a similar history of tobacco
growing, remnants of the Spanish tobacco monopoly. The center of tobacco production in
the Philippines is squarely located in Luzon for all types, and any policy measures should
be focused on northern Luzon, particularly the Virginia tobacco growing provinces. Any
tobacco farming policy measure would be felt at a proportionally higher rate for the
northeastern Luzon provinces as well.
Of the methods used, I believe that dynamic visualization would be the most
useful for policy makers. The most important difference between stationary and dynamic
visualization is that for dynamic visualization, data are presented easily on a single output
map, whereas stationary visualization requires an animation or static time series. In this
case, dynamic visualization clearly shows the skewed importance of tobacco farming in
Northern Luzon, and the relative magnitude and direction of change by year.
Stationary visualization techniques did, however, provide deeper insight into
provincial level changes that could be further investigated through supplemental graphs
and statistics. For example, it would have been difficult to attribute the downturn from
1992 to 1994 to the Mount Pinatubo eruption, without first visualizing the production dip
in surrounding provinces.

68

CHAPTER 6: CONCLUSIONS
This thesis set out to demonstrate the efficacy of spatiotemporal visualization and
analysis as a policy support tool by performing a case study on tobacco farming in the
Philippines. The topic of the case study was chosen precisely due to its controversial
nature; so that the results could provide novel, spatially driven insights for policy makers
to make better informed decisions. Thus, the research was framed to determine how
spatial variation in tobacco farming has changed over time in the Philippines.
Five key questions were developed to better address challenges faced from the
perspective of a national level policy maker tasked with deciding the future of tobacco
farming policy in the Philippines. These questions served as the basis for analysis by
means of stationary and dynamic visualization techniques. From these questions, it was
hypothesized that clear spatiotemporal trends would emerge and that those trends could
be linked to potential causes. The key questions were:
1. What provinces and regions are most dependent on tobacco farming and how has
this changed spatially over time?
2. What provinces and regions produce the highest total volume of tobacco
production and how has this spatially changed over time?
3. Is there any spatial variation in farm gate prices and, if so, how has this changed
spatially over time?
4. Considering the first three key questions, is there any spatial variation between
the three types of tobacco grown in the Philippines: Native, Virginia, and Burley?
If so, how has this changed over time?
5. If clear spatiotemporal trends are identified, can possible causes also be reasoned?
69

For question 1, it was found that in 1990, La Union was most dependent on
tobacco farming. By 2012 however, Ilocos Sur was the most dependent on tobacco
farming. Overall, the Ilocos Region proved to be the most dependent on tobacco farming
throughout the timeframe despite substantial declines in tobacco hectares planted since
1990.
Question 2 resulted in similar conclusions to question 1. Because the data were
not normalized to provincial area, the Cagayan Valley Region gained more significance
in the analysis of tobacco volume of production. Additionally, the increasing dominance
of Ilocos Sur over the time period became visible by using dynamic visualization
techniques.
For Question 3, farm gate pricing varied by type, but not necessarily by location,
with the notable exception of Native type tobacco in the southern provinces. Virginia
type prices showed the least volatility, and also the lowest annual rate of growth in
average farm gate price. The annual rate of growth in farm gate prices for Native type
tobacco was quite a bit higher, likely driving the observed increase in production in
northern Luzon.
For question 4, it was found that Native type tobacco production totaled much less
than Virginia type yet was grown in many more provinces. Native type tobacco
production was spread across many regions however the number of provinces where
Native type tobacco was grown decreased over time. Virginia type production was very
much centered in only a couple provinces in the Ilocos Region. In the last couple years
however, a few other provinces began growing Virginia type tobacco.
70

For question 5, large trends were picked up and linked to potential causes.
Notably, there was a regionalized downturn in production from 1992 to 1994 that may
have been attributed to the Mount Pinatubo eruption. Additionally, there was a
countrywide downturn from 2003 to 2007 that may have been a result of the Tobacco
Regulation Act of 2003 and the ensuing five year program aimed at transitioning tobacco
farmers into alternative farming systems and livelihoods.
On all accounts, I believe that this thesis has achieved its goal. The spatiotemporal
analysis provides empirical evidence for policy makers to better understand regional and
provincial trends in tobacco farming over time. Tobacco farming has drastically reduced
in some provinces, yet remained steady or increased in others. Understanding where and
why the variability occurs is important for determining the future direction of tobacco
farming policy in the Philippines.

