Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Essays on inflation, stock market and borrowing constraints
(USC Thesis Other)
Essays on inflation, stock market and borrowing constraints
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
ESSAYS ON INFLATION, STOCK MARKET AND BORROWING CONSTRAINTS
by
Hüseyin Günay
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
August 2010
Copyright 2010 Hüseyin Günay
Dedication
to my mom Sevim, my dad ¸ Seref, my sister Ay¸ senur, my aunt Naciye,
and to my uncle Ali, who passed away but not forgotten...
specially to my beloved wife Nuray and my daughter Efnan who is not born yet...
also to all my family members and everyone that I love.
ii
Acknowledgments
I would like to thank and extend my heartfelt gratitude to the following persons
who helped me through my education in the University of Southern California and
have made the completion of this dissertation possible.
I am grateful to Professor Selahattin
˙
Imrohoro˘ glu and Professor Christopher S.
Jones for their continuous support, encouragement, help and invaluable comments
for my dissertation. I would not have been able to complete this dissertation without
them. I am also grateful to Professor Fernando Zapatero for his comments and advice
on my work.
I am thankful to Professor Ay¸ se
˙
Imrohoro˘ glu for her constant trust on and support
to me. I thank Professors Cheng Hsiao, Robert Dekle, Vincenzo Quadrini and Geert
Ridder for their comments and advice. I also would like to thank all the Department of
Economics faculty members and staff. I also thank to my friends Cenk Cevat Karahan
and Mustafa Kılınç.
I specially thank to my parents for everything they have sacrificed to support me.
Responsible for the remaining errors, only I am.
iii
iv Table of Contents
Dedication ii
Acknowledgments iii
List of Tables vi
List of Figures ix
Abstract x
Chapter 1: Heterogeneous Expected Inflation and the Stock Market 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.1 Two-Period Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.2 Risk Neutral Agent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3.3 Heterogeneous Expected Inflation for the Case of Risk Neutral
Agents in Two Period Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.5 Empirical Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
40
Chapter 2: Effects of Inflation in Predicting the Stock Market
Returns: International Evidence 42
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.4 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
v Chapter 3: Boom-Bust Cycles in Turkey: Capital Market
Imperfections and Asymmetries in Sector Based Investment 55
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.2 Literature Review. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.3 Macro Co-movements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.3.1 Financial Liberalization and Boom-Bust Cycles . . . . . . . . . . . . 63
3.3.2 Macro Co-movements and Credit . . . . . . . . . . . . . . . . . . . . . . . 67
3.4 Micro Evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.4.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.4.3 Results for Company Accounts Survey Data Set . . . . . . . . . . . 75
3.4.4 Results for Istanbul Stock Exchange Firms Data Set . . . . . . . . 81
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Bibliography 88
Appendices 93
Appendix A: Supplementary Materials for Chapter 1 93
Appendix B: Supplementary Materials for Chapter 2 102
Appendix C: Supplementary Materials for Chapter 3 109
C.1 Data Set Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
C.2 Additional Data Information and Regressions . . . . . . . . . . . . . . . . 114
Appendix D: Another Model for Chapter 3 121
D.1 Accelerator Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
vi List of Tables
1.1 Data Downloaded for Return Regressions . . . . . . . . . . . . . . . . . . . . . . . . 23
1.2 Variables Used in Return Regressions . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.3 Descriptive Statistics for Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.4 Forming Inflation Expectations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.5 Descriptive Statistics for the Rest of the Variables . . . . . . . . . . . . . . . . . 29
1.6 Replicating Fama and Schwert 1977 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.7 Fama and Schwert 1977 Model for 1950-2008 . . . . . . . . . . . . . . . . . . . . 32
1.8 Regression Results for VWRETX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
1.9 Return Regression Results for VWRETX cont.’d . . . . . . . . . . . . . . . . . . 35
1.10 Return Regression Results for VWRETX with Dummy . . . . . . . . . . . . 36
1.11 Return Regression Results for VWRETX with DYMA . . . . . . . . . . . . 38
1.12 Return Regression Results for VWRETX with DYMA cont.’d . . . . . . 39
1.13 Return Regression Results for VWRETX with Dummy and DYMA . . 40
2.1 Forming Inflation Expectations for Int.’l Data Set . . . . . . . . . . . . . . . . . 45
2.2 Fixed Effect Return Regression Results for XSR . . . . . . . . . . . . . . . . . . 48
2.3 Fixed Effect Return Regression Results for XSR cont.’d . . . . . . . . . . . . 49
2.4 Fixed Effect Return Regression Results for XSR with Dummy . . . . . . . 50
vii 2.5 Fixed Effect Return Regression Results for XSR . . . . . . . . . . . . . . . . . . 51
2.6 Fixed Effect Return Regression Results for XSR cont.’d . . . . . . . . . . . . 52
2.7 Fixed Effect Return Regression Results for XSR with Dummy . . . . . . . 53
3.1 Macro Data Regression Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.2 Variable Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.3 Summary Statistics for CBRT Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.4 Regression Results for CBRT Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.5 Pooled Regression Results for CBRT Data Set . . . . . . . . . . . . . . . . . . . . 85
3.6 Summary Statistics for ISE Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.7 Pooled Regression Results for ISE Data Set . . . . . . . . . . . . . . . . . . . . . . 87
A.1 Return Regression Results for VWRETD . . . . . . . . . . . . . . . . . . . . . . . 93
A.2 Return Regression Results for VWRETD cont.’d . . . . . . . . . . . . . . . . . 94
A.3 Return Regression Results for VWRETD with Dummy . . . . . . . . . . . . 95
A.4 Return Regression Results for SP500RET . . . . . . . . . . . . . . . . . . . . . . . 96
A.5 Return Regression Results for SP500RET cont.’d . . . . . . . . . . . . . . . . . 96
A.6 Return Regression Results for SP500RET with Dummy . . . . . . . . . . . . 97
A.7 Return Regression Results for VWRETD with DYMA . . . . . . . . . . . . . 98
A.8 Return Regression Results for VWRETD with DYMA cont.’d . . . . . . . 98
A.9 Return Regression Results for VWRETD with Dummy and DYMA . . 99
A.10 Return Regression Results for SP500RET with DYMA . . . . . . . . . . . 100
A.11 Return Regression Results for SP500RET with DYMA cont.’d . . . . . 100
viii A.12 Return Regression Results for SP500RET with Dummy and DYMA . 101
C.1 Survey and Observation Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
C.2 Sector Names . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
C.3 Number of Firms in Tradable (Manufacturing) Sectors . . . . . . . . . . . . . 117
C.4 Number of Firms in Non-tradable (Non-manufacturing) Sectors . . . . . . 117
C.5 Sectors and Number of Firms in each Sector . . . . . . . . . . . . . . . . . . . . . 118
C.6 Regression Results for CBRT Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . 119
C.7 Pooled Regression Results for CBRT Data Set . . . . . . . . . . . . . . . . . . . 120
D.1 Variable Definitions and Data Source for the Micro Data Sets . . . . . . . 128
D.2 Regression Results with Time Fixed Effects . . . . . . . . . . . . . . . . . . . . . 132
D.3 Pooled Regression Results with Time Fixed Effects . . . . . . . . . . . . . . . 133
D.4 Regression Results with Time Fixed Effects . . . . . . . . . . . . . . . . . . . . . 134
D.5 Pooled Regression Results with Time Fixed Effects . . . . . . . . . . . . . . . 135
D.6 Summary Statistics for Additional Variables . . . . . . . . . . . . . . . . . . . . . 136
D.7 Pooled Regression Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
D.8 Pooled Regression Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
ix List of Figures
1.1 Annualized Monthly Inflation in Percentage. . . . . . . . . . . . . . . . . . . . . . 24
2.1 Annualized Monthly Inflation in Percentage-5% . . . . . . . . . . . . . . . . . . 46
3.1 Percentage Deviations of GDP from HP trend . . . . . . . . . . . . . . . . . . . . 65
3.2 Percentage Deviations from HP trend: Consumption and Investment. . . . 65
3.3 Non-tradable to Tradable Output Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.4 % Private Credit / GDP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.5 Investment over Capital ratio for Tradable and Non-tradable sectors . . . 76
3.6 Cash flow over Capital ratio for Tradable and Non-tradable sectors . . . 77
3.7 Investment over Capital ratio for Tradable and Non-tradable sector . . . 82
3.8 Cash flow over Capital ratio for Tradable and Non-tradable sector . . . . 82
B.1 Canada Annualized Monthly Inflation in Percentage . . . . . . . . . . . . . . 102
B.2 France Annualized Monthly Inflation in Percentage . . . . . . . . . . . . . . . 103
B.3 Germany Annualized Monthly Inflation in Percentage . . . . . . . . . . . . . 104
B.4 Italy Annualized Monthly Inflation in Percentage . . . . . . . . . . . . . . . . . 105
B.5 Japan Annualized Monthly Inflation in Percentage . . . . . . . . . . . . . . . . 106
B.6 UK Annualized Monthly Inflation in Percentage . . . . . . . . . . . . . . . . . 107
B.7 US Annualized Monthly Inflation in Percentage . . . . . . . . . . . . . . . . . . 108
Abstract
In this dissertation, firstly I model the relation between the stock market and infla-
tion, and provide empirical evidence of my theory in the US equity market. The second
chapter has international evidence of effects of inflation on the equity markets for the
G7 countries. In the third chapter, I look at the effects of business cycle fluctuations
on the equity market and explain them through the credit channel via borrowing con-
straints on different sectors in a developing country..
Both recent and previous research mentioned the relationship among inflation,
stock return and dividend yield (or price to earnings ratio). These papers empirically
show that there is a positive correlation between dividend yield and inflation, and a
negative link between expected inflation and stock returns. Nevertheless, these prop-
erties had not been used for modelling asset prices up to now. I model the relation
between inflation and stock prices from a standard asset pricing model stand point
by introducing nominal prices for a share of stock. Furthermore, by introducing het-
erogeneity among agents in terms of inflation expectations, the model investigates the
implications of having differences among the agents on the stocks. Using the model
x
provided, it is shown that heterogeneity in inflation expectations can cause underval-
uation or overvaluation in the stock market resulting in predicted returns to be posi-
tive or negative, respectively. These predictions are corroborated by the results of the
empirical study on the United States equity market. Results do not only show that
stock returns are on average negatively affected by expected inflation on average but
also that the effect is reversed during high inflationary periods, in accordance with the
predictions of the model. The second chapter of this dissertation presents a similar
panel regression study on a group of developed countries; Canada, France, Germany,
Italy, Japan, the United Kingdom and the United States, the so called G7 countries.
International empirical study also yields results comparable to the US study in the first
chapter.
The capital market imperfections have important consequences for aggregate cycles,
especially for financially developing countries. The research on the relation between
imperfections and output dynamics at the macro level are ample, but the lack of wide
coverage micro data sets for developing countries limit the study of aggregate implica-
tions of the micro level capital market imperfections. The final part of my dissertation
first documents the boom-bust cycles in Turkey and shows that non-tradable sector
is more volatile over the business cycles than tradable sector. Moreover, this sector
based asymmetry is strongly correlated with aggregate credit movements. To establish
the connection between the sector based asymmetries and the credit markets further,
secondly it constructs two micro data sets. Using structural estimation, the study finds
xi
that non-tradable sector is financially more constrained than tradable sector. With non-
tradable sector being more constrained, credit movements become an important deter-
minant of boom-bust cycles. Therefore, it can be established that the asymmetry in the
financial constraints of the different sectors at the micro level can generate the observed
asymmetrical aggregate response of sectors over the business cycle.
xii
Chapter 1
Heterogeneous Expected Inflation
and the Stock Market
1.1 Introduction
In this chapter, I use a two-period model to establish the relationship between
expected inflation and stock price levels and dividend yield. I do not only show that
the stock returns are affected by inflation expectations but also that stock price levels
move from their fundamental value i.e., they become undervalued or overvalued. In
addition, this move for the stock prices change direction depending on the nature of
the heterogeneity in expected inflations. To the author’s best knowledge, this is the
first attempt in the literature to incorporate a nominally priced stock, which pays real
dividends, in to a standard Consumption Based Asset Pricing Model to see the effects
of inflation and inflation expectations on the stock market. Furthermore, evidence from
the empirical study supports the predictions of the two period model.
The stock market has been long thought of as a hedge against inflation due to the
fact that stocks are claims on real assets in the future and their value should not be
affected by the changing price level in the economy. Although this comes very natu-
rally at first, both recent and previous research has found a relation among inflation,
1
stock returns and dividend yield that challenges this basic intuition. In these papers,
it is empirically shown that there is a positive relation between dividend yield and
inflation and a negative link between expected inflation and stock returns, but neither
of these properties has been used for modeling asset prices as of yet. Moreover, the
effect of inflation on stock market returns has been practically neglected more recently
in the United States. due to the prolonged periods of low inflation over the last couple
of decades
1
. With the United States government injecting billions of dollars into the
economy to stabilize the markets in the aftermath of the recent financial crisis, there is
a renewed interest in how increasing expected inflation affects the stock market
2
. This
chapter of the dissertation models the relation between inflation and stock prices in a
standard two-period consumption based asset pricing model by introducing nominal
prices for a share of stock. Although the stock can be bought or sold at nominal prices,
it still pays real dividends. By also introducing heterogeneity among agents in terms of
inflation expectations, it investigates the implications of differences among the agents
on the pricing of stocks and other claims. The model not only predicts the negative
link between returns and expected inflation but also gives an idea about how the het-
erogeneity among agents may cause overpricing during low inflationary periods and
underpricing during high inflationary periods. Motivated by the simple two-period
model, I empirically examine the effects of expected inflation on the United States stock
1
Using Robert Shiller’s data set on CPI from his website, www.econ.yale.edu/~shiller/data.htm , aver-
age annualized monthly inflation in the United States is 2.9% for the 1993-2008 period.
2
The United States money supply increase from January 2008 to April 2010 is $341.2 bil-
lions, 24.81%, and $1,074.5 billions, 14.29%, in terms of M1 and M2 measures, respectively.
Data set used for above calculations is Monthly Historical Money Stock Tables under Federal
Reserve Statistical Release, H6 Money Stock Measures, Historical Data released on June 3, 2010.
http://www.federalreserve.gov/releases/h6/hist
2
market returns. I show in the statistical analysis part of the chapter that the effect of
the inflation on future returns is negative on average but the effect is positive during
high inflationary periods, i.e. stock prices are undervalued during a high inflationary
period predicting future positive returns and vice versa.
The outline of the chapter is as follows; section 1.2 gives a brief literature review
about the past and recent research on the subject. In section 1.3, two-period consump-
tion based asset pricing model is introduced with the real dividend paying nominal
priced stock to show the effects of expected inflation itself and heterogeneous expected
inflations on the stock market. Section 1.4 explains the data set used for the empirical
study. Then section 1.5 presents the results of the empirical study regarding the effects
of inflation on the stock market returns followed by the conclusion in section 1.6.
1.2 Literature Review
Effects of inflation both on the stock market returns and dividend yield have been
extensively studied for the high inflationary periods of late 70’s and early 80’s. How-
ever, they are evidently forgotten for the recent years in the U.S. due to the low infla-
tionary periods of the last couple of decades. Although the recent crisis of the Ameri-
can economy sow deflationary worries in to the minds of the investors at first, an edu-
cated economist looking at the FED’s balance sheet would predict high inflation in the
future and start working on more pressing effects of inflation on the equity market. In
this section, I will provide background information on the past research in somewhat
chronological order.
3
Fama and Schwert (1977) documents that realized stock returns are negatively
related to the expected inflation using both equally weighted and value weighted port-
folios of all NYSE stocks in 1953-1971 period, during which the average annual infla-
tion rate was around 2.2%. They use the Fisher equation, E
E
, and assum-
ing real interest rate not changing through out time, they predict the expected inflation
as a constant plus the short term interest rate, E
E
constant
, in the
economy. Plugging it in to the linear regression model of Equation 1.1, they find out
that the stock returns are negatively related to not only expected but also unexpected
part of the inflation and changes in expected inflation rate, though these results are
found to be less consistent. In addition, Gultekin (1983) prove the negative effect of
unexpected inflation on stock returns for the 1952 to 1979 period.
(1.1)
Modigliani and Cohn (1979) uses a present value discounted model to find out
the relation between low values of stock prices and high inflation during 70’s. They
claim that in the presence of unexpected and fluctuating inflation some investors are
affected by certain forms of "money illusion" or "inflation illusion". As a result of this
effect, they make errors in pricing equities as if real payoffs are discounted at nomi-
nal interest rate. According to Modigliani and Cohn (1979), investors systematically
undervalued the stock prices during 70’s because of these inflation-induced errors.
They pointed out that "Rationally valued, the level of the S&P 500 at the end of 1977
should have been 200. Its actual value at that time was 100". If investors are making
4
errors by means of discounting real values by nominal interest rate, then for a ratio-
nal investor this does not only mean that stock prices are undervalued during high
inflation but also overvalued during low inflation periods. Modigliani and Cohn also
checked two more possibilities; (i) inflation being accompanied by a significant deteri-
oration of profits and (ii) the capitalization factor to be applied to earnings decreasing
with the rate of inflation, but neither of these found support in their research. With
their idea of money illusion they become the first researchers to talk about heterogene-
ity in terms of expected inflation.
In a recent study, Campbell and Vuolteenaho (2004)’s results provide strong sup-
port to Modigliani and Cohn (1979). They used the log-linear dynamic valuation
framework of Campbell and Shiller (1988) to allow for time varying discount rates
and Gordon growth model of Williams (1938) and Gordon (1962) to define the divi-
dend price ratio as being equal to both the objective and subjective differences of long
term discount and dividend growth rates, and
, i.e.
. Using the model in Equa-
tion 1.2 and taking the residual of the model as the mispricing term, they show its
relation with the inflation as an evidence of Modigliani and Cohn hypothesis in a VAR
framework.
(1.2)
Sharpe (2002) also use the framework of Campbell and Shiller (1988) with some
modifications to have the price to earnings per share ratio instead of dividend yield as
5
the dependent variable. Rather than using econometric models for expected values, he
used survey data for these variables. He discovers a strong relation between price to
earnings per share ratio and expected inflation claiming that inflation’s effect on stock
prices is a result of its impact on both earnings growth and required returns.
There are also general equilibrium models in the literature which used these ideas.
In an earlier version of their forthcoming paper, Basak and Yan (2009) model money
illusion in a continuous time general equilibrium model framework by introducing a
price effect in the Constant Relative Risk Aversion (CRRA) utility function of the risk
averse investor. They find a negative coefficient for the marginal effect of expected
inflation on dividend yield in their model. They later changed their approach of pre-
senting inflation illusion in to the economy and let the agents, who are affected by
money illusion, use the nominal interest in their discounting by means of introducing
a different pricing factor.
There is another growing branch of the literature that seeks to establish the relation
between nominal assets and real assets. Piazzesi and Schneider (2008a) model the infla-
tion illusion in a two-period model to understand the effects of inflation on a particular
class of real assets, namely on housing market. They observe that if a portion of the
economy is under the effect of inflation illusion, i.e. if there is a disagreement about
the real interest rate between the investors with inflation illusion and the rest of the
economy, the smart investors to be exact, this will cause a boom in the housing market.
In addition, Piazzesi and Schneider (2008b) try to explain the opposite movements of
house and stock prices in the 1970’s. They document the differences between young
6
and old households in terms adjusting their inflation and by means of modelling it
through a general equilibrium model, they find out that changes in inflation make real
assets, housing in their case, more attractive due to capital gains taxes on stocks and
mortgage deductibility. They also claim that if agents expect higher inflation in the
future, they take this as a sign of negative returns for the stock market. Piazzesi and
Schneider (2008b) is also first in the literature to incorporate the Michigan Expected
Inflation Survey results for different cohorts in to their calibration.
Similar to the findings of this study, Titman and Warga (1989) finds a positive rela-
tion between inflation and stock returns for a fairly short sample of October 1979 to
October 1982 period.
I also like to call attention to the answer ex-chairman of Federal Reserve Bank Alan
Greenspan gave when he was asked about the increasing probability of a future reces-
sion period for the United States during a talk on C-NBC in 2007. He answered that the
probability of the country entering into a recession was more than it was thought and
continued with pointing out that the most important thing for the United States in the
coming years would be the fight against the possible inflation that might result from
the policies in the recession period. Shortly after that, during last two years, in order to
stabilize the effects of the crisis US government increased the money supply drastically
at the rate of 24.81% and 14.29% for M1 and M2 measures, respectively. On the other
hand, FED keeps the interest rates at a fairly low values, close to zero, and causes some
investors to have deflationary and others to have inflationary expectations.
7
To the contrary, Fama (1981) says that the relation between the stock returns and
the expected inflation is spurious but cannot explain the decline in expected real stock
returns in the period after 1953. Bodie (1976) states that stocks can be used as a hedge
against the inflation with certain diversification but finds out that the investors should
short sell the stocks to hedge themselves.
In addition, several empirical studies, such as Ritter and Warr (2002), Sharpe (2002),
Cohen, Polk, and Vuolteenaho (2005), and Chordia and Shivakumar (2005), provides
empirical evidence for the positive effect of inflation on dividend yield. These results
are interpreted as inflation’s impact on stock prices since dividend payments are quite
stable over time. Similar results are also shown by Brunnermeier and Julliard (2008)
for housing prices.
Finally, some other papers that study the relation between inflation and stock mar-
ket returns are Firth (1979), Cohn and Lessard (1981), Gordon (1983), Day (1984) and
Hasbrouck (1984).
Given all these studies, there seems to be a lack of interest in the literature to use this
relation in asset pricing framework. The aim of this study is to model this relation from
a standard asset pricing model view and investigate the implications of heterogenous
expected inflation among agents in terms of asset pricing.
8
1.3 Model
Akerlof and Yellen (1985) has showed that when agents move away from rational-
ity, with their losses being in second-order, if the effect of their deviation from ratio-
nality does not cancel each other on the average, then the movement away from the
equilibrium will be in first order. This shows us that divergence from rationality, or
some other type of heterogeneity, can cause divergence from equilibrium. We can see
similar occurrences in the stock market. In this section, initially the effects of inflation
on stock prices are presented in a two-period model set up. Secondly, implications of
having heterogeneous expectations about inflation is shown by means of embedding
differences in expected inflations of the agents in the economy in to the model. There
is only one real consumption good and households can trade real bonds and a stock
that pays real dividends with each other as well as the rest of the economy. Although a
share of the stock pays real dividends in terms of the consumption good in both peri-
ods, it is priced nominally in this economy. We can call this type of an asset a Lucas
tree with nominal share price.
Households have two measures of saving; a real bond and the nominal priced stock
that pays real dividends. It can be seen from Equation 1.3 that the one period ahead
cum dividend real gross return on the stock is not only dependent on the dividend
yield
, and the growth of the nominal stock price, the nominal capital return
, but also on the growth rate of the price level
, as well. Since the return
of the stock is affected by the changes in the price level of the economy, in other words
the inflation, a household should consider the expected inflation when making the
9
choice of stock holding. In addition, any heterogeneity in terms of inflation expecta-
tions, ceteris paribus, will cause differences in subjective valuation of both nominal and
real prices of stock
3
. Therefore, along with expected nominal stock price growth and
the dividend yield
4
, the inflation expectations affect the nominal value of a share of the
stock.
(1.3)
1.3.1 Two-Period Model
The two-period model used in this study has a different structure for the stock in
the economy and is a variation of Consumption Based Asset Pricing Model.
There are two dates, and
, and one consumption good in the model. Households
in this economy can only get utility from consuming the real consumption good and
the future utility from consumption is discounted by a factor of . They can buy or sell
one-period real bonds and shares of a stock. Bonds pay a net real return of , where
is the real interest rate from period to
, in other words bonds trade at a price of
during time period . A share of stock trades at the nominal price
and
3
Since the real price of the stock share is equal to nominal value of the stock price divided by the
price level of the economy, i.e. , if we keep the price level fixed for period or given , any
difference in subjective valuations of nominal stock price will bring different valuations in terms of
real stock price as well.
