1002’

LIBRARY
Michigan State
University

This is to certify that the
dissertation entitled

Essays on Business Cycles in Emerging Economies

presented by

Tao Peng

has been accepted towards fulfillment
of the requirements for the

degree” In Economics

 

 

PhD/ha M0 W

Major Professor’s Signature

5) Show

Date

MSU is an affinnative-action, equal-opportunity employer

- 4..—.—.—--O-I-t-l-0-l-I-n-l—-¢-.-n-o-I-l-l-I-l-I-l-I-I-O-'-'-l-0-I-I-C-I.

4 —.-.—.-.—.-.-.— 4

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5I08 K.IPro;/Acc&Pres/ClRC/DateDue.mdd

ESSAYS ON BUSINESS CYCLES IN EMERGING ECONOMIES

By

Tao Peng

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY

Department of Economics

2008

ABSTRACT
ESSAYS ON BUSINESS CYCLES IN EMERGING ECONOMIES
By

Tao Peng

In the last several decades, business cycles were significant in emerging economies. An
important observation is that real interest rates are negatively, correlated with output,
while there is no such pattern in developed economies. This raises the question: Do real
interest rate shocks drive business cycles in emerging economies? Extant literature
modifies the standard small open economy RBC model by introducing a working capital
constraint and Shows that international interest rate shocks can drive business cycles in
emerging economies through their effects on the labor market. This thesis aims to Show
that financial constraints can play a role in transmitting interest rate shocks and
international interest rate shocks can drive business cycles in emerging economies
through their effects on the credit market.

Chapter 1: Real Interest Rates, Credit Constraints and Business Cycles in
Emerging Economies (with Raoul Minetti)

This essay studies how international real interest rate Shocks can drive business cycles in
emerging economies. We present evidence that, in emerging economies, real interest
rates and real estate prices are negatively correlated and real interest rates are
countercyclical. Motivated by this evidence, we develop a model of a small open
economy, where the entrepreneur is borrowing constrained and the domestic lenders is

lending constrained. We Show that a positive interest rate shock leads to a fall in the real

estate price and, through the credit constraints, to a decline in output. The model is
calibrated to Argentina data. Simulation results Show that our model can explain the
countercyclicality of real interest rates and the output pattern better than a model with the
borrowing constraint but without the lending constraint.

Chapter 2: Real Interest Rates and Business Cycles in Emerging Economies

This essay studies the relationship between real interest rates and business cycles in
emerging economies in a dynamic general equilibrium setting. I modify the standard
small open economy RBC model by introducing a net worth constraint. The net worth
constraint serves as a financial accelerator which can amplify and prolong the effects of
interest rate shocks. I calibrate the model economy to Argentina data. Quantitative results
Show that real interest rates are negatively correlated with output and this negative
relationship is stronger for a more constrained economy than for a less constrained
economy.

Chapter 3: The Determinants of Country Risk Spreads: Evidence from Argentina
This essay studies the determinants of the country risk spread by using Argentina’s
monthly data during the period January 1998 to December 2006. We find that the country
risk spread is not a perfect measure of the default risk. Our estimation results suggest that
although international investors tend to take into account the default risk. Other factors
which may not represent the existence of the default risk, such as contagion effect,
investors’ memory, IMF bailout expectations are also important in determining the

country risk spread.

ACKNOWLEDGMENTS

I am deeply indebted to my advisor, Raoul Minetti, for generous advice, inspiration, and
continuous support. In every stage of this dissertation, I benefited a lot from numerous

discussions with him.

I am also grateful to Rowena Pecchenino, Antonio Doblas-Madrid for several very
helpful discussions. I appreciate Luis Araujo, Long Chen, ‘Matteo Iacoviello, Thomas

Jeitschko, and Andrei Shevchenko for valuable comments.

Finally, I would like to thank Luke Feltham and Rong Liu for providing me with some

data and the Matlab code for interpolation.

TABLE OF CONTENTS

LIST OF TABLES
LIST OF FIGURES
CHAPTER 1
Real Interest Rates, Credit Constraints and Business Cycles in Emerging Economies
Introduction ......................................................................................................... 1
Related Literature ................................................................................................ 3
Stylized Facts ...................................................................................................... S
The Empirical Regularities of Real Interest Rates, Real Estate ............... -. ..... 6
Prices and Output in Emerging and Developed Economies
The Relationship between Bank Lending and Bank Capital in ..................... 10
Emerging Economies
The Model Economy ........................................................................................... 11
Households .............................................................................................. 12
Entrepreneurs ............................................................................................. 15
Discussion .................................................................................................. 19
Equilibrium .............................................................................................. 21
Real Interest Rates ................................................................................... 21
A Model without the Lending Constraint ......................................................... 23
Shock Processes and Parameterization ............................................................. 24
Shock Processes ........................................................................................ 24
Parameterization ...................................................................................... 24
Characterization of Equilibrium and Results .................................................... 28
The Transmission Mechanism of Interest Rate Shocks ............................. 28
Impulse Response Functions ..................................................................... 3O
Cyclical Properties .................................................................................... 37
Conclusion ........................................................................................................ 38
CHAPTER 2
Real Interest Rates and Business Cycles in Emerging Economies
Introduction ...................................................................................................... 39
The Model ......................................................................................................... 44
Entrepreneurs .............................................................................................. 44
Households .................................................................................................. 48
Equilibrium and Dynamics ............................................................................... 51
Equilibrium ................................................................................................. 51
Steady State and Dynamics ......................................................................... 52
Calibration ........................................................................................................ 54
Results ............................................................................................................... 56

Impulse Response Functions ...................................................................... 56

Cyclical Properties ....................................................................................... 60
Sensitivity Analysis ..................................................................................... 61
The relative importance of the net cash flow and the
undepreciated capital stock as debt collateral .............................................. 62
Conclusion ........................................................................................................ 65
CHAPTER 3
The Determinants of Country Risk Spreads: Evidence from Argentina
Introduction ................................................................................. 7 ..................... 66
The Determinants of Country Risk Spreads ..................................................... 69
Econometric Analysis and Results .................................................................... 73
Conclusion ......................................................................................................... 79
Appendix
Appendix A. Data Sources ............................................................................ 80
Appendix B. The Steady State ....................................................................... 84
Appendix C. Loglinearized Equations ........................................................... 85
Appendix D. Serial Correlation Test ............................................................. 87
Appendix E. Unit Roots Tests ...................................................................... 88
Bibliography ......................................................................................................... 89

vi

LIST OF TABLES

Table 1.1 Business cycles in emerging and developed economies ....................... 9

Table 1.2 Regression Results ................................................................................... 11
Table 1.3 Parameterization ...................................................................................... 27
Table 1.4 Cyclical properties ................................................................................... 37
Table 2.1 Parameter values ..................................................................................... 55
Table 2.2 Cyclical properties ................................................................................... 61
Table 2.3 Results for sensitivity analysis ................................................................ 62
Table 3.1 Spreads Function Estimation 1998-2006 ............................................. 76
Table 3.2 Error Correction Model Estimation ........................................................ 78
Table 3.3 Unit Roots Tests ...................................................................................... 88

vii

Figure 1.1

Figure 1.2

Figure 1.3

Figure 1.4

Figure 1.5

Figure 1.6

Figure 1.7

Figure 1.8

Figure 2.1
Figure 2.2

Figure 2.3
Figure 2.4

Figure 2.5

Figure 2.6

LIST OF FIGURES

Impulse response functions of y and q to a shock to Rw ................ 31
Impulse response functions of h and h' to a shock to Rw ................. 32
Impulse response functions of c and c' to a shock to Rw .................... 32
Impulse response functions of k and n to a Shock to Rw ..................... 33
Impulse response functions of y and q to a shock Rw ........................ 34
Impulse response functions of h and h' to a shock to Rw .................... 35
Impulse response functions of y and q to a shock to At ( fi > O) ........... 36
Impulse response functions of y and q to a shock to A t ( 77 = O) .......... 36
Impulse Responses to a shock to the world interest rate ..................... 57
Impulse Responses to a Shock to the country risk spread ..................... 58
Impulse Response Function to a Shock to productivity (7] > O) ............ 59
Impulse Response Function to a shock to productivity (77 = O) ............ 60
Impulse Response to a shock to the world interest rate ........................ 64

(net cash flow as debt collateral)
Impulse Response to a Shock to the world interest rate ......................... 64

( undepreciated capital stock as debt collateral)

viii

Chapter 1
Lending Constraints, Real Estate and Business Cycles

in Emerging Economies

1. Introduction

In the last several decades, emerging economies experienced significant business
cycles. Associated with output fluctuations, a number of facts can be observed: (1)
Credit and property price experience large fluctuations. For instance, over the 1980-
95 period, the volatilities of bank credit to private sector for Argentina, Brazil,
Mexico were 34.4, 32.8, 22.1, respectively, while they were 3.4, 2.5, 3.1 for the
United States, Japan and Germany, respectively (Goldstein and Turner, 1996). In
addition it is found that during the last four decades, rising (falling) property price
was strongly associated with the build up (declining) phase of credit boom in
emerging economies (Mendoza and Terrones, 2008). (2) Real interest rates are
countercyclical. For example, during 1994 to 2001, the average correlation coefficient
between output and the real interest rate for the five emerging economies (Argentina,
Brazil, Korea, Mexico, and Philippines) was -0.55, while this correlation coefficient
for the five small open developed economies (Australia, Canada, Netherlands, New
Zealand, and Sweden) was 0.19 (Neumeyer and Perri, 2005).

The objective of this paper is to develop a small open economy model with credit
constraints tied to real estate prices to account for the countercyclicality of real

interest rates in emerging economies. In our model, entrepreneurs borrow from

domestic households and foreign lenders. Both entrepreneurs and domestic
households face credit frictions. Entrepreneurs must pledge collateral for borrowing.
As in Iacoviello and Minetti (2006), to capture the feature that domestic households
have more information about liquidating collateral, we assume that domestic
households incur smaller transaction costs for deploying collateral and, hence, for
given collateral values, they can grant larger loans to entrepreneurs than foreign
lenders. However, unlike in Iacoviello and Minetti (2006), in our model domestic
households also face credit frictions. In particular, domestic households’ lending is
constrained by the value of their real estate holdings. This modeling of the lending
constraint is supported by empirical evidence. Chiuri et a1. (2001) document that
during the last decade an increasing number of emerging economies have adopted a
bank capital adequacy requirement following the spirit of the Basel I Capital Accord.
The capital adequacy requirement imposes that banks maintain a minimum capital-
risk weighted asset ratio. Using bank-level data from 16 emerging economies, Chiuri
et a1. (2001) find that the enforcement of bank capital adequacy requirements
significantly curtailed credit supply in these countries.

We Show that a positive real interest rate Shock causes a reduction in the real
estate price. As a result, both the borrowing constraint and the lending constraint
tighten, which leads to a decrease in entrepreneurs’ investment in real estate and
variable capital, and a subsequent output drop. We compare quantitatively our model
with a benchmark model with the borrowing constraint but without the lending

constraint. In order to carry out quantitative comparisons, we calibrate the model to

Argentina data. The impulse response functions Show that a positive interest rate
Shock induces a 0.8% decline in output in the model with both credit constraints (our
preferred model ), while it induces a 0.4% output drop in the model without the
lending constraint (the comparison model). Simulation results Show that in the
preferred model the correlation coefficient between real interest rate and output is -
0.72, which is close to the data -0.75. This suggests that a model incorporating both
the borrowing constraint and the lending constraint can better explain the regularities
of real interest rates and business cycles in emerging economies.

The remainder of this paper is organized as follows. Section 2 discusses the
related literature. Section 3 documents the empirical regularities of real interest rates,
real estate prices and output in emerging and in developed economies. In addition, we
provide empirical evidence about the relationship between bank lending and bank
capital. Section 4 presents our preferred model. Section 5 presents the model without
the lending constraint. Section 6 specifies the shock processes and parameterization.
Section 7 discusses the transmission mechanism and presents the results. Section 8

concludes.

2. Related Literature

This paper most closely relates to Neumeyer and Perri (2005), Kiyotaki and
Moore (1997) and Iacoviello and Minetti (2006). Neumeyer and Perri (2005) study
the countercyclical pattern of real interest rates in emerging economies. They modify

the standard small open economy RBC model by introducing a working capital

constraint and Show that a positive interest rate shock has a negative effect on labor
demand, causing a decline in output. Our paper differs from Neumeyer and Perri
(2005) in that the transmission mechanism operates through the credit market rather
than through the labor market.

Kiyotaki and Moore (1997) analyze the dynamic interaction between a borrowing
constraint and asset prices and how this interaction can transmit small shocks to large,
persistent output fluctuations. Iacoviello and Minetti (2006) extend Kiyotaki and
Moore (1997) to an open economy environment. They assume that domestic and
foreign lenders have different liquidation technologies, which induces entrepreneurs
to adjust their debt exposure and allocation of collateral between domestic and foreign
lenders in response to productivity shocks. The objective of Iacoviello and Minetti
(2006) is to explain the comovement of output across countries. Although our paper
shares a similar transmission mechanism of shocks, it is distinct from Kiyotaki and
Moore ( 1997) and Iacoviello and Minetti (2006) as it introduces a lending constraint
and studies the aggregate consequences of interest rate shocks rather than technology
shocks.

The novelty of our approach is that we consider an economy in which not only
entrepreneurs are borrowing constrained but also lenders are lending constrained. The
extant literature on the financial accelerator focuses either on the firm’s balance sheet
channel (Bernanke et a1. , 1998, Kiyotaki and Moore, 1997, Iacoviello, 2005, etc.) or
the bank lending channel (Minetti, 2007, Van and Heuvel, 2007, etc.). An exception is

Holmstrom and Tirole (1997) who consider moral hazard problems when a firm

manages its investment and when financial intermediaries monitor firms. Holmstrom
and. Tirole (1997) Show that a credit crunch (a reduction in the financial
intermediary’s capital) or a collateral squeeze (a reduction in firms’ net worth) raises
firms’ threshold of net worth (above which the firm can obtain extema] finance
without violating the film’s and the financial intermediary’s incentive compatible
constraints and the investor’s participation constraint) and leads to a decrease in
investment. The model in Holmstrom and Tirole (1997) is static and the analysis is
qualitative. By contrast, our model is dynamic and the analysis is quantitative.