71

REFERENCES
Alechnowicz, K.., and S. Chapman. 2004. The Philippine tobacco industry: “the strongest
tobacco lobby in Asia”. Tobacco Control 13:71-8.
Andrienko, N., G. Andrienko, and P. Gatalsky. 2003. Exploratory spatio-temporal
visualization: An analytical review. Journal of Visual Languages & Computing 14
(6):503-41.
Berisso, Z.A., and W.T. de Vries. Exploring charactersitics of GIS adoption decisions
and type of induced changes in developing countries: the case of Ethiopia. The
Electronic Journa of Information Systems in Developing Countries 40 (2):1-16
Brown, T., and G. Moon. 2012. Geography and global health. The Geographical Journal
178 (1):13-7.
Cabrera, O., and L. Gostin. 2011. Human rights and the framework convention on
tobacco control: mutually reinforcing systems. International Journal of Law in
Context 7 (3):285-303
Campbell, M., and D. Ballas. 2013. A spatial microsimulation approach to economic
policy analysis in Scotland. Regional Science Policy & Practice 5 (3):263-288.
Citi Research. 2014. Annual Cigarette Synopsis.
Coppock, J. T. 1965. Tobacco growing in Nigeria (Tabakanbau in Nigeria). Erdkunde 19
(4) (Nov.): 297-306.
Crouch, H.A. 1985. Economic change, social structure, and the political system in
Southeast Asia: Philippine development compared with the other ASEAN countries.
Institute of Southeast Asian Studies.
Dahly, D.L., P. Gordon-Larsen, M. Emch, J. Borja, and L.S. Adair. 2013. The spatial
distribution of overweight and obesity among a birth cohort of young adult Filipinos
(Cebu Philippines, 2005): an application of the Kulldorff spatial scan statistic.
Nutrition and Diabetes 3 (80).
De Jesus, E.C. 1980. The tobacco monopoly in the Philippines: Bureaucratic enterprise
and social change, 1766-1880. Ateneo de Manila University Press Quezon City.
Deng, M., Q. Liu, J. Wang, and Y. Shi. 2013. A general method of spatio-temporal
clustering analysis. Science China Information Sciences 56 (10): 1-14.
Desai, A., R.T. Greenbaum, and Y. Kim. 2009. Incorporating policy criteria in spatial
analysis. The American Review of Public Administration 39 (1):23-42.
72