4
Dividend yield in the literature used in different ways from time to time. Some researchers used
and others used . In this case we are using the for the dividend yield. can also be
written as which gives us as the dividend yield.
10
pays a real dividend
during time period. It is also expected to pay nominal value
and real dividend
at time period
. Households enter the initial period with an
endowment of goods
, initial stock holdings
, which pays
consumption goods of
dividend and are sold for
, and an amount of goods
from previous bond market
activity.
Given these, to solve its maximization problem, a household chooses consumption
levels for period and
,
and
, the quantity of shares of stock to hold
5
, and the
amount of goods to be invested in real bonds
for given real interest rate and nom-
inal market price level
6
.
is determined endogenously in the model. During their
decision process households also consider expected nominal values for stock price and
market price level,
and
, respectively. Stock holdings also pay a dividend,
,
in terms of consumption good at time period
, which also affects the decision of the
household. Additionally, households are endowed with
units of consumption good
for period
and they are allowed to borrow up to
in the initial period. This
rules out the infinite leverage opportunity in case the expected return of the stock is
greater than that of the real bond. Then the maximization problem of the household
can be written as follows:
5
There is no short selling for the stock introduced in the economy for the time being and hence .
6
, or the price level in the first period, is in the information set of the houshold, i.e. they know the
price level before they make their decission about consumption and saving or borrowing.
11
E
,
subject to
(1.4)
As it can easily be inferred from the maximization problem in 1.4, stock hold-
ings enter the maximization problem of the household in a slightly different manner.
Although it pays real dividend proportional to the share holding of the household,
when it comes to trading stock shares, households should start thinking in nominal
values. Hence the level of expected inflation very much affects the valuation of the
stock. As it is mentioned earlier in the chapter, we can think these stocks as the Lucas
tree that pays the fruit of the tree as dividend but when it comes to trading the tree
itself that transaction is completed in nominal terms, i.e. the stock has a nominal price
and can only be bought by money.
The rest of the economy is a combination of the government, foreign and business
sectors that help to clear out the bond and goods market. It pays all the bonds with
which households enter to period and issues new bonds needed by them. In addition
to that, there is no initial public offering for the stock, i.e. there is no new tree planted
12
in the economy for the time being, hence the shares of the stock adds up to in both
periods. These conditions are summarized in 1.5.
(1.5)
The timing of the events are as follows. In the initial period, first the endowment
is provided to the households and they are informed about the price level, interest
rate and the dividend payment for that period. Then households make their choices
regarding consumption and asset holdings, bond and stock, with respect to their expec-
tations about the nominal stock growth, next period’s real dividend payment and infla-
tion. The price of the nominal stock is determined endogenously in the market and we
move to the next period.
First order conditions for the maximization problem of an household can be written
as in Equations 1.6 and 1.7. We have the stochastic discount factor equal to inverse of
and we can also write the real stock price in time period as the sum of the
discounted value of the dividend payment and the stock price from period
, where
is the stochastic discount factor.
13
E
(1.6)
E
E
E
E
(1.7)
E
E
E
E
(1.8)
First order condition for the share of stock is a little more complicated in a
general utility case and needs more attention. First we can use the equation
EEECov. Applying this property to Equation 1.8, we reach Equa-
tion 1.9, where we can write the real value of the stock price in period as the sum of
the expected real payoff from the stock in period
multiplied by the expected stochas-
tic discount factor plus the covariances of stochastic discount factor with dividend and
real stock price.
14
E
E
E
(1.9)
Cov
Cov
is provided to households at the beginning of time period before they make
their choices, i.e. price level is in the information set of the household during their
choice process. Therefore, it can be moved into the expectations and covariances with-
out any problem. Let’s multiply both sides of the first order equation with
and
define inflation from period to
as
. As it is seen from the first order
condition for stock holding below, inflation term shows up in the denominator. Using
the following approximation,
, we can move it to the
numerator but not without an overestimating second order error that increases with
inflation
7
. This approximation makes it a lot convenient to work with the inflation and
helps us see the relation between inflation and stock price
8
.
7
There is 1% overestimation for 10% inflation.
8
A better way of approximating the inflation in denominator is to use exponential. This approximation
method will be used in the future research.
15
E
E
E
Cov
Cov
E
E
Cov
Cov
E
E
E
E
Cov
Cov
Cov
Cov
E
E
E
Cov
Cov
Cov
We can use EEECov property one more time to bring inflation
term out of the expectation where it is multiplied by the period
stock price and write
the first order condition for the share of stock as the following.
E
E
E
E Cov
(1.10)
Cov
Cov
Cov
16
As it can be seen from Equation 1.10, the nominal stock price depends on the
expected dividend, stock price and inflation. It also depends on the covariance
between stock price and the inflation and the covariances of dividends, nominal stock
prices and the stock price times the inflation with the stochastic discount factor.
To simplify things even further, the risk neutral agent case will be examined for the
effects of expected inflation on the stock price.
1.3.2 Risk Neutral Agent
When the households are risk neutral, i.e. in its most general form,
then the marginal utility from the changes in consumption is constant and the sto-
chastic discount factor, , becomes a constant, , as well. In addition, because is
constant, all the covariances including is equal to zero, CovCov.
Also, it is assumed that Cov
9
. We can write the first order conditions of the
households under the risk neutral utility assumption as in Equations 1.11 and 1.12.
E E
E
(1.11)
9
Using the S&P500 and CPI data provided by Robert Shiller in his website,
http://www.econ.yale.edu/~shiller/data.htm , the covariance and correlation between nominal stock
prices and inflation for the 1950-2009 period are as follows; Cov
and Corr
.
17
E
E
E
Cov
Cov
Cov
Cov
E
E
E (1.12)
After making the proper cancellations, we reach to the initial period stock price that
depends on the expected future dividend, expected future nominal stock price and the
expected inflation in the case of risk neutral agent. It’s seen in Equation 1.12 that the
effect of the inflation shows up as a multiple of the expected value of the period
stock price. Suppose E
and E
are exogenously given and not effected by the
expected inflation of the household from initial period to the second period. In this
case the only thing that is endogenously determined in the model is the nominal stock
price in period. Then the derivative of the initial stock price with respect to expected
inflation is negative since both and E
should be greater than , and hence the
initial nominal stock price decreases with the increasing expected inflation at a rate of
E
.
18
E
E
E
E
E
Dividend yield can also be deduced from the first order condition of the risk neu-
tral agent for the stock holding. E
gives us the next period’s expected dividend
payment. Since this is a real value, we have to divide it by a real value to reach the div-
idend yield. So we divide it by the real value of the stock price in period zero,
,
which will give us the dividend yield
10
.
E
E
E
E
E
E
E
E
E
When we take the derivative of the dividend yield with respect to expected infla-
tion, we can see that it is positive and depends on the expected nominal gross capital
return on the stock.
E
E
E
10
As it was stated earlier in the paper, dividend yield in the literature used in different ways from time
to time. One way of presenting it is and in our two-period model case it is
.
19
1.3.3 HeterogeneousExpectedInflationfortheCaseofRiskNeutralAgents
in Two Period Model
Effects of having heterogeneous expected inflation in the economy will be inspected
in the simple setting of a two period model with risk neutral agents in this part of the
chapter. First, define
as the first period stock market price. Assume that there
are two agents in the economy and allow one of them to change his/her expectation
about the inflation and let the other one to keep it the same as the previous case making
him/her the rational agent. Denote the valuation of the latter agent as
and if there
is no heterogeneity,
. In other words,
is the stock market price in the
case of no heterogeneity. Also define
and
as the subjective stock price
valuation of the agent with lower and higher expectations for inflation compared to
the rational investor, respectively.
Suppose we have the expected nominal stock price and the real dividend payment
for the next period exogenously given. Then we can write the subjective first period
stock prices for both agents as follows.
E
E
E
E
E
E
for !"
The valuation of the agent keeping his/her expectations unchanged has the same
valuation we had in the previous case. On the other hand, the valuation of the stock for
20
the agent with changed expected inflation depends on how the inflations are formed.
Let’s assume here the agent exaggerates the inflation expectations, i.e. the agent
expects to have a higher inflation in a high and a lower inflation in a low inflation-
ary period. In mathematical notification, we assume the following for the expected
inflation heterogeneity, E
E in a low inflationary case and E E
in
a high inflationary case. Then we can easily infer that in a high inflationary period
the subjective valuation of the agent with higher expected inflation will be lower than
that of the agent who did not change his/her expectation. The agent with changed
expectations will see the stock overvalued and decide to sell it eventually pulling the
price of the stock down to a lower value. This will yield
and cause undervaluation of the stock in a high inflationary period as presented in the
model. For the low inflationary period case, the subjective valuation of the agent with
changed expectations is higher than that of the agent with unchanged expectations. In
this case, the resulting stock market price will have
and hence
the overvaluation of the stock in a low inflationary period is shown using the model.
In case of undervaluation due to a heterogeneity, result of a deviation form ratio-
nality, eventually the stock price will revert to its fundamental value. This will give us
a predicted positive return for the future in the equity market. Similarly a low infla-
tionary period causing overvaluation will gives us a predicted positive return for the
future in the stock market.
21
1.4 Data
In this section, data used for this study is introduced. I used data from five
different data sources namely, CRSP data set from WRDS at Wharton
11
, Robert J.
Shiller’s and Kenneth R. French’s data sets from their websites
12
, University of Michi-
gan Consumer Survey data
13
, and Federal Reserve Board data set
14
. Value and Equal
Weighted Returns of the S&P500 companies including and excluding dividend pay-
ments (VWTRETX, VWTRETD, EWTRETX and EWTRETD) and S&P500 index are
taken from CRSP data set. Real Dividend Payments (RDIV) and Real Stock Price (RS)
for S&P500 companies, Monthly Consumer Price Index (CPI) and the Long Term Inter-
est Rate (LTR) are taken from Robert J. Shiller’s data set used for Irrational Exuber-
ance. Risk Free rate (RF) is taken from Kenneth R. French’s data set for Fama French
factors. One of the expected inflation measures (IMF_M) that will be used in the empir-
ical study is directly downloaded from University of Michigan Consumer Survey data
for expected inflation. Finally, the 3 Month Treasury Bill rate (3MTB), Moody’s Aaa
and Baa interest rates are taken from Federal Reserve Board historical data sets. All
variables used in this study are monthly in frequency and the sample period for the
empirical work is from January 1950 to August 2008 except expected inflation data
downloaded from University of Michigan Consumer Survey data, for which sample
11
http://wrds.wharton.upenn.edu
12
www.econ.yale.edu/~shiller/data.htm
http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
13
http://www.src.isr.umich.edu/
14
http://www.federalreserve.gov/econresdata/default.htm
22
Table 1.1: Data Downloaded for Return Regressions
Variable Explanation Source
VWRETX S&P500 Value Weighted Return CRSP data set from WRDS
Excluding Dividend Payments
VWRETD S&P500 Value Weighted Return CRSP data set from WRDS
Including Dividend Payments
EWRETX S&P500 Equal Weighted Return CRSP data set from WRDS
Excluding Dividend Payments
EWRETD S&P500 Equal Weighted Return CRSP data set from WRDS
Including Dividend Payments
SP500 S&P500 Index CRSP data set from WRDS
RDIV Real Dividend Payment for S&P500 Robert Shiller’s website
RS Real Stock Price for S&P500 Robert Shiller’s website
CPI Monthly Consumer Price Index Robert Shiller’s website
RF Risk Free Rate Ken French’s website
LTR Long Term Interest Rate Robert Shiller’s website
INF_M Expected Inflation Univ. of Michigan C. S.
3MTB 3 Month Treasury Bill Rate Federal Reserve website
Aaa Moody’s Aaa interest rate Federal Reserve website
Baa Moody’s Baa interest rate Federal Reserve website
C. S. means Consumer Survey
period starts from January 1978. Table 1.1 gives the summary of variables, their expla-
nation and the source they are downloaded from.
Table 1.2 displays the variables used in and generated for the empirical study with
their explanation and how they are formed. I use VWTRETX, VWTRETD, EWTRETX
and EWTRETD directly from CRSP data set. I also generate a fifth return for S&P500
(SP500RET) using SP500 index, also from CRSP data set. All five returns are used to
compare the data set with Fama and Schwert (1977) results but only value weighted
and S&P500 returns are used in the return regression analysis. Real monthly dividend
payments for S&P500 companies (RDIV) can be found from the Robert J. Shiller’s web-
site. To be consistent, I used the Real Stock Price for S&P500 companies (RS) from the
23
-30
-20
-10
0
10
20
30
50 55 60 65 70 75 80 85 90 95 00 05 10
Figure 1.1: Annualized Monthly Inflation in Percentage
same website while calculating both Dividend Yield (DY) and 1 Year Moving Aver-
age Dividend Yield (DYMA). Inflation (INF) and the square of the inflation (INF
2
)
are generated using CPI but Expected Inflations (INF_E) are formed by a regression
model which is explained later in this section. Additionally, I use the Expected Infla-
tion from University of Michigan Consumer Survey data (INF_M) in my return regres-
sions. Although I use the RF as the right hand side variable, 3MTBR is used to construct
excess returns for the left hand side of the regressions. Term Spread (TS) is generated
as the difference between the long term and the short term interest rates. Moreover,
Default Spread (DS) is formed as the difference between Baa and Aaa interest rates.
24
Table 1.2: Variables Used in Return Regressions
Variable Explanation
VWRETX S&P500 Value Weighted Return
Excluding Dividend Payments
VWRETD S&P500 Value Weighted Return
Including Dividend Payments
EWRETX S&P500 Equal Weighted Return
Excluding Dividend Payments
EWRETD S&P500 Equal Weighted Return
Including Dividend Payments
SP500RET S&P500 Return SP500RET=
SP500
SP500(-1)
DY Dividend Yield DY
RDIV
RS
DYMA 1 Year Moving Avg. Dividend Yield DYMA
RDIV(-j)
RS
INF Monthly Inflation INF
CPI
CPI(-1)
INF
2
Square of Monthly Inflation
INF_E Expected Inflation from Inflation Model in Eq’n 1.13
INF_M Expected Inflation from Univ. of Michigan Consumer Survey
RF Risk Free Rate
3MTB 3 Month Treasury Bill Rate
TS Term Spread TSLTR3MTB
DS Default Spread DSBaaAaa
D Dummy variable for the inflationary period of 1972-1983
D
if 1973-1982
otherwise
I also used a dummy variable (D) for the period of 1973-1982 to see whether the
effect of the variables, especially the one expected inflation, on stock returns change
from high inflationary periods to low inflationary periods. The reason why 1973-1982
period is chosen for the high inflationary dummy can be seen easily from Figure 1.1 and
Table 1.3. Figure 1.1 shows that the inflation is relatively higher in the 70’s and early
80’s compared to the rest of the sample period. According to Descriptive Statistics for
Inflation in Table 1.3, annualized percentage rate of monthly inflation from January
1973 to December 1982 happens to be 8.351%, which is three to four times higher than
25
Table 1.3: Descriptive Statistics for Inflation
Variable 1950-1972 1973-1982 1983-2008 1950-2008
INF Mean 2.566 8.351 3.232 3.844
SD 3.624 4.513 3.313 4.201
INF
2
Mean 0.001 0.001 0.001 0.001
SD 0.030 0.067 0.024 0.043
INF_E Mean 2.772 7.379 3.461 3.865
SD 1.812 2.675 1.624 2.510
All the means and standard deviations are annualized and converted to percentage changes.
the two other subsample inflation rates, 2.566% and 3.232 %, for the January 1950 to
December 1972 and January 1983 to August 2008, respectively. The average inflation
rate for the whole sample is 3.844% and it’s way below the inflation rate of the 1973-
1982 period.
INF
INF
INF
INF
INF
(1.13)
3MTB
3MTB
3MTB
#
Mean and standard deviation for inflation expectations, generated from the model
in Equation 1.13, are also given in Table 1.3. Means of the expected inflation for all
subperiods and the whole sample are quite close to the actual inflation average, how-
ever standard deviations are considerably lower compared to the actual ones. The
model gives us a smoother expectation in contrast to higher volatility of the inflation.
Since the households do not have the inflation information for a given time period $,
a measure for inflation expectations is needed for the return regression. Geske and
Roll (1983) states that changing real interest rate will make using short term rate as a
26
proxy for the inflation expectation really risky, since small changes in real interest rate
can cause large percentage changes in stock prices in the opposite direction. That’s
why, to see the effects of inflation expectations on equity returns, I use Two-Stage Least
Squares (2SLS), which is a very common procedure in the literature. First, I regress the
monthly inflation to lagged monthly inflations from the first to the fourth lag. The lags
of the 3 month treasury bill rate, from the first lag up to the third, are also included in
the first stage regression
15
. The expected inflation obtained from this first stage regres-
sion is used in the return regression, which is the second stage regressions in this case.
Results of the first stage regression conducted to form expected inflation are presented
in Table 1.4.
Table 1.4 shows that lagged inflation affects inflation expectations in all lags pos-
itively and significantly except three months. Short term treasury bill rate affects
expected inflation positively for the first two lags but negatively for the third lag. All
three coefficients are significant for lagged short term interest rate. The fourth lag of
the short term interest rate is also used in finding expected inflation measure but it
is dropped later due to its low significance. Descriptive statistics for the rest of the
variables are provided in Table 1.5.
1.5 Empirical Study
In this section, I first replicate the empirical work by Fama and Schwert (1977) and
try to see whether the new data set from CRSP gives results similar to the results of
15
The fourth lag of the 3 month T-Bill rate was also included in the regression analysis for forming
expected inflation but it’s dropped due to the very low significance of the variable.
27
Table 1.4: Forming Inflation Expectations
Sample:1950-2008
Dep. Var. Monthly Inflation formed from Shiller’s CPI data
INF
INF(-1) 0.361***
(0.053)
INF(-2) 0.133**
(0.053)
INF(-3) -0.020
(0.040)
INF(-4) 0.112***
(0.036)
3MTB(-1) 0.422*
(0.217)
3MTB(-2) 0.609*
(0.369)
3MTB(-3) -0.803***
(0.248)
Constant 0.0004*
(0.0002)
R
2
0.358356
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
January 1953 to July 1971 sample period used in their study. Comparable results will
motivate to further the study on the effects of inflation on stock market. Later, the
results for my empirical study in a broader sample, 1950-2008, are presented.
Fama and Schwert (1977) estimated expected inflation through Fisher equation, i.e.
if the nominal return is known and the real return can be assumed as constant for an
asset, such as the three month treasury bill, then E
E
%
and hence
expected inflation can be written as E
%
. Regressing the following
model in Equation 1.14,
&
(1.14)
28
Table 1.5: Descriptive Statistics for the Rest of the Variables
Variable 1950-1972 1973-1982 1983-2008 1950-2008
VWRETX Mean 9.038 4.300 9.577 8.466
SD 42.707 61.399 50.849 49.843
VWRETD Mean 12.905 8.783 12.096 11.849
SD 42.881 61.489 51.002 49.973
SP500RET Mean 9.274 3.127 9.708 8.416
SD 42.658 57.965 50.204 48.858
DY Mean 0.039 0.046 0.025 0.034
SD 0.012 0.008 0.010 0.014
DYMA Mean 0.036 0.042 0.023 0.031
SD 0.011 0.008 0.009 0.012
RF Mean 3.090 8.155 4.908 4.748
SD 1.588 8.155 2.248 2.815
3MTB Mean 3.245 8.422 4.958 4.877
SD 1.601 3.164 2.237 2.842
TS Mean 0.962 0.943 1.836 1.340
SD 0.598 1.577 1.142 1.148
DS Mean 828.565 1623.100 1167.974 1112.489
SD 260.784 677.167 359.573 488.046
All the means and standard deviations, except for DY and DYMA, are annualized and converted
to percentage changes.
&
(1.15)
with the propositions of and E
&
%
, will show that the best measure
for expected inflation and unexpected inflation are
and
, respectively. The
return regression in Equation 1.15 above is run to compare Fama and Schwert (1977)
results for their value and equal weighted portfolio with the results from data set used
in this study. Results, which can be seen in Table 1.6, show that numbers from the
inflation and return regressions, with all five returns from CRSP data set, are not that
different from those of Fama and Schwert (1977). Before going in to detail with my
29
empirical study, first I regress the model in Equation 1.15 through out my whole sam-
ple with and without dummy variable for each of the five returns from CRSP data set.
Results in Table 1.7 show that not only the effect of short term interest rate, which is
the proxy for expected inflation, on the stock returns is negative and significant but
also the difference between short term interest rate and inflation, which is the proxy
for unexpected part of inflation, is negative and significant in all cases. The first result
of this chapter is that finding of Fama and Schwert (1977) for the negative effect of
expected inflation on stock returns persists over time and the negative effect of unex-
pected inflation becomes even more significant over a longer horizon. Although these
results are promising, their explanatory power is not that high and remains around 110
to 130 basis points. Motivated by these results, I continue with the empirical findings
for return regressions.
Aim of the study here is to provide an empirical evidence of the effects of inflation
on the stock market prices. In order to do this, returns are regressed against a con-
stant and standard return regression variables such as past dividend yield, risk free
rate, term spread, and default spread. I subsequently add one of the monthly inflation
measures among realized inflation (INF), one month lagged realized inflation (INF(-
1)), inflation expectation obtained from Equation 1.13 (INF_E) and monthly expected
inflation from Michigan Expected Inflation Survey (INF_M) into the model to assess
the effects of inflation and expected inflation on the stock market returns. I also want
to express the differences among the high and low inflationary periods, that’s why a
30
Table 1.6: Replicating Fama and Schwert 1977
Sample:January 1953to July 1971
Comparison of Inflation Regressions
INF
Fama and Schwert
3MTB 0.927*** 0.98***
(0.106) (0.10)
Constant -0.001** -0.001**
(0.0003) (0.0003)
Adj. R
2
0.212 0.29
Comparison of Return Regressions using Fama and Schwert (1977) Model
VWRETX FS S
EWRETX FS S
3MTB -5.546*** -5.52*** -5.475* -5.70***
. (1.848) (1.85) (2.935) (2.17)
INF-3MTB 0.068 -0.77 -1.732 -2.35
(0.923) (1.22) (1.351) (1.484)
Constant 0.023*** 0.023*** 0.023*** 0.024***
(0.005) (0.006) (0.007) (0.006)
Adj. R
2
0.032 0.03 0.021 0.03
VWRETD EWRETD SP500RET
3MTB -5.826*** -5.948** -5.641***
. (1.847) (2.915) (1.736)
INF-3MTB -0.096 -1.813 0.589
(0.923) (1.350) (0.914)
Constant 0.026*** 0.027*** 0.023***
(0.005) (0.007) (0.005)
Adj. R
2
0.036 0.026 0.035
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
dummy variable (D) is utilized for the 1972-1983 period
16
. This variable helps to distin-
guish different levels of marginal effects of dividend yield, inflation expectation, risk
free rate, term spread and default spread. I first present the return regression results
with monthly dividend yield (DY) and then with 1 year moving average dividend
yield (DYMA) from Table 1.8 to Table 1.13. The left hand side variable used for these
16
The average annualized monthly inflation for the 1972-1983 period is 8.35%. On the otherhand the
average annualized monthly inflation for our sample is 3.84% including the high inflationary period of
1972-1983.
31
Table 1.7: Fama and Schwert 1977 Model for 1950-2008
Sample:January 1950 to August 2008
Return Regressions using Fama and Schwert (1977) Model
VWRETX VWRETX VWRETD VWRETD
3MTB -1.678** -2.034** -1.589** -2.134**
. (0.719) (1.019) (0.724) (1.026)
D*3MTB 0.390 0.596
(0.889) (0.896)
INF-3MTB -1.387*** -1.465*** -1.377** -1.496***
(0.532) (0.558) (0.542) (0.566)
Constant 0.013*** 0.014*** 0.015*** 0.017***
(0.003) (0.004) (0.003) (0.004)
Adj. R
2
0.013 0.012 0.012 0.011
SP500RET SP500RET
3MTB -1.595** -1.637*
. (0.679) (0.984)
D*3MTB 0.046
(0.839)
INF-3MTB -1.336*** -1.345**
(0.513) (0.541)
Constant 0.012*** 0.012***
0.003) (0.003)
Adj. R
2
0.012 0.011
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
regressions is the excess return formed from value weighted return excluding divi-
dend payments (VWRETX) and short term interest rate or three month treasury bill
rate (3MTB). The most general model used for these regressions is given in Equation
1.16 where INF. M. can be realized inflation, lagged inflation and inflation expectations
from either Equation 1.13 or University of Michigan Consumer Survey data set.