There exists a growing literature on business cycles in emerging economies.
Aguiar and Gopinath (2007) identify shocks to trend growth as the primary source of
fluctuations in emerging economies in a Standard RBC model. Uribe and Yue (2006)
investigate the intricate relation linking the world interest rate, country risk spreads
and emerging market fundamentals and the roles of world interest rates and country
risk spreads in driving business cycles in emerging economies. Oviedo (2005)
explores the relation among interest rates, borrowing-lending spreads and business

cycles in emerging economics.

3. Stylized Facts

In this section, we document the empirical regularities of real interest rates, real
estate prices and output in seven emerging economies (Argentina, Hong Kong, Korea,
Malaysia, Mexico, Thailand and Turkey) and contrast them with those of seven small

open developed economies (Austria, Australia, Canada, Denmark, Netherlands, New

Zealand and Sweden). We also explore the relationship between bank lending and
bank capital.
3.1 The Empirical Regularities of Real Interest Rates, Real Estate Prices and
Output in Emerging and Developed Economies
3.1.1 Data Overview

The key variables we are interested in are real interest rates, real estate prices
and real output. There is no single database containing real estate prices for different
countries. For all the above countries except Argentina, we use the house price index
(l-IPI) obtained from various sources as a proxy for real estate prices. For Argentina,
since the house price index is not available, we use the Global Property Research
(GPR) Argentina General Index as a proxy for real estate prices. The GPR Argentina
General Index is calculated based on total return (price return and dividend return) on
shares of IRSA, the largest real estate company in Argentina. The availability of the
HPI (or the GPR index) is a major factor why we choose the above countries. For the
seven developed economics, the sample starts in the first quarter of 1994 and ends in
the fourth quarter of 2006. For Argentina, Hong Kong, Korea, Thailand and Turkey,
the sample Starts in the first quarter of 1994. For Malaysia, the sample starts in the
first quarter of 1999. For emerging economies, the sample has at least 32 quarters of
data. Appendix A provides details on the source of data for each economy.

We choose the year 2000 as a base year and compute the real house price index
(or the real GPR) by dividing the house price index (or the GPR) by the consumer

price index. We construct real interest rates as in Neumey and Perri (2005). For

developed economics, the real interest rate is obtained by subtracting the expected
inflation rate from the short term nominal interest rate reported in OECD statistics.
The expected inflation rate is computed as the average of inflation in the current
period and in the three preceding periods. For Argentina, Korea and Thailand, the
nominal interest rates are constructed as the 90-day U.S. T- bill rate (proxy for the
world interest rate) plus the JP Morgan EMBI (Emerging Market Bond Index)
spread. The real interest rates are obtained by subtracting expected U.S. inflation rate
from the nominal dollar rate. As argued by Neumey and Perri (2005), constructing
real interest rates in this way has the advantage of not relying on the local inflation
rate, which is usually volatile and difficult to compute. Moreover, it has the advantage
of reflecting the true intertemporal terms of trade faced by local private agents during
financial crises. In fact, during crises most of the new borrowing of emerging
countries in the international financial markets occurs through official institutions.
For Malaysia and Turkey, although EMBI spread are available, there are many
missing data in our sample, so we construct the real interest rate by subtracting the
expected inflation rate from the local currency denominated short term interest rate

(3-month T-bill rate for Malaysia and interbank interest rate for Turkey).

3.1.2 Observations
We test each series for a seasonal component and deseasonalize the series when a

significant seasonal component is discovered. We then filter the series using the

Hodrick-Prescott filter with smoothing parameter 1,600 to derive business cycle
movements.

Table 1.1 reports the correlation coefficients among real interest rates, real estate
prices and output for emerging economies and for developed economies. The first
column of Table 1.1 reveals that there is a negative comovement between real interest
rates and real estate prices in emerging economies, while there is no such pattern in
developed economies. In emerging economies, on average the correlation coefficient
is - 0.36 and it ranges from - 0.70 in Korea to - 0.14 in Thailand. In developed
economies, on average the correlation coefficient is 0.19 and it ranges from 0.50 in
Canada to 0.06 in New Zealand.

The second column of Table 1.1 documents the correlation of real estate prices
with output. In both sets of economies real estate prices are procyclical. However, the
average correlation coefficient for emerging economies is 0.34, while it is 0.29 for
developed economies, signaling that the relationship between real estate prices and
output in emerging economies is stronger than in developed economies.

The third column of Table 1.1 reports the correlation of real interest rates with
output. Our results are consistent with those in existing literature. In emerging
economies, real interest rates are countercyclical. The correlation coefficient ranges
from - 0.75 in Argentina to - 0.21 in Thailand, with an average of - 0.44. In
developed economies, real interest rates are procyclical or acyclical. The correlation
coefficient ranges from 0.46 in Canada to - 0.11 in Australia, with an average of

0.23.

Table 1.1 Business cycles in emerging and developed economies

 

Corr (R,HPI) Corr ( HPI,Y) Corr (R, Y)

Emerging economies

Argentina - 0.42 0.45 - 0.75
Hong Kong (China) - 0.18 0.49 0.05
Koreal - 0.70 0.52 - 0.45
Malaysia - 0.34 0.39 - 0.42
Mexico - 0.27 0.34 - 0.44
Thailand - 0.14 0.27 - 0.21
Turkey - 0.20 0.08 - 0.36
Average - 0.32 0.36 - 0.37
Developed economies

Austria 0.10 0.16 0.11
Australia 0.02 0.17 - 0.01
Canada 0.50 0.22 0.46
Denmark 0.26 0.64 0.44
New Zealand 0.06 0.26 0.13
Netherlands - 0.13 0.59 0.13
Sweden 0.05 0.62 0.48
Average 0.12 0.38 0.25

 

For developed countries, the procyclicality or acyclicality of real interest rates
and the positive relation between real interest rates and real estate prices could stem
from the endogenous monetary policy stance. However, a detailed exploration of

these relationships in developed economies is beyond the purpose of the present

papen

3.2 The Relationship between Bank Lending and Bank Capital in Emerging
Economies

Since the lending constraint is the critical feature in our model, in this subsection,
we conduct empirical analysis to test whether the amount of lending and the level of
bank capital are closely related in emerging economies. To investigate the effects of
bank capital on bank lending, we estimate the following equation,

loant = a + ,6 capitalt _ + 8t

1

Where a is a constant, fl captures the effects of bank capital at period t-l on bank
lending at period t. 8t is the error term. The data on loans and bank capital are
aggregate data. The data has monthly frequency except the data of Korea, for which
only quarterly data is available.

Table 1.2 shows the regression results. In general, the results are satisfactory from

the perspective of the Signs and level of significance of the coefficients. For example,

10

for Argentina, an increase in bank capital by 1 peso will lead to an increase in bank
lending by 3.5 pesos. This effect is strongly statistically significant (t = 27.51). For
Thailand, an increase in bank capital by 1 baht is expected to cause an increase in
bank lending by 6.02 baht. This effect is also statistically Significant (t = 8.62). Thus,
our analysis provides additional evidence that there is a close relationship between

bank lending and bank capital in emerging economies.

Table 1.2 Regression Results

 

 

 

Country ,3 t statistic
Argentina 3 .5 27.51
Korea 9.93 8.00
Malaysia 8.72 21.74
Mexico 1 .9 24.61
Slovak 8.08 0.64
Thailand 6.02 8.62
Turkey 6.4 29.55
4. The Model Economy

This section describes a small open economy in which the empirical regularities
established in the preceding section can be interpreted as the equilibrium of an
economy subject to shocks to the world interest rate. We build on Iacoviello and
Minetti (2006) but impose a lending constraint and model the real interest rate

following Neumeyer and Perri (2005).

11

Time is discrete and infinite, t = 0,1,2 ...... There are three types of agents in this
economy: entrepreneurs, households and foreign lenders. There is a final good which
is invested and produced by entrepreneurs and consumed by entrepreneurs and
households. There are two kinds of assets. One is a fixed amount of real estate that
can be used either by entrepreneurs as input for production and collateral for loans or
by households as a consumption good. The other asset is a non—contingent bond that
can be traded by entrepreneurs, households and foreign lenders in the domestic and
international financial markets. In each period, the economy is subject to interest rate
shocks and productivity Shocks, which are revealed at the beginning of the period.

4.1. Households
Preferences and constraints There are a large number of identical infinitely lived
households in the model economy. Households derive utility from consumption, real
estate services and leisure. The representative household’s expected lifetime utility is

givenby
°° t .. ._3 .77
E0 Zoflanct+jlnht ”n! ) (1)
1‘:

Where E0 denotes the expectation operator conditioned on the information available

in period 0. ,6 is the household’s subjective discount factor, c't is the household’s

consumption, h't is its real estate holdings, n't is its labor supply, to is the utility
weight on the household’s real estate. 7] is the labor supply elasticity and {is the

weight parameter on the disutility of labor.

In each period, the household enters with real estate h‘t _ and bonds coming to

l

maturity B't— It rents labor to entrepreneurs, purchases bonds issued by

1.
12

entrepreneurs, invests in real estate and consumes. The household’s period budget

constraint is
(2)

cr +Bt +qt(ht —ht_1 )= (0 nt +R B

t t—l t—l

where qt is the real estate price in terms of the consumption good, cut is the real

wage rate, Rt is the real interest rate.

When households lend to entrepreneurs, their lending is constrained. In

particular, the household’s lending is subject to the following lending constraint

BtScrqtht_1 (3)

Where 0' is a constant satisfying 0505.1. In period t, the household’s lending cannot
exceed a fraction of the value of its real estate holdings at the beginning of the period.
The parameter 0 can be thought of as the inverse of the capital adequacy ratio. We
tie the household’s lending to the value of its previous period real estate holdings
since this value can represent the major component of its net worth. Indeed, the price
of real estate and bank capital are closely related. “A reduction in the price of real
estate will decrease bank capital directly by reducing the value of the bank’s own real
estate assets. It will also reduce the value of loans collateralized by real estate and
may lead to defaults, which will further reduce bank capital” (Herring and Watcher,

1999).

13

The Household ’s problem In each period, the household choose consumption, real
estate holdings, labor supply, and the amount of lending to maximize its expected
lifetime utility (1) subject to the budget constraint (2) and the lending constraint (3).
Before we solve the household’s problem, we make the following assumption:
Assumption 1 The household’s subjective discount factor, ,6 , satisfies

flR>L
In conjunction with assumption 2 below, this assumption ensures that the household
will lend to the entrepreneur in equilibrium since it is more patient than the
entrepreneur. This assumption will also ensure that the household’s lending
constraints is binding in steady state. Under assumption 1, we can treat that the
lending constraint is binding around the steady state of the economy.
Let A,‘ be the multiplier associated with the lending constraint. The first order

conditions for the household’s problem are:

i 1

 

 

[ct’]: +2 '= E ,6 R (4)
c' t t c 't
t+1
[h’l° i =i+Efl< 1 +2 '0 ) (5)
" C'qt )z' t c vqi+1 r+1 qr+1
t t t+I
[us]; Tn "7‘1=iw (6)
t C' t

The household chooses consumption, real estate holdings and labor supply by
equating marginal utility losses and marginal utility gains. Equation (6) is the
standard first order condition for labor supply. The first order conditions for bond

holdings (equation (4)) and the first order condition for household’s real estate

14

demand (equation (5)) are different from Standard ones due to the presence of the
lending constraint.

4.2. Entrepreneurs

Preferences and Constraints There are a large number of identical infinitely lived

entrepreneurs who maximize the following expected lifetime utility
Et 2 y‘ In ct , (7)
t = 0
where y is the representative entrepreneur’s subjective discount factor, and c, denotes
consumption.
The entrepreneur’s production function is a Cobb-Douglas given by
= . = ,u v 1 - ,u - v

yr F(kr’ht’nt’At) Atkt-l ht—l "t ’ (8)

where yt is output, kt is the variable capital stock that can be reproduced from the

final good, ht is the entrepreneur’s real estate, nt is the labor input and At is the
exogenous stochastic productivity. ,uandvare the variable capital share parameter

and the real estate share parameter, respectively.
Variable capital accumulation follows the law of motions

kt=it+(l—6)kt_l, (9)
where it is the investment expenditure in period t, and 6 is the depreciation rate.
In each period, the entrepreneur borrows from the household and the foreign lender,
invests in variable capital and real estate, produces using the variable capital, real
estate and labor, pays wages, repays debt, and consumes. The entrepreneur’s period

budget constraint is given by

15

V 1— —V H F_
AHkt_1/1h_1nt # +Bt +Bt —ct+kt—(l—6)kt_l+

(10)

H F
+qt(ht—ht—l)+Rt—lBt—l +Rt—lBt—I ”or":

where BI” is the entrepreneur’s domestic borrowing while B: is the entrepreneur’s
foreign borrowing.

When the entrepreneur borrows from the household and the foreign lender, she
faces a borrowing constraint, i.e. she cannot borrow more than the expected present
value of her pledgeable resources net of any recovery cost. We assume that the
entrepreneur can use only real estate as collateral for loans. We also assume that the
household pays less transaction costs for disposing of the collateral than the foreign
lender. This assumption wants to capture the idea that households have relatively
more local experience or knowledge necessary for the recovery and redeployment of

the collateral. In the event of debt repudiation, the household expects to pay a

transaction cost Et ((1 — m H )q h) and expects to receive a recovery value of

t+lt

E (q k): E (qu h) (11)

(l_m”)q t+l t

t+lht r+1t

where m H can be thought of as a proxy for domestic loan to value ratio. The foreign

lender expects to pay a transaction cost Et ((1 — 171,, )q h) and expects to receive a

t+lt

recovery value of

E (q h)= E (mpq h ) (12)

_l(—mF)q t+1t

I+Iht t+It

where m: < mH can be thought of as a proxy for the foreign loan to value ratio. Since
mp < mH, the entrepreneur would like to borrow from the household as much as

possible. However, she may not be able to do so Since the household is lending

l6

constrained. In this case, the entrepreneur will borrow partly from the foreign lender

in order not to waste collateral. Let at be the share of real estate ht used by the
entrepreneur as domestic collateral and l - at the share used as international

collateral. The entrepreneur’s borrowing constraints are

H
RtBt SEt(atm”qt+lht) (13)

RIB; s Et((1— at)qut+1ht) (14)
The entrepreneur’s problem The entrepreneur chooses factor inputs (real estate, variable
capital, labor), the amount of borrowing and consumption to maximize her lifetime expected
utility (7) subject to the budget constraint (10) and the borrowing constraints (13) and (14).
Before we solve the entrepreneur’s problem we make the following assumption.
Assumption 2 The entrepreneur’s subjective discount factor satisfies

7 R < 1

Under assumptions 1 and 2, we can treat the borrowing constraints are binding around
the steady State of the economy.