Espino, R.R, D. Evangelista, and E.U. Dorotheo. 2009. Survey of the tobacco growing
areas of the Philippines. Southeast Asia Tobacco Control Alliance.
ESRI. 2011. Working with temporal data in ArcGIS. ESRI Regional User Group
Conferences, Redlands CA.
Euromonitor International. 2013. Passport: Tobacco in the Philippines.
Fisher, L. 2000. Tobacco farming and tobacco control in the United States. Cancer
Causes & Control 11 (10):977-9.
Gleditsch, K.S., and N.B. Weidmann. 2012. Richardson in the information age:
Geographic information systems and spatial data in international studies. Annual
Review of Political Science 15:461-81.
Kayaalp, Ebru. 2009. Livelihoods up in smoke: Tobacco farming in turkey. Middle East
Report 252:34-7.
King, B.A., S. Dube, and M.A. Tynan. 2012. Current tobacco use among adults in the
United States: Findings from the national adult tobacco survey. American Journal of
Public Health 102 (11):93-100.
Launio, C.C., G.O. Redondo, J.C. Beltran, and Y. Morooka. 2008. Adoption and spatial
diversity of later generation modern rice varieties in the Philippines. Agronomy
Journal 100 (5): 1380-1389.
McKinsey Global Institute. 2012. Urban world: cities and the rise of the consuming class.
Monastiriotis, V. 2007. Patterns of spatial association and their persistence across socio-
economic indicators: the case of the Greek regions. The Hellenic Observatory.
Moog, F. 2006. Country pasture/forage resource profiles: Philippines. World Health
Organization.
Moon, G., and R. Barnett. 2003. Spatial scale and the geography of tobacco smoking in
New Zealand: A multilevel perspective. New Zealand Geographer 59 (2): 6-15.
Philippine Statistics Authority. 2010. Data kit of official Philippine statistics.
Republic Act No. 9211. 2003. An act regarding the packaging sale, distribution and
advertisements of tobacco products and for other purposes.
Sahu, S.K., and K.V. Mardia. 2005. Recent trends in modeling spatio-temporal data.
Paper presented at Proceedings of the special meeting on Statistics and Environment.
73

Southeast Asia Initiative on Tobacco Tax. 2011. Cigarette tax and price: affordability and
impacts on consumption and revenue.
Southeast Asia Initiative on Tobacco Tax, and Southeast Asia Tobacco Control Alliance.
2013. ASEAN tobacco tax report card: regional comparisons and trends.
Southeast Asia Tobacco Control Alliance. 2008. Cycle of poverty in tobacco farming:
tobacco cultivation in Southeast Asia.
Southeast Asia Tobacco Control Alliance. 2013. Child labour in tobacco cultivation in
the ASEAN region.
University of the Philippines College of Law Development Foundation, Inc. 2010.
Manual for the implementation of republic act 9211 (tobacco regulation act of 2003)
and other related laws.
Von Eichborn, S., and L. Norger. 2012. Alternative livelihoods to tobacco: approaches
and experiences.
World Bank. 1999. Curbing the epidemic: Governments and the economics of tobacco
control. Tobacco Control 8 (2) (June 01): 196-201.
World Bank. 2013. Press Release: New partnership seeks to bring benefits of open data to
developing countries.
World Health Organization. 2005. WHO framework convention on tobacco control.
World Health Organization. 2008. Tobacco and poverty in the Philippines.
World Health Organization. 2010. Economically sustainable alternatives to tobacco
growing (in relation to articles 17 and 18 of the WHO framework convention on
tobacco control).
World Health Organization. 2012. 2012 global progress report on implementation of the
WHO framework convention on tobacco control.
Worrall, L., and D. Bond. 1997. Geographical information systems, spatial analysis and
public policy: The British experience. International Statistical Review 65 (3): 365-
79.

74

APPENDIX A: PERCENT OF TOTAL PROVINCIAL AREA DEVOTED TO
GROWING TOBACCO, BY TYPE AND YEAR

Figure 21 All Tobacco Types

75

Figure 21 All Tobacco Types (continued)

76

Figure 21 All Tobacco Types (continued)
77

Figure 21 All Tobacco Types (continued)

78

Figure 22 Native Tobacco Type

79

Figure 22 Native Tobacco Type (continued)

80

Figure 23 Virginia Tobacco Type

81

Figure 23 Virginia Tobacco Type (continued)

82

APPENDIX B: TOTAL VOLUME OF TOBACCO PRODUCTION IN METRIC
TONS, BY TYPE AND YEAR

Figure 24 All Tobacco Types
83

Figure 24 All Tobacco Types (continued)
84

Figure 24 All Tobacco Types (continued)
85

Figure 24 All Tobacco Types (continued)

86

Figure 25 Native Tobacco Type

87

Figure 25 Native Tobacco Type (continued)
88

Figure 25 Native Tobacco Type (continued)