Stock Return
3MTB
D+
DY
INF. M.
RF
TS
DS
D*DY
D*INF. M. (1.16)
D*RF
D*TS
D*DS
&
32
Table 1.8: Regression Results for VWRETX
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETX and 3MTB
VWRETX
3MTB
1 2 3456 7
DY(-1) 0.216* 0.277** 0.348*** 0.335*** 0.321*** 0.362*** 0.369
(0.115) (0.112) (0.119) (0.121) (0.119) (0.126) (0.246)
RF(-1) -1.777*** -0.687
(0.675) (0.751)
Constant -0.004 0.001 -0.0002 -0.002 -0.002 0.001 0.003
(0.004) (0.005) (0.005) (0.004) (0.004) (0.005) (0.006)
INF. M. INF INF INF(-1) INF_E INF_M
-1.888*** -2.069*** -1.761*** -3.162*** -3.466*
(0.525) (0.498) (0.435) (0.786) (1.792)
Adj. R
2
0.005 0.012 0.030 0.030 0.022 0.024 0.004
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 7: January 1978 to August 2008
In Table 1.8, results for simple regressions are given. First I regress excess return
only on a constant and dividend yield and then add the risk free rate. Inclusion of
the risk free rate increases the adjusted R
2
considerably. Afterwards, the realized infla-
tion is added to the regression to see whether inflation has any effects on the stock
returns. Not only the explanatory power of the regression has gone up by 180 basis
points but also the significance of the risk free rate disappears while the inflation has
a negative and highly significant coefficient. Dropping the risk free from the model
did not change the results for the regression in model 3
17
. Since realized inflation is
not observable, it cannot be used in return regressions. First, I used the 1 lagged infla-
tion as the expected inflation and then the expected inflation from Equation 1.13 and
17
Regression models from 5 to 7 are also run with risk free included. Since the results are not that
different, only the ones without risk free is reported in Table 1.8.
33
University of Michigan Consumer Survey data. In all cases inflation has a highly sig-
nificant negative effect with the expected inflation from Equation 1.13 providing the
highest adjusted R
2
at 0.024, among the inflation expectation measures. The nega-
tive effect of inflation is in line with the findings of previous research, such as Fama
and Schwert (1977), Modigliani and Cohn (1979), Campbell and Vuolteenaho (2004),
etc. Additionally, dividend yield has a positive and highly significant effect on stock
returns for every regression model, which is also in accordance with the literature. On
the other hand, expected inflation from University of Michigan Consumer Survey data
gives comparably low number for the adjusted R
2
and an insignificant dividend yield.
This may be due to the different sample period of University of Michigan Consumer
Survey data, which includes the last few years of the inflationary period in the US, or
to unreliability of the survey data itself.
Table 1.9 continues with the results for return regressions with the inclusion of term
and default spreads. Addition of term and default spreads increases the explanatory
power compared to the regression results in Table 1.8. Effect of inflation on stock
returns is always negative and highly significant. The coefficient of dividend yield
is significantly positive while the risk free affects stock returns negatively and signifi-
cantly except for model 13 where it is insignificant. Again the expected inflation from
Equation 1.13 provides the highest explanatory power among the inflation expectation
measures. Default spread is positive and significant except for model 8 and 14 where
it shows up insignificant. On the other hand, term spread is insignificant except for
34
Table 1.9: Return Regression Results for VWRETX cont.’d
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETX and 3MTB
VWRETX
3MTB
8 9 10 11 12 13 14
DY(-1) 0.232** 0.244** 0.323*** 0.311*** 0.301*** 0.321*** 1.495***
(0.110) (0.111) (0.119) (0.118) (0.116) (0.124) (0.559)
RF(-1) -2.824*** -1.984** -1.919** -2.017** -1.141 -6.924***
(0.993) (0.977) (0.985) (0.981) (1.134) (2.299)
TS(-1) 3.994** 0.809 -0.436 -0.083 0.286 -0.031 -9.485**
(1.806) (1.884) (1.995) (1.882) (1.888) (1.909) (4.495)
DS(-1) 0.0004 0.012* 0.013** 0.012** 0.012** 0.012** 0.004
(0.005) (0.006) (0.006) (0.006) (0.006) (0.006) (0.008)
Constant -0.010 -0.006 -0.006 -0.006 -0.006 -0.005 0.021**
(0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.011)
INF. M. INF INF INF(-1) INF_E INF_M
-1.389** -1.829*** -1.447*** -3.020*** -6.745**
(0.674) (0.522) (0.437) (1.024) (3.106)
INF
2
-54.497
(80.803)
Adj. R
2
0.009 0.020 0.036 0.037 0.030 0.029 0.022
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 14: January 1978 to August 2008
model 8 and 14 where it has different signs. I also controlled for the effect of infla-
tion volatility on stock returns by means of including the square of the inflation in the
model in regression 10 but it turns out that the inflation volatility is insignificant in
explaining excess returns and is not included for the rest of the regressions. Looking
at Tables 1.8 and 1.9, it can be seen that inflation and expected inflation affect equity
returns significantly in a negative way which provides evidence for the first part of the
theory in Section 1.3.
To see whether some kind of heterogeneity in inflation expectations can cause
movements in the stock prices and returns, I included the dummy variable (D), which
35
Table 1.10: Return Regression Results for VWRETX with Dummy
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETX and 3MTB
VWRETX
3MTB
15 16 17 18
DY(-1) 0.238** 0.280** 0.238** 0.281**
(0.118) (0.129) (0.119) (0.131)
INF_E -3.899*** -3.887***
(1.197) (1.207)
RF(-1) -1.920 -0.278 -2.264** -0.713
(1.259) (1.259) (1.151) (1.187)
TS(-1) -0.162 -0.030 -0.782 -0.325
(2.165) (2.069) (2.000) (1.991)
DS(-1) 0.009 0.009 0.013** 0.014***
(0.007) (0.007) (0.006) (0.005)
D*DY(-1) 1.673*** 1.336** 1.701*** 1.675***
(0.462) (0.569) (0.490) (0.467)
D*INF_E 8.549** 6.111**
(3.610) (2.784)
D*RF(-1) -4.274 -6.305** -2.938* -5.616***
(3.141) (3.073) (1.788) (2.054)
D*TS(-1) -2.253 3.273
(5.892) (7.519)
D*DS(-1) 0.011 0.013
(0.011) (0.011)
D -0.070*** -0.092*** -0.068*** -0.079***
(0.024) (0.025) (0.022) (0.018)
Constant -0.004 -0.001 -0.006 -0.004
(0.006) (0.025) (0.006) (0.006)
Adj. R
2
0.033 0.045 0.035 0.044
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
takes the value of 1 for the high inflationary period of the US 1973 to 1982 and 0 other-
wise, in to the regressions. Table 1.10 provides the results for return regression with the
inclusion of dummy variable. I first keep the inflation expectation out of the model and
provide the results for the return regression with dividend yield, risk free rate, term
and default spreads and with all these variables’ interaction terms with the dummy
variable. Results show that the level of the effect of dividend yield increase from low
36
to high inflationary periods significantly while the rest of the variables are insignifi-
cant. Including the inflation expectation from Equation 1.13 with its interaction term
with the dummy variable increases the adjusted R
2
from 0.035 to 0.042. Additionally,
the effect of inflation expectations is negative and significant at the level of -3.899 for
the low inflationary periods. On the other hand, its effect in high inflationary peri-
ods is significantly different by 8.549 which gives a positive effect of expected inflation
during high inflationary period at the level of 4.66. As a result we can see a signifi-
cant change from negative to positive in the effect of expected inflation on the stock
returns as the economy moves from low to high inflationary periods. Similar results
can also be observed when the term and default spreads insignificant interaction terms
are dropped from the model. These provide evidence for the second part of the the-
ory which was presented in Section 1.3.3. According to Section 1.3.3, heterogeneity
in expected inflation can cause undervaluation or overvaluation in the stock market
during high and low inflationary periods, respectively. In high inflationary periods,
investors who exaggerate the rate of inflation and expect higher than rational inflation
expectation can cause an undervaluation in the economy predicting positive returns in
the future and vice versa. Finding significant alternating positive and negative effects
of expected inflation from high to low inflationary periods in our predictive return
regressions supports the proposition of the theory Section 1.3.3.
Although the results are promising for the validity of the theory, there are some
drawbacks coming from the dividend data in Shiller’s website. Dividend data is not
available every month for the equity market since none of the firms pays dividend in
37
Table 1.11: Return Regression Results for VWRETX with DYMA
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETX and 3MTB
VWRETX
3MTB
1 2 3456 7
DYMA(-1) 0.226* 0.304** 0.374*** 0.353*** 0.344** 0.390*** 0.398
(0.129) (0.128) (0.136) (0.137) (0.135) (0.143) (0.268)
RF(-1) -1.815*** -0.763
(0.676) (0.749)
Constant -0.004 0.001 0.0002 -0.001 -0.002 0.001 0.004
(0.004) (0.004) (0.005) (0.004) (0.004) (0.005) (0.006)
INF. M. INF INF INF(-1) INF_E INF_M
-1.842*** -2.038*** -1.744*** -3.154*** -3.468*
(0.525) (0.501) (0.434) (0.786) (1.792)
Adj. R
2
0.004 0.011 0.029 0.029 0.022 0.024 0.004
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 7: January 1978 to August 2008
some months. For his monthly data, Shiller used a way of extrapolation to get the
values of dividend payments. To resolve this issue, I repeat the regressions in Tables
1.8, 1.9 and 1.10 with a new dividend yield calculated form 1 year moving average real
dividend payments and the real stock price. Results are shown in Tables from 1.11 to
1.13.
Tables 1.11 and 1.12 provide quite similar results to Tables 1.8 and 1.9, respec-
tively Expected inflation has a negative and significant effect for all cases increasing
the adjusted R
2
when added to the regression. Dividend yield is positive and signifi-
cant in most of the cases. Term spread is not showing significance in most of the cases
while default spread shows more positive and significant effect.
Table 1.13 also provides similar results to Table 1.10. The negative and significant
effect of expected inflation during low inflationary period changes to significantly pos-
itive during high inflationary period, supporting the theory of Section 1.3.3. Similar
38
Table 1.12: Return Regression Results for VWRETX with DYMA cont.’d
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETX and 3MTB
VWRETX
3MTB
8 9 10 11 12 13 14
DYMA(-1) 0.245** 0.263** 0.340** 0.328** 0.322** 0.342** 1.732***
(0.124) (0.126) (0.134) (0.133) (0.131) (0.138) (0.063)
RF(-1) -2.820*** -2.030** -1.965** -2.051** -1.195 -6.974***
(0.987) (0.965) (0.975) (0.971) (1.112) (2.349)
TS(-1) 4.006** 0.825 -0.414 -0.078 0.290 -0.014 -10.135**
(1.806) (1.883) (1.995) (1.880) (1.886) (1.907) (4.627)
DS(-1) 0.0002 0.012** 0.012** 0.012** 0.011** 0.012** 0.003
(0.005) (0.006) (0.006) (0.006) (0.006) (0.006) (0.008)
Constant -0.009 -0.006 -0.005 -0.005 -0.006 -0.004 0.024**
(0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.011)
INF. M. INF INF INF(-1) INF_E INF_M
-1.376** -1.795*** -1.429*** -2.960*** -7.258**
(0.675) (0.523) (0.435) (1.024) (3.192)
INF
2
-51.686
(81.480)
Adj. R
2
0.009 0.019 0.035 0.036 0.029 0.028 0.024
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 14: January 1978 to August 2008
to the results in Table 1.10, dividend yield’s positive and significant effect increases
significantly from low inflationary periods to high inflationary periods.
For robustness control, I regressed my models from 1 through 18 using value
weighted returns including dividend payments and S&P500 returns with both divi-
dend yield and 1 year moving average dividend yield. The negative effect of inflation
on average over the whole sample and its sign change from low to high inflationary
periods continue to show up. Another result that remains is the positive and significant
effect of dividend yield and 1 year moving average dividend yield on stock returns and
the significant increase in the level of this effect from low to high inflationary periods.
Tables including the results for all these regressions can be found in the appendix.
39
Table 1.13: Return Regression Results for VWRETX with Dummy and DYMA
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETX and 3MTB
VWRETX
3MTB
15 16 17 18
DYMA(-1) 0.250* 0.286** 0.250* 0.286**
(0.135) (0.144) (0.135) (0.145)
INF_E -3.770*** -3.776***
(1.191) (1.201)
RF(-1) -1.928 -0.389 -2.199* -0.723
(1.244) (1.123) (1.138) (1.164)
TS(-1) -0.165 -0.074 -0.666 -0.256
(2.166) (2.069) (2.007) (1.991)
DS(-1) 0.009 0.009 0.012** 0.013**
(0.007) (0.007) (0.005) (0.005)
D*DYMA(-1) 2.003*** 1.644** 2.051*** 2.018***
(0.506) (0.646) (0.534) (0.519)
D*INF_E 8.008** 5.809**
(3.633) (2.807)
D*RF(-1) -4.458 -6.310** -3.430* -5.931***
(3.188) (3.116) (1.808) (2.085)
D*TS(-1) -1.803 3.217
(5.964) (7.518)
D*DS(-1) 0.009 0.011
(0.011) (0.010)
D -0.073*** -0.094*** -0.073*** -0.083***
(0.023) (0.025) (0.021) (0.018)
Constant -0.004 -0.0003 -0.005 -0.003
(0.006) (0.006) (0.006) (0.006)
Adj. R
2
0.035 0.045 0.037 0.045
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
1.6 Conclusion
Using a standard two-period consumption based asset pricing model, this study
models the effects of inflation on the equity market via introducing nominal prices
for a share of stock. By means of introducing heterogeneity among agents in terms of
expected inflation, it also explains the undervaluation and overvaluation of the stock
40
market during high and low inflationary periods, respectively. Empirical results show
that the negative effect of the expected inflation, found by Fama and Schwert (1977),
is still existent. Additionally, the negative effect of the unexpected part of inflation
becomes significant over time. Using the new techniques for expected inflation, the
same negative effect of expected inflation on stock returns can still be found. Moreover,
the effect of the inflation switches from negative to positive when the economy moves
from low to high inflationary period, which supports the proposition of the model
for heterogeneous expected inflation causing overvaluation and undervaluation in the
market.
41
Chapter 2
Effects of Inflation in Predicting the
Stock Market Returns: International
Evidence
2.1 Introduction
In this part of the dissertation, I will provide international evidence of the previous
chapter’s findings. The theory of the effects of inflation on the equity market is the
same as the one shown in the previous chapter. What is expected from the international
data is to show negative effect of inflation and possible change in the level of this effect
from low to high inflationary periods. A sign change in the effect of the inflation is also
anticipated.
2.2 Literature Review
It is hard to find in the literature many studies on international empirical evidence
for effects of inflation on the stock market returns. We may add Branch (1974), Solnik
(1983) and Hess and Lee (1999) to the previous chapter’s literature review,. Checking
42
whether stocks are a good hedge against inflation in Germany, Japan, UK and the US
for a sample period of 1953 to 1969, Branch (1974) finds that stocks appear to be a
partial but not complete long-run inflation hedge. Solnik (1983) finds that real returns
are not independent of inflationary expectations for all the major stock markets of the
world. On the other hand, Hess and Lee (1999) shows that depending on the relative
importance of the supply and demand shocks the relation between stock returns and
inflation differs over time and across countries.
The aim of this part of the dissertation is to provide the effect of inflation on inter-
national stock market returns and corroborate the findings of the previous chapter.
2.3 Data
The entire data set used in this chapter is taken from the website of Global Finan-
cial Data (GFD)
1
. Monthly Stock Returns (SR) are calculated from the monthly Stock
Exchange Indices (SEI) data. I used TSX300 for Canada, SBF250 for France, CDAX for
Germany, BIC for Italy, Tokyo SE for Japan, FTSE All for the UK and S&P500 for the
US. The stock indices chosen for this study is selected to give the longest time horizon
and to access available dividend yield data in GFD for the corresponding index used to
find stock return. Dividend yield is given directly in the GFD set. For Canada, Japan,
UK and US, the dividend yield is available for the stock indices I used to calculate the
stock return but for the remaining countries I used the data that is available to give
the longest sample period. The excess return is calculated as the difference between
1
www.globalfinancialdata.com
43
the stock return and the short term interest rate which is taken as the 3 Month Trea-
sury Bill rate (3MTB). Three month treasury rate is also used as the risk free rate for
the countries. Term Spread (TS) is the difference between the Government Bond for 10
Year (GB10Y) and 3MTB. Default Spread (DS) is calculated from the difference between
GB10Y and Corporate Bond Yield (CB) except for the US where I used the Aaa-Baa of
the previous chapter as the default spread
2
. Inflation is calculated from the Consumer
Price Index (CPI) data as the net growth rate of the CPI. As Inflation Measure (INF. M.),
I used realized inflation (INF), lagged inflation (INF(-1)) and the Inflation Expectation
(INF_E) formed from the model of previous chapter which is also given in Equation
2.1 for any country . Results for the first stage inflation regressions are given in Table
2.1. Sample periods are taken to be mostly from 1950 to 2008 but for some countries,
such as Germany, it is almost 10 years shorter due to the unavailability of data. Similar
results to the those of previous chapter are found for the US with short term interest
rate losing its explanatory power for the first three lags but gaining it for the forth. This
may be the result of missing CPI data in the GFD set for the US or different CPI calcula-
tions, since two series are not quite the same when compared. In order to be consistent,
I continue with the results from Table 2.1 to calculate the inflation expectations for this
chapter. Looking at Table 2.1, it can be observed that generally the first two lags of
inflation effect inflation expectations positively. Moreover, some lags of the 3MTB are
also important for inflation expectations. On the other hand, the explanatory power
for Germany and Japan are fairly low compared to the results for other countries.
2
Corporate bond yileds for differnet rated companies is not available for any country other than the
US.
44
Table 2.1: Forming Inflation Expectations for Int.’l Data Set
Sample:1950-2008
Dep. Var. Monthly Inflation
INF CAN FRA GER ITA JAP UK US
INF(-1) 0.138*** 0.562*** 0.164*** 0.254*** 0.211*** 0.239*** 0.350***
(0.047) (0.101) (0.046) (0.064) (0.041) (0.054) (0.038)
INF(-2) 0.161*** -0.165 0.085* 0.210*** -0.188*** 0.106*** 0.116***
(0.042) (0.169) (0.050) (0.054) (0.042) (0.031) (0.040)
INF(-3) 0.091** 0.104 -0.023 0.085 0.129*** 0.036 -0.021
(0.037) (0.068) (0.044) (0.053) (0.042) (0.047) (0.040)
INF(-4) 0.128*** 0.069 -0.019 0.106** -0.007 0.099** 0.107***
(0.045) (0.065) (0.037) (0.049) (0.042) (0.040) (0.038)
3MTB(-1) 0.311 0.806** 0.408 0.468 3.789** 0.710 0.642**
(0.356) (0.331) (0.400) (0.434) (1.678) (0.531) (0.278)
3MTB(-2) 0.256 -0.243 0.129 0.296 -2.304 -0.240 -0.059
(0.497) (0.541) (0.526) (0.371) (2.426) (0.872) (0.415)
3MTB(-3) -0.511 -0.812 -1.119** -0.832* 1.793 0.550 0.335
(0.434) (0.698) (0.514) (0.485) (2.424) (0.607) (0.415)
3MTB(-4) 0.170 -0.482 0.978** 0.286 -2.381 -0.650 -0.648
(0.170) (0.420) (0.405) (0.310) (1.671) (0.497) (0.277)
Constant 0.0004 0.0005 0.0003 0.0003 -0.0003 0.0002 0.0003
(0.0003) (0.0004) (0.0003) (0.0002) (0.0005) (0.0005) (0.0002)
R
2
0.220 0.360 0.093 0.457 0.171 0.214 0.355
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
INF
(2.1)
INF
INF
INF
INF
3MTB
3MTB
3MTB
3MTB
#
Figure 2.1 displays how inflation changes among countries. In this figure, the annu-
alized monthly inflation in percentage minus 5% data is presented for all G7 countries.
45
-40
-20
0
20
40
60
50 55 60 65 70 75 80 85 90 95 00 05 10
CAN
-40
-20
0
20
40
60
50 55 60 65 70 75 80 85 90 95 00 05 10
FRA
-40
-20
0
20
40
60
50 55 60 65 70 75 80 85 90 95 00 05 10
GER
-40
-20
0
20
40
60
50 55 60 65 70 75 80 85 90 95 00 05 10
ITA
-40
-20
0
20
40
60
50 55 60 65 70 75 80 85 90 95 00 05 10
JAP
-40
-20
0
20
40
60
50 55 60 65 70 75 80 85 90 95 00 05 10
UK
-40
-20
0
20
40
60
50 55 60 65 70 75 80 85 90 95 00 05 10
US
Figure 2.1: Annualized Monthly Inflation in Percentage-5%
This figure helps us easily observe the periods where countries have inflation higher
than 5% annually. Looking at Figure 2.1 it’s seen that for Canada, France, UK and US
there is a higher inflation period between early 70’s and early 80’s. On the other hand,
Italy has a longer inflationary period from early 70’s to late 80’s while Germany and
Japan have quite low but volatile inflations
3
. Since a dummy variable has been defined
for the US in the previous chapter as taking the value of 1 for the period between 1972
3
Country by country infaltion figures can be found in the appendix for chapter 2.
46
and 1983 and 0 otherwise, and the inflation data of the Canada, France and UK is sim-
ilar to that of US, I use the same dummy variable in the empirical part of this chapter.
First, I firstly present the empirical results for all G7 countries and then for Canada,
France, UK and US.
2.4 Empirical Results
In this section, regression results for the G7 countries are presented. The model in
Equation 2.2 is the most general model for the empirical study of this part. For the
time being it’s assumed that D
D for every country, where D is the 1972 to 1983
dummy variable. Results for the pooled country fixed effect regressions including all
G7 countries are presented in Tables 2.2, 2.3 and 2.4.
SR
3MTB
D
+
DY
INF. M.
RF
(2.2)
TS
DS
D
*DY
D
*INF. M.
D
*RF
D
*TS
D
*DS
&
Table 2.2 shows that the inclusion of the short term interest rate is significant but
not with the expected inflation acquired via Equation 2.1. Dividend yield’s effect is
positive and somewhat significant. Effect of realized inflation is not significant but
47
Table 2.2: Fixed Effect Return Regression Results for XSR
Dep. Var. Excess Return formed from GFD data and 3MTB
Countries: CAN, FRA, GER, ITA, JAP, UK, US
Balanced Sample: February 1972 to May 2006
XSR 1 2 3 4 5 6 7
DY(-1) 0.001 0.002 0.002* 0.002 0.002 0.002* 0.003*
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
3MTB(-1) -1.101** -0.981* -0.267
(0.564) (0.580) (0.754)
Constant -0.003 -0.0004 -0.001 -0.003 -0.003 0.053*** 0.046*
(0.005) (0.005) (0.005) (0.005) (0.005) (0.021) (0.026)
INF. M. INF INF INF(-1) INF_E INF_E
-0.202 -0.367 -0.736*** -1.495*** -1.293*
(0.255) (0.251 (0.220) (0.512) (0.666)
Adj. R
2
0.001 0.002 0.002 0.001 0.004 0.004 0.004
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
that of expected inflation is highly significant and negative. Although the inclusion of
inflation increases the explanatory power, overall adjusted R
2
’s are relatively low.