Let xltfljttF denote the multipliers associated with the domestic and foreign

borrowing constraints, respectively. The first order conditions for the entrepreneur’s

 

 

problem are:

[q]: i=2”)? +E 7 R (15)
c t t t C t
t t+l

[c,]: —1—=/1.FR +13 7 R (16)
C t 1 ’6 I
t t+l

17

l

 

 

_ H
:‘h‘Eer athqt+l+
t
[ht]: y (17)
F 7 t+l
+/it (l—at)qut+l+C (V +q,+,))
t+l t
no: i=1: 7 (awn-6) (18)
c ’c
t t+l t
[m]: (l-fl-vw, = wt", (19)

The entrepreneur chooses factors demands and borrowing by equating marginal
utility costs and marginal utility gains. Equations (18) and (19) are standard first order
conditions for variable capital stock and labor demand. Equations (15) and (16) are
first order conditions for consumption. Equation (17) is the first order condition for
the entrepreneur’s real estate demand. Equations (15), (16) and (17) are different from
the standard first order conditions due to the presence of the borrowing constraints.

If the maximum present value of the entrepreneur’s domestic debt limit,

Ethqt +111! lRt, IS less than the household 5 lending lImIt, 0' qtht—l , then the

household’s lending constraint has no effect on the entrepreneur’s borrowing. To rule

out this uninteresting case, we assume that in equilibrium

Ethq h IR: >0qtht—

t + l t holds and check that this is true in the steady state

1

. . . <1) . . . . . = . .
equrlrbnum . This means that In equrlrbnum Et (athqt + lht / Rt) 0 q th I _1, re.

the domestic borrowing equals domestic lending (Recall that we assumed that the

domestic borrowing and the lending constraints are binding in equilibrium). This

condition determines a, , the fraction of borrowing from the household.

 

(1): Given mH and R are close to 1, this condition is satisfied as long as the gap between h and h’ is

not large.

18

4.3. Discussion

Novelty of the paper The novelty of this paper is that we consider an economy in
which the entrepreneur’s borrowing constraint and the household’s lending constraint
can bind Simultaneously in equilibrium. In a closed model economy this cannot be
true. In fact, if one constraint binds, the other will not bind (unless the debt capacity
and the lending capacity are the same). However, in our small open economy
environment, this can be true since when the lending constraint is binding,
entrepreneurs can still borrow from foreign lenders up to her debt capacity, although
they can only borrow less from foreign lenders.

Colletaral-based constraints in emerging economies Although most of the existing
models with collateral-based borrowing constraints aim to explain business cycles in
developed economies, recent empirical studies show that collateral-based credit is a
more common phenomenon in emerging economies. Based on a panel dataset of 3199
manufacturing establishments from 1984 to 1994 in Mexico, Gelos et al. (2002)
explore the role of real estate as collateral by examining the relationship between
firms’ investment and the value of their real estate holdings. They find that the value
of real estate has a significant positive effect on the investment decisions of all firms
and this effect is stronger after the financial liberalization in 1989. Using a dataset of
560 credit files of Thai commercial banks, Menkhoff et al. (2005) document that the
incidence and degree of real estate based collateralization are higher in Thailand than
in developed countries. The relatively important role of collateral in emerging

economies’ credit activities is easily understood. Theoretically, collateral is regarded

19

as a means to solve the potential enforcement problem in the credit relationship and
the problem of credit rationing due to asymmetric information between lenders and
borrowers. Since these problems are more severe in emerging economies, the need for
collateral in credit activities is higher in emerging economies than in developed
economies.

Real interest rate As in Neumeyer and Perri (2005), we assume that all agents in the
model economy face the same real interest rate, which is charged by foreign lenders.
This is justified by the following three assumptions. First, we have assumed that there
is only one financial asset, a noncontingent bond in the model economy. Second, we
assume that that the emerging economies are net debtors to the rest of the world. This
assumption can be justified by the fact that most economies in Latin America are net
debtors to the rest of the world. Third, we assume the emerging economy we study is
a small open economy. The one real interest rate assumption simplifies the modeling
but does not change the general results of the issue we are interested in.

Household ’s budget constraint In a standard small open economy RBC model, a
bond holding adjustment cost is commonly modeled in the household’s budget
constraint to ensure stationarity. In a Standard small open economy model, since the
real interest rate is exogenous, the steady State of the model depends on the
household’s initial bond holdings and the equilibrium dynamics of the endogenous
variables possess a random walk component. Thus, a bond holding adjustment cost is
needed to guarantee stationary. However, in our model, because of the lending

constraint, the steady state household’s initial bond holdings are tied to the steady

20

states of real estate price and real estate holdings. Thus, we do not add a bond holding

adjustment cost in the household’s budget constraint.

4.4 Equilibrium

Given initial variable capital stock (kO ), initial real estate holdings (h0,h'O ), initial
bonds ( 80’ 8'0 )and a sequence of interest rates R t and productivity At , an
equilibrium of this economy is a sequence of allocations {Ct’ht’ B t’k t’nt’at } for
entrepreneurs, a sequence of allocations { c' t ,h't ,B' t ,n' t } for households and prices

{ qt, (at } such that:

(i) The entrepreneur solves her optimization problem.
(ii) The household solves its optimization problem.

(iii) The real estate market, the labor market, and the domestic credit market clear.

4.5. Real Interest Rate
Our model economy is subject to interest rate and productivity shocks. We follow the
literature to model real interest rates in emerging economies as a product of world
interest rates and country risk spreads.

R t = R tht (20)
Where Rtw denotes the world interest rate and Dr denotes the country risk spread.
The world interest rate is identified as the 3—month T—bill rate in the United States.
Country risk spreads capture foreign lenders responses to investment risks and

uncertainty in emerging economies. There is no single theory about how to model the

21

country risk spread. Uribe and Yue (2006) model country risk spreads as a function of
current and past values of a set of endogenous variables ( output, investment and
balance of payments) and world interest rates. Li (2007) models country risk spread
as a function of net external debt position. Neumey and Perri (2005) model country
risk spreads as a function of expected future productivity. All of them link the country
risk spreads to economic fundamentals since empirical analyses show that
fundamentals in emerging economies has effects on country risk spreads ( see Uribe
and Yue (2006) for a literature review). We follow Neumey and Perri’s (2005)
approach to model country risk Spreads. That is,

Dr :f(EtAt+l) (2])
where f is a decreasing function. Although this modeling has the problem that
productivity is exogenous, it is simple and we believe that it still captures some
features of country risk behavior in emerging economies. For instance, according to
Kehoe (2003), after Argentina launched Convertibility Plan to reduce inflation in
1991, it experienced significant interest rate spikes during the early days of the plan,
during the Tequila Crisis and in late 2000. This is because foreign (domestic)
investors fear that the Convertibility plan would be abandoned during these periods
and inflation would be serious. Empirical evidence has shown that inflation and TFP
are negatively correlated in Argentina (Kehoe 2003). Thus, we can think of fears on
the part of investors about the high inflation in an emerging economy as their

expectations of a low future productivity in that country. In other words, country risk

spread and expected future productivity are negatively correlated.

22

5. A Model without the Lending Constraint

For quantitatively comparison purpose, we introduce a model with borrowing
constraint but without lending constraint. In this case, the entrepreneur would borrow
from the household up to her debt capacity, i.e. at =1. Since the household does not
face lending constraint now, we add a bond holding adjustment cost in the
household’s budget constraint to guarantee stationarity. The household’s budget
constraint becomes

ct'+Bt'+qt(ht'—ht_1')+§(BI'—B')2 = wtnt '+Rt_13t_1'

(22)

Where §(Bt '—B')2 is the bond holding adjustment costs. B’ is the steady state of B't
and K is a constant parameter. The household’s problem is choosing decision

variables to maximize lifetime utility subject to the budget constraint ( 9 ).

Given initial variable capital stock (kO ), initial real estate holdings (ho, h'O ), initial
bonds ( 80,8'0 )and a sequence of interest rates Rt and productivity At , an
equilibrium of this economy is a sequence of allocations { Ct’ht’Bt’kt’nt } for
entrepreneurs, a sequence of allocations { c't ,h't ,B't ,n't } for households and prices

{ qt’ (at } such that:

(iv) The entrepreneur solves his optimization problem.
(v) The household solves his optimization problem.

(vi) The real estate market and labor market clear.

23

6. Shock processes and parametization
6.1 Shock processes

To close the model, we need to specify the processes for productivity, the world

interest rate and the country risk spread. Let the variable 5? t denote the log deviation
of xt from an HP filtered trend. We assume that total factor productivity, At , and the

world interest rate, Rt , follow AR( 1) processes given by ( 23) and (24 )
At =’oAAt-l +£A,t
(23)
Rt = prRt _1 +8R,t (24)

where p A is the autocorrelation coefficient for the productivity process. pr is the

autocorrelation coefficient for world interest rate process, 8 are normally

A,t 81a:
distributed independent productivity and world interest rate shocks with standard

deviations 0' and , res ctivel .
A,t aRwJ pe y

The process for the country risk spread is given by
Dr = _77EtAt +1 +804 ’
(25)

Where 77 >0 is a constant capturing how much the country risk spread responds to

expected productivity shocks and ED t is a normally distributed independent shock

9

with standard deviation 0'0 t

, O

6.2 Parameterization

24

To quantitatively analyze the model, we need to set values for the following
parameters or exogenous variables: the parameters describing preferences (7, ,6, j,77 ),
the parameters describing technology ( ,u,v,c3 ), the loan to value ratios for domestic
and foreign debts (mH, mp), the inverse of the capital adequacy ratio (0' ), the steady
state value for the exogenous real interest rate R, and the parameter values for the

Stochastic processes for productivity and the world real interest rate

( p A’ p Rw’fi’ 0 AVG Rw,t’0 D,t)' For the model without the lending constraint, we
also need to set values for the household’s steady state bond holdings ( B’ ) and the
bond holding adjustment cost parameter (K). The parameter values are set either by
referring to existing literature on Argentina or by calibration and estimation.
Following Kehoe (2003), we set the labor share (1— ,u — v ) at 0.6. In
Kehoe(2003), the capital share and the land share are set at 0.3 and 0.1, respectively.
Since in our model, the real estate does not mean land only, we choose the variable
capital share to be 0.2 and the real estate share to be 0.2. The depreciation rate for the
variable capital 6 is set at 11.3% per year as in Kydland and Zarazaga (2002). This
gives us a quarterly depreciate rate of 2.8%. We set the steady state real interest rate
equal to 1.1. This value is consistent with an average US. interest rate of about 4%
and an average country risk spread of about 7% in Argentina. Kydland and Zarazaga
(2002) choose the same real interest rate Steady state value when they study the
driving forces of Argentina’s depression in 19803. Because of the presence of the

lending constraint, the household’s subjective discount factor ,8 cannot be determined

in the usual way (i.e. R ,6 ¢ 1). We choose ,6 at 0.94, which is between the value

25

(0.9445) set in Kehoe (2003) and the value (0.93) set in Neumey and Perri (2005).

. 1 1 .
Since we have assumed —<R<— In the steady State, we set the entrepreneur’s

subjective discount factor 7 at 0.9.

The domestic and foreign loan to value ratios are set equal to m” =0.8 and
m F = 0.6. The inverse of capital adequacy ratio is set at 0.3, implying that the share
of the entrepreneur’s borrowing from the household is about 0.6. The utility weight
on the household’s real estate is set at 0.2, implying that entrepreneurs and
households approximately equally Split the real estate stock. AS in Iacoviello and
Minetti (2006), we set labor supply elasticity equal to 0.05 so that the response of
output to Shocks depends almost entirely on the behavior of technology and on
changes in entrepreneur’s real estate holdings in stead of on labor supply.

For the model without the lending constraint, the household’s Steady state of
bond holdings is not uniquely pinned down. We set it at 0.01. The bond holding
adjustment cost K is set at 105, the minimum value that guarantees that the
equilibrium solution is Stationary.

We estimate the parameters for the productivity process based on Solow
residuals estimated by Maia and Kweitel (2003) for Argentina from 1960 to 2000. We
first take logarithm of these residuals and detrend them using HP filter to get the

series A. We then estimate equation (23) and get the estimated correlation coefficient

p A =0.81. Given the estimated correlation coefficient, we construct series for the

shocks, 8 , and use them to find the estimated standard deviation 0 t: 3.8%.

A,t A.