89

Figure 26 Virginia Tobacco Type
90

Figure 26 Virginia Tobacco Type (continued)

91

Figure 26 Virginia Tobacco Type (continued)

92

APPENDIX C: FARM GATE PRICE OF TOBACCO BY PROVINCE, PESOS
PER KILOGRAM

Figure 27 Native Tobacco Type
93

Figure 27 Native Tobacco Type (continued)
94

Figure 27 Native Tobacco Type (continued)

95

Figure 27 Native Tobacco Type (continued)

96

Figure 28 Virginia Tobacco Type
97

Figure 28 Virginia Tobacco Type (continued)

98

Figure 28 Virginia Tobacco Type (continued)
99

Figure 28 Virginia Tobacco Type (continued)
100

Figure 29 Burley Tobacco Type
101

Figure 29 Burley Tobacco Type (continued)
102

Figure 29 Burley Tobacco Type (continued)
103

Figure 29 Burley Tobacco Type (continued)

Abstract (if available)

Abstract This study demonstrates the utility of visualization‐based spatiotemporal analysis as a policy support tool in the agricultural sector through a case study analyzing changes in the spatial distribution of tobacco farming in the Philippines from 1990 through 2012. Tobacco farming remains divisive in the Philippines

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Integrating spatial visualization to improve public health understanding and communication

PDF

Visualizing sporting event temporary flight restrictions

PDF

Address points and a maser address file: improving efficiency in the city of Chino

PDF

Spatiotemporal patterns of salt and nutrient contamination in Los Angeles County's groundwater basins

PDF

Cartographic approaches to the visual exploration of violent crime patterns in space and time: a user performance based comparison of methods

PDF

Generating bicyclist counts using volunteered and professional geographic information through a mobile application

PDF

Using pattern oriented modeling to design and validate spatial models: a case study in agent-based modeling

PDF

Investigating electoral college reform: geography's impact on elections, and how maps influence our perception of election outcomes

PDF

Congestion effects on arterials as a result of incidents on nearby freeway: When should you get off the highway?

PDF

Analyzing earthquake casualty risk at census block level: a case study in the Lexington Central Business District, Kentucky

PDF

3D visualization models as a tool for reconstructing the historical landscape of the Ballona Creek watershed

PDF

Modeling open space acquisition in Boulder, Colorado

PDF

Determining the utility of GIS in border disputes case study: Sudan and South Sudan

PDF

Modeling patient access to point-of-care diagnostic resources in a healthcare small-world network in rural Isaan, Thailand

PDF

The role of GIS in asset management: integration at the Otay Water District

PDF

Social media canvassing using Twitter and Web GIS to aid in solving crime

PDF

GIS data curation and Web map application for La Brea Tar Pits fossil occurrences in Los Angeles, California

PDF

An evaluation of soil sampling methods in support of precision agriculture in Northeastern North Carolina

PDF

Semi-automated visualization of spatial information in unstructured text

PDF

Using volunteered geographic information to model blue whale foraging habitat, Southern California Bight

Asset Metadata

Creator Rubinyi, Steven Louis (author)

Core Title Spatiotemporal visualization and analysis as a policy support tool: a case study of the economic geography of tobacco farming in the Philippines

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 07/02/2014

Defense Date 05/30/2014

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

Tag economic geography,GIS,OAI-PMH Harvest,Philippines,Public Policy,spatiotemporal,tobacco farming

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kemp, Karen K. (committee chair), Chiang, Yao-Yi (committee member), Vos, Robert O. (committee member)

Creator Email slrubinyi@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-431667

Unique identifier UC11287183

Identifier etd-RubinyiSte-2608.pdf (filename),usctheses-c3-431667 (legacy record id)

Legacy Identifier etd-RubinyiSte-2608.pdf

Dmrecord 431667

Document Type Thesis

Format application/pdf (imt)

Rights Rubinyi, Steven Louis

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