While the term and default spreads are introduced to the model in Table 2.3, again
inflation expectations are negative and significant but realized inflation is not signifi-
cant at all. Although dividend yield is significantly positive, term and default spreads
are not significant except for model 8 where term spread is positively significant.
The short term interest rate is negative and significant but loses its significance once
the expected inflation from the first stage regression is introduced in to the model.
Adjusted R
2
are higher compared to the ones in Table 2.2 but not high enough to be
conclusive.
In Table 2.4, results for the regressions with dummy variable are presented. Divi-
dend yield is significant and positive but there is no significant change at the level of
48
Table 2.3: Fixed Effect Return Regression Results for XSR cont.’d
Dep. Var. Excess Return formed from GFD data and 3MTB
Countries: CAN, FRA, GER, ITA, JAP, UK, US
Balanced Sample: February 1972 to May 2006
XSR 8 9 10 11 12
DY(-1) 0.001 0.003* 0.003* 0.004** 0.004**
(0.002) (0.002) (0.003) (0.002) (0.002)
3MTB(-1) -1.772** -1.601* -1.307** -0.900
(0.903) (0.940) (0.944) (1.131)
TS(-1) 2.686*** 0.623 0.720 0.903 -0.798
(2.686) (1.403) (1.418) (1.419) (1.422)
DS(-1) -1.022 -1.341 -1.453 -1.714 -1.654
(3.922) (3.921) (3.912) (3.923) (3.897)
Constant -0.006 -0.0002 -0.001 -0.738 0.046
(0.007) (0.007) (0.007) (0.295) (0.030)
INF. M. INF INF(-1) INF_E
-0.282 -0.738** -1.297*
(0.310) (0.295) (0.787)
Adj. R
2
0.001 0.006 0.006 0.009 0.007
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
its effect from low to high inflationary periods. The rest of the variables are all signifi-
cant including the inflation expectation both for high and low inflationary periods. The
only variable that is significant other than the dividend yield is the dummy variable
for models 13 and 14. Adjusted R
2
increased with the dummy interaction terms but
they are not that high for a promising results.
Looking at the results for the pooled country fixed effect regressions for all G7 coun-
tries, expected inflation shows some negative effect on the stock returns but this effect
is not strong. This can be resulting from the three countries that have quite different
inflation processes, namely Germany and Japan with their low but volatile inflation
and Italy with her prolonged high inflationary period. Excluding these three countries
49
Table 2.4: Fixed Effect Return Regression Results for XSR with Dummy
Dep. Var. Excess Return formed from GFD data and 3MTB
Countries: CAN, FRA, GER, ITA, JAP, UK, US
Balanced Sample: February 1972 to May 2006
XSR 13 14 15 16
DY(-1) 0.008*** 0.007*** 0.007*** 0.007***
(0.002) (0.002) (0.002) (0.002)
INF_E -0.350 -0.157
(0.828) (0.844)
3MTB(-1) -1.500 -1.680 -1.045 -1.098
(1.113) (1.239) (1.016) (1.145)
TS(-1) -0.785 -0.841 1.127 1.183
(2.098) (2.118) (1.459) (1.483)
DS(-1) 1.153 1.016 -0.551 -0.575
(3.562) (3.535) (3.705) (3.618)
D*DY(-1) -0.002 -0.001 -0.0002 -0.0002
(0.002) (0.003) (0.003) (0.003)
D*INF_E 0.009 0.011
(0.020) (0.018)
D*3MTB(-1) 0.800 0.743 -0.557 -0.620
(0.003) (1.622) (1.174) (1.227)
D*TS(-1) 4.557 4.778
(3.116) (3.083)
D*DS(-1) -2.956 -2.384
(4.468) (4.759)
D -0.022* -0.023* -0.011 -0.012
(0.013) (0.013) (0.009) (0.010)
Constant -0.007 -0.020 -0.010 -0.016
(0.008) (0.033) (0.008) (0.033)
Adj. R
2
0.015 0.014 0.014 0.013
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
from the regressions and regressing the same models for Canada, France, UK and US
pooled sample gave the following results presented in Table 2.5, 2.6 and 2.7.
Results in Table 2.5 shows that the expected inflation effects equity market nega-
tively. This effect is negative and appears for only inflation expectation measures, such
as lagged inflation and inflation expectations from the first stage inflation regression.
Adjusted R
2
’s are quite high compared to the previous regressions except for model
50
Table 2.5: Fixed Effect Return Regression Results for XSR
Sample:1950-2008
Dep. Var. Excess Return formed from GFD data and 3MTB
Countries: CAN, FRA, UK, US
XSR 1 2 3 4 5 6 7
DY(-1) 0.002* 0.004** 0.004** 0.003* 0.003** 0.004*** 0.004***
(0.001) (0.002) (0.001) (0.001) (0.001) (0.002) (0.002)
3MTB(-1) -1.749*** -1.743*** -1.145*
(0.532) (0..544) (0.653)
Constant -0.007 -0.003 -0.003 -0.008 -0.008 0.120*** 0.074*
(0.006) (0.006) (0.006) (0.006) (0.006) (0.037) (0.042)
INF. M. INF INF INF(-1) INF_E INF_E
-0.013 -0.252 -0.716*** -1.957*** -1.209*
(0.263) (0.264) (0.258) (0.557) (0.647)
Adj. R
2
0.003 0.013 0.013 0.003 0.010 0.015 0.017
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
1. Inflation expectation generated from Equation 2.1 brings the highest adjusted R
2
s.
Dividend yield’s effect on equity returns is positive and highly significant. Short term
rate also shows significance in explaining stock returns and its effect is negative as
expected.
Results for the models with term and default spreads are given in Table 2.6. Includ-
ing term and default spreads in the models do not change the findings presented
in Table 2.5 significantly. Inflation expectations have negative and highly significant
effect on stock returns and the results from the first stage regression yields the highest
adjusted R
2
. Realized inflation does not have any explanatory power for the interna-
tional data set either. Dividend yield affects stock returns positively and significantly.
On the other hand, the short term interest rate has a significantly negative effect on
equity returns, except for model 12 where it is insignificant just like the previous cases.
51
Table 2.6: Fixed Effect Return Regression Results for XSR cont.’d
Sample:1950-2008
Dep. Var. Excess Return formed from GFD data and 3MTB
Countries: CAN, FRA, UK, US
XSR 8 9 10 11 12
DY(-1) 0.002 0.005** 0.006** 0.006** 0.007***
(0.941) (0.003) (0.003) (0.003) (0.003)
3MTB(-1) -2.897** -2.689** -2.414** -1.385
(1.130) (1.181) (1.156) (1.298)
TS(-1) 2.273** -1.168 -1.088 -0.931 -0.998
(1.148) (1.585) (1.600) (1.593) (1.603)
DS(-1) -3.536 -4.910 -5.317 -5.930 -5.947
(3.642) (3.765) (3.689) (3.765) (3.757)
Constant -0.010 -0.005 -0.006 -0.008 0.171***
(0.009) (0.008) (0.007) (0.008) (0.056)
INF. M. INF INF(-1) INF_E
-0.440 -0.972*** -2.802***
(0.351) (0.345) (0.868)
Adj. R
2
0.007 0.014 0.015 0.020 0.022
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Table 2.7 not only shows negative and significant effect of inflation in predicting
future returns but also confirms the positive significant difference between low and
high inflationary periods for coefficient of expected inflation both in model 14 and 16.
Although there exists a significant difference, it is not enough to change the sign of the
coefficient of expected inflation from negative to positive. Effect of the dividend yield
is positive and significant without changing with the dummy. Another observation
from Table 2.7 is that adjusted R
2
s are fairly high compared to the those in previous
tables.
52
Table 2.7: Fixed Effect Return Regression Results for XSR with Dummy
Sample:1950-2008
Dep. Var. Excess Return formed from GFD data and 3MTB
Countries: CAN, FRA, UK, US
XSR 13 14 15 16
DY(-1) 0.005** 0.005** 0.005** 0.005**
(0.002) (0.002) (0.002) (0.002)
D -0.034* -0.041* -0.038* -0.047**
(0.021) (0.002) (0.021) (0.025)
INF_E -1.984** -1.776**
(0.868) (0.866)
3MTB(-1) 1-.292 -0.282 -1.645 -0.758
(1.166) (1.297) (1.130) (1.243)
TS(-1) -0.142 0.199 -1.138 -0.929
(1.868) (1.892) (1.757) (1.729)
DS(-1) -0.762 -1.559 -3.727 -4.196
(2.691) (2.666) (3.255) (3.162)
D*DY(-1) 0.010 0.012 0.009 0.011
(0.007) (0.007) (0.006) (0.007)
D*INF_E 0.056** 0.063**
(0.025) (0.028)
D*3MTB(-1) -4.138 -4.595 -2.400 -2.637
(3.257) (3.317) (1.873) (1.973)
D*TS(-1) -4.545 -5.236
(5.060) (5.002)
D*DS(-1) -9.840 -9.255
(8.914) (8.601)
Constant -0.007 0.121** -0.008 0.107*
(0.008) (0.057) (0.007) (0.057)
Adj. R
2
0.030 0.035 0.029 0.034
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
2.5 Conclusion
Results show that for countries with similar inflation processes the negative effect
of expected inflation on stock returns can be proven not for the entire set of G7 coun-
tries but a subsample of it, Canada, France, UK and US. However, the alternating
53
nature of the effect of inflation on equity returns from low to high inflationary peri-
ods, which was captured for the US data in the previous chapter, could not be found
even for the subsample of G7 countries. Although there is a significant difference in
the level of the effect of inflation, it is not enough to change the sign of the coefficient.
Using different period dummy variables for different countries can increase the sig-
nificance of the results and may even bring the desired result of alternating effect of
inflation on equity returns.
54
Chapter 3
Boom-Bust Cycles in Turkey:
Capital Market Imperfections and
Asymmetries in Sector Based
Investment
3.1 Introduction
The financing of investment has important implications for the business cycle and
economic growth of the countries
1
. Even though there is ample research on aggregate
investment and business cycles, there exists limited micro evidence to prove the pos-
sible effects of financing decisions of the firms on aggregate output dynamics. Lack
of detailed micro data is more evident, especially for developing countries, which
impedes the proper analysis of the connections between financial effects at the micro
and aggregate levels. In this paper, we construct two micro level data sets for Turkey
1
This chapter of my dissertation is co-authored with Mustafa Kılınç, Ph.D. in Economics, who works
as an economist in the Central Bank of the Republic of Turkey. Most of the work has been completed
during my visits to the Central Bank of the Republic of Turkey in 2008 and 2009.
55
to see the financing conditions at a more disaggregate level and demonstrate the con-
nection of these findings with the aggregate output dynamics.
Turkey has experienced highly volatile business cycles after the financial liberaliza-
tion of the economy in 1989, similar to the experience of other developing countries.
These macro dynamics mostly resemble boom-bust cycles where expansionary periods
are followed by sharp contractions. We show that there are asymmetries among sectors
over the business cycles, and link these asymmetries to the structure of capital markets
in Turkey. Then we present micro level evidence to support the proposed relationship
between credit markets and sector based asymmetries.
The paper first documents the boom-bust cycles experience of Turkey after finan-
cial liberalization. The sector based asymmetries in the business cycles are revealed
afterwards, i.e. non-tradable sector is more volatile than tradable sector over the busi-
ness cycle. Subsequently, it is shown that there are strong correlations between the
asymmetrical response of the sectors and aggregate credit movements. The second
part of the paper uses two data sets, a detailed data set at sector based level and a firm
level data set of stock market , to look at the financial constraints in the economy. This
data set is assumed to be representative for the whole economy, while stock market
data is used as a control group. We find that non-tradable sector is financially more
constrained for sector level data set whereas in the stock market data we could not
find any difference between sectors in terms of financial constraints.
56
With non-tradable sector being more constrained in terms of access to the finan-
cial markets, credit market developments become an important determinant of boom-
bust cycles and the asymmetrical response of different sectors over the business cycle.
Therefore, with an inflow of credit into the economy non-tradable firms will bene-
fit more from availability of financing and increase their investment and output at a
faster rate than tradable firms. When faced with a drying of the credit during the bust,
however, non-tradable will be affected more disproportionately leading to asymme-
tries across sectors. As a result, we can establish that the asymmetry in the financial
constraints of the different sectors at the micro level can generate the observed asym-
metrical aggregate response of sectors over the business cycle.
Structural factors, financial and real frictions and business cycle properties are dra-
matically different in developed and developing countries. Analysis of structural fac-
tors and frictions in developing countries constitute a sizable empirical and theoretical
research area in economics.
2
Identification of these structural factors are instrumental
in analyzing their implications on aggregate dynamics of developing economies. In
a detailed study, Tornell and Westermann (2005) documents the stylized facts about
business cycles in middle income developing countries, and they name these cycles as
boom-bust cycles due to the common pattern of up and down swings. Two important
results stem from their analysis. First, there is significant sector based asymmetry over
2
For example, Neumeyer and Perri (2005) put working capital channel and risk premium variations as
important sources of fluctuations in Argentina; Augiar and Gopinath (2007) considers structural factors
that produce variations in long-term growth trends as an important part of business cycles for Mexico;
and Garcia-Cicco, Pancrazi and Uribe (2010) using more than a century of data for Mexico and Argentina
find that financial frictions provide a good account of business cycles in developing countries. Ranciere,
Tornell, and Westermann (2008) analyze a large set of developing countries and find that countries with
a moderate level of contract enforceability experience a higher mean growth but also greater incidence of
crises.
57
the business cycle, namely non-tradable sector output grows faster during the boom
and it falls harder during the bust. Second, the sector based asymmetry seems to have
a strong correlation with the credit movements in the economy. This relationship can
be linked to the fact that non-tradable firms are more constrained than tradable firms.
Although Tornell and Westermann (2005) provide some indirect evidence about
these financial constraints in the non-tradable sector, the scarcity of micro level data
and the absence of detailed sectors in these data sets for developing countries prove
that a more detailed micro level analysis remains to be an important step to motivate
connections between financial conditions and aggregate dynamics. This point is also
emphasized in Hubbard (1988), where author reviews the literature on capital market
imperfections and investment. Most of the studies on financial constraints and market
imperfections analyze limited data either from public firms or companies in a small
set of sectors such as manufacturing. Hubbard (1988) mentions that a multi sector
based analysis would provide helpful for micro examination of aggregate investment
and output dynamics. In this paper, we will try to establish a connection between
these two research areas by constructing a wide micro level data set to see the financial
constraints and also be looking at the macro level implications of these credit market
imperfections.
Section 3.2 gives brief summary of the literature, and section 3.3 demonstrates
the aggregate boom-bust cycles and asymmetries across sectors and connects these
to aggregate credit movements. Then section 3.4 shows the differences in the credit
58
constraints between tradable and non-tradable sectors followed by the conclusion in
section 3.5.
3.2 Literature Review
Neoclassical theory of investment, as put in detail by Modigliani and Miller (1958),
posits that under perfect capital markets a firm’s investment decision is related only
to the future expected profit opportunities and the user cost of capital. Furthermore,
this decision is independent of the financial structure of the firm. In other words, since
with complete markets, external and internal financing are perfect substitutes, financial
conditions are irrelevant in a firm’s investment decision. However, if capital markets
are not functioning perfectly, such as due to information asymmetries, there will be
a premium for external funds over internal funds. This difference arises due to pos-
sible moral hazard and adverse selection problems between lenders and borrowers.
Stiglitz and Weiss (1984) and Myers and Majluf (1984) present models where adverse
selection, under imperfect information regarding the project returns of the borrowers,
leads to external finance premium. Jensen and Meckling (1976) present a model with
moral hazard under costly monitoring and incentive problems and show that lenders
require a premium for compensation of moral hazard. Under such structures, invest-
ment decision and financial conditions of the firms are related so that investment will
be positively correlated with the changes in the internal funds.
In a seminal study, Fazzari, Hubbard, and Petersen (1998) tested the effects of finan-
cial conditions of firms on the investment decision using a Tobin’s Q framework. They
59
classify a sample of US manufacturing firms a priori according to dividend payout
ratios (a proxy for financial constraints) and show that low dividend payout firms
have higher investment responsiveness for internal funds than high dividend payout
firms. A large literature has followed this seminal study to analyze the financial con-
straints and investment decisions.
3
Tobin’s Q approach was the first empirical way
to test financial frictions. However, the problems with the measurement of Tobin’s Q
and also the sample of firms that are not traded in stock markets (so that one cannot
estimate Tobin’s Q) have led to other empirical approaches. Gilchrist and Himmel-
berg (1995) generate a ‘Fundamental Q’ and show that cash flow is still relevant for
financially constrained firms. Contrary to the main findings in the literature, Kaplan
and Zingales (1997), using Tobin’s Q framework, claims that cash flow is more rel-
evant for the least constrained firms. Cleary (1999) and Kadapakkam, Kumar, and
Riddick (1998) find similar results for US and other developed countries where least
constrained firms have the lowest sensitivity to the cash flow. However, Allayannis
and Mozumdar (2004) show that these results are conditional on influential observa-
tions and negative cash flows. Once these are excluded from the estimation, cash flows
become more correlated with investment for constrained firms.
Another approach is the accelerator approach, where instead of Tobin’s Q one uses
sales or change in sales as a variable to control for investment and profit opportuni-
ties. In these regressions, significant coefficient on internal funds is also interpreted as
evidence of financial constraints. One other approach is the Euler equation approach
3
Following papers provide a detailed review of the literature: Chrinko (1993), Schiantarelli (1996),
Hubbard (1988), Mairesse,Hall and Mulkay (1999) and Bond and Reenen (2007).
60
where one derives a structural equation from the optimization problem of the firm
without needing Tobin’s Q. Whited (1992), Hubbard, Kashyap, and Whited (1995) and
Bond and Meghir (1994) employ this approach and reject the perfect-capital markets
model. Coefficients of this equation have structural interpretations so that the equa-
tion controls for the expectations and the investment opportunities in a proper way.
Using the different methods listed above, there have been several studies analyzing
the investment behavior of firms in different countries.
4
Aforementioned financial market imperfections have macro implications as well.
Carlstrom and Fuerst (1997) and Bernanke, Gertler, and Gilchrist (1999) study closed
economy general equilibrium model with financial frictions; and Gertler, Gilchrist, and
Natalucci (2007) studies the same mechanism in a small open economy framework. In
these models, financial frictions amplify the macro effects of small shocks, so these
models are called “financial accelerator” models. Schneider and Tornell (2004) uses a
framework where non-tradable sector faces credit constraints and shows that boom-
bust cycles arise in aggregate economy due to movements in credit availability. As
these models illustrate, financial frictions that might lead to external finance premiums
or credit constraints can have important aggregate results.
Moreover, there have been empirical studies looking at the relationship between
financing constraints and selected micro and aggregate variables. However, most of
4
Among these are Hoshi, Kashyap and Scharfstein (1991) for Japan, Shin and Park (1999) for Korea,
Gelos and Werner (2002) for Mexico, Athey and Laumas (1994) for India, Bond et al. (2003) for Belgium,
France, Germany and the UK, Calomiris and Hubbard (1995) for US, Carpenter and Guariglia (2008) for
UK, Poncet, Steingress and Vandenbussche (2009) for China, Jaramillo, Schiantarelli, and Weiss (1996) for
Ecuador, Becker and Sivadasan (2006) for Europe, Gilchrist and Himmelberg (1995) for US, Carpenter and
Guariglia (2008) for UK, Harhoff (1998) for Germany, Galindo and Schiantarelli (2002) for Latin America,
Svejnar, Talavera and Gorodnichenko (2009) for Germany, and Schaller (1993) for Canada.
61
the studies in the investment-cash flow literature use either stock market firms or lim-
ited surveys of firms in some sub-sectors of the aggregate economy such as manufac-
turing.
5
Publicly traded firms constitute a very biased sample of the economy, partic-
ularly for developing countries. These firms are usually very large with conglomerate
and bank connections and some with foreign ownership. Therefore it would be mis-
leading to arrive at aggregate conclusions for developing countries merely from the
analysis of publicly traded firms. Limited surveys covering few sectors can also pro-
duce very valuable information, but again it will be difficult to infer about the aggre-
gate stance of the investment dynamics in these countries. Few papers connect the
financial constraints to aggregate investment cycles. Brown, Fazzari, and Petersen
(2009) show that for the US, financial factors for young high-tech firms can explain
most of the boom and bust in aggregate R&D in 1990s.
6
Authors use Euler equation
framework to estimate financial frictions and show that shifts in the internal and exter-
nal financing can account for the aggregate R&D cycles. In a similar fashion, in this
study we will first demonstrate the differences in aggregate cycles across sectors and
establish a link between these asymmetries and the aggregate credit movements in
5
For example, about financial liberalization and financial development, Love (2003) and Islam and
Mozumdar (2007) show that financial development decreases the level of financial constraints, Harrison,
Love and McMillan (2004) analyses the relationship between capital flows and financing constraints and
Leaven (2003) finds that financial liberalization reduces financing constraints for small firms. Almeida
(2004) checks the effect of cash holdings on financial constraints, Agca and Mozumdar (2008) examines the
relationship between investment-cash flow sensitivity and factors related to capital market imperfections
and Brown and Petersen (2009) studies the effect of changing decomposition of investment and increasing
public equity on investment-cash flow sensitivities.
6
Goyal and Yamada (2004) uses stock market data from Japan and find that investment-cash flow
sensitivities are correlated with the boom and busts in aggregate asset prices. This study connects the
aggreagte cycles to the financial frictions. Gorodnichenko and Schnitzer (2010) study how financial con-
straints affect macroeconomic outcomes of a country like level of income and export intensity. They
find that financial constraints restrict the ability of firms to innovate and export, thereby hindering the
catching-up process of development.
62
Turkey. Then using a detailed data set, we will examine the financial constraints across
sectors and conclude that differences in the financial structure of the sectors are closely
in line with the aggregate differences for sector based movements.
3.3 Macro Co-movements
After Turkey opened up its trade markets in early 1980s, liberalization of financial
accounts followed in 1989 as the natural second step of economy-wide liberalization.
Opening the domestic markets to financial flows was followed by the economy expe-
riencing two main boom-bust cycles that were closely related to the financial account.
The first boom ended in the current account crises of 1994 and second boom has ended
in the twin crises (current account and banking) of 2001. Here in this section, we will
document the basic boom-bust cycles and co-movements in Turkish economy for the
period after financial liberalization and also demonstrate sector based asymmetries
between non-tradable and tradable sectors over the business cycle in a macro level
analysis.
3.3.1 Financial Liberalization and Boom-Bust Cycles
Figure 3.1 shows the percentage deviations of real GDP from HP trend for the
period of 1990-2004. There are two major negative deviations from trend in the years
1994 and 2001. First one corresponds to the current account crises of 1994 and sec-
ond one corresponds to the twin crises (current account and banking) of 2001. It is
63
clear from Figure 3.1 that both of these crises are preceded by a boom, and the corre-
sponding boom ends with the crises. So, we will consider these two boom-busts after
financial liberalization. First one is a mild boom from 1991 to 1993 and the bust in 1994.
It is followed by the second cycle, boom expanding from 1995 to 2000 and the crises
occurring in 2001. The second cycle was larger in magnitude for both boom and bust
periods as can be seen from Figure 3.1.
Following the expenditure approach to GDP , we will first decompose the deviations
in real GDP into the parts arising from investment and consumption. Later, we will
decompose the deviations in real GDP into parts stemming from different sectors of the
economy following the production approach to GDP . Taking the expenditure approach
into consideration, Figure 3.2 shows the deviations in consumption and investment
for the same period. Similar to the cases in most of the other developing countries,
first inference from the graph is that investment is more volatile than consumption.
Consumption moves in the
interval for the sample period. In contrast, investment
makes swings as large as
and. This implies that investment volatility is
an important source of fluctuations.
As an alternative way, we will also look at the differences in the production side.