Finally, the parameter values determining the process for the world interest rate and

26

Table 1.3 Parameterization

 

 

Parameter values symbol core model comparison model
household discount ,5 0.94 0.91
entrepreneur discount 7 0.90 0.90
h utility weight j 0.2 0.2
Labor wage elasticity l/(l — 77) 0.05 0.05
Capital share ,u 0.3 0.3
Real estate Share V 0.1 0.1
Depreciate rate 5 0.028 0.028
Real interest rate R 1.1 1.1
Domestic LTV for h my 0.8 0.8
Foreign LTV for h m; 0.6 0.6
Household’s bond holdings B’ 0.01
Bond holding adjustment cost K 10 '5
Inverse of capital requirement 0' 0.3
Productivity pA 0.81 0.81
UAJ 3.8% 3.8%
World interest rate p Rw 0.81 0.81
0' Rw,t 0.63% 0.63%
Country risk 77 1.04 1.04
0' D, t l .7% 1.7%

 

27

the country risk Spread are taken from Neumeyer and Perri (2005). Table 1.3
summarizes our parameterization results.
7. Characterization of Equilibrium and Results

In this section we conduct qualitative analysis to study the transmission
mechanism of real interest rate Shocks to the aggregate economy implied in our
model. We then report the main results of impulse response functions and cyclical
properties of the model economy.
7.1. The Transmission Mechanism of Interest Rate Shocks

In our economy, the real estate price is determined in the real estate market in
which the supply of real estate is fixed and both entrepreneurs and households
demand the real estate. To better grasp the forces to work, we momentally assume mH
=mF =m. The necessary condition for the entrepreneurs’ real estate demand in log-

linearized form can be stated as

l A ma m ,. ,. ,. m ..
(l—V)(1—(E-7)m_7)ht=-—R +(I—EXct—ch—q,+(r+;—mr)q

R t t+l

1 .. 1 A
+(1-(E—r)m—r)At+l+(1—(E-7)m-7)#k,

1 A
+(“(;")’"‘7>(“”“’)".+1

(26)
where the coefficients for all the variables are positive under our parameterization.

First, suppose that real estate price does not respond to interest rate Shocks, i.e.

A

q t = q t +1 = 0. Since productivity is exogenous, it does not respond to interest rate

shocks, i.e. A, +1 =0. Variable capital could fall or not change much on impact of

28

interest rate Shocks and by assumption we made employment respond little to interest
rate shocks. Then a positive interest rate shock has a negative direct effect on real
estate demand through the first term and has an indirect effect by inducing changes in
consumption. Since entrepreneurs are borrowers, the substitution effect and the

income effect of interest rate shocks work in the same direction. A positive interest

rate shock reduces current consumption, i.e. 6t --6 < 0. Intuitively, when current

[+1

consumption decreases, the marginal cost of investing in real estate increases and
entrepreneurs’ real estate demand decreases.
The necessary condition for households’ real estate demand in log-linearized form

can be stated as
i); =i(5 ._ ~ )_i.§ +—q wrap +fl/i'0'q/i' (27)
t ' t t+l C't C. t+l t+l t+l

where the coefficients for all the variables are positive under our parameterization.
We can ignore the last two terms since the coefficients are small. If the real estate
price does not respond to interest rate shocks, then a positive interest rate shock can
have a negative effect on the household’s real estate demand. Since households are
lenders, the substitution effect and the income effect of interest rate Shocks work in
opposite directions. The substitution effect implies a reduction in current
consumption, while the income effect can be neglected since labor supply responds

little to interest rate shocks.

29

Thus, if the real estate price does not respond to interest rate shocks, both the
entrepreneur’s and the household’s real estate demand holdings tend to fall. However,
this cannot be true since the real estate supply is fixed. If the. entrepreneur’s
(household’s) real estate decreases, the real estate price must fall to induce the
household (entrepreneur) to increase real estate holdings. Note that our model implies
that it is possible for the entrepreneur to increase real estate holdings in response to a
positive interest rate shock, while in Kiyotaki and Moore(1997), the entrepreneur’s
real estate holdings always decline when there is a negative technology shock.

The decline in real estate price tightens both the borrowing constraint and the
lending constraint. Consequently, the entrepreneur cuts its investment in variable

capital and real estate. This causes a subsequent fall in output

7.2. Impulse Response Functions

Figure 1.1 depicts the impulse response functions to a positive world real interest
rate shock for the core model economy. On impact, the price of real estate falls as
expected. A 1% increase in the world interest rate leads to about a 4% decrease in the
real estate price. In the meantime, there is a slight increase in output. This is because
although we restrict the elasticity of labor supply to a small value, we do not
eliminate the effect of interest rate shocks on the labor supply. When there is an
increase in interest rate, people tend to work more and hence output increases.
However, this effect is very small and output falls shortly after the interest rate

shocks. A 1% increase in the interest rate can induce about a 0.8% fall in output in the

30

second quarter and then output gradually approaches its steady state. Figures 1.2-1.4
report the impulse response functions of the entrepreneur’s and the household’s real
estate holdings, consumption and variable capital and employment to a real interest
rate shock. These results are consistent with our analysis in section 7.1. In short, a
positive interest rate Shock leads to a decrease in the entrepreneur’s real estate
holdings, an increase in the household’s real estate holdings, a decrease in both the
entrepreneur’s and the household’s consumption, a decrease in variable capital and a

very limited increase in employment.

Impulse responses to a shock In Ftw

I I I 7 T r I

 

     

world real interest rate

  

P
or

 

 

O

s5
an

' output

I
—L
l
1

I
N
I
l

I"
or

real estate price

Percent deviation from steady state
a:

I
a
I

 

 

.—

an
N
O

 

4 l l l l l
0 1 2 3 4 5

Years after shock

Figure 1.1 Impulse response functions of y and q to a shock in RW

31

Percent deviation from steady state

Percent deviation from steady state

 

    
   
 

Impulse Responses to a shock In Rw

I

ousehold's real estate

world real interest rate

 

 

 

entrepreneur's real estate

5..
y.—

l l I

y—

>—

 

 

Figure 1.2 Impulse response functions of h and h' to a shock to RW

1 2 3 4 5 6
Years after shock

 

P
or

Impulse responses to a shock In My

I I

world real interest rate

 

O

:5
‘."

)-

I
d

E
d
“I

I

r1:

!'°
on

do

 

 

ntrepreneur's consumption

household's consumption

1 l l l l 1

.—

 

sh
cm

Figure 1.3 Impulse response functions of cand c' to a shock to RW

1 2 3 4 5 6
Years after shock

32

 

 

Impulse responses to a shock in 8w

1 I i T i I fl I

 

world real Interest rate

labor

Percent deviation from steady state

 

 

 

 

_2 l 1 I l 1 I l
0 1 2 3 4 5 6 7 8
Years after shock

Figure 1.4 Impulse response functions of k and n to a shock in RW

Figure 1.5 depicts the impulse response functions of output and the real estate
price to a world interest rate shock for the model economy without the lending
constraint. In this case, the negative effect of interest rate shocks to output is smaller
than that in the core model. Our explanation is the following. In the core model, a
positive interest rate shock leads to a decrease in the real estate price, a decrease in
the entrepreneur’s real estate holdings, and an increase in the household’s real estate
holdings but the extent of this increase is less than that of the decrease in the real
estate price. This implies that both the borrowing constraint and the lending constraint
are tightened in response to the interest rate shock. Consequently, output experiences
a relatively large fall. By contrast, in the model without the lending constraint, a
positive interest rate shock leads to a decrease in the real estate price, a decrease in

the household’s real estate holdings and an increase in the entrepreneur’s real estate

33

holdings but the extent of this increase is less than that of the decrease in the real
estate price (see Figure 1.6). In this case, although the entrepreneur’s borrowing
constraint is tightened, the large reduction in the value of the household’s real estate
has no effect on the entrepreneur’s borrowing. Consequently, output experiences a
relatively small reduction.

Figure 1.7 shows the impulse response functions of output and the real estate
price to a productivity shock. A 1% increase in productivity induces a 5.5% rise in the
real estate price on impact of the shock. It also induces a 1.7% increase in output in
the second quarter following the shock. Figure 1.8 shows the impulse response

functions of output and the real estate price to a productivity shock when the country

risk spread is not a function of productivity (i.e. 77 = 0 ). In this case, a 1% increase in

Impulse responses to a shock in Rw

1 I I I I I I T

 

world real interest rate j

 

 

  
 

real estate price

Percent deviation from steady state
5.’

 

 

 

-2.5 ~
.3 _
-3.5 1
4 I I I 1 I I I
0 1 2 3 4 5 6 7 8
Years after shock

Figure 1.5 Impulse response functions of y and q to a Shock Rw

34

Impulse responses to a shock In Rw

 

 

 

 

 

 

 

 

2 I 7 I T I I H
1.5 ~
household's real estate
5 ‘ ‘
§ 0.5“ (__ “
92 ‘ world real Interest rate
~
g _
S .o 5. J
3
>
'8 -1 - entrepreneur's real estate r
E
-2 ~ 1
-2.5 I I I I I I I
0 1 2 3 4 5 6 7 8

Years after shock
Figure 1.6 Impulse response functions of h and h' to a shock to RW
productivity induces a 1.8% rise in the real estate price on impact of the shock. It also
induces a 1.1% increase in output in the second quarter following the shock. Our
explanation for the relatively larger impulse responses of output and the real estate
price in our model is the following. When there is a positive productivity shock in
period t, people will expect an increase in productivity in period. t+1 since
productivity exhibits persistence. This will reduce the country risk spread as well as
the real interest rate. Equation (26) shows that a positive productivity shock will lead

to a larger increase in entrepreneur’s demand for real estate when the country risk

Spread is negatively correlated with productivity (since At + > 0 and RI < 0 ), while

1

when the country risk spread is not a function of productivity, the increase in

entrepreneurs’ demand for real estate is smaller ( since A t + > 0 and R t = 0). Thus,

1

in our model the real estate price experiences a larger increase in response to a

35

positive productivity shock. This relaxes both the borrowing constraint and the
lending constraint. Consequently, output experiences a larger increase.

Impulse responses to a shock in A
6 T I I I I I I

 

real estate price

Percent deviation from steady state

 

 

 

 

 

 

0 1 2 3 4 5 6 7 8
Years after shock

Figure 1.7 Impulse response functions of y and q to a shock in At (5 > 0)

Impulse responses to a shock In A

I l

 

1.8 s r

d
I
O
I
l

.s
b
r

real estate price ~

9 .° ° 7‘
:- aa a .. N
1 J l L 1

Percent deviation from steady state

0
N
I

 

 

 

 

O

O 1 2 3 4 5 6 7
Years after shock
Figure 1.8 Impulse response functions of y and q to a shock in At (77 > 0)

36

7.3 Cyclical properties

We Simulate the model with an HP-filtered series to obtain the second moments
properties of the model economy. Table 1.4 reports the results. The model economy
predicts that the correlation coefficient between the real interest rate and output in
Argentina is -0.72, which is very close to the data -0.75. While the correlation
coefficients between the real interest rate and the real estate price and that between
the real estate price and output have the same signs as the data, they are significantly
higher than the data. These differences could result from the fact that we use the GPR

Argentina general index as a proxy for the real estate price.

Table 1.4 Cyclical properties

 

Model Argentina Data
Corr ( R,q) - 0.9 - 0.42
Corr (Y ,q) 0.83 0.45
Corr (R,Y) - 0.72 - 0.75

 

37

8. Conclusion

In this paper we present a simple model of a small open economy with credit
constraints to explain the countercyclicality of real interest rates in emerging
economies. We emphasize that both the borrowing constraint and the lending
constraint can play a role in accounting for the countercyclicality of real interest rates
in emerging economies. An extension of this paper could be to explicitly model
bank’s behavior. This would lead to a richer interaction between entrepreneurs’ net
worth and banks’ capital and provide a better explanation for the observed regularities

in emerging economies.

38

Chapter 2

Real Interest Rates and Business Cycles

in Emerging Economies

1. Introduction

In the last decade or so, an important empirical regularity in emerging market
economies was that business cycles were associated with real interest rates
fluctuations. The real interest rate was observed to be countercyclical in emerging
market economies, while it was acyclical in developed economies. For example,
during 1994 to 2001, the average correlation coefficient between output and the real
interest rate for the five emerging economies (Argentina, Brazil, Korea, Mexico, and
Philippines) was —0.55, while this correlation coefficient for the five small open
developed economies (Australia, Canada, Netherlands, New Zealand, and Sweden)
was 0.19 (Neumeyer and Perri, 2005). A natural question is: Do real interest rate
shocks drive business cycles in emerging economies? If the answer is yes, then how
does it occur? The present paper attempts to address this question and it emphases the
role of net worth borrowing constraints as the financial accelerator in transmitting real
interest rate shocks.

The paper is closely related to Neumeyer and Perri’s paper (2005, henceafter NP
) but the transmission mechanisms in the two papers are different. NP (2005) studies

the relationship between real interest rates and business cycles in emerging economies

39

by introducing a working capital constraint into the standard small open economy
RBC model. The working capital constraint implies that the real interest rate has an
impact on labor demand. They also assume that households have GHI-I preferences
(proposed by Greenwood, Hercovitz and Huffman), which implies that the labor
supply is independent of consumption and the real interest rate. Thus, a positive real
interest rate shock will decrease the labor demand and hence output falls. Following
NP (2005), other studies (for example, Uribe and Yue, 2006) also introduce a working
capital constraint to study the relationship between real interest rates and business
cycles in emerging economies.

Oviedo (2005) argues that including a working capital constraint in the standard
small open economy model is not an effective mechanism to align the interest-rate
macro volatility predicted by the model with what is observed in emerging
economies. Oviedo (2005) Shows that when a working capital constraint is added into
the standard small open model economy, the equilibrium condition for labor demand
has the property that a very large fall in interest rate is required to make up for a small
increase in the wage rate to keep output unchanged. In contrast, a small rise in
productivity is enough to induce the same result. This implies that for the working
capital constraint to add a significant amplification mechanism to interest rate Shocks,
both the interest rate level and swings must be large.

In this paper, we propose another transmission mechanism, i.e., a credit channel,
to explain the impact of the real interest rate shocks on the aggregate economy in

emerging economies. In our model economy, entrepreneurs face net worth borrowing

40

constraints when they borrow from domestic and foreign investors. A positive interest
rate shock will increase debt repayment, decrease entrepreneurs’ net worth and tighten
the borrowing constraint. Consequently, investment and output fall. The decrease in
output will further tighten entrepreneurs’ net worth and cause a further decrease in
entrepreneurs’ investment and output. Our motivation to include a borrowing
constraint to the small open model economy comes from two considerations. The first
consideration is the fact that in emerging (developing) countries, financial market
imperfections are more pervasive and deeper than in developed countries. The second
consideration is due to a significant role of financial constraints in business cycle
models. That is, they can act as a financial accelerator which transmits small shocks to
large fluctuations in aggregate variables.

When we set up the model, an important issue is how to model the real interest
rate in emerging economies. We follow NP’s (2005) approach to model the real
interest rate. First, in the model economy all the agents face the same real interest rate,
which is charged by foreign lenders. This is justified by the following three
assumptions. The first assumption is that we assume that there is only one financial
asset, a noncontingent bond. The second assumption is that we assume that the
emerging economies are net debtors to the rest of the world. This assumption can be
justified by the fact that most economies in South American are net debtors to the rest
of the world. The third assumption is that the model economy is a small open
economy. These three assumptions imply that all the agents in the model economy

face the same real interest rate. Second, the real interest rate in emerging economies

41

is modeled as a product of the world real interest rate and a country risk spread. The
world real interest rate is identified as the rate on non-investment grade bonds in the
United States. When foreign investors invest in emerging economies, they face
country risks which may come from the possibility of expropriation of assets,
entrepreneurs’ debt default etc. In response to the country risks, foreign investors
charge a country risk Spread

When dealing with the effects of the real interest rate on the aggregate economy
in emerging economies, an interesting question is whether the fundamentals of the
emerging economies also have an impact on the real interest rate. Since the emerging
economies are small open economies, the world real interest rate is not affected by the
fundamentals of emerging economies. However, a large body of empirical literature
shows that the country risk spread is related to the fundamentals of the emerging
economies. (see Uribe and Yue, 2006). Following NP (2005), we model the country
risk spread as a function of productivity. Thus, the real interest rate shocks will affect
the aggregate economy through a borrowing constraint. On the other hand, the
fundamentals of the economy, represented by the productivity, may affect the real
interest rate and aggravate the effects of the productivity shocks on the aggregate
economy.

We calibrate the model to Argentine data. Quantitative results Show that the
model predictions are consistent with the data in general. In particular, the real
interest rate is countercyclical and it has a larger and longer impact on output than in

the NP model. The correlation between the real interest rate and output is -0.54, In

42

addition, the model predicts that output exhibits high volatility and that consumption
is more volatile than output. Sensitivity analysis Shows that the negative relationship
between the real interest rate and output are stronger for more borrowing constrained
firms than for less borrowing constrained firms.

The paper is built on the NP model. The key difference between the NP model
and our model is that we impose a borrowing constraint in the small open RBC
model. This means that the transmission mechanism of interest rate shocks as well as
the productivity shocks to the economy is through the credit market instead of the
labor market. The paper is also related to the literature on the credit channel of
business cycles (Benanke et al, 1998, Kiyotaki and Moore, 1997, etc.). The
transmission mechanism is similar to that in Aghion et a1 (2004). Aghion et al (2004)
study a general equilibrium currency crisis model in emerging economies in which the
possibility of currency crisis is driven by the interplay between firms’ net worth
constrains and nominal rigidities. Our paper is distinct from their paper in that the
entrepreneur’s net worth is not affected by the fluctuations of exchange rates but by
the fluctuations of real interest rates. Although there is a large body of literature on
the effects of credit market imperfections on business cycles, to our knowledge, no
paper introduces firm borrowing constraints to analyze the impacst of real interest rate
Shocks on business cycles in emerging economies.

The remainder of the paper is organized as follows. Section 2 describes the model.
Section 3 defines the equilibrium and characterizes the dynamics. Section 4 calibrates

the model to Argentine Data. Section 5 presents the results. Section 6 Concludes.

43

2. The Model

Time is discrete and infinite. There are three types of agents in this small open
economy: entrepreneurs, households and foreign lenders. There is one good, which is
produced by entrepreneurs and consumed by entrepreneurs and households. The only
asset traded by entrepreneurs, households and foreign investors in the domestic and
international financial markets is a noncontingent bond. The economy is subject to
interest rate shocks and production Shocks, which are revealed at the beginning of

each period.
2.1 Entrepreneurs

There are a large number of entrepreneurs. We assume that entrepreneurs are
identical and normalize the total number of entrepreneurs to 1. The assumption of a
large number of entrepreneurs implies that entrepreneurs behave competitively, that
is, their actions do not affect market prices in the economy. The representative

entrepreneur has lifetime utility over consumption, which is given by

U(cl,c2 ,...ct....)= Et 207’ lnct (1)
t:

where ct is consumption at period t, 7 is the discount factor. At each period, the

entrepreneur consumes, invests, borrows from households and foreign lenders, and
produces using capital and labor. The entrepreneur’s production function is Cobb-

Douglas

_ a I-a
W—fih—IW (3

where yt is output, kt— is capital, "t is labor input, At is the exogenous stochastic

1

productivity, and a is the elasticity of output with respect to capital.

The Stock of capital evolves according to

kt =it +(1—6)kt_1 (3)

where 5 is the depreciation rate.

The period budget constraint for the entrepreneur is

a 1—a _ _ _
Atkt_1 ”t +Bt—Ct+kt (1 §)kt_1+a)tnt+Rt_lBt_1 (4)

where Bt is bond issued by the entrepreneur and (at is the real wage rate. At period t,

the entrepreneur spends the output and borrowed funds on wage payment, debt
repayment, investment and consumption.

The financial market in which entrepreneurs, domestic households and foreign
lenders participate is not frictionless. The friction comes from the possibility that
entrepreneurs may default on their debt obligation. To prevent this enforcement

problem, a borrowing constraint is imposed on the financial contract. The

entrepreneur can only borrow an amount up to a fraction 10 of his net worth, Wt’

45

which is defined by period t output net of debt repayment, wage payment,

consumption and the nondepreciated capital stock. Thus

(5)

_ a l—a _ _ _ _
where Wt _Atkt—1 nt ct mint Rt—lBt—l +(1 §)kt_1and0<tp<1.
The entrepreneur’s investment is funded by borrowing ( Br) and net cash flows (
*
Wt )

it = Br +Wt * (6)

*_ a 1—a_ _ _
where Wt _Atkt—l nt Ct (otnt Rt—lBt—l

Combining equations (3) (5) (6), we obtain

B s—"’ k 26k (7)

where 9 is a constant.

The entrepreneur’s problem is to choose consumption ct, labor demand n t’

capital stock kt and bond holdings Bt to maximize his expected lifetime utility

subject to the budget constraint and the borrowing constraint. That is, the entrepreneur

solves the following problem:

46

max E0 :0); 1n Ct
t :

. a l—a _ _ _
s.t Ak n +Bt—ct+kt (1 6)kt_

tt_1 t +wnt+R B

1 t t-l t—l
B $19k
t t

Before we solve the entrepreneur’s problem, we make two assumptions.

Assumption 1: The subjective discount factor for the entrepreneur” is less than that
for the household :6 . This assumption ensures that the entrepreneur will not postpone
production and has to borrow from the household, since he is less patient than the
household.

Assumption 2: the parameters a, .3, 7,995 are restricted by the following expression

a[a—e>/r+;¢r6+w—m >1

am ] a ‘1 +1- 6 — ,6 . This assumption ensures that at the

 

steady state, the rate of return to investment is greater than the real interest rate, so
that the entrepreneur will invest as much as possible.

Under Assumptions 1 and 2, the borrowing constraint is binding in equilibrium.
Let 4* be the Lagrangian multiplier associated with the budget constraint. Let Z t be
the Lagrangian multiplier associated with the borrowing constraint. The Lagrangian

for the entrepreneur’s problem can be written as follows:

_ 00 a l-a _ _ _
L—EO Zy’{lnct+/it[Atkt_ n +Bt ct kt+(l 5)k,_1 H

1 t
t=0

_Rt—lBt—l]+Zt(Bt —6kt)}

47

The first order conditions with respect to consumption c t’ labor demandnt , capital

stock kt and Br are:

[c ]: —22 (8)
r 1‘
Ct
a -a _
["t] At(1"a)kt_1 "t —wt (9)
_ a—l l—a _
[kt] lt+ZtBTfitflt+Umt+lkt "t+1 +1 6 (10)
[BI]: Ar+lr =}Er’1r+1Rr (11)

Combining (9), (10) and (11), we obtain,

 

i_ 7’ 1 61—1 l—a _ _
ct_1_9 tct+l[w’t+1kt "t+1 +1 6 gRt]
(12)

Equation (9) defines the entrepreneurs’ labor demand schedule, by equating the

marginal productivity of labor in period t to the real wage rate cut . Equation (12) is the

entrepreneur’s Euler condition for consumption. It is different from the usual Euler

condition due to the presence of borrowing constraint.

2.2 Households

There are a large number of households in the economy and the number of
households is normalized to 1. The representative household’s period preference is

described by GHH preference

48

. . _ l . . (l—O')
u(ct,nt)-l—_;[ct-I/Intv] (13)

where c't is the household’s consumption, n't is the labor supply. The intertemporal
elasticity of substitution is given by 1/0. The elasticity of labor supply is given by
l/(l—v) , wherev>1. 1” captures the labor weight in the household’s preference,
V>0. GHH preferences are commonly used in small open economy business cycles
models. They have the property that labor supply does not depend on consumption
and thus is independent of the real interest rate.

In each period, the household supplies labor to the entrepreneur, receives labor
payment and the principal and interest of last period’s bonds. The household also buys
new bonds, pays bond holding adjustment costs and consumes at the end of each
period.

The household’s period budget constraint is given by

r r E l_ r*2_ I r
ct+Bt+2(Bt B ) —cotnt+Rt_lBt_1 (14)

E ,
where 3'! is the household bond holding at period t. '5 (B't - B *) 2 is the bond
holding adjustment costs. 3' * is the steady state of 8'! and 5 is a constant

parameter. At period t, the household spends the proceeds from bond holdings and
labor income on consumption, bond purchase and bond holding adjustment costs. A
Bond holding adjustment cost is commonly used to guarantee stationarity in small
open economy models when asset market is incomplete. As we shall see, the steady

state of the model depends on the household’s initial bond holdings, whose rate of

49

return is exogenously determined abroad. This induces long—run effects of transient
Shocks on the steady state of the economy, i.e. the equilibrium dynamics of the
endogenous variables possess a random walk component. However, all available
techniques are valid locally around a given stationary path. Adding bond holding
adjustment costs is one of the methods to solve the problem of nonstationarity.

(Schmitt-Grohe and Uribe, 2003)

The household’s problem is to choose consumption c; , labor supply n't, bond

holdings B; to maximize the expected lifetime utility subject to the budget constraint.

That is, the household solves the following problem:

 

so I _
max Eotzoflt { 1 _ U[c't ‘W n't t)](1 0.)

r v E r_ l*2_ I I
S.t. Ct+Bt+2(Bt B ) _wtnt+Rt—lBt—1
The Lagrangian of the household’s problem is
L-E E p’{——1—[ ' - 'v](1—0')+7r [0) '+R B' — ' —B' —
_ O 0 1-0'Ct Wt t tnt t-l t—l ct t
t:

8 . . 2
_ _ _ *
2 (B t B ) ]}
where ”t is the Lagrangian multiplier associated with the budget constraint.

The first order conditions with respect to consumption c; , labor supply n; , bond

holdings B; are:

50

I . ’ - ’ v -0 -

[Ct]' [Ct IA nt ] — It” (15)
., '-I//'v-oI/Iv'v-l_flw

[nt]. [Ct nt ] nt t t (16)
’ E ’ _ B' _- ,6

[Bt] ”t [1+ (BI *)] Et 7ft 1Rt (17)

Equation (15) States that consumers choose consumption such that the marginal utility
of consumption equals the marginal utility of wealth. Equation (16) defines
consumers’ labor supply schedule, by equating marginal disutility of labor to the
marginal utility of wealth brought by real wage. Equation ( 17) is the non-arbitrage
condition for bond holdings in equilibrium.

Combining equations (15) and (16), we obtain the optimal labor supply condition

I” U ’ U" =
nt (at (18)
Combining equations (15) and (17), we obtain

[c't-V/n',v]_°[1+e(B',—B'*)1=Etlc'Hl-Wn',“v1_0(flR,) . (19)

which is the household’s Euler condition for consumption.

3. Equilibrium and Dynamics
3.1. Equilibrium

Given kO , 8'0 and a sequence of interest rates Rt and productivity At ,
equilibrium of this economy is a sequence of allocations {ct’nt’kt’Bt } for

entrepreneurs, a sequence of allocations {c't ,n't ,B't } for households and prices

{ wt } such that:

(i) The entrepreneur solves his optimization problem.

(ii) The household solves its optimization problem.

51

(iii) The labor market clears.
The equilibrium of this economy is characterized by equations (4), (7) (binding),

(9), (12), ( 13), (18), (19) and the labor market clearing condition nt = n't

3.2. Steady State and Dynamics

Assume R* ,6 :1 in the steady state. Set A*=1 in the steady state. The steady
states of the endogenous variables n*,k*, y*,c*,c'*,B*,w * are given by:

1— _ _
ind/(a 1)]l/(v l),

n*=[

y*=”aKa—Dn*,
k*=.1_y*,

n
B*=6k*,

c*=ay*+(6—5—%6)k*,

c'*=w*n*+(i—1)B'*,

 

a/(a—l)
aI*=(1-a)fl .
where n=<1-6)fl+>fi6+26—2fl
tnfi

Loglinearize the equilibrium conditions around their steady states, we obtain the
following system of equations: (where we use the condition nt = n't in equilibrium)

yt = At +akt—1+(1_a)nt (24)

52

A _A = *a—I *1-0! A _ *a_] *l—a A
Etct+1 ct {(ak n )EtAt+1+(a(a 1)k n )Etkt+
+(a(1-a)k*a_1n*1—a)Etfit+l-t9(1/,B)Eth}/(ak*a-ln*’-a+
+1—5-0(1/,B)) (25)
At+akt_1—ant=a)t (26)

4:" *2 _ an“ *A _ _ *A :1: Sc" :1: as"
y yt+B Bt—C ct+k kt (1 63k kt—l+w n nt+w n (ot+
(27)
+(1/fl)B*Bt_1+(1/,B)B*Rt_1
Bt:kt (28)
_ —_ _ —A _ A '*A' : — _— — —A' —
0'/(c an))[cct I/lvnvnt]+£B Bt ( 0'/(c I/lnv))EtccH_1
(29)
-0'/(c —r//n v))y/vn vEtnt+1+ Eth
(v— 1) nt = wt (30)

c' at +B'*B't = (0* n * at + a)* n * at + (1/fl)B'*B't _1+(1/,B)B'*Rt _1 (31)

Equations (24) —(31) are the log-linearization equations for the production function
(24), the entrepreneur’s Euler equation for consumption (25), the optimal labor
demand condition (26), the entrepreneur’s budget constraint (27), the entrepreneur’s
borrowing constraint (28), the household’s Euler equation for consumption (29), the

optimal labor supply condition (30), the household’s budget constraint (31).