In order to do this, we will decompose the GDP into the value-added by different
sectors. Viewing Turkey as a small open economy, it is natural that first way of aggre-
gating output would be to separate production to tradable and non-tradable parts as
T and NT, respectively. Tradable refers to the industry sector of the economy, which
includes manufacturing, electricity, gas and water, and mining. For non-tradable, we
64
-10
-8
-6
-4
-2
0
2
4
6
8
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Figure 3.1: Percentage Deviations of GDP from HP trend
-35
-25
-15
-5
5
15
25
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Consumption
Investment
Figure 3.2: Percentage Deviations from HP trend: Consumption and Investment
will include transportation and communication, commerce and construction. We omit
financial sector and agriculture in our analysis. To depict the different movements
more starkly, we also plot the ratio of NT to T value in Figure 3.3. Firstly, it is easily
noticeable that this ratio is not a constant or a random walk but it has clear cycles. Sec-
ondly, the cycles closely correspond to the boom-bust cycles in real GDP . During the
boom-bust cycle of 1991-1994, NT over T ratio follows the same trend with an initial
increase and a decrease later. The second cycle in GDP of 1995-2001 is also followed
65
1.4
1.5
1.6
1.7
1.8
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Figure 3.3: Non-tradable to Tradable Output Ratio
closely. This picture shows that there is an asymmetry in the sector based movements
of the output. In other words to say, aggregate cycle masks an asymmetrical move in
the sector based output.
We also analyze the size of the credit markets in Turkey after financial liberaliza-
tion. Figure 3.4 presents the ratio of private sector credit to GDP . There is a startling
resemblance between this credit data and the Non-tradable to Tradable output ratio in
Figure 3.3. Evidently, during a boom, credit to GDP ratio increases along with non-
tradable to tradable ratio. Similarly, credit to GDP ratio and non-tradable to tradable
ratio decrease in tandem during a crisis. This might be due to the fact that non-tradable
sector is more responsive to the credit markets than the tradable sector. We suggest that
there is a credit channel that works asymmetrically across sectors. In the next subsec-
tion, we will try to establish this conjecture using our macro data set.
66
10
15
20
25
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Figure 3.4: % Private Credit / GDP
3.3.2 Macro Co-movements and Credit
Main mechanism we offer for the macro level co-movements is based on Tornell
and Westermann (2005). They explain the cycles in developing economies in terms of
cycles in financial markets, more specifically in credit markets. They decompose an
economy into tradable and non-tradable sectors. Tradable sector is assumed to have
better access to external finance than non-tradable sector. As a result, non-tradable
sector will be financially constrained and more responsive to the changes of credit in
the economy. In general, their mechanism can explain the basic cycles in developing
countries and as a further step it can also account for the sector based asymmetries.
The authors effectively prove their mechanism for developing countries using macro
level data in Tornell and Westermann (2005).
We too show these macro correlations for Turkey. Moreover, as an extra effort to dis-
play the relationship more concretely, we connect these macro correlations to a micro
67
level data set in subsequent sections. By doing so, we would like to show that non-
tradable sector is indeed more constrained than tradable sector. Then we can argue
more boldly that credit is an important factor for asymmetrical business cycles. We can
suggest that non-tradable sector is more responsive to the credit and thus any move-
ment in credit can induce movements in NT over T ratio as well as overall economy.
We use annual data from 1970 to 2004 to run the regression model in Equation 3.1.
Our aim is to quantify the relationship between Credit over GDP and NT over T ratios
after controlling for other explanatory factors. We used the lagged trade liberalization
dummy because trade liberalization is a structural change. Compared to financial lib-
eralization, it’s a real sector adjustment, which means it should take some time before
we notice its effects. We also include the crises and the lagged crises dummy variables
because we expect their impact on the non-tradable sector to be more pronounced and
hence we expect a negative coefficient for these dummies. Table 3.1 provides the results
for our macro data regression analysis. Since both Credit over GDP and NT over T
ratios are not rejected to be unit root we use the first differences of the corresponding
variables in the regressions.
'
(
)* $
)* $
(3.1)
!+,,* (+* ! *+- .+$ /0
1 0+0 +- ! *+- .+$ /0
) 2*2
!+,,* ) 2*2 &
68
Our results for macro analysis show that, for all five models Credit to GDP ratio
affects NT to T ratio significantly with one lag. The coefficient of the lagged first differ-
ence of the Credit to GDP ratio is highly significant, at least level. This shows that
there is a strong relation between the credit market and sector based asymmetries. We
see that trade liberalization dummy affects NT over T ratio negatively in the sense that
it helps more to T sector. On the other hand, financial liberalization affects NT over
T ratio positively. We expect that with financial liberalization,credit in the economy
will increase and this will benefit non–tradable sectors more profoundly due the fact
that they are more financially constrained than tradable sectors. Both of the coefficients
are quite significant as expected. Tornell and Westermann (2005) showed NT over T
ratio to decrease with trade liberalization and increase with financial liberalization for a
large sample of developing countries. Crises of 1994 and 2001 have negative effects on
NT over T ratio supporting the observation that non-tradable sector is hit more heavily
by the crises but the crises dummy is only significant for our most general model, i.e.
model five.
Given the fact that non-tradable firms are more constrained than tradable firms,
we can generate the boom-bust cycles we see in the data. During a credit boom, firms
that are more responsive to the credit (non-tradable) will increase their investment and
correspondingly their output more than other firms (tradable). So NT over T ratio will
increase over the boom. And during a credit bust or decrease, non-tradable firms will
decrease investment and output more than other firms, and we will see NT over T ratio
decreasing. This will potentially indicate that credit market is an important part of the
69
Table 3.1: Macro Data Regression Results
Macro Data: Sample:1970-2004
Dependent Variable: D(N/T)
12 34 5
D(Cr/GDP) -0.0004 0.0012 -0.0011 -0.0046 -0.0060
(0.0063) (0.0067) 0.0061 (0.0063) (0.0065)
D(Cr/GDP)(-1) 0.0187*** 0.0172** 0.0170** 0.0188*** 0.0173**
(0.0067) (0.0071) (0.0066) (0.0065) (0.0067)
TL(-1) -0.0053 -0.0779* -0.0773* -0.0798*
(0.0328) (0.0442) (0.0429) (0.0432)
FL 0.0874** 0.1020** 0.1133**
(0.0424) (0.0422) (0.0442)
Crises -0.1049 -0.1169*
(0.0646) (0.0662)
Crises(-1) -0.0591
(0.0660)
Constant -0.0051 -0.0045 0.0028 0.0027 0.0038
(0.0151) (0.0264) (0.0238) (0.0231) (0.0232)
R
2
0.2640 0.2340 0.3202 0.3590 0.3540
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
D refers to differenced variable and (-1) refers to one lag.
N/T is the output ratio of non-tradable and tradable sectors.
Cr/GDP the ratio of private sector credit to GDP .
TL(-1) is the trade liberalization dummy with one lag, which is 1 for the 1985-2004 period.
FL is the financial liberalization dummy, which is 1 for the 1990-2004 period.
Crises is the dummy for 1994 and 2001 crises and Crises(-1) is its one lag.
All regressions include a significant dummy for 1980 where there was a military coup.
boom-bust cycles in the economy and this effect is asymmetric across sectors. Next
section of the paper will try to establish this claim in a more structural way using the
micro level data sets.
70
3.4 Micro Evidence
3.4.1 Model
To test the capital market imperfections across sectors, we use the Euler equa-
tion approach to investment.
7
Following Bond and Meghir (1994) and Bond, Elston,
Mairesse, and Mulkay (2003), we assume that there is a firm maximizing the stream of
net cash flows under the presence of symmetric, quadratic adjustment costs. The firm
solves the following problem:
3
!
4
!
"
(3.2)
subject to the capital accumulation constraint,
3
53
4
(3.3)
where
13
!
4
3
6
!
4
is the current net cash
flow. !
denotes variable factor inputs, 3
denotes the beginning of the period capi-
tal stock, 6
is the price of variable factors,
is the price of output,
is the price of
investment,
is the discount factor between period $ and period $7, 5 is the rate of
depreciation, 1 is the production function and
is the adjustment cost function. The
first order conditions yield the following Euler equation:
7
For robustness, the results of accelerator approach are given in the appendix, in section D.1. Results
are very similar to Euler equation approach.
71
4
5
4
3
(3.4)
Under the assumption of competitive markets and that 1 is constant returns to scale,
and with the adjustment cost functional form
4
3
43
3
we
can write Equation 3.4 as follows:
4
3
4
3
4
3
#
3
8
$
. (3.5)
Here,
13
!
4
3
6
!
is the gross current profit and 8
is the real cost of capital. This equation implies that current investment is positively
related to the expected future profit and to the current gross profits, and negatively
related to the cost of capital. There are two points to be emphasized. Firstly, the expec-
tation is captured by one-period ahead investment forecast. Secondly, profits that are
likely to be correlated with cash flows enter the equation even without financial con-
straints. Therefore, we can have structural interpretation of the basic Euler equation
for investment-cash flow regressions. By using realized 43
plus an error term
instead of expectation and rearranging Equation 3.5 we can arrive at the following
empirical specification:
4
3
4
3
4
3
3
(3.6)
9
3
:
72
where the cost of capital is replaced by time and firm fixed effects, and output-capital
ratio is added to account for the possibility of non-constant returns to scale or monop-
olistic competition. Under the case of no financial constraints, we have the following
properties for the coefficients:
and
;
The main advantage of this structural Euler equation approach is that it controls for
the effects of expectations on investment decisions. In a non-structural estimation like
accelerator model, when financial variables enter the regression with significant coeffi-
cients, one cannot be very conclusive due to the possibility that these financial variables
might be correlated with future profitability. The Euler equation approach eliminates
this problem and shows that cash flow or profits enter the specification even without
financial constraints. However, the model implies that coefficient on profits
will
be negative. However, if firms face financial constraints, Bond et al (2003) write that,
"investment spending is positively related to cash flow or profits through the effect of
financial constraints." And " since the gross operating profits term 3
in Equa-
tion 3.6 will be highly correlated with cash flow, the prediction of a negative sign on
this term may be expected to fail in the presence of financial constraints." As a result,
a first test of the financial constraints would be to look at the sign of the lagged cash
flow or profits variable. Also under the financial constraints, one would expect posi-
tive relationship between investment and current cash flow which is the basic internal
finance option for the firm. As in Brown, Fazzari, and Petersen (2009), we also add
contemporaneous cash flow or profits. Due to the correlation between sales and cash
flow, contemporaneous sales are added as an additional control for firm demand to
73
avoid the omitted variable bias. Therefore, we will also use the following modified
specification as in Brown, Fazzari, and Petersen (2009):
4
3
4
3
4
3
3
3
(3.7)
9
3
9
3
:
3.4.2 Methodology
We estimate Equations 3.6 and 3.7 using both fixed and random effects models,
and the first difference GMM method developed by Arellano and Bond (1991). In
order to compare the efficiency of random and fixed effects, we conduct Hausman test.
If the test favors fixed effects, we only report fixed effects results. Conversely, if it
favors random effects, we report both fixed and random effects to save space. For the
first difference GMM estimates, we assume that all the right hand side variables are
possibly endogenous and use the appropriate lagged levels as the instruments.
Our goal in this study is to examine the sector based asymmetries in terms of cap-
ital market imperfections and determine the relation between these asymmetries and
boom-bust cycles. To understand whether the internal funds of firms matter for invest-
ment decision or not, we can start with Equation 3.6, which specifies a model without
capital market imperfections, i.e. frictionless capital markets. Suppose, due to credit
market imperfections, firms have some kind of credit constraints or high information
74
Table 3.2: Variable Definitions
Variable Definition
3
Capital stock at the beginning of period $.
4
Investment in period $.
3$ 3$
Net sales in period $
)
Change in sales
$ $
)1
Cash flow in period $.
Net Profit+Owners Capital+
Depreciation Allowances
Dummy variable
1 for non-tradable sector
0 for tradable sector
costs or the cost of external financing exceeds that of internal one, then changes in net
worth would affect the investment decision of such firms .
For our analysis of the micro data sets, first we run the cash-flow regressions for
tradable and non-tradable sectors separately. Secondly, we pooled the data and intro-
duce a dummy variable that takes the values of for non-tradable sectors and for
tradable sectors. The definition of the variables used in our regression are given in the
following Table 3.2.
3.4.3 Results for Company Accounts Survey Data Set
Table 3.3 gives the summary statistics of main variables used for the non-tradable
and tradable sectors. At first glance, the variables show that investment normalized
by capital stock has a higher average and a much more volatility in non-tradable sec-
tor, even tough the mean and standard deviations of cash flows are not so different
for the two sectors. This implies that for given similar cash flow levels and volatilities,
75
I / K(-1)
-0.1
0.1
0.3
0.5
0.7
1991
1993
1995
1997
1999
2001
2003
2005
T NT
Figure 3.5: Investment over Capital ratio for Tradable and Non-tradable sectors
we observe a higher volatility in the investment of non-tradable sector. The ratio of
standard deviation of investment to cash flow is 0.4879 in non-tradable sector which
is much higher than the corresponding value of 0.2670 in tradable sector. We can see
the same pattern in the Figures 3.5 and 3.6 where the cash flow figures for both sec-
tors resemble each other closely, whereas the investment figure is much more volatile
for non-tradable sector over the whole period. For very similar levels of cash flows,
non-tradable sector experiences a higher investment during boom years and a lower
investment during bust years. These results highlight that, even though both sectors
might have similar internal funds, structural differences like access to financing can
produce pronounced differences in investment behavior. This hypothesis of difference
in the investment behavior across sectors is analyzed in a more structural way through
the regression results
Table 3.4 displays the separate cash flow regressions for both tradable and non-
tradable sectors using fixed effects and first-differenced GMM methods. Model 1 is the
Equation 3.6 corresponding to the perfect markets case, and Model 2 is the Equation
76
CF / K(-1)
-0.1
0.3
0.7
1.1
1.5
1.9
1991
1993
1995
1997
1999
2001
2003
2005
T NT
Figure 3.6: Cash flow over Capital ratio for Tradable and Non-tradable sectors
Table 3.3: Summary Statistics for CBRT Data Set
CBRT Company Accounts Survey Weighted Average Data Set
Tradable Sector
Mean Std. Dev. Min Max Std. Dev. Ratio*
I/K(-1) 0.0806 0.1421 -0.1803 0.6677 0.2670
CF/K(-1) 0.3738 0.5323 -0.4692 3.3668
CS/K(-1) 0.3760 0.8468 -3.2910 5.7014
Non-tradable Sector
Mean Std. Dev. Min Max Std. Dev. Ratio*
I/K(-1) 0.1542 0.2813 -0.2076 1.7014 0.4879
CF/K(-1) 0.3984 0.5765 -0.5948 2.9687
CS/K(-1) 0.6869 1.3283 -0.9691 8.1573
* Standard deviation ratio of I/K(-1) and CF/K(-1).
3.7 corresponding to the financial constraints case. Neither non-tradable nor tradable
sectors fully satisfy the coefficient restrictions of perfect capital markets case. In Model
1, coefficient of lagged sales is positive and coefficient of lagged cash flows is negative
for tradable sector; and these signs are in line with the perfect capital markets hypoth-
esis. However, for non-tradable sectors, none of the coefficients are in line with the
perfect markets case, with insignificant numbers for lagged sales and cash flow vari-
ables. Therefore, we can strongly reject the perfect capital markets for non-tradable
77
Table 3.4: Regression Results for CBRT Data Set
CBRT Company Accounts Weighted Average Data Set
Dep Var. Tradable Non-tradable
I/K(-1) FE GMM FE GMM
Variable 1 2 1 2 1 2 1 2
I(-1)/K(-2) -0.059 0.051 -0.050 0.079 -0.078 -0.120** -0.070 -0.083
0.134 0.124 0.113 0.095 0.061 0.052 0.091 0.078
I(-1)/K(-2)
2
0.276 0.273 0.213 0.209 -0.076* -0.080** -0.076 -0.084*
0.297 0.266 0.183 0.143 0.039 0.034 0.053 0.046
S/K(-1) 0.012 0.020** -0.032** -0.034***
0.009 0.009 0.016 0.008
S(-1)/K(-2) 0.027*** 0.010 0.048*** 0.019** 0.003 0.007 0.005 0.016*
0.008 0.009 0.008 0.008 0.007 0.009 0.008 0.008
CF/K(-1) 0.132*** 0.156*** 0.355*** 0.378***
0.017 0.014 0.072 0.046
CF(-1)/K(-2) -0.065*** -0.050*** -0.039* -0.028* -0.054 0.078 0.009 0.089
0.021 0.017 0.021 0.017 0.053 0.048 0.064 0.055
Constant 0.103* -0.051 0.002 -0.002 0.149** 0.281** 0.015** -0.007
0.053 0.042 0.002 0.002 0.067 0.125 0.007 0.007
# of Obs. 414 414 384 384 188 188 172 172
R
2
0.217 0.376 0.168 0.293
m1 0.000 0.000 0.000 0.000
m2 0.701 0.170 0.827 0.658
Sargan 0.011 0.126 0.918 0.996
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
FE and RE (random effects) give very similar results, but Hausmann test favors FE.
Therefore we do not report RE.
In GMM estimation t-2 to t-5 lags are used as instrumental variables.
FE and RE results are the same.
p-values are provided for m1, m2 and Sargan tests.
sectors. We then add current sales and cash flows to the regressions to quantify the
financial market imperfections as in Brown, Fazzari, and Petersen (2009). Current cash
flow coefficient in Model 2 is significant for both tradable and non-tradable sectors at
one percent level. This might suggest that both sectors face financial constraints. How-
ever, there is a startling difference between the current cash flow coefficients across sec-
tors. The value is around 0.13-0.16 for tradable compared to 0.36-0.38 for non-tradable
78
sector. As a result, for similar amounts of variance in the cash flows, the response of
investment will be much higher for non-tradable sector. This finding is in line with
Figures 5 and 6, where average cash flow profiles are very similar over time but invest-
ment is more volatile for non-tradable sector.
Table 3.5 combines the data for both sectors and includes a dummy for non-tradable
sector for the coefficient of cash flows. We use fixed effects, random effects and first-
differenced GMM estimations in the regressions. When we conduct fixed effects and
random effects regression for both perfect capital markets case (Model 1) and financial
constraints case (Model 2), Hausmann test chooses random effects as more efficient
estimation than fixed effects. In Model 1, where we test perfect markets assumption,
random effects show that lagged cash flow dummy is negative for tradable sector as
proposed by the theory but the dummy for non-tradable sector is significantly positive
meaning that non-tradable sector does not satisfy perfect capital markets. In the case of
GMM estimation lagged cash flow is not significant for both sectors thereby rejecting
the perfect markets for both sectors. In both estimations, we discard the perfect access
to capital markets assumption for tradable sector. Then in Model 2 we include current
sales and cash flows as additional variables to see the financial constraints. Hausmann
test favors random effects regression over fixed effects regression yet again. And both
random effects and GMM estimations find that current cash flow is significant for non-
tradable and tradable sectors. This shows that financial constraints can be binding
for both sectors. We also see that the coefficient on cash flow for non-tradable sector
is significantly higher than tradable’s. As in Table 3.4, this difference shows that for
79
very similar amounts of volatility in cash flows, investment response in non-tradable
sector is significantly larger than that of tradable sector. Therefore, any movements in
financing sources will be reflected in higher volatility of investment for non-tradable
sectors. This finding is also in accordance with the asymmetrical aggregate movements
across sectors.
We can infer from these results that in the economy, non-tradable sectors are always
more financially constrained and its responsiveness to business cycle is higher. Trad-
able sectors are classified as constrained with respect to fixed investment but tradable
firms are less constrained than non-tradable firms. Since most of the cash flow studies
in the literature use either stock market data, which is not a representative sample of
the overall economy, or limited firm level data from manufacturing sector, our result is
also important for aggregate investment dynamics and business cycles. For example,
Gelos and Werner (2002) consider the investment behavior for manufacturing firms
in Mexico for the period 1984-1994. They find that firms are financially constrained.
However, this result would not be enough to infer about the aggregate investment and
output dynamics. As stated by Hubbard (1988), estimating investment equations for
other sectors of the economy would provide helpful for aggregate investment dynam-
ics. In the Turkish case, we have estimated the investment equations for non-tradable
and tradable sectors of the economy, and gotten micro evidence for the aggregate
investment and output dynamics. In our estimation we have a higher sensitivity of
non-tradable sector. This implies that non-tradable firms will be more responsive to
external finance or bank credit. So when there is a lending boom, non-tradable firms
80
will be investing and producing more, and when credit decreases non-tradable sec-
tor will decrease investment and output more than tradable sector. Then this frame-
work presents a micro evidence for the macro correlations between NT/T ratio and
Credit/GDP ratio, and establishes the link between credit markets and output dynam-
ics.
3.4.4 Results for Istanbul Stock Exchange Firms Data Set
We conduct cash-flow regression analysis on stock market firms in this section.
Our period covers 1987 to 2003. We exclude financial and utility firms. We again cat-
egorize our firms into non-tradable and tradable sectors. In Table 3.6, we first check
the summary statistics. In contrast with the summary statistics in survey of sectors,
stock market firms there are do not display much difference in levels and variances of
both investment and cash flow variables. Our analysis yields very similar values for
the ratio of standard deviations of investment and cash flows for tradable and non-
tradable sectors, namely 0.4799 and 0.4543 respectively. Graphical representation of
investment and cash flows over the sample period in Figures 3.7 and 3.8 further rein-
forces this result and shows that for the given similar levels of cash flows, we get very
similar movements of investment in both tradable and non-tradable sectors. There-
fore, from the first diagnostics of the data, we do not find significant differences in the
investment and financial variables.
Tables 3.7 provides the results for several models of cash flow regressions for
pooled data of all firms. We estimate fixed effects, random effects and first-differenced
GMM regressions. Hausmann test favors fixed effects as the efficient estimator over
81
I / K(-1)
-0.05
0.1
0.25
0.4
1989
1991
1993
1995
1997
1999
2001
2003
T NT
Figure 3.7: Investment over Capital ratio for Tradable and Non-tradable sector
CF / K(-1)
-0.05
0.35
0.75
1.15
1.55
1989
1991
1993
1995
1997
1999
2001
2003
T NT
Figure 3.8: Cash flow over Capital ratio for Tradable and Non-tradable sector
random effects. For the case of perfect capital markets setup, in Model 1, coefficients
of the regressions do not satisfy the restrictions. Coefficient of lagged cash flow, which
is predicted to be negative under perfect markets case, is insignificant in fixed effects
estimation and positive in first-differenced GMM estimation. Moreover the dummy
for non-tradable sector is not significant in either case. Hence, we can conclude that
perfect capital markets hypothesis is rejected for both sectors under an Euler equation
approach. To further quantify the market imperfections, we add current sales and cash
82
flows to the benchmark regression as in Brown, Fazzari, and Petersen (2009) in Model
2. The coefficient of current cash flow in the imperfect capital markets case is significant
and positive for all estimations and the number is around 0.11-0.13. For our purpose of
analyzing the sector based differences in access to financial markets, we also look at the
dummy for non-tradable sector cash flow. This coefficient is insignificant in all cases.
In contrast to the survey data set, where we have found significant differences in cash
flow coefficients across sectors, for the stock market data there are no differences in
capital market imperfections. Even though, both tradable and non-tradable firms are
financially constrained, their response to cash flow movements do not diverge. This
finding is in line with Figures 3.7 and 3.8, where both investment and cash flow profiles
are very similar across sectors.
Comparing these results to the previous section for a larger and more representa-
tive data set for economy, we conclude that using stock market firms we can assess
whether they are constrained or not, but it does not help with studying the big macro
picture.
From these micro results, we can conclude that market imperfections are of para-
mount importance for the economy and there is an asymmetry across sectors that non-
tradable sectors are more constrained than tradable sector.
83
3.5 Conclusion
In this study, we pointed to the macro relationship between non-tradable to trad-
able ratio and private credit to GDP ratio. We demonstrated that there is a close rela-
tionship between NT/T ratio and Credit/GDP ratio. We concluded from this study
that credit market is an important part of the boom-bust cycles in the economy and
this effect is asymmetric across sectors. To support these findings we checked two sep-
arate micro data sets. For stock market firms, we get that there is no difference in terms
of credit constraints for the firms. But we noticed that this data set is not representa-
tive of the economy, since these firms are very large with respect to average firms and
have better access to credit markets. To see more representative results, we checked a
larger data set for different sectors. From this data set, we concluded that non-tradable
sectors are financially more constrained than tradable sectors.
Given the fact that non-tradable sectors are more constrained than tradable firms,
we can generate the boom-bust cycles we see in the data. During a credit boom, firms
which are more responsive to the credit (non-tradable) will increase their investment
and correspondingly their output more than other firms (tradable). So NT/T ratio will
increase over the boom. And during a credit bust or decrease, non-tradable firms will
decrease investment and output more than other firms, and we will see NT/T ratio
decreasing.
84
Table 3.5: Pooled Regression Results for CBRT Data Set
CBRT Company Accounts Weighted Average Data Set
Dep Var.
I/K(-1) 1 2
Variable FE RE GMM FE RE GMM
I(-1)/K(-2) 0.018 0.074* 0.019 0.060 0.100* -0.013
0.050 0.045 0.051 0.063 0.060 0.044
I(-1)/K(-2)
2
-0.070* -0.040 -0.069** -0.053 -0.005 -0.071**
0.037 0.026 0.035 0.042 0.042 0.031
S/K(-1) 0.017 0.015 -0.027***
0.012 0.013 0.005
S(-1)/K(-2) 0.009** -0.001 0.011** -0.003 -0.018 0.019***
0.004 0.002 0.005 0.007 0.011 0.005
CF/K(-1) 0.131*** 0.143*** 0.186***
0.017 0.016 0.020
CF(-1)/K(-2) -0.065*** -0.055*** 0.002 -0.033* -0.015 0.006
0.020 0.021 0.025 0.017 0.016 0.021
D*CF/K(-1) 0.173*** 0.153** 0.187***
0.066 0.061 0.035
D*CF(-1)/K(-2) 0.018 0.062*** -0.021 0.064* 0.023 0.054
0.035 0.033 0.042 0.034 0.031 0.037
Constant 0.070** 0.300*** 0.004 -0.068 0.093 -0.007***
0.032 0.076 0.003 0.070 0.062 0.002
# of Obs. 602 602 556 602 602 556
R-sq. 0.083 0.147 0.211 0.374
Hausman 0.996 0.988
m1 0.000 0.000
m2 0.678 0.260
Sargan 0.033 0.252
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
We used t-2 to t-12 lags for GMM estimation.
Coefficients don’t change when we used all the lags available.
p-values are provided for m1, m2 and Sargan tests.
85
Table 3.6: Summary Statistics for ISE Data Set
ISE Firms Data Set
Tradable Sector
Mean Std. Dev. Min Max Std. Dev. Ratio*
I/K(-1) 0.1062 0.4219 -0.5096 2.5669 0.4799
CF/K(-1) 0.5355 0.8792 -1.1222 4.4564
CS/K(-1) 0.2274 1.3943 -4.2843 4.8296
Non-tradable Sector
Mean Std. Dev. Min Max Std. Dev. Ratio*
I/K(-1) 0.1097 0.4180 -0.3418 1.4469 0.4543
CF/K(-1) 0.7144 0.9201 -0.4706 3.6901
CS/K(-1) 0.3612 1.2484 -1.8577 4.2099
* Standard deviation ratio of I/K(-1) and CF/K(-1).
86
Table 3.7: Pooled Regression Results for ISE Data Set
ISE Firms Data Set
Dep Var.
I/K(-1) 1 1 2 2
Variable FE RE GMM FE RE GMM
I(-1)/K(-2) -0.101** 0.053 -0.128* 0.048 0.112** -0.040
0.046 0.045 0.069 0.049 0.046 0.077
I(-1)/K(-2)
2
0.006 -0.011 0.037 -0.026 -0.017 -0.021
0.018 0.014 0.028 0.018 0.012 0.024
S/K(-1) 0.071*** 0.024** 0.102***
0.016 0.011 0.019
S(-1)/K(-2) 0.027*** 0.003* 0.014*** 0.016* -0.007 0.042**
0.006 0.002 0.005 0.009 0.009 0.020
CF/K(-1) 0.108*** 0.126*** 0.134***
0.032 0.029 0.046
CF(-1)/K(-2) 0.002 0.036** 0.075** 0.008 -0.009 0.100**
0.024 0.018 0.032 0.024 0.021 0.043
D*CF/K(-1) 0.022 -0.008 -0.092
0.064 0.060 0.077
D*CF(-1)/K(-2) 0.042 0.004 0.040 0.046 0.004 0.002
0.034 0.022 0.056 0.040 0.041 0.070
Constant -0.033 0.107 -0.010** -0.282*** -0.033 -0.002
0.069 0.063*** 0.004 0.081 0.058 0.005
# of Obs. 1581 1581 1398 1300 1300 1051
R
2
0.031 0.072 0.082 0.131
Hausman (p-value) 0.002 0.000
m1 (p-value) 0.000 0.000
m2 (p-value) 0.205 0.574
Sargan 0.265 0.090
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
87
88 Bibliography
Akerlof, G. and J. Yellen (1985). Can small deviations from rationality make
significant differences to economic equilibria? American Economic Review
75, 708–720.
Allayannis, G. and A. Mozumdar (2004). The impact of negative cash flow and
influential observations on investment-cash flow sensitivity estimates.
Journal of Banking & Finance 28(5), 901–930.
Arellano, M. and S. Bond (1991). Some tests of specification for panel data:
Monte Carlo evidence and an application to employment equations. Review of
Economic Studies 58(2), 277–297.
Basak, S. and H. Yan (2009, July). Equilibrium asset prices and investor
behavior in the presence of money illusion. Yale ICF Working Paper No. 08-
23, Forthcoming in Review of Economic Studies.
Bernanke, B., M. Gertler, and S. Gilchrist (1999). Handbook of
Macroeconomics, Chapter The financial accelerator in a quantitative business
cycle framework, pp. 1341–1393. North-Holland.
Bodie, Z. (1976). Common stocks as a hedge against inflation. Journal of
Finance 31(2), 459–470.
Bond, S., J. A. Elston, J. Mairesse, and B. Mulkay (2003). Financial factors and
investment in belgium, france, germany, and the united kingdom: A
comparison using company panel data. Review of Economics and Statistics
85(1), 153–165.
Bond, S. and C. Meghir (1994). Financial constraints and company investment.
Fiscal Studies 15(2), 1–18.
Branch, B. (1974). Common stock performance and inflation: An international
comparison. Journal of Business 47(1), 48–52.
89 Brown, J. R., S. M. Fazzari, and B. C. Petersen (2009). Financing innovation and
growth: Cash flow, external equity, and the 1990s R&D boom. Journal of
Finance 64(1), 151–185.
Brunnermeier, M. and C. Julliard (2008). Money illusion and housing frenzies.
Review of Financial Studies 21, 135–180.
Campbell, J. Y. and R. J. Shiller (1988). The dividend-price ratio and
expectations of future dividends and discount factors. Review of Financial
Studies 3, 195–228.
Campbell, J. Y. and T. Vuolteenaho (2004). Inflation illusion and stock prices.
American Economic Review Papers and Proceedings 94, 19–23.
Carlstrom, C. T. and T. S. Fuerst (1997). Agency costs, net worth, and business
fluctuations: A computable general equilibrium analysis. American Economic
Review 87(5), 893–910.
Chordia, T. and L. Shivakumar (2005). Inflation illusion and post-earnings
announcement drift. Journal of Accounting Research 43, 521–556.
Cleary, S. (1999, April). The relationship between firm investment and financial
status. Journal of Finance 54(2), 673–692.
Cohen, R., C. Polk, and T. Vuolteenaho (2005). Money illusion in the stock
market: The Modigliani-Cohn hypothesis. Quarterly Journal of Economics
120, 639–668.
Cohn, R. A. and D. R. Lessard (1981). The effect of inflation on stock prices:
International evidence. Journal of Finance 36(2), 277–289.
Day, T. E. (1984). Real stock returns and inflation. Journal of Finance 39(2),
493–502.
Fama, E. F. (1981, Sept). Stock returns, real activity, inflation, and money.
American Economic Review 71(4), 545–565.
Fama, E. F. and G.W. Schwert (1977). Assset returns and inflation. Journal of
Financial Economics 5, 115–146.
Fazzari, S., G. Hubbard, and B. Petersen (1998). Financing constraints and
corporate investment. Brookings Paper on Economic Activity 1, 141–206.
90 Firth, M. (1979). The relationship between stock market returns and rates of
inflation. Journal of Finance 34(3), 743–749.
Gelos, R. G. and A. M. Werner (2002). Financial liberalization, credit
constraints, and collateral: investment in the Mexican manufacturing sector.
Journal of Development Economics 67(1), 1–27.
Gertler, M., S. Gilchrist, and F. M. Natalucci (2007). External constraints on
monetary policy and the financial accelerator. Journal of Money, Credit and
Banking 39(2-3), 295–330.
Geske, R. and R. Roll (1983). The fiscal and monetary linkage between stock
returns and inflation. Journal of Finance 38(1), 1–33.
Gilchrist, S. and C. P. Himmelberg (1995). Evidence on the role of cash flow for
investment. Journal of Monetary Economics 36, 541–572.
Gordon, M. (1962). The investment, financing, and valuation of corporation.
Homewood.
Gordon, M. J. (1983). The impact of real factors and inflation on the
performance of the U.S. stock market from 1960 to 1980. Journal of Finance
38(2), 2.
Gultekin, N. B. (1983). Stock market returns and inflation forecasts. Journal of
Finance 38(3), 663–673.
Hasbrouck, J. (1984). Stock returns, inflation, and economic activity: The survey
evidence. Journal of Finance 39(5), 1293–1310.
Hayashi, F. (1982, January). Tobin’s marginal q and average q: A neoclassical
interpretation. Econometrica 50(1), 213–224.
Hess, P. J. and B.-S. Lee (1999). Stock return and inflation with supply and
disturbances. Review of Financial Studies 12(5), 1203–1218.
Hubbard, G. (1988, March). Capital market imperfections and investment.
Journal of Economic Literature 36, 193–225.
Hubbard, R. G., A. K. Kashyap, and T. M. Whited (1995). Internal finance and
firm investment. Journal of Money, Credit and Banking 27, 683–701.
91 Jaramillo, F., F. Schiantarelli, and A. Weiss (1996). Capital market
imperfections before and after financial liberalization: An Euler equation
approach to panel data for Ecuadorian firms". Journal of Development
Economics 51, 367–386.
Jensen, M. C. and W. H. Meckling (1976). Theory of the firm: Managerial
behavior, agency costs and ownership structure. Journal of Financial
Economics 3(4), 305–360.
Kadapakkam, P.-R., P. Kumar, and L. A. Riddick (1998). The impact of cash
flows and firm size on investment: The international evidence. Journal of
Banking and Finance 22, 293–320.
Kaplan, S. and L. Zingales (1997). Do investment-cashflow sensitivities provide
useful measures of financing constraints? Quarterly Journal of Economics
112(1), 169–215.
Modigliani, F. and R. Cohn (1979). Inflation, rational valuation, and the market.
Financial Analysts Journal 35, 24–44.
Modigliani, F. and M. H. Miller (1958, June). The cost of capital corporation
finance and the theory of investment. American Economic Review 48(3), 261–
297.
Myers, S. C. and N. S. Majluf (1984). Corporate financing and investment
decisions when firms have information that investors do not. Journal of
Financial Economics 13, 187–221.
Neumeyer, A. and F. Perri (2005). Business cycles in emerging economies: the
role of interest rates. Journal of Monetary Economics 55(2), 163–184.
Piazzesi, M. and M. Schneider (2008a). Asset Prices and Monetary Policy,
Chapter Inflation Illusion, Credit, and Asset Pricing, pp. 147–181. Chicago
University Press.
Piazzesi, M. and M. Schneider (2008b, April). Inflation and the price of real
assets. Working Paper.
Ranciere, R., A. Tornell, and F. Westermann (2008). Systematic crisis and
growth. Quarterly Journal of Economics 123(1), 359–406.
Ritter, J. R. and R. S. Warr (2002). The decline of inflation and the bull market
of 1982-1999. Journal of Financial and Quantitative Analysis 37, 29–61.
92 Schneider, M. and A. Tornell (2004). Balance sheet effects, bailout guarantees
and financial crises. Review of Economic Studies 71(3), 883–913.
Sharpe, S. A. (2002). Reexamining stock valuation and inflation: The
implications of analysts’ earnings forecasts. Review of Economics and
Statistics 84(4), 632–648.
Solnik, B. (1983). The relation between stock prices and inflationary
expectations: The international evidence. Journal of Finance 38, 35–48.
Stiglitz, J. and A. Weiss (1984). Information imperfections in the capital market
and macroeconomic fluctuations. American Economic Review 74, 194–199.
Titman, S. and A. Warga (1989). Stock returns as predictors of interest rates and
inflation. Journal of Financial and Quantitative Analysis 24(1), 47–58.
Tornell, A. and F. Westermann (2005). Boom-Bust Cycles and Financial
Liberalization. MIT Press.
Whited, T. M. (1992). Debt, liquidity constraints, and corporate investment:
Evidence from panel data. Journal of Finance 47, 1425–1460.
Williams, J. B. (1938). The theory of investment value. Harvard University Press.
Appendix A
Supplementary Materials for
Chapter 1
Table A.1: Return Regression Results for VWRETD
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETD and 3MTB
VWRETD
3MTB
12 3456 7
DY(-1) 0.293*** 0.354*** 0.427*** 0.414*** 0.397*** 0.438*** 0.444*
(0.115) (0.112) (0.120) (0.122) (0.120) (0.126) (0.247)
RF(-1) -1.776*** -0.661
(0.678) (0.756)
Constant -0.004 0.001 0.0001 -0.002 -0.002 0.001 0.003
(0.004) (0.005) (0.005) (0.004) (0.004) (0.004) (0.006)
INF. M. INF INF INF(-1) INF_E INF_M
-1.930*** -2.105 -1.737*** -3.123 -3.441*
(0.520) (0.494) (0.435) (0.787) (1.788)
Adj. R
2
0.008 0.011 0.035 0.035 0.029 0.028 0.006
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 7: January 1978 to August 2008
93
Table A.2: Return Regression Results for VWRETD cont.’d
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETD and 3MTB
VWRETD
3MTB
8 9 10 11 12 13 14
DY(-1) 0.310*** 0.322*** 0.401*** 0.390*** 0.377*** 0.398*** 1.572***
(0.110) (0.111) (0.120) (0.119) (0.117) (0.124) (0.558)
RF(-1) -2.811*** -1.945** -1.885* -2.020** -1.172 -6.617***
(0.993) (0.977) (0.985) (0.983) (1.135) (2.298)
TS(-1) 3.978** 0.808 -0.431 -0.105 0.294 -0.013 -9.489**
(1.810) (1.886) (1.992) (1.884) (1.890) (1.910) (4.487)
DS(-1) 0.0003 0.012* 0.013** 0.012** 0.012** 0.012** 0.004
(0.005) (0.006) (0.006) (0.006) (0.006) (0.006) (0.007)
Constant -0.010 -0.006 -0.006 -0.005 -0.006 -0.004 0.021**
(0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.011)
INF. M. INF INF INF(-1) INF_E INF_M
-1.467** -1.873*** -1.419*** -2.946*** -6.729**
(0.664) (0.517) (0.439) (1.039) (3.089)
INF
2
-50.239
(79.003)
Adj. R
2
0.013 0.023 0.040 0.041 0.032 0.032 0.037
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 14: January 1978 to August 2008
94
Table A.3: Return Regression Results for VWRETD with Dummy
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETD and 3MTB
VWRETD
3MTB
15 16 17 18
DY(-1) 0.315*** 0.357*** 0.316*** 0.358***
(0.118) (0.129) (0.119) (0.131)
INF_E -3.750*** -3.738***
(1.203) (1.212)
RF(-1) -1.914 -0.338 -2.251** -0.761
(1.261) (1.265) (1.153) (1.191)
TS(-1) -1.192 -0.067 -0.783 -0.348
(2.160) (2.067) (1.999) (1.991)
DS(-1) 0.009 0.009 0.013** 0.014***
(0.007) (0.007) (0.006) (0.005)
D*DY(-1) 1.666*** 1.340** 1.699*** 1.674***
(0.463) (0.568) (0.490) (0.467)
D*INF_E 8.248** 5.855**
(3.574) (2.780)
D*RF(-1) -4.217 -6.172** -2.950* -5.512***
(5.918) (3.077) (1.786) (2.052)
D*TS(-1) -2.102 3.246
(5.918) (7.518)
D*DS(-1) 0.010 0.013
(0.011) (0.010)
D -0.069*** -0.091*** -0.068*** -0.078***
(0.024) (0.025) (0.022) (0.019)
Constant -0.004 -0.001 -0.006 -0.004
(0.006) (0.006) (0.006) (0.006)
Adj. R
2
0.037 0.047 0.039 0.047
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
95
Table A.4: Return Regression Results for SP500RET
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from SP500RET and 3MTB
SP500RET
3MTB
1 2 3456 7
DY(-1) 0.215* 0.275** 0.344*** 0.330*** 0.317*** 0.354*** 0.384*
(0.112) (0.651) (0.114) (0.116) (0.115) (0.120) (0.230)
RF(-1) -1.762*** -0.712
(0.651) (0.722)
Constant -0.004 0.005 -0.0002 -0.002 -0.003 0.001 0.004
(0.004) (0.005) (0.005) (0.004) (0.004) (0.004) (0.005)
INF. M. INF INF INF(-1) INF_E INF_M
-1.817*** -2.005*** -1.702*** -3.067*** -3.921**
(0.509) (0.478) (0.416) (0.737) (1.674)
Adj. R
2
0.004 0.012 0.030 0.030 0.022 0.027 0.007
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 7: January 1978 to August 2008
Table A.5: Return Regression Results for SP500RET cont.’d
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from SP500RET and 3MTB
SP500RET
3MTB
8 9 10 11 12 13 14
DY(-1) 0.239** 0.250** 0.329*** 0.314*** 0.304*** 0.322*** 1.447***
(0.108) (0.109) (0.115) (0.114) (0.113) (0.120) (0.552)
RF(-1) -2.529*** -1.737* -1.659* -1.755* -0.905 -5.757**
(0.967) (0.950) (0.961) (0.959) (1.109) (2.269)
TS(-1) 3.938** 1.085 -0.200 0.228 0.583 0.283 -9.137**
(1.718) (1.818) (1.894) (1.793) (1.808) (1.828) (4.458)
DS(-1) -0.001 0.009 0.010* 0.010* 0.009* 0.009* 0.001
(0.004) (0.006) (0.006) (0.005) (0.006) (0.006) (0.007)
Constant -0.009 -0.005 -0.006 -0.005 -0.005 -0.004 0.022**
(0.006) (0.006) (0.006) (0.006) (0.006) (0.006) (0.011)
INF. M. INF INF INF(-1) INF_E INF_M
-1.227* -1.760*** -1.387 -2.890*** -7.388**
(0.688) (0.507) (0.423) (0.987) (3.081)
INF
2
-65.895
(80.218)
Adj. R
2
0.009 0.017 0.033 0.034 0.027 0.026 0.020
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 14: January 1978 to August 2008
96
Table A.6: Return Regression Results for SP500RET with Dummy
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from SP500RET and 3MTB
SP500RET
3MTB
15 16 17 18
DY(-1) 0.272*** 0.308** 0.273** 0.309**
(0.114) (0.123) (0.116) (0.126)
INF_E -3.461*** -3.446***
(1.160) (1.172)
RF(-1) -1.394 0.068 -1.856* -0.476
(1.225) (1.228) (1.126) (1.163)
TS(-1) 0.169 0.290 -0.509 -0.104
(2.098) (2.014) (1.935) (1.932)
DS(-1) 0.006 0.006 0.012** 0.012**
(0.007) (0.007) (0.005) (0.005)
D*DY(-1) 1.330*** 0.984* 1.417*** 1.396***
(0.446) (0.543) (0.449) (0.432)
D*INF_E 8.369*** 5.398**
(3.312) (2.540)
D*RF(-1) -4.081 -5.955** -2.671 -5.038
(3.091) (3.023) (1.760) (1.969)
D*TS(-1) –2.014 3.839
(5.607) (7.177)
D*DS(-1) 0.014 0.016
(0.011) (0.010)
D -0.061*** -0.085*** -0.059*** -0.069***
(0.021) (0.023) (0.019) (0.017)
Constant -0.005 -0.002 -0.007 -0.006
(0.006) (0.006) (0.006) (0.006)
Adj. R
2
0.031 0.040 0.031 0.038
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
97
Table A.7: Return Regression Results for VWRETD with DYMA
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETD and 3MTB
VWRETD
3MTB
1 2 3456 7
DYMA(-1) 0.311** 0.389*** 0.460*** 0.440*** 0.427*** 0.473*** 0.479*
(0.130) (0.129) (0.137) (0.138) (0.136) (0.143) (0.268)
RF(-1) -1.825*** -0.752
(0.680) (0.757)
Constant -0.004 0.001 0.0004 -0.001 -0.002 0.001 0.004
(0.004) (0.005) (0.005) (0.005) (0.004) (0.005) (0.006)
INF. M. INF INF INF(-1) INF_E INF_M
-1.879*** -2.072*** -1.722*** -3.118*** -3.448*
(0.522) (0.497) (0.434) (0.788) (1.787)
Adj. R
2
0.007 0.015 0.034 0.034 0.025 0.027 0.006
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 7: January 1978 to August 2008
Table A.8: Return Regression Results for VWRETD with DYMA cont.’d
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETD and 3MTB
VWRETD
3MTB
8 9 10 11 12 13 14
DYMA(-1) 0.331*** 0.349*** 0.427*** 0.415*** 0.407*** 0.426*** 1.817***
(0.125) (0.127) (0.135) (0.135) (0.132) (0.139) (0.626)
RF(-1) -2.811*** -1.997** -1.937** -2.058** -1.229 -6.972***
(0.987) (0.966) (0.976) (0.973) (1.116) (2.347)
TS(-1) 4.000** 0.829 -0.400 -0.094 0.305 0.012 -10.148**
(1.812) (1.885) (1.993) (1.882) (1.888) (1.909) (4.618)
DS(-1) 0.0001 0.011* 0.012** 0.012** 0.012** 0.011** 0.003
(0.005) (0.006) (0.006) (0.006) (0.006) (0.006) (0.008)
Constant -0.009 -0.005 -0.005 -0.005 -0.005 -0.004 0.024**
(0.006) (0.006) (0.005) (0.005) (0.006) (0.006) (0.011)
INF. M. INF INF INF(-1) INF_E INF_M
-1.450** -1.833*** -1.400*** -2.882*** -7.253**
(0.666) (0.519) (0.438) (1.028) (3.175)
INF
2
-47.180
(79.870)
Adj. R
2
0.012 0.023 0.039 0.040 0.032 0.031 0.025
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 14: January 1978 to August 2008
98
Table A.9: Return Regression Results for VWRETD with Dummy and DYMA
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from VWRETD and 3MTB
VWRETD
3MTB
15 16 17 18
DY(-1) 0.336** 0.370** 0.336** 0.370**
(0.136) (0.145) (0.136) (0.146)
INF_E -3.611*** -3.617***
(1.203) (1.213)
RF(-1) -1.927 -0.452 -2.190* -0.775
(1.246) (1.239) (1.141) (1.172)
TS(-1) -0.199 -0.112 -0.665 -0.278
(2.162) (2.069) (2.008) (1.993)
DS(-1) 0.008 0.009 0.012** 0.013**
(0.007) (0.007) (0.005) (0.005)
D*DY(-1) 1.996*** 1.651** 2.049*** 2.017***
(0.506) (0.643) (0.533) (0.518)
D*INF_E 7.683** 5.529**
(3.595) (2.807)
D*RF(-1) -4.392 -6.167** -3.444* -5.823***
(3.188) (3.119) (1.804) (2.083)
D*TS(-1) -1.616 3.209
(5.988) (7.513)
D*DS(-1) 0.008 0.010
(0.010) (0.010)
D -0.073*** -0.092*** -0.073*** -0.082***
(0.023) (0.024) (0.021) (0.018)
Constant -0.003 -0.0001 -0.004 -0.002
(0.006) (0.006) (0.006) (0.006)
Adj. R
2
0.038 0.047 0.040 0.048
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
99
Table A.10: Return Regression Results for SP500RET with DYMA
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from SP500RET and 3MTB
SP500RET
3MTB
1 2 3456 7
DYMA(-1) 0.225* 0.302** 0.369*** 0.348*** 0.339*** 0.382*** 0.412*
(0.125) (0.123) (0.130) (0.130) (0.129) (0.136) (0.250)
RF(-1) -1.801*** -0.791
(0.653) (0.721)
Constant -0.004 0.001 0.0002 -0.002 -0.002 0.001 0.004
(0.004) (0.005) (0.005) (0.004) (0.004) (0.004) (0.005)
INF. M. INF INF INF(-1) INF_E INF_M
-1.768*** -1.972*** -1.683*** -3.055*** -3.909**
(0.508) (0.481) (0.416) (0.737) (1.669)
Adj. R
2
0.003 0.012 0.029 0.029 0.021 0.023 0.006
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 7: January 1978 to August 2008
Table A.11: Return Regression Results for SP500RET with DYMA cont.’d
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from SP500RET and 3MTB
SP500RET
3MTB
8 9 10 11 12 13 14
DYMA(-1) 0.255** 0.270** 0.348*** 0.333*** 0.327*** 0.344*** 1.671***
(0.122) (0.123) (0.129) (0.129) (0.127) (0.134) (0.619)
RF(-1) -2.536*** -1.796* -1.716* -1.802* -0.975 -6.109***
(0.960) (0.938) (0.950) (0.947) (1.086) (2.317)
TS(-1) 3.956** 1.095 -0.178 0.229 0.584 0.298 -9.743**
(1.719) (1.815) (1.894) (1.790) (1.805) (1.826) (4.588)
DS(-1) -0.001 0.009 0.010* 0.010* 0.009* 0.009* 0.001
(0.004) (0.006) (0.005) (0.005) (0.006) (0.006) (0.007)
Constant -0.008 -0.005 -0.005 -0.004 -0.005 -0.004 0.025**
(0.006) (0.006) (0.006) (0.005) (0.006) (0.006) (0.011)
INF. M. INF INF INF(-1) INF_E INF_M
-1.211* -1.720*** -1.364*** -2.828*** -7.837**
(0.690) (0.507) (0.420) (0.972) (3.163)
INF
2
-62.790
(80.888)
Adj. R
2
0.008 0.017 0.032 0.033 0.026 0.025 0.022
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Sample for regression 14: January 1978 to August 2008
100
Table A.12: Return Regression Results for SP500RET with Dummy and DYMA
Sample: January 1950 to August 2008
Dep. Var. Excess Return formed from SP500RET and 3MTB
SP500RET
3MTB
15 16 17 18
DYMA(-1) 0.289** 0.318** 0.289** 0.318**
(0.130) (0.138) (0.132) (0.140)
INF_E -3.309*** -3.316***
(1.149) (1.161)
RF(-1) -1.421 -0.063 -1.811 -0.509
(1.208) (1.198) (1.113) (1.930)
TS(-1) 0.150 0.234 -0.433 -0.069
(2.098) (2.015) (1.939) (1.930)
DS(-1) 0.006 0.006 0.011** 0.011**
(0.007) (0.007) (0.005) (0.005)
D*DYMA(-1) 1.636*** 1.262** 1.751*** 1.724***
(0.491) (0.620) (0.492) (0.482)
D*INF_E 7.832** 5.106**
(3.332) (2.561)
D*RF(-1) -4.300 -5.993** -3.154* -5.357***
(3.124) (3.052) (1.764) (1.938)
D*TS(-1) -1.712 3.659
(5.658) (7.175)
D*DS(-1) 0.012 0.014
(0.010) (0.010)
D -0.065*** -0.087*** -0.064*** -0.073***
(0.021) (0.023) (0.019) (0.017)
Constant -0.004 -0.001 -0.006 -0.004
(0.006) (0.007) (0.006) (0.006)
Adj. R
2
0.032 0.040 0.033 0.039
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
101
Appendix B
Supplementary Materials for
Chapter 2
-20
-10
0
10
20
30
40
50 55 60 65 70 75 80 85 90 95 00 05 10
Figure B.1: Canada Annualized Monthly Inflation in Percentage
102
-40
-20
0
20
40
60
80
50 55 60 65 70 75 80 85 90 95 00 05 10
Figure B.2: France Annualized Monthly Inflation in Percentage
103
-40
-30
-20
-10
0
10
20
30
40
50
50 55 60 65 70 75 80 85 90 95 00 05 10
Figure B.3: Germany Annualized Monthly Inflation in Percentage
104
-30
-20
-10
0
10
20
30
40
50 55 60 65 70 75 80 85 90 95 00 05 10
Figure B.4: Italy Annualized Monthly Inflation in Percentage
105
-80
-60
-40
-20
0
20
40
60
50 55 60 65 70 75 80 85 90 95 00 05 10
Figure B.5: Japan Annualized Monthly Inflation in Percentage
106
-30
-20
-10
0
10
20
30
40
50
60
50 55 60 65 70 75 80 85 90 95 00 05 10
Figure B.6: UK Annualized Monthly Inflation in Percentage
107
-30
-20
-10
0
10
20
30
50 55 60 65 70 75 80 85 90 95 00 05 10
Figure B.7: US Annualized Monthly Inflation in Percentage
108
Appendix C
Supplementary Materials for
Chapter 3
C.1 Data Set Preparation
In this study, we have used one macro data set and two different micro data sets,
one of which is the Company Accounts Survey conducted by the Central Bank of the
Republic of Turkey (CBRT) and the other is a firm level data set for the Istanbul Stock
Exchange (ISE) firms. For all these data sets, we defined the tradable and non-tradable
sectors as the manufacturing sector and the non-manufacturing sector, respectively.