53

To close the model, we need to specify the processes for the productivity At’ the

world real interest rate RIW and the specification for the real interest rate in emerging

economies Rt' We assume that the percentage deviations from the trend of At and

Rtw follow AR (1) processes, that is,

A: :plAt—l +£A,t

(33)

.w_ .
Rt _’02Rt—1+8R

w, t
(34)

Where the disturbances 8 are normally distributed and serially

,e
A,t Rw,t

uncorrelated. The real interest rate in emerging economies Rt is the product of the

world real interest rate Rtwand the country risk premium Dt , i.e. Rt =Rtht .
Following NP (2005), we model fit relating to the productivity as follows:

Dr = —nEtAt+l +814

where 7] >0 is a constant capturing how much country risk premium responds to

expected productivity Shocks, 81 is a normally distributed independent shock. It

I

A

follows that the process for Rt is

.. _.W_ .
R —Rt nEtA

t (35)

+8
t+1 I,t

Equations (24) - (35) describe the dynamics of the economy.

4. Calibration

To quantitatively analyze the model, we calibrate the model to Argentina’s

economy. The parameters are chosen to mimic some of the empirical regularities in

54

Argentina's economy. For comparison purposes, we share the parameter values
chosen in the NP model. A period in the model is assumed to be a quarter. The

remaining parameters need to be chosen are 7,6and B'*. We have assumed 7 < ,6 .

,3 is chosen to be 0.93 to match an average real interest rate of 14.8% in Argentina
(Neumeyer and Perri, 2005). We choose 7 to be 0.90. AS a benchmark case, we
choose 6’ to be 0.33, implying the net worth constraint multiplier ¢ is 0.5. Later, we
will change the value of 9 for sensitivity analysis. The value of B'* is not uniquely
pinned down. We choose it to be - 0.3. This is consistent with the fact as well as the
assumption in our model that the emerging economies are net debtors for the rest of

the world. Table 2.1 summarizes the parameter values.

Table 2.1 Parameter values

 

Preference parameters Symbol Value
Discount factor B 0.93
Utility curvature 0 5
Labor curvature v 1.6
Labor weight 11’ 2.48
Capital share a 0.38
Depreciation rate 5 0.044
Borrowing constraint parameter 9 0.33
Bond holding cost 8 10'5
Shocks
Productivity ,0: 0.95
0 (5 A) 1.98 %
World interest rate p 2 0.81
0 (8 Rw) 0.63%
Induced country risk 77 1.04
0' ( 8 1) 1.7%

 

55

5. Results

In this section, we present impulse response functions and cyclical properties of
the model economy. We also conduct sensitive analysis to test our hypothesis that the
borrowing constraint plays a role in transmitting shocks. The results Show that, in
general, the model economy can predict the empirical regularities observed in

Argentina.

5.1. Impulse Response Functions

Figure 2.1 depicts the impulse response function to a shock in the world interest
rate. On impact, the interest rate shock has no effect on output, because debt
repayment in period t is made at the real interest rate in period t-l (RH). Starting
from period t+1, output declines. A 1% increase in interest rate induces about a 0.9%
fall in output and about 0.6% fall in employment. In the NP model, a positive interest
rate shock decreases labor demand and through the production function, output falls.
So, the degree of the impact of interest rate shock on output is a fraction 1— a of that
on employment. The latter depends on vanda. By assuming that firms need to
borrow all the working capital, they obtain the maximum impact of the world real
interest rate on output is about 0.6%. The impacts of interest rate on output and
employment peak at period t+1 ( the second quarter), then output and employment
quickly approach to their steady states. In our model, an interest rate shock not only
affect labor demand, but also affect investment through the net worth constraint, thus

a positive interest rate shock can impact on output more than on employment. In

56

addition, it will take more time for output and employment to go back to their steady
states. Changing the values of 7,19and B’* will not change the general results. We
conclude that in our model the net worth constraint acts as a financial accelerator,
which amplifies and prolongs the effects of real interest rate shocks.

Impulse responses to a shock In FM

1 I f I

 

0.8 ~ 4

0 6 _ world real Interest rate .

0.4 ~ «

Percent deviation from steady state
o

 

 

 

‘0-2 ' employment ‘
-o.4 - .
-0.6 r
-0.8 output ~
_1 I I I I I I I
0 1 2 3 4 5 6 7 8
Years after shock

Figure 2.1 Impulse Responses to a shock to the world interest rate

Figure 2.1 shows the impulse response functions of output and employment to a
country risk shock. A 1% increase in country risk can induce a 0.3% decline in output
and about a 0.2% decline in employment. Thus, in our model economy, country risk
shocks alone can have nonnegligible effects on the aggregate economy but these

effects are smaller than the effects of world interest rate shocks.

57

Impulse responses to a shock in Rinnov
1 I T I I I r

 

0.8 a

0.6 ~ I

0.4 ~ «

0-2 " country risk spread 3

 

 

 

Percent deviation from steady state

 

 

 

o - 1
employment
-0.2 ~
output
_o.4 l l I 1 l l I
0 1 2 3 4 5 6 7 8

Years after shock

Figure 2.2 Impulse Responses to a shock in the country risk spread

Using a VAR approach, Uribe and Yue (2006) study the impulse response of
aggregate variables to world interest rate shocks. The data set consists of a dynamic
panel data for, including Argentina, seven emerging economies. Our model
predictions of the impulse responses of output and employment to world interest rate
shocks and country risk shocks are similar to their empirical result.

Figure 2.3 shows the impulse response function to a shock in productivity. On
impact, a 1% increase in productivity induces a 1.7% increase in output. The response
of output peaks at 3.6% in year 3. The relatively large response of output to
productivity shocks comes from the specification that country risk spreads are a
function of productivity. Since productivity generally‘exhibits persistence, a high

productivity in period t induces economic agents to expect a high productivity in

58

period t +1. This will reduce country risk spreads as well as the real interest rate, and
relax the borrowing constraint. Thus, the effects of productivity shocks are amplified
by changes in country risk spreads through the borrowing constraint. If productivity
has no effect on country risk spreads, that is, if 7] =0, Figure 2.4 shows that the
impulse response of output to productivity shocks peaks at 2 %, which is 44% lower

than that when country risk spreads are a function of productivity.

Impulse responses to a shock In A
4 r I I I fl

 

      
 

employment

productivity

Percent deviation from steady state
to

 

 

 

 

o I I I L g I I
0 1 2 3 4 5 6 7 8

Years after shock

Figure 2.3 Impulse Response Function to a shock to productivity (17 > 0)

59

Impulse responses to a shock in A
2.5 I I I I T I

 

output

1.5~ *

 

     
    

employment

Percent deviation from steady state

productivity

 

 

 

0 1 2 3 4 5 6 7 8
Years after shock

Figure 2.4 Impulse Response Function to a shock to productivity (7] = 0)

5.2 Cyclical Properties

We simulate the model with HP-filtered series to obtain the second moments
properties of the model economy. Table 2.2 shows the results. For comparison, we
also report the second moments properties of data in Argentina and the small open
developed economies. The Simulation results Show that the correlation between the
real interest rate and output in the model economy is -0.54. In the data, this correlation
is -0.63. The standard deviation of output is 5.9%. In the data, this value is 4.22% in
Argentina and 1.37% in developed economies. The ratio of standard deviation of

consumption to the Standard deviation of output is 1.5. In the data, this ratio is 1.17 in

60

Argentina and 0.9 in developed economies. The volatility of real interest rate in the
model is 3.12%. In the data, this volatility of the real interest rate is 3.87% in
Argentina and 1.66% in developed economies. Thus, our model predicts that the
volatility of output in emerging economies is larger than that in developed eConomies,
that consumption is more volatile than output in emerging economies, and that the
real interest rate has larger volatility in emerging economies than in developed

economies. All these are consistent with the regularities in emerging economies.

Table 2.2 Cyclical properties

 

Model Argentina Developed economies
Correlation
Between Y and R -0.54 -0.63 0.20
%Standard deviation
of Y 5.9 4.22 1.37
%Standard deviation 3.12 3.87 1.66
of R
%Standard deviation 1.5 1.17 0.9

of C / %Standard
deviation of Y

 

notes: data in the table are taken from Neumeyer and Perri (2005). Data in developed
economies are average data for the five small open economies: Australia, Cananda,
Netherlands, New Zealand and Sweden.

5.3. Sensitivity Analysis

In this subsection, we conduct sensitivity analysis to test our hypothesis that the
borrowing constraint plays a role in transmitting shocks to aggregate variables. In the

benchmark model, we set 9 equals 0.33, which corresponds to the parameter for net

61

worth constraint (0 being 0.5. Now, we set ‘9 equals to 0.44 ((0 :08) and 0.5 ((0:1)
respectively. We redo the simulation and compare the results we obtain from different
values of 9. Table 3 summarizes the results. We find that the negative correlation
between the real interest rate and output is larger for more borrowing constrained
economy than less borrowing constrained economy, which are the expected results.
However, the volatility of output is large for less borrowing constrained economy
than more borrowing constrained economy. This may be because more production

activities are associated with higher volatility of output.

Table 2.3 Results for sensitivity analysis

 

Correlation between Y and R % Standard deviation of Y
9 = 0.33 -054 5.9
0 = 0.44 -o.43 7.29
6 = 0.5 -0.38 8.38

 

5.4 The relative importance of the net cash flow and the undepreciated capital
stock as debt collateral

To assess the relative importance of different components of the entrepreneur’s

net worth, i.e. the net cash flow (W,*) and the undepreciated capital Stock ( (1- 5 ) km
) as debt collaterals in generating the negative correlation between real interest rates

and business cycles, we consider two cases. First, we assume that the entrepreneur’s

62

borrowing is a fraction ((PI) of his net cash flow only. Second, we assume that the
entrepreneur’s borrowing is a fraction ((02) of the undepreciated capital stock only.
These modifications change the budget constraints, the Euler equations as well as the
borrowing constraints for the entrepreneur’s problem correspondingly. Assuming $1:

0.5 and ¢2=0.5, we redo the quantitative analysis and find that the undepreciated

capital stock as the debt collateral plays an important role in generating the negative
correlation between real interest rates and business cycles, while the role of the net
cash flow as the debt collateral is insignificant. Figure 2.5 and Figure 2.6 show the
impulse response functions to a world interest rate shock in these two cases
respectively. A 1% increase in the real interest rate can only induce a less than 0.2%
decline in output in the first case, while it can induce about a -1.6% decline in output
in the second case. However, when the undepreciated capital stock is the debt
collateral only, the real interest rate is slight procyclical at the beginning of the
impact. In additon, the magnitudes of out and employment responses to a world real
interest rate shock are relatively larger. For our model to better replicate the
relationship between real interest rates and business cycles in emerging economies,
assuming that both the net cash flow and the undepreciated capital stock function as

the debt collateral is more appropriate.

63

Impulse responses to a shock In Rw
1.2 . . fl . . .

 

0.8 ~ -

world real Interest rate
0.6 ~ I

Percent deviation from steady state
9
fl

 

 

 

 

 

0.2 ~ —
employment
0 a
V
. \ output

_0.2 l l I I I I I

, 0 1 2 3 4 5 6 7 8
Years after shock

Figure 2.5 Impulse Response to a shock to the world interest rate

(net cash flow as debt collateral

Impulse responses to a shock In Rw

1 I f i f l I l

 

P
O
1
I

world real Interest rate

.0
or
I

P
b
I

P
N
1
1

 

 

O

    
 

.6
to

employment

Percent deviation from steady state
.6
‘3

.6
9

 

 

b—
l-

l

1 2 3 4 5 6
Years after shock

b—

 

-0.8
0

Ni-
Q

Figure 2.6 Impulse Response to a shock to the world interest rate

( undepreciated capital stock as debt collateral)

64

6. Conclusion

In this paper, we introduce a net worth constraint into the small open economy
RBC model to study the effect of real interest rate shocks on the real economy in
emerging economies. We find that real interest rate shocks have significant effects on
the real economy due to the two way link between the real interest rate and the real
economy and the presence of the net worth constraint. The model also predicts that
output, consumption and real interest rates are more volatile in emerging economies

than in developed economies.

65

Chapter 3
The Determinants of Country RiSk Spreads:

Evidence from Argentina

1. Introduction

Since 1980s, it has become a common phenomenon for many emerging
economies to participate in the international financial markets. An important fact is
that when these economies raise funds through issuing bonds or obtaining loans, they
are charged by foreign investors an interest rate which is higher than the world
interest rate. The difference between the interest rate faced by emerging economies
and the world interest rate is called the country risk spread. Since different countries
face different country risk spreads, it raises the question about what determine the
country risk spread. Edwards (1984) shows that the country risk Spread is determined
by the default risk of a country. In principle, a higher default risk is associated with a
higher country risk Spread. Naturally, the factors that determine the default risk are
the ones that determine the country risk spread. In this paper, we use monthly data
during the period January 1998 to December 2006 to study the determinants of the
country risk Spread in Argentina. Following most of the current literature, we choose
the J .P. Morgan EMBI+ spreads as a measure of country risk spreads. We find that
broadly speaking, international investors tend to take into account the default risk of

the specific emerging countries, where the extent of the risk is associated with the

66

condition of the fundamentals. However, the country risk spread is not a perfect
measure of the default risk. Other factors which may not represent the existence of the
default risk, such as contagion effect, memory, IMF bailout expectation, etc, also play
a role in determining the country risk spread.

The empirical Study of the determinants of the country risk Spread is important
for several reasons. First, it provides a measure of the market’s ability to pricing the
default risk in emerging economies. It is argued that recent debt crises in emerging
countries (as well as LDCs) may be due to foreign lenders’ inability to assess these
countries’ default risk appropriately. By studying the determinants of the country risk
Spread, we can investigate to what extent the international financial market has taken
into account the risk characteristics of emerging economies. Second, empirical studies
suggest that there is a close relationship between real interest rates and business
cycles in emerging economies. Since most emerging countries are open to
international financial markets, real interest rates in these economies are closely
determined by country risk spreads. Thus, understanding the determinants of the
country risk spread can help us to make progress on understanding business cycles in
emerging economies. Third, understanding the determinants of the country risk spread
is useful for emerging economies to take positive steps towards reducing the country
risk spread.

The issue on the determinants of the country risk spread has been addressed by a
number of papers. Edwards (1984) studies the role of a set of fundamentals in

determining the country risk spread. He finds that debt-output ratio is an important

67

determinant of the country risk spread. Eichengreen and Mody (1998) study the
determinants of emerging economies’ bond launch spreads in the years 1991-1996.
They find that although fundamentals are related to the country risk spread, changes
in market sentiment instead of changes in fundamentals explain most of changes in
the country risk spread. Dell’Ariccia et a1 (2006) analyze the effects of non-bailout of
Russian in 1998 on sovereign bond spreads in emerging markets. They observe that
the Russian crisis was followed by significant increases in the levels of spreads as
well as significant increase in the cross-sectional dispersion of spreads. Jahjah and
Yue (2007) analyze the impact of exchange rate policy on the sovereign bond spreads.
Nogues and Grander (2001) study the determinants of country risk spread in
Argentina for the period 1994 to 1998.

The present paper differs from previous literature in several aspects. First, most
papers (Edwards(1984), Eichengreen and Mody (1998), Dell’Ariccia et a1 (2006),
Jahjah and Yue (2007), etc) on the determinants of country risk spreads are cross-
country studies. The advantage of using cross-country data is that the sample is larger.
Thus, more information can be used in analysis. However, since the country risk
spread may be affected by country specific factors or events, it might be more
appropriate to use country specific data. Second, our paper uses an updated data (Jan
1998 to Dec 2006) in Argentina. Since IMF refused to bailout Argentina in 2001
when it went into debt crisis, using the updated data allows us to examine the effects

of bailout expectations on the country risk spread. Third, to our knowledge, our paper

68

is the first one which considers the effects of foreign investors’ memory on the
country risk spread.

The remainder of the paper is organized as follows. Section 2 analyzes the factors
that affect the country risk spread. Section 3 specifies the econometric model and

presents the results. Section 4 concludes.

2. The Determinants of Country Risk Spreads

We argue that the country risk spread in Argentina is affected by a set of
fundamentals and nonfundamentals variables. In particular, the country risk spread is
a function of the following variables:

(1’ (2’ This variable can be considered as an

1) the external debt — output ratio.
indicator of the degree of solvency in Argentina. It is expected that an increase in this
ratio will result in a higher probability of default and have a positive effect on the
country risk spread. The data on external debt was obtained from Joint External Debt
Hub developed by BIS, IMF, OECD, and the World Bank. The data on GDP was
obtained from International Financial Statistics (IFS).

2) the import - output ratio. This variable measures the degree of openness. In
general, a more open economy is more vulnerable to foreign shocks. It is expected

that this ratio is positively related to the default risk and the country risk spread. This

variable was constructed from data obtained from IFS.

 

(1) We define the external debt as the sum of Brady bonds, loans from foreign controlled and
foreign located banks, international debt securities, and officially supported trade credits.
(2) Data on external debt, import and GDP are only available quarterly. We employ cubic spline

to transform them into monthly data.

69

3) GDP growth rate. It has been argued that a higher rate of growth of output will
result in a lower probability of default, and thus a lower country risk spread.

4) Inflation rate. This variable is used as a proxy for sound macroeconomic
management. A lower inflation rate indicates effective fiscal and monetary policies. We
expect that an economy with low inflation will have a lower probability of default, and
thus lower country risk Spread. This variable was constructed from data obtained from
IFS.

5) The country risk Spreads in Brazil and Russia. These two variables are chosen
to capture the contagion effects. It is documented that country risk spreads comove
across countries (Mauro et al., 2002). This suggests that there exist contagion or
spillover effects from one country’s spreads to other countries’ spreads. Contagion
effects can be differentiated as fundamentals-based contagion and nonfundamentals-
based contagion. The fundamentals-based contagion could arise due to both trade
links and financial links between two countries. The nonfundamentals-based
contagion could result from herd behavior. Due to incomplete information in the
international financial markets, when foreign investors see a higher probability of
default in one country, they tend to infer that the probability of default of other
countries with similarities will increase as well even though the fundamentals of other
countries are good. We expect that an increase in Brazil’s or Russia’s spread will

have a positive effect on Argentina’s spread. We also expect that the contagion effect

70

from Brazil’s spread to Argentina’s spread will be larger than that from Russia’s
spread to Argentina’s spread. Argentina and Brazil are both Latin American countries.
Their economies are closely linked. For example, Brazil is Argentina’s main trading
partner. Brazil is the destination of about 30 percent of Argentina’s exports. On the
contrary, the similarities and the degree of economic link between Argentina and
Russia are much less.
6) The one-period lag of the country risk spread in Argentina. This variable
captures the memory effect. We expect the Sign of the coefficient of this variable is
positive. Foreign investors tend to hold memories of past performances of emerging
economies. A high probability of default in the previous period leads foreign lenders
to think that the probability of default in the current period might be high. However,
the probability of default in the current period is determined by the economic
conditions in the current period. In other words, there is no direct link between the
previous period spread and the default risk.
7) The 10-year US. Treasury bond rate. This variable is a proxy for global
economic condition. This Sign of the coefficient of this variable in the regression
model could be positive or negative. This can be seen from analyzing the supply and
demand for the emerging bond market.

In this market, emerging countries issue bonds, so they are suppliers. Foreign
investors purchase bonds so they are demanders. The 10-year US. Treasury bond is a
substitute for the emerging bond. When the 10-year US. Treasury bond rate

increases, for the supplier, this means the cost of borrowing increases (since interest

71

rate in emerging economies equals the world interest rate plus the country risk
spread). It will lead to the supply curve of emerging bonds market shifting left,
resulting in an increase in the price of the emerging bond and a decrease in the
country risk spread (we call it the cost effect). For the demander, the increase in the
US. Treasury bond rate make them switch to purchase the US. Treasury bond and
decrease the demand for emerging bond. Thus the demand curve for the emerging
bond market shifts to the left and results in a decrease in the price of the emerging
bond and an increase in the country risk spread (we call it the substitution effect).
The ultimate effect of an increase in the US. Treasury bond rate on the country risk
spread depends on which of the above two effects are stronger.

8) IMF bailouts expectations. IMF bailouts are not a determinant of the default risk.
However, IMF bailouts are expected to reduce investors’ losses in case of default.
When foreign investors believe that IMF will bailout emerging countries in case of
default, they tend to think that the risk of holding emerging market bonds is less than
that without IMF bailout. This is the so called investors’ moral hazard. Thus, an
increase in the belief of IMF bailout is expected to be associated with a decrease in
the country risk spread.

All through the end of 1990s, Argentina did not experience non-bailout case.
Argentina was regarded by IMF as important to the region of Latin America and as a
major test-case for the liberalization program, the “Washington Consensus” policies
advocated by IMF (Noy, 2003). In 2001, Argentina went into a serious debt crisis due

to huge budget deficit, political turmoil and high interest burden. Although IMF

72

committed to bailout Argentina throughout 2001, it refused to provide loan
installment in December 2001. IMF'resumed its bailout to Argentina in late 2002. We
believe that at least in the first half of 2002, foreign investors’ expectations about
IMF bailout were significantly lower than the rest of time in our sample. Hence, we
set IMFbtex = 0 for the period 2002ml-2002m6 and IMFbtex = l for the rest of the

period.

3. Econometric Analysis and Results
To analyze the effects of fundamentals and nonfundamentals on the country risk

spread in Argentina, consider the following econometric model;

log sprdst = '60 + '61 log(exdt/ GDP)! + ,52 log(imp/ GDP)t + '63 (GDPgr)t

+ ,64 inf It + ,85 log Brazilspt + ,6 6 log Russiaspt + ,67 log sprdst _1 (1)

+ ’68 log USTRBI + flglMFbtext + at

where at is the identically distributed stochastic error term. The dependent variable

and independent variables are in logarithmic forms so that the coefficients of the
independent variables measure the elasticities of the country risk spread. However,
some of the data on GDP growth rate and inflation rate are negative and cannot be
transformed into logarithmic data. The coefficients of these two variables have a

semi-elasiticty interpretation.

73

Since the presence of the one-period lag of spreads in the model, we suspect that
the errors are serially correlated. We conduct serial correlation test for the errors and
find that we cannot reject the null hypothesis that the errors are serially correlated
(see appendix A). This means that we cannot use the usual t test to make inference
about the coefficients. To correct the problem of serial correlation, we estimate the
coefficients of the explanatory variables by OLS, but use the heteroskedastic and
autocorrelation consistence (HAC) standard error (Newey-West standard error).

Table 3.1 shows the estimation results. In general, the results are satisfactory
from the perspective of the signs and level of significance of the coefficients. As

expected, increases in the ratio of external debt to GDP and the ratio of import to

GDP lead to increases in the country risk spread ( ,6] =0.63,,62 =1.27 ). These

effects are statistically Significant (t . = 2.25, t .. = 2.66 ). The effect of inflation on
1 2

the country risk spread is economically large. A 1% increase in the inflation rate

causes a 34% increase in the country risk spread. However, this effect is only

marginally statistically significant (t ,B = 1.55 ). The effect of the GDP grth rate
4

on the country risk spread has an unexpected sign ( ,63 = 0.057 ). This may be because

the monthly GDP data is obtained by interpolation and it incurs distortions when the

GDP growth rate is calculated. Since the GDP growth rate is statistically very

insignificant (t )3 = 0.3 ), we can ignore this variable without changing the regression
3

results.
Most of the nonfundamentals variables have significant effects on Argentina’s

country risk spread. As expected, both Brazil’s spread and Russia’s spread have

74

positive effects on Argentina’s spreads. In particular, a 1% increase in Brazil’s spread
leads to a 0.24% increase in Argentina’s spread. A 1% increase in Russia’s spread

leads to a 0.13% increase in Argentina’s spread. The Brazil’s spread is statistically

significant (t . =2.82 ), while the Russia’s spread is not statistically significant

5
(t A = 1.14). This suggests that contagion effects are more likely to occur between

6

countries with similarities.
The previous period spread and IMF bailout expectations also have significant
effects on Argentina’s country risk spread. A 1% increase in the previous period

spread leads to a 0.84% increase in the current period spread and this effect is very

statistically significant (t x = 12.12 ). This indicates that foreign investors’ memory
7

plays an important role in determining the country risk spread. The coefficient on the
variable, IMFbtex, is -0.18, meaning that with IMF bailout expectation, the country

risk spread is lower by 18% than that without IMF bailout expectation. The variable

IMF btex is statistically significant 0,8 = —3.01), indicating that investors’ moral
9

hazard does exist in the international financial market. The coefficient of the US

Treasury bill rate has a positive Sign and it is marginally Statistically significant

(t . = 1.48 ), which suggests that the substitution effect dominates the cost effect.
8

In summary, our estimation results suggest that factors affecting the default risk
have nonneglible effects on the country risk spread. On the other hand, factors which
are not related to the default risk also play a significant role in determining of the

country risk spread.

75

To ensure that the regression results in Table 3.1 is not a spurious regression; we
first conduct unit roots tests for all the variables and find that except for the external
debt to GDP ratio, we cannot reject the null hypothesis that these variables have unit
roots (see Appendix B). We then check the existence of a long-run relationship

Table 3.1 Spreads Function Estimation 1998-2006

 

 

Variable coefficient t-statistic probability

log(exdt/GDP) 0.63 2.25 0.026
(0.2794)

log(imp/GDP) 1.27 2.66 0.009
(0.4779)

GDPgr 0.058 0.30 0.764
(0.1916)

Infl 0.347 1.55 0.124
(0.2232)

log(BrazilSp) 0.24 2.82 0.006
(0.0849)

Log(Russiasp) 0.13 1.14 0.255
(0.1 143)

log Sprds(t-l) 0.842 12.12 0.000
(0.0695)

log USTRB 0.426 1.48 0.142
(0.2877)

IMFbtex - 0.195 - 3.01 0.003

 

between the dependent variable and the independent variables by using the Engle-

Granger cointegration test. The t-statistic on the residual tit_l is —7.345. It is less
than the critical value —3.34 (5% Significant level). Thus, we reject the null

hypothesis that the dependent variable, log sprdst, and the independent variables are

not cointegrated. In other words, our test verifies that the regression results reported

76

in Table 1 represents a long-run relationship between the dependent variable and the
independent variables.

In order to study the short-run dynamics in the relationship between the
dependent variable and its determinants, we estimate the following error correction

model:

Alogsprdst=a0+6iit_1+7AXt+£t (2)

where X t represents all the independent variables except the dummy variable

IMFbtex. ft is an independently identically distributed error. tit is the one period

—1
lag residual obtained from the estimation of equation (1). Sfit_l is the error

correction term. It is expected that 5 to be negative so that when the dependent

variable, log Sprds in the previous period has been above the equilibrium level, i.e.

A

u >0, the error correction term works to push logSprds back towards the

r—l
equilibrium. Similarly, when logSprds in the previous period has been below the

equilibrium level, i.e. I? <0, the error correction term induces a positive change in

t —1
logSprds back towards the equilibrium. Table 3.2 reports the estimation results. As
expected, the error correction coefficient is a negative number, -0.96, and it is
statistically significant. This means that if the dependent variable, logSprds, in the

previous period is above the equilibrium by 1%, then 0.96% of the gap is closed in

one month.