We used the production price index (PPI)
1
to control for inflation.
Our macro data consists of real GDP , total credit in the economy, aggregate con-
sumption and aggregate investment of Turkey between the years of 1970 and 2004.
Aggregate production and relative price ratios between tradable and non-tradable sec-
tors are also included in our macro data. We use these variables to show the boom-bust
cycles in the Turkish economy and their relation to credit markets. Also, by using the
1
PPI is taken from CBRT website www.tcmb.gov.tr .
109
Our main data set for this study, is the Company Accounts Survey
3
, which has
been gathered by the CBRT since 1992. The CBRT collected this data set each year by
means of asking thousands of firms to bring their balance sheets for the preceding three
years. For example, firms in 1997 survey data set
4
are asked to provide their balance
sheets for 1994, 1995 and 1996. That’s why our main data set covers the years between
1989 and 2006. The publicly available part of the data set gives us the aggregate balance
sheet for each sector. We divided this data set into two main sectors as well, tradable
and non-tradable. Both tradable and non-tradable sectors have further information
available for their sub-sectors. The names of the sub-sectors used in this study can be
found in Table C.2 of Appendix in section C.2.
We use our main data set in two different ways to run our regressions. As men-
tioned earlier, we have the data of the preceding three years for every year the survey
is conducted and this actually provides us with three different observations of a vari-
able for each period, each coming form three different surveys. Suppose, we have a
variable
!
for sector 2 in year $ from the survey conducted in year . Then, we can
easily see that $$
$ or $
and $. The
first way we use our Company Accounts survey data is that, we pick the last observa-
tion from each survey and if a lagged variable is needed in our regression we take it
from the same survey year. In other words, we use
!
and when we need one lag
of this variable, we use
!
not
!
. This provides us within survey consis-
tency when using our data. Secondly, we take a different strategy to form continuous
3
This data set is publicly available in the CBRT website under Periodic Publications.
4
See Table 1C of Appendix C in section C.2.
110
variables from our data set in order to understand the dynamic properties of the data
in a better way. For this purpose we have formed weighted average variables, i.e. we
used the average of three observations for a period, coming from three different sur-
veys. Let’s assume that we have a variable
!
and the number of the firms ,
'
!
,
for sector 2 in year $ from the survey conducted in year . We calculated the weighted
average value of our variable as,
!
!
'
!
!
'
!
.
We say weighted average because our observations not only differ in the value but also
in the sample size as well, i.e. number of the firms in our sample changes from year
to year
5
, and we don’t have any information about the firms entering or leaving the
Company Accounts Survey due to the survival or selection. In other words to say, we
have
'
!
'
!
'
!
but
'
!
'
!
for 7
. A firm taking
place in all $$
$ surveys have triple the weight of a firm that take place in
only one of those surveys for the year $ observation. That’s why we form our new data
set by taking the weighted average of each variable and name it as weighted average
data set. We will discuss the reasons behind forming this data set in detail in section
3.4.2.
5
You may find the number of firms for each subsector in CBRT Company Accounts survey data set in
Table 3C and 4C of Appendix C in section C.2 for tradable and non-trdable sectors, respectively.
111
The ISE firm level data set between the years of 1987 and 2007 is the third data set
used in this study. This data set is bought from a private data management company,
Finnet
6
, and it is used to test our findings among the large firms of the ISE in a firm
level analysis. There are 175 firms used from this data set. We also divide them in to
two groups, namely tradable and non-tradable, depending on their sectors
7
.
We have used Company Accounts Survey data set of the CBRT and the ISE firms
data set to form investment, capital stock, cash flow and sales variables. Detailed
description of the data will be given in section 3.4.2.
We run our regressions with sector level averages for Company Accounts Survey
data set of the CBRT due to data availability. Moreover, as we mentioned in section C.1,
we have three different observations of a variable for each year coming form three dif-
ferent surveys. As a result, we define our variables not only according to a sector 2 and
year $ but also according to survey year , and use our data set in two different ways.
First, we use the last observation from each survey, e.g.
)1
!
, and for the lagged
variables we do not change the survey year but the observation year, e.g.
3
!
or
3
!
8
. The reason behind this is that the size of the survey samples change from
year to year and it is more logical to use the same samples’ lagged variable to form our
variables that are going to be used in the regression analysis, e.g.
)1
!
3
!
.
So, our model for the Company Accounts Survey data can be written as follows:
6
Wed address for Finnet is www.finnet.com.tr .
7
For ISE firm level data set, you may find the names of the sectors and the number of firms in each
sector in Table 5A of appendix A.
8
Remember that
.
112
4
!
3
!
4
!
3
!
4
!
3
!
!
3
!
(C.1)
!
3
!
9
!
3
!
9
!
3
!
!
:
!
Although we have the lagged observation for the variables, they are not the lag of
the variables as it is seen in Equation C.2.
-+,
4
!
3
!
4
!
3
!
4
!
3
!
(C.2)
To deal with this problem and see the dynamic properties of the model in a better way,
we formed the weighted average data mentioned above. Then, our model used for this
data set can be written as follows,
4
!
3
!
4
!
3
!
4
!
3
!
!
3
!
!
3
!
(C.3)
9
!
3
!
9
!
3
!
!
:
!
and hence the model in Equation D.13 has the desired dynamic properties. The prob-
lem of this data set is that it gives more weight to the firms that report more frequently
during the process of the preparation of the variables needed for the regression.
113
Our third data set is the firm level data for ISE firms. We can also control for the
size of the firm with our <=, which is defined as the ratio of the real sales of a firm
to median real sales of the sector it belongs to. This variable is calculated for each year
and each firm, which means the value changes over time depending on the relative
size of the firm in its sector. We also have the !> , leverage to debt ratio, to control
for the level of indebtedness of a firm. As a result our model for the ISE firms can be
written as the following,
4
3
4
3
4
3
3
3
(C.4)
9
3
9
3
<=
!>
:
C.2 Additional Data Information and Regressions
114
Table C.1: Survey and Observation Relation
CBRT Company Accounts Data Set
Survey Year
Year ... 1996 1997 1998 1999 2000 2001 2002 2003 ...
...
1992 ...
1993 ... X
1994 ... X X
1995 XXX
1996 XXX
1997 XXX
1998 XXX
1999 XXX
2000 XXX
2001 X X ...
2002 X ...
... ...
This table shows us how the Company Accounts survey of CBRT is collected.
115
Table C.2: Sector Names
CBRT Company Accounts Survey Data Set
Tradable Sector
Sector Code Sector Name
DA-15 Manufacture of Food Products & Beverages
DA-16 Manufacture of Tobacco
DB-17 Manufacture of Textiles
DB-18 Manufacture of Wearing Apparel, Dressing and Dyeing of Fur
DC Manufacture of Leather and Leather Products
DD Manufacture of Wood and Wood Products
DE Manufacture of Pulp, Paper & Paper Products, Publ. & Printing
DG Manufacture of Chemicals, Chemical Prod. & Man-made Fibres
DH Manufacture of Rubber and Plastic Products
DI Manufacture of other Non-metallic Mineral Products
DJ Manufacture of Basic Metals and Fabricated Metal Products
DK Manufacture of Machinery and Equipment N.E.C.
DL Manufacture of Electrical and Optical Equipment
DM Manufacture of Transport Equipment
DN Manufacture of Furniture, Manufacturing N.E.C.
Non-Tradable Sector
Sector Code Sector Name
F-451 Site Preparation
F-4520 Building of Compl. Constr. or Parts Thereof, Civil Engineering
F-4521 General Constr. of Buildings and Civil Engineering Works
F-453 Building Installation
F-454 Building Completion
G Wholesale and Retail Trade
H Hotels and Restaurants
I Transport, Storage and Communication
116
Table C.3: Number of Firms in Tradable (Manufacturing) Sectors
CBRT Company Accounts Survey Data Set
Survey Year
Sector 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
DA-15 546 665 670 691 700 689 581 599 612 615 680 628
DA-16 21 22 20 22 23 22 21 19 14 14 14 13
DB-17 505 644 583 641 649 648 509 538 507 514 549 580 576 615 663
DB-18 419 556 428 513 531 434 308 310 307 318 345 351 329 321 298
DC 52 72 76 87 93 102 84 84 79 77 81 82 71 73 72
DD 101 144 144 148 169 151 104 101 104 106 109 96 71 76 83
DE 132 161 165 176 182 172 129 138 144 154 158 150 132 133 18
DG 112 238 282 282 304 311 261 259 262 263 281 289 255 216 249 260
DH 140 166 184 184 217 233 181 184 197 210 222 203 195 209 224
DI 202 238 247 257 275 258 229 237 240 256 265 250 224 241 277
DJ 149 343 416 417 469 476 422 315 320 329 337 366 335 298 378 403
DK 119 214 266 268 300 319 293 241 248 248 253 273 269 248 285 287
DL 162 200 201 204 225 207 160 166 165 166 181 176 151 155 174
DM 142 174 185 196 231 205 177 189 193 213 214 210 207 224 250
DN 55 63 65 90 129 111 89 91 109 124 135 136 126 122 134
Total 377 3272 4069 3935 4282 4530 4208 3387 3486 3511 3638 3881 3734 2838 3081 3143
Table C.4: Number of Firms in Non-tradable (Non-manufacturing) Sectors
CBRT Company Accounts Survey Data Set
Survey Year
Sector 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
F-451 364 69 66 103 112 110 105 79 67 57 49
F-4520 388 609 440 274 308 330 329 328 283 324 453
F-4521 364 269 256 177 159 138 129 118 98 70 1504
F-453 81 215 99 83 79 83 79 77 76 73 64 66
F-454 262 395 334 330 344 341 338 294 226 167 145 127
G 792 2259 2497 2245 2701 2797 2170 1618 1567 1436 1482 1522 1460 1471 1504 1527
H 145 343 455 463 746 480 332 289 293 282 296 316 273 220 266 272
I 125 323 450 459 471 542 427 310 318 332 348 394 319 307 357 361
Total 1062 2925 3402 3167 5377 5307 4127 3147 3171 3054 3111 3154 2814 2699 2846 2402
117
Table C.5: Sectors and Number of Firms in each Sector
ISE Firms Data Set
Tradable Sector
Sector Number of Firms
Food and Beverages 23
Textile 25
Furniture 2
Paper 14
Chemicals 20
Metal 25
Metal Products 15
Other Manufacturing 1
Total 125
Non-tradable Sector
Sector Number of Firms
Rocks and Sand 25
Electricity 3
Construction 2
Whole Sale 3
Retail Sale 7
Hotels and Restaurants 5
Transportation 4
Communication 1
Total 50
118
Table C.6: Regression Results for CBRT Data Set
CBRT Company Accounts Survey Data Set
Dep Var. Tradable Non-tradable
I/K(-1) FE FE
Variable 1 2 1 2
I(-1)/K(-2) 0.099 0.178 0.263 0.032
0.148 0.152 0.196 0.172
I(-1)/K(-2)
2
0.225 0.190 -0.167* -0.092
0.245 0.234 0.098 0.077
S/K(-1) 0.016 -0.040
0.010 0.026
S(-1)/K(-2) 0.019** 0.003 -0.013 0.010
0.009 0.011 0.011 0.014
CF/K(-1) 0.049* 0.310***
0.026 0.116
CF(-1)/K(-2) -0.019 -0.029 0.059 0.134**
0.028 0.024 0.070 0.067
Constant -0.007 -0.016 -0.413
0.059 0.056 0.518
# of Obs. 222 222 103 103
R
2
0.475 0.497 0.429 0.428
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Random effects results are very similar and Hausmann test prefers FE results.
119
Table C.7: Pooled Regression Results for CBRT Data Set
CBRT Company Accounts Survey Data Set
Dep Var.
I/K(-1) 1 2
Variable FE RE FE RE
I(-1)/K(-2) 0.268** 0.275*** 0.446*** 0.465***
0.113 0.101 0.106 0.099
I(-1)/K(-2)
2
-0.153** -0.102** -0.238*** -0.201***
0.064 0.044 0.056 0.046
S/K(-1) 0.028*** 0.026***
0.005 0.005
S(-1)/K(-2) -0.007 -0.005 -0.020*** -0.027***
0.011 0.004 0.007 0.005
CF/K(-1) 0.043 0.056*
0.030 0.033
CF(-1)/K(-2) 0.001 0.014 -0.006 0.017
0.029 0.024 0.029 0.025
D*CF/K(-1) 0.223*** 0.192***
0.078 0.074
D*CF(-1)/K(-2) 0.044 0.069* 0.105*** 0.045
0.044 0.041 0.035 0.035
Constant 0.124 0.065 0.035
0.086 0.049 0.033
# of Obs. 325 325 325 325
R
2
0.309 0.323 0.526 0.611
Hausman (p-value) 1.000
1.000
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
Generalized Hausman Test is used instead of Hausman Test.
120
Appendix D
Another Model for Chapter 3
D.1 Accelerator Approach
Following the model of Hubbard (1988), we assume that a firm wants to maximize
the expected value of its stream of profits from investment after subtracting the adjust-
ment costs of investment,
> 3
%
!
!
3
!
!
)4
!
3
!
!
4
!
&
(D.1)
subject to the capital accumulation constraint,
3
!
53
!
4
!
(D.2)
where and $ denote the firm and time period, respectively. 3
is the beginning of the
period capital stock, 4
is investment, and )
is the adjustment cost of investment for
firm at time $. Also, is the profit function,
is the firm specific discount rate,
is
the relative price of capital good at time $, and 5 is the constant rate of depreciation.
;
is the expectations conditioned on the information set available to firm at
time $. First order condition of this maximization problem yields:
121
)
4
3
?
(D.3)
where
?
%
!
!
5
"
3
!
!
)
"
3
!
&
. (D.4)
Equation D.4 is similar to the marginal ? in Hayashi (1982) which is defined as the
present discounted value of profits from new investment. If we assume the following
quadratic form for the cost function of firm ,
)4
3
4
3
+
3
(D.5)
then the first order condition in Equation D.3 turns out to be:
4
3
+
?
&
(D.6)
where the last term in Equation D.6, &
, is a maximization error. So far, a firm maxi-
mizes its profit stream form investment and marginal ? is the main determinant of the
investment and hence it will control the investment opportunities for the firm. On the
other hand, it is really difficult to calculate ?. Hubbard (1988) provided that average
@ constructed from financial market data can be used as a proxy for the difference
between marginal ? and , the price of investment good, if the the following assump-
tions hold; perfect competition in the factor and product markets, homogeneity of
fixed capital, linear homogeneity of technologies for production and adjustment costs,
122
and independence of financing and investment decisions. Then the relation between
investment and average @ can be written as the following equation where .
4
3
+
@
&
(D.7)
Our aim in this part of the study is to examine the sector based asymmetries in
terms of capital market imperfections and determine the relation between these asym-
metries and boom-bust cycles via accelerator model. To understand whether the inter-
nal funds of firms matter for investment decision or not, we can start with Equation
D.7, which specifies a model without capital market imperfections, i.e. frictionless
capital markets. Suppose, due to credit market imperfections, firms have some kind
of credit constraints or high information costs or the cost of external financing exceeds
that of internal one, then changes in net worth would affect the investment decision
of such firms according to Fazzari, Hubbard, and Petersen (1998) and Hubbard (1988).
Hence, if one regresses 43 on @ for these firms, then one should expect the residuals
of this regression to be correlated with changes in net worth. This kind of a correla-
tion would reject the frictionless @ model presented in Equation D.7. Then, we have
to introduce change in net worth term in our regression and to do that we are going to
use cash flow as a proxy for the changes in net worth. We define the cash flow of a firm
as the following,
)1 Net Profit+Owners Capital+Depreciation Allowances.
123
Although it is an imperfect one, most studies used cash flow because for many firms
it is the only measure available that can be used as a proxy to changes in net worth.
We also have the end of the year capital stock data not the beginning of the year in our
micro level data sets, that’s why we are going to use 3
to be consistent with our
model’s notation. As a result, we can write our cash flow equation as the following,
4
3
+
@
)1
3
&
(D.8)
where )1
is the cash-flow of firm at time $ and @
is equal to ?
. Equation
D.8 is called the cash-flow regression which is widely used in finance literature. We
use it to examine the sector based asymmetries between the non-tradable and trad-
able sectors and their implication on credit market imperfections. Also, by using cash
flow regressions, we depict the relation between the boom-bust cycles and sector based
asymmetries.
Let’s define our reasoning behind using the cash-flow regression in a better way. If
we have the null hypothesis of efficient financial markets and the set up of our model
in Equation D.8, then under Hayashi (1982)’s assumptions @ is a sufficient explanatory
variable for investment decision. Cash flow should not have any effect on investment
once the profitability of the firm is controlled for. In other words to say, when we
regress Equation D.8, we anticipate an insignificant coefficient, , for )13 to support
frictionless capital markets. A significantly positive results in rejecting the frictionless
model and suggests the presence of financial constraints and capital market imperfec-
tions.
124
We run our regressions with sector level averages for Company Accounts survey
data set of the CBRT due to data availability. This gives us whether a sector in our study
is financially constrained or not depending on the coefficient of )13. If different
sectors have different coefficients for cash flow term then the level of the constraint
is different for them, and if the size of the difference between them changes with the
boom-bust cycles then we can also say the imperfections change with booms and busts
of the economy. Also because we have sector level average data and most of these
firms are not publicly available, there is no way of computing @ for this data set. One
of the proxies that is used for the profitability of the firm, or sector in this case, in the
literature is the changes in sales over capital stock ratio, and we are going to use it as
well. Moreover, as we mentioned in section C.1, we have three different observations
of a variable for each year coming form three different surveys. As a result, we defined
our variables not only according to a sector 2 and year $ but also according to survey
year , and used our data set in two different ways. First, we use the last observation
from each survey, e.g.
)1
!
, and for the lagged variables we did not change the
survey year but the observation year, e.g.
3
!
or
3
!
1
. The reason behind
this is that the size of the survey samples change from year to year and it is more
logical to use the same samples lagged variable to form our variables that are going
to be used in the regression analysis, e.g.
)1
!
3
!
. So, our model for the
Company Accounts survey data can be written as the following,
1
Remember that
.
125
4
!
3
!
+
)
!
3
!
)1
!
3
!
&
!
. (D.9)
Though this way of forming our data is reasonable, it has some caveats. When we want
to include a lagged variable of our regression model, dependent or independent, that
is going to appear like another explanatory variable and its not going to give us any
dynamic effect at all, e.g. in Equation D.10 although we have the lagged observation
for the dependent variable, it is not the lag of the dependent variable as it is seen in
Equation D.11.
4
!
3
!
+
)
!
3
!
)1
!
3
!
4
!
3
!
*
)
!
3
!
&
!
(D.10)
-+,
4
!
3
!
4
!
3
!
4
!
3
!
(D.11)
To deal with this problem and to see the dynamic properties of the model in a better
way, we formed the weighted average data mentioned in section C.1. Then, our model
used for this data set can be written as follows,
4
!
3
!
+
)
!
3
!
)1
!
3
!
&
!
. (D.12)
If we include lagged terms it will be as the following,
4
!
3
!
+
)
!
3
!
)1
!
3
!
4
!
3
!
*
)
!
3
!
&
!
(D.13)
126
and hence the model in Equation D.13 has the desired dynamic properties. The prob-
lem of this data set is that it gives more weight to the more occurring firms during the
process of the preparation of the variables needed for the regression.
Finally we can also compute the Market to Book Ratio, =(, which gives us the
profitability of the firm, for the firms in ISE data set since all these firms are publicly
available. We also include <= and !> to our model. As a result our model for the
ISE firms for the accelerator approach can be written as the following,
4
3
+
=(
)1
3
)
3
*<=
A!>
&
. (D.14)
We also used the lagged independent and dependent variables in our mode for the ISE
firm data to see the dynamic properties of our regression.
For our analysis of the micro data sets with accelerator approach, first we run the
cash-flow regressions for tradable and non-tradable sectors separately. Secondly, we
pooled the data and introduce a dummy variable that takes the values of for non-
tradable sectors and for tradable sectors. As we explained above, we introduced
the lagged dependent and independent variables to see the dynamic properties of the
model. We used Fixed Effects and Random Effects to run our cash-flow regressions
and provide the p-values for Hausman test of specification as needed. You may find
the definition of the variables used in our regression in Table D.1.
Table D.2 runs the separate cash flow regressions for both tradable and non-
tradable sectors. Cash flow seems to be an important determinant of investment for
127
Table D.1: Variable Definitions and Data Source for the Micro Data Sets
Variable Definition
=(
Market to book ratio*
<=
Sales over median sales ratio
real sales/median real sales
!>
Firm leverage*
Total Borrowing/Total Asset
*MTBR and LEV are directly from ISE data set, which has been bought from Finnet.
firms in both sectors; it is always significant at one percent level. This tells that nei-
ther tradable nor non-tradable sectors have access to perfect capital markets and are
constrained by their internal funds. Then as we compare the coefficient for sectors
we see that non-tradable sector is always more constrained than tradable sectors. The
number for non-tradable sector is as large as three times of the counterpart in trad-
able sector. To check the significance of this difference Table D.3 pools the sector data
together and runs fixed and random effects regressions. We conduct a test and get that
the coefficient for non-tradable is always significantly bigger than tradable coefficient.
The interaction term between cash flow coefficient and non-tradable sector dummy is
always significant at one percent, validating the difference of constraints.
We also conduct the same analysis with the extended data set for survey of sectors
in Tables D.4 and D.5. Again for separate regressions and for the pooled regression, we
get the result that both sectors are financially constrained and the non-tradable sector
has a significantly larger constraint than tradable sector.
128
From these results we can see that in the economy, non-tradable sectors are always
financially constrained and their responsiveness is higher. Tradable sectors are clas-
sified as constrained with respect to fixed investment but tradable firms are less con-
strained than non-tradable firms. This result is important for aggregate investment
dynamics as well as business cycles, because most of the cash flow studies in the lit-
erature use either stock market data, which is not a representative sample of the over-
all economy, or firm level data from manufacturing sector. For example, Gelos and
Werner (2002) consider the investment behavior for manufacturing firms in Mexico for
the period 1984-1994. They find that firms are financially constrained, but this result
would not be enough to infer about the aggregate investment and output dynamics.
As stated by Hubbard (1988), estimating investment equations for other sectors of the
economy would provide helpful for aggregate investment dynamics. In the Turkish
case, we have estimated the investment equations for non-tradable and tradable sectors
of the economy, and gotten micro evidence for the aggregate investment and output
dynamics. In our estimation we have a higher sensitivity of non-tradable sector. This
implies that non-tradable firms will be more responsive to external finance or bank
credit. So when there is a lending boom, non-tradable firms will be investing and pro-
ducing more, and when credit decreases non-tradable sector will decrease investment
and output more than tradable sector. Then this framework presents a micro evidence
for the macro correlations between NT/T ratio and Credit/GDP ratio, and establishes
the link between credit markets and output dynamics.
129
We also conduct cash-flow regression analysis on stock market firms. Our period
covers 1987 to 2007. We exclude financial and utility firms in accordance with the liter-
ature. Again we categorize our firms into non-tradable and tradable sectors. In Table
D.6 we first check the summary statistics for the additional variables added to the ISE
firm level data set. Since these are the firms in the stock market we also use Market-
to-Book value as another control variable for investment opportunities. Because we do
not have enough detail to estimate better value for Tobin’s Q ourselves, using replace-
ment costs or some other methodology, we take reported market-to-book values. We
estimate both fixed effects and random effects regressions with Hausman test compar-
ing the both estimates. In all regressions, fixed effects is chosen to be the efficient one,
so we only report the fixed effects regressions. We use both market to book ratio and
change in sales as measures of profitability. As in most other studies, market to book
ratio performs poorly as a proxy for Tobin’s Q. With different model specifications that
also controls for the size and leverage of firms, we always get the coefficient of cash
flow.
Tables D.7 and D.8 provide several models of cash flow regressions for pooled data
of all firms. We estimate both fixed effects and random effects regressions with Haus-
man test comparing the both estimates. In all regressions, fixed effects is chosen to be
the efficient one, so we only report the fixed effects regressions. We use both market to
book ratio and change in sales as measures of profitability. As in most other studies,
market to book ratio performs poorly as a proxy for Tobin’s Q. With different model
130
specifications that also controls for the size and leverage of firms, we always get the
coefficient of cash flow.
Comparing these to the previous section for a larger and more representative data
set for economy, we conclude that using stock market firms can tell us whether they
are constrained or not, but it does not help with studying the big macro picture.
From these micro results, we can conclude that market imperfections are important
for the economy and there is an asymmetry across sectors that non-tradable sectors are
more constrained than tradable sector.
131
Table D.2: Regression Results with Time Fixed Effects
CBRT Company Accounts Data Set
Dep. Var. Tradable Sector (FE)
I/K(-1) 1 2 3 4
I(-1)/K(-2)
0.2458***
(0.0569)
0.2545***
(0.0576)
CS/K(-1)
0.0098
(0.0099)
0.0089
(0.0104)
0.0100
(0.0095)
0.0073
(0.0099)
CS(-1)/K(-2)
-0.0026
(0.0086)
-0.0082
(0.0083)
CF/K(-1)
0.0741***
(0.0221)
0.0750***
(0.0223)
0.0650***
(0.0212)
0.0674***
(0.0213)
Constant
0.0850
(0.0647)
0.0858
(0.0649)
0.0579
(0.0622)
0.0596
(0.0622)
# of Obs. 222 222 222 222
R
2
0.4631 0.4620 0.5257 0.5258
Dep. Var. Non-tradable Sector (FE)
I/K(-1) 1 2 3 4
I(-1)/K(-2)
0.0244
(0.1305)
0.0257
(0.1298)
CS/K(-1)
-0.0760***
(0.0253)
-0.0826***
(0.0256)
-0.0756***
(0.0255)
-0.0822***
(0.0259)
CS(-1)/K(-2)
-0.0340
(0.0252)
-0.0340
(0.0254)
CF/K(-1)
0.2292***
(0.0776)
0.2314***
(0.0772)
0.2292***
(0.0781)
0.2314***
(0.0777)
Constant
-0.5350***
(0.1721)
-0.4766***
(0.1766)
-0.5389***
(0.1744)
-0.4806***
(0.1789)
# of Obs. 103 103 103 103
R
2
0.5426 0.5640 0.5437 0.5651
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
3 is the end of the period capital stock and 4 is the investment.
is the net sales and) is the change in sales.
)1 is the sum of net profit, owners capital, and depreciation allowance.
(-1) refers to one lag and (-2) refers to two lags of the corresponding variable.
132
Table D.3: Pooled Regression Results with Time Fixed Effects
CBRT Company Accounts Data Set
Dep. Var. 1 2
I/K(-1) FE RE FE RE
CS/K(-1)
0.1338***
(0.0120)
-0.0479***
(0.0115)
-0.0466***
(0.0119)
-0.0498***
(0.0112)
CS(-1)/K(-2)
-0.0371***
(0.0098)
-0.0400**
(0.0089)
CF/K(-1)
0.0775***
(0.0330)
0.0788***
(0.0306)
0.0866***
(0.0324)
0.0893***
(0.0297)
D*CF/K(-1)
0.1608***
(0.0448)
0.1342***
(0.0347)
0.1587***
(0.0438)
0.1545***
(0.0338)
Constant
0.0963
0.0804)
0.0614
(0.0793)
0.0944
(0.0786)
0.3100***
(0.0627)
# of Obs. 325 325 325 325
R
2
0.3538 0.3589 0.3978 0.3989
Hausman Test 0.1824 1.0000
Dep. Var. 3 4
I/K(-1) FE RE FE RE
I(-1)/K(-2)
0.1338**
(0.0646)
0.1810***
(0.0611)
0.1604***
(0.0632)
0.2138***
(0.0593)
CS/K(-1)
-0.0388***
(0.0120)
-0.0479***
(0.0114)
-0.0469***
(0.0118)
-0.0493***
(0.0110)
CS(-1)/K(-2)
-0.0396***
(0.0098)
-0.0436***
(0.0088)
CF/K(-1)
0.0705**
(0.0330)
0.0706***
(0.0302)
0.0788**
(0.0322)
0.0803***
(0.0292)
D*CF/K(-1)
0.1723***
(0.0449)
0.1320***
(0.0339)
0.1723***
(0.0437)
0.1557***
(0.0330)
Constant
0.0774
(0.0804)
0.0378
(0.0787)
0.0716
(0.0783)
0.3185***
(0.0615)
# of Obs. 325 325 325 325
R
2
0.3688 0.3768 0.4205 0.4236
Hausman Test 0.0005 -
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
D is a dummy variable defined as 1 for non-tradable sector and 0 for tradable sector.
3 is the end of the period capital stock and 4 is the investment.
is the net sales and) is the change in sales.
)1 is the sum of net profit, owners capital, and depreciation allowance.
(-1) means one lag and (-2) means two lags of the corresponding variable.
p-value is provided for Hausman test. Significant p-value means FE is better than RE.
133
Table D.4: Regression Results with Time Fixed Effects
CBRT Company Accounts Weighted Average Data Set
Dep. Var. Tradable Sector (FE)
I/K(-1) 1 2 3 4
I(-1)/K(-2)
0.1120**
(0.0454)
0.1301***
(0.0467)
CS/K(-1)
0.0167***
(0.0056)
0.0173***
(0.0060)
0.0186***
(0.0058)
0.0165***
(0.0060)
CS(-1)/K(-2)
-0.0054
(0.0057)
-0.0093
(0.0058)
CF/K(-1)
0.1317***
(0.0139)
0.1337***
(0.0144)
0.1393***
(0.0144)
0.1387***
(0.0144)
Constant
0.0545*
(0.0287)
0.0410
(0.0289)
0.0379**
(0.0295)
0.0101
(0.0307)
# of Obs. 444 414 414 414
R
2
0.3814 0.3864 0.4059 0.4091
Dep. Var. Non-tradable Sector (FE)
I/K(-1) 1 2 3 4
I(-1)/K(-2)
-0.0284
(0.0756)
-0.0321
(0.0754)
CS/K(-1)
-0.0445***
(0.0162)
-0.0573***
(0.0179)
-0.0487***
(0.0178)
-0.0556***
(0.0184)
CS(-1)/K(-2)
-0.0237
(0.0169)
-0.0240
(0.0170)
CF/K(-1)
0.3088***
(0.0442)
0.3165***
(0.0470)
0.3134***
(0.0483)
0.3122***
(0.0482)
Constant
0.0268
(0.1128)
0.1178
(0.1201)
0.0599
(0.1168)
0.1102
(0.1218)
# of Obs. 206 188 188 188
R
2
0.3659 0.3859 0.3613 0.3838
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
3 is the end of the period capital stock and 4 is the investment.
is the net sales and) is the change in sales.
)1 is the sum of net profit, owners capital, and depreciation allowance.
(-1) refers to one lag and (-2) refers to two lags of the corresponding variable.
134
Table D.5: Pooled Regression Results with Time Fixed Effects
Table 5D: Pooled Regression Results with Time Fixed Effects
CBRT Company Accounts Weighted Average Data Set
Dep. Var. 1 2
I/K(-1) FE RE FE RE
CS/K(-1)
-0.0108*
(0.0065)
-0.0169***
(0.0063)
-0.0155**
(0.0071)
-0.0205***
(0.0066)
CS(-1)/K(-2)
-0.0112*
(0.0066)
-0.0174***
0.0062)
CF/K(-1)
0.1397***
(0.0196)
0.1398***
(0.0183)
0.1434***
(0.0203)
0.1397***
(0.0191)
D*CF/K(-1)
0.1680***
(0.0313)
0.1293***
(0.0246)
0.1590***
(0.0329)
0.1396***
(0.0260)
Constant
0.0426
(0.0555)
0.0241
(0.0551)
0.0404
(0.0560)
0.0995***
(0.0357)
# of Obs. 650 650 602 602
R
2
0.2943 0.2994 0.3028 0.3079
Hausman Test 0.0311 -
Dep. Var. 3 4
I/K(-1) FE RE FE RE
I(-1)/K(-2)
0.0131
(0.0396)
0.0821**
(0.0384)
0.0171
(0.0396)
0.0869**
(0.0381)
CS/K(-1)
-0.0131*
(0.0069)
-0.0212***
(0.0066)
-0.0159**
(0.0071)
–0.0227***
(0.0066)
CS(-1)/K(-2)
-0.0114*
(0.0067)
-0.0191***
(0.0061)
CF/K(-1)
0.1448***
(0.0204)
0.1446***
(0.0191)
0.1442***
(0.0204)
0.1427***
(0.0190)
D*CF/K(-1)
0.1606***
(0.0331)
0.1264***
(0.0251)
0.1600***
(0.0330)
0.1363***
(0.0251)
Constant
0.0367
(0.0564)
0.0049
(0.0561)
0.0378
(0.0563)
0.0756**
(0.0373)
# of Obs. 602 602 602 602
R
2
0.2939 0.3026 0.3049 0.3141
Hausman Test 0.0000 0.0000
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
D is a dummy variable defined as 1 for non-tradable sector and 0 for tradable sector.
3 is the end of the period capital stock and 4 is the investment.
is the net sales and) is the change in sales.
)1 is the sum of net profit, owners capital, and depreciation allowance.
(-1) means one lag and (-2) means two lags of the corresponding variable.
p-value is provided for Hausman test. Significant p-value means FE is better than RE.
135
Table D.6: Summary Statistics for Additional Variables
ISE Firms Data Set
Tradable Sector
Mean Std. Dev. Min Max
MTBR 2.4637 2.0930 0.4795 15.4121
SOM 4.5141 15.7881 0.0911 139.1361
LEV 55.1470 20.5555 15.8940 126.4945
Non-tradable Sector
Mean Std. Dev. Min Max
MTBR 3.4517 4.3412 0.6200 22.2195
SOM 2.4874 4.8363 0.0556 25.1094
LEV 50.5649 25.6188 16.1043 94.6748
=( is the market to book ratio and <= is the size over median size ratio.
!> is the firm leverage.
(-1) means one lag and (-2) means two lags of the corresponding variable.
136
Table D.7: Pooled Regression Results
ISE Firms Data Set
Dep. Var. FE
I/K(-1) 1 2 3 4 5 6 7
I(-1)/K(-2) -0.0107 0.0085 -0.0178 -0.0327 -0.0548**
(0.0238) (0.0233) (0.0240) (0.0253) (0.0252)
MTBR -0.0002 0.0002 -0.0008 -0.0007 -0.0004 -0.0009 -0.0045*
(0.0027) (0.0027) (0.0026) (0.0025) (0.0027) (0.0026) (0.0027)
CS/K(-1) 0.0006 0.0114 0.0064
(0.0080) (0.0084) (0.0084)
MTBR(-1) 0.0012 0.0009 0.0003 -0.0009
(0.0031) (0.0031) (0.0032) (0.0032)
CS(-1)/K(-2) 0.0128* 0.0134*
(0.0079) (0.0078)
CF/K(-1) 0.2202*** 0.2121*** 0.2144*** 0.2091*** 0.1949*** 0.2007*** 0.2489***
(0.0169) (0.0170) (0.0168) (0.0168) (0.0178) (0.0186) (0.0201)
D*CF/K(-1) -0.0123 -0.0370 -0.0297 -0.0394 -0.0034 -0.0262 -0.0451
(0.0323) (0.0326) (0.0329) (0.0329) (0.0350) (0.0382) (0.0378)
SOM -0.0001
(0.0027)
LEV 0.0067***
(0.0011)
SOM*
CF
K(-1)
-0.0025
(0.0016)
Constant -0.0935 0.0362 -0.0455 0.0203 -0.0429 0.0010 -0.3221***
(0.0830) (0.0552) (0.0821) (0.0589) (0.0692) (0.0605) (0.0574)
# of Obs. 1523 1467 1400 1361 1426 1276 1271
R
2
0.1468 0.1395 0.1586 0.1609 0.1330 0.1578 0.1800
Hausman T. 0.0002 0.0000 0.0040 0.0000 0.0000 0.0000 0.0000
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
D is a dummy variable defined as 1 for non-tradable sector and 0 for tradable sector.
3 is the end of the period capital stock and 4 is the investment.
is the net sales and) is the change in sales.
)1 is the sum of net profit, owners capital, and depreciation allowance.
=( is the market to book ratio and <= is the size over median size ratio.
!> is the firm leverage.
(-1) refers to one lag and (-2) refers to two lags of the corresponding variable.
RE results are not presented here because they are pretty similar to FE ones.
p-value is provided for Hausman test. Significant p-value means FE is better than RE.
137
Table D.8: Pooled Regression Results
ISE Firms Data Set: SOM10
Dep. Var. FE
I/K(-1) 1 2 3 4 5 6 7
I(-1)/K(-2) -0.0231 -0.0014 -0.0285 -0.0364 -0.0584**
(0.0251) (0.0245) (0.0252) (0.0263) (0.0261)
MTBR -0.0012 -0.0009 -0.0016 -0.0015 -0.0009 -0.0012 -0.0043
(0.0030) (0.0030) (0.0029) (0.0028) (0.0031) (0.0029) (0.0029)
CS/K(-1) -0.0001 0.0132 0.0131
(0.0087) (0.0091) (0.0093)
MTBR(-1) 0.0010 0.0003 0.0006 -0.0012
(0.0037) (0.0036) (0.0037) (0.0036)
CS(-1)/K(-2) 0.0169** 0.0208**
(0.0086) (0.0086)
CF/K(-1) 0.2289*** 0.2182*** 0.2248*** 0.2173*** 0.2035*** 0.2073*** 0.2731***
(0.0182) (0.0185) (0.0182) (0.0182) (0.0189) (0.0195) (0.0253)
D*CF/K(-1) 0.0003 -0.0253 -0.0137 -0.0232 -0.0116 -0.0288 -0.0513
(0.0344) (0.0354) (0.0361) (0.0360) (0.0365) (0.0397) (0.0397)
SOM -0.0425
(0.0292)
LEV 0.0069***
(0.0011)
SOM*
CF
K(-1)
-0.0146
(0.0096)
Constant -0.0663 0.0053 -0.0514 -0.0079 0.0046 0.0456 -0.2316***
(0.0708) (0.0617) (0.0767) (0.0648) (0.0612) (0.0516) (0.0855)
# of Obs. 1401 1349 1285 1250 1327 1189 1184
R
2
0.1545 0.1452 0.1686 0.1693 0.1385 0.1686 0.1789
Hausman T. 0.0081 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
* 10%, ** 5%, *** 1% Significance levels. Standard deviations are in parenthesis.
D is a dummy variable defined as 1 for non-tradable sector and 0 for tradable sector.
3 is the end of the period capital stock and 4 is the investment.
is the net sales and) is the change in sales.
)1 is the sum of net profit, owners capital, and depreciation allowance.
=( is the market to book ratio and <= is the size over median size ratio.
!> is the firm leverage.
(-1) refers to one lag and (-2) refers to two lags of the corresponding variable.
RE results are not presented here because they are pretty similar to FE ones.
p-value is provided for Hausman test. Significant p-value means FE is better than RE.
138
Abstract (if available)
Abstract
In this dissertation, firstly I model the relation between the stock market and inflation, and provide empirical evidence of my theory in the US equity market. The second chapter has international evidence of effects of inflation on the equity markets for the G7 countries. In the third chapter, I look at the effects of business cycle fluctuations on the equity market and explain them through the credit channel via borrowing constraints on different sectors in a developing country.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Essays on financial markets
PDF
Three essays on macro and labor finance
PDF
Which of stock returns or dividend growth is predictable? A defense of stock return predictability
PDF
Essays on delegated asset management in illiquid markets
PDF
Share repurchases: how important is market timing?
PDF
Two essays on major macroeconomic shocks in the Japanese economy: demographic shocks and financial shocks
PDF
Essays on political economy of privatization
PDF
Three essays in derivatives, trading and liquidity
PDF
Essays on delegated portfolio management under market imperfections
PDF
Emerging market financial crises, investors and monetary policy
PDF
Essays on interest rate determination in open economies
PDF
Three essays on heterogeneous responses to macroeconomic shocks
PDF
Essays in macroeconomics
PDF
Essays on consumer conversations in social media
PDF
Essays in financial intermediation
PDF
Two essays on the mutual fund industry and an application of the optimal risk allocation model in the real estate market
PDF
Essays on monetary policy and international spillovers
PDF
Housing market crisis and underwater homeowners
PDF
Risks, returns, and regulations in real estate markets
PDF
Four essays on how policy, the labor market, and age relate to subjective well-being
Asset Metadata
Creator
Günay, Hüseyin
(author)
Core Title
Essays on inflation, stock market and borrowing constraints
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Economics
Publication Date
08/07/2010
Defense Date
06/16/2010
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
boom-bust cycles,borrowing constraint,cash-flow regression,dividend yield,financial asymmetry,G7 countries,inflation,inflation heterogeneity,OAI-PMH Harvest,stock market,Turkey
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
İmrohoroǧlu, Selahattin (
committee chair
), Jones, Christopher S. (
committee member
), Konchitchki, Yaniv (
committee member
), Zapatero, Fernando (
committee member
)
Creator Email
gunay@usc.edu,gunayhus@yahoo.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-m3337
Unique identifier
UC1454678
Identifier
etd-Gunay-3986 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-378180 (legacy record id),usctheses-m3337 (legacy record id)
Legacy Identifier
etd-Gunay-3986.pdf
Dmrecord
378180
Document Type
Dissertation
Rights
Günay, Hüseyin
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Repository Name
Libraries, University of Southern California
Repository Location
Los Angeles, California
Repository Email
cisadmin@lib.usc.edu
Tags
boom-bust cycles
borrowing constraint
cash-flow regression
dividend yield
financial asymmetry
G7 countries
inflation
inflation heterogeneity
stock market