77

Table 3.2 Error Correction Model Estimation

 

Variable

A

uH

Alog(exdt/GDP)

Alog(imp/GDP)

A(GDPgr)

Ainfl

AlogBrazilsp

AlogRussiaSp

AlogUSTRB

Cons

coefficient t-statistic
- 0.96 - 3.69
(0.2626)

- 0.117 - 0.08
(1.4927)

0.866 0.34
(2.549)

- 0.357 ’ - 0.65
(.55)

- 0.637 - 1.09
(0.5829)

0.388 1.81
(0.214)

0.195 0.98
(0.199)

0.406 0.9
(0.453)

0.015 0.64
(0.023)

probability

0.000

0.938

0.735

0.518

0.277

0.073

0.332

0.373

0.523

 

78

4. Conclusion

In this paper, we use monthly data during the period 1998-2006 to study the
determinants of the country risk spread in Argentina. We find that the country risk
spread is not a perfect measure of the default risk. Our estimation results suggest that
factors affecting the default risk have nonneglible effects on the country risk spread.
On the other hand, factors which are not related to the default risk also play 'a
significant role in determining the country risk Spread.

Our analysis implies several potential ways to reduce the country risk spread. If
increases in the external debt or imports have the extemality of increasing the country
risk spread, then imposing external debt tax or import tax are expected to reduce the
country risk spread. Moreover, given the significant effects of contagion and memory
on the country risk spread, a more transparent information revelation on the
performance of emerging economies will be helpful for foreign investors to access the

default risk appropriately.

79

Appendix A: Data Sources

Data Sources (I)

 

 

Quarters HPI
Emerging Economies
Argentina 1994Q1-2006Q4 EcoWin (GPR)
Korea 1994Q1-2005Q1 the Bank of Korea
Malaysia 1999Q1-2006Q4 Ministry of Finance
Mexico 1 994Q1-2006Q4 OECD
Slovak l995Q1-2006Q4 OECD
Thailand 1994Ql -2003Q4 Bank of Thailand
Turkey 1994Q1-2006Q1 Central Bank of Turkey

Developed Economies
Austria

Australia

Canada

Denmark

Netherlands

New Zealand

Sweden

1994Q1-2006Q4
1994Q1-2006Q4
1 994Q1-2006Q4
1994Q1-2006Q4
1994Q1-2006Q4
1994Q1-2006Q4

1994Q1-2004Q4

Oesterreichische National Bank

Australia Bereau of Statistics

Statistics Canada

Statbank Denmark

NVM

Reserve Bank of New Zealand

Statistics Sweden

 

Notes: RVD = Ratings and Valuation Department

NVM = Nederlandse Vereniging van Makelaars

80

 

Data Sources (11)

 

Quarters

Interest rate

GDP GDP deflator CPI

 

Emerging Economies

 

Argentina 1994Q1-2006Q4 EMBI IFS IFS IFS
Korea 1994Q1-2005Q1 EMBI OECD IFS IFS
Malaysia 1999Q1-2006Q4 CBM IFS IFS IFS
Mexico 1994Q1-2006Q4 EMBI OECD IFS IFS
Slovak 1995Q1-2006Q4 IFS IFS IFS IFS
Thailand 1994Q1-2003Q4 EMBI IFS IFS IFS
Turkey 1994Q1—2006Q1 IFS CBT IFS IFS
Developed Economies

Austria 1994Q1-2006Q4 OECD OECD IFS IFS
Australia 1994Q1-2006Q4 OECD OECD IFS IFS
Canada 1994Q1-2006Q4 OECD OECD IFS IFS
Denmark 1994Q1—2006Q4 OECD OECD IFS IFS
Netherlands 1994Q1-2006Q4 OECD OECD IFS IFS
New Zealand 1994Q1-2006Q4 OECD OECD IFS IFS
Sweden 1994Q1-2006Q4 OECD OECD IFS IFS
Notes

 

IFS = International Financial Statistics (IMF)

EMBI = J .P. Morgan Emerging Market Bond Index

81

CBM 2 Central Bank of Malaysia CBT = Central Bank of Turkey

 

 

 

Data Sources (111)
Countries Periods Bank Balance Sheet Data
Argentina 2003m9-2007m12 Banco Central De La Republic Argentina
Korea 1999Q2-2006Q1 Monthly Financial Statistics Bulletin
Korea Financial Supervisory Service
Malaysia 2004m1-2007m12 Bank Negara Malaysia
Mexico 2001m1-2007m12 Bank of Mexico
Slovak 2003m1-2007m12 National Bank of Slovakia
Thailand 1997m1-2003m12 Bank of Thailand
Turkey 1994m1-2006m 12 Quarterly Bulletin .
Central Bank of the Republic of Turkey
Notes:

(1) For Argentina, bank balance Sheet data is drawn from “Private Banks’ Balance
Sheet”. The item “Net Worth” is regarded as capital.

(2) For Korea, bank balance sheet data is drawn from “Summarized Balance Sheet of
Bank Account of Commercial Banks”. The item “Loans and Discounts in Won” is
regarded as loans. The item “Total Shareholders’ Equity ” is regarded as capital.

(3) For Malaysia, bank balance Sheet data is drawn from “Commercial Banks’
Statement of Assets and Liabilities”. The item “Capital and Reserves” is regarded as

capital.

82

 

is regarded as loans. The item “Other Liabilities and Capital, Net from Other Assets”
is regarded as capital.

(5) For Slovak, bank balance sheet data is drawn from “Monthly Report on Balance
Sheet Items of Credit Institutions”.

(6) For Thailand, bank balance sheet data is drawn from “All Commercial Banks’
Assets and Liabilities (1997-2003)”. The item “Credits” is regarded as loans. The
item “Stockholders’ equity” is regarded as capital.

(7) For Turkey, bank balance sheet data is drawn—from “Deposit Money Banks

Balance Sheet”. The item “Credits” is regarded as loans.

83

Appendix B. The Steady State
We normalize the steady state real estate price to q = 1.

The steady state is described by the following equations:

 

 

¢
1+ l-Ro
c' l—y—v
E 4” (29)

 

c'zl—fl—tflR—IW

l

1 mp
—— R 1——+
(R 7) 0'( m ) Wh'

 

3= H (30)
” 1—(1- )m —
R 7’ F 7’
h+h'=1 (31)
Roh'
: 32
h ( )
RBH =amHh (33)
F _
RB 41—00th (34)
wn=(1—,u—V)y (35)
1= 7(fl%+1-5) (36>
y+BH+BF =c+dk+RBH +RBF+wn (37)
R=oh' as
c'+ 3': RB'+am (39)

84

Appendix C. Loglinearized Equations

am (1— a)m

 

 

L ,. _‘ _ i- _ _ c __ H F c
”’qu hr) ovqh aer (1 “)mF7+i’)cr+l (1+ R + R )c,
cm: (1—a)m am
— E— ‘ — H F " _ _ _ __1i
+(mH mFxR 7am, ( R + R >R, q, (mHarHl oomFr R
(l—a')mF ..
-——-§—-—-r)q,+1 =0 (40)
A A H_e _,. _A :
Rt+Bt at qt+1 ht 0 (41)
A .F a A A _a _
Rt+B t 1_aa't qt+1 ht—O (42)
at+qt+1+ht—qt—ht_l—Rt=0 (43)
. ~_ _ a ._ . H“ H HA F“ F
cct+kkt (l 5)kkt_1+qhht qhht_1+RB Bt_l +RB Rt_1+RB Bt_1
+RBFR +comi +0)n(?) —y9 —BHBH—BFBF=O (44)
{—1 t I t t I
yt _wt —nt =0 (45)
gm“. . 1 ”yr, 1 fl]; =0 (46)
‘ py+1—5k —y—+1—6k

85

 

461

_£*1 ia_~ _ie _.
ht+c'(ct 'BCHI) C'qt+c.q

h. + E330 614'

t+1 t+1

 

 

RB'B '+RB'Rt _

t—l 1

hht+hht =0

86

+anc?) +ami -c'E' —qh'h '+qh'h'
t t t t t-

1

+ ,6/1'0 q/i't +

A

—B'B' = 0
t

(47)

(48)

(49)

(50)

(51)

(52)

(53)

(54)

(55)

(56)

 

Appendix D Serial Correlation Test

We test the serial correlation for the error terms in two steps:

Step 1 : Run the OLS regression of log(sprds)t on
log(exdt / GDP)t , log(imp / GDP), , (GDPgr)t , Inflt , log(Brazilsp)t , log(Russiasp)t ,

log Sprds log USTRBt,IMFbtext and obtain the OLS residuals lit for all t = 1,

t-l’
2....n.

Step 2: Run the regression of tit on log(exdt/GDP)t , log(imp/ GDP) t’ (GDPgr)t,

Inf lt’ log(Braztlsp)t , log(Russrasp)t, log Sprdst_1, 10g USTRBt' IMFbtext' ut_1

forallt=2, 3, ...... n.
The t-statistic for the coefficient on [it—l is 3.58. Hence, we reject the null

hypothesis that the coefficient on 12 is 0, which means that the error terms follows

t—l

AR(1) serial correlation.

87

 

Appendix E Unit Roots Tests
We conduct Augmented Dickey-Fuller (ADF) test for all the variables to check
whether these variables contain unit roots. Since we don’t find significant trends for

all the variables, we use ADF test based on the following specification:
n
Ayr =a+0yt_1+ .ZlyjAyt-j+£t
J 3:

Table 3.3 shows the results

Table 3.3 Unit Roots Tests

 

Variable optimal lags t-value (H0 : 9 = 0)

log(sprds) 1 - 1.32
log(exdt/GDP) 4 - 8.81
log(imp/GDP) 5 - 2.11
GDPgr 4 - 1.62
Infl 5 - 1.99
log(Brazilsp) 1 - 1.15
log(Russiasp) 1 _ - 0.79
log(sprds(t-l)) 2 - 1.14
log(USTRB) 2 - 1.8

 

( 1) The critical value at 5% level is - 2.89.
(2) The optimal lags are chosen by examining the t statistic value for lagged variables.

If the lagged variables are insignificant, we drop them.

88

 

Bibliography

[1] Aghion, P., P.Bacchetta, A.Banerjee (2004) A Corporate Balance-Sheet
Approach to Currency Crisis, Journal of Economic Theory, 119, pp. 6-30

[2] Aguiar, M. and Gopinath, G. (2007). Emerging Market Business Cycles: The
Cycle is the Trend, Journal of Political Economy, 115(1), 69-102

[3] Bemanke, B., Gertler, M. and Gilchrist. S. (1998). The financial Accelerator in a
Quantitative Business Cycle Framework, NBER Working Paper No.6445.

[4] Bemanke, B., Lown, C. and Friedman, B. (1991). The Credit Crunch, Brookings
Paper on Economic Activity. 1991(2), 205-247

[5] Caballero, R. and Krishnamurthy, A. (2001). International and Domestic
Collateral Constraints in a Model of Emerging Market Crises, Journal of Monet_ary_
Economics, 48(3), 513-548

[6] Chen, N. (2001) Bank Net Worth, Asset Prices and Economic Activity, Journal
of Monetary Economics . 48 (2)

[7] Chiuri, M., Ferri, G. and Majnoni, G. (2001). The Macroeconomic Impact of
Bank Capital Requirements in Emerging Economies: Past Evidence to Assess the

Future, Working Pamr.

[8] Chiuri, M., Ferli, G. and Majnoni, G. (2001). Enforcing the 1988 Basel Capital
Requirements: Did it Curtail Bank Credit in Emerging Economies? Economic Notes,
30(3), 399-419.

[9] Cooley, T. (Editor) (1995) Frontiers of Business Cycle Research, Princeton
University Press.

[10] Correia, I., Neves, J .C. and Rebelo, S. (1995). Business Cycles in a Small Open
Economy. European Economic Review , 39 (6), 1089-1113.

[11] Dell’Ariccia, G., Schnabel, I., and Zettelmeyer, J. (2006) How Do Official
Bailouts Affect the Risk of Investing in Emerging Markets? Journal of Money,
Credit, and Banking, 38(7), 1689-1714.

[12] Edwards, S. (1984) LDC Foreign Borrowing and Default Risk: An Empirical
Investigation, 1976-80. Americg Economic Review. 74(4), 726-734.

89

 

[l3] Eichengreen, B. and Mody, A. (1998) What Explains Changing Spreads on
Emerging Market Debt: Fundamentals or Market Sentiment? NBER Working Papgr,
No. 6408.

[14] Farmer, R. (1999) Macroeconomics of Self—fulling Prophecies, The MIT Press.

[15] Goldstein, M. and Turner, P. (1996) Banking Crises in Emerging Economies:
OrginS and Policy Option, BIS Economic Papers. No.46, October 1996.

[16] Herring, R]. and Wachter, S. (1999) Real Estate Booms and Banking Busts:
An International Perspective, Working Papgr, The Wharton School, University of
Pennsylvania

[17] Holmstrom, B. and Tirole, J. (1997). Financial Interrnediation, Loanable Funds,
and The Real Sector, The Quarterly Journal of Economics, Vol.CXII, Issue 3

[18] Iacoviello, M. and Minetti, R. (2006). International Business Cycles with
Domestic and Foreign Lenders, Journal of Monetm Economics, 53(8), 2267-2282

[19] Iacoviello, M. (2005). Housing Prices, Borrowing Constraints, and Monetary
Policy in the Business Cycle, American Economic Reviewm, 95(3), 739-764

[20] International Monetary Fund (2004) Debt-Related Vulnerabilities and Financial
Crises—An Application of the Balance Sheet Approach to Emerging Market
Countries.

[21] Jahjah, S. and Yue, V. (2007) Exchange Rate Policy and Sovereign Bond
Spreads in Developing Countries, Working Papgr, IMF.

[22] Kaminsky, G. and Reinhart, C. (2000) On Crises, Contagion, and Confusion.
Journal of International Economics. 51(2000), 145-168.

[23] Kehoe, T. (2003). What Can We Learn from the Current Crisis in Argentina?
ScottiaLloumal of Political Economy. 50(5), 609-633.

[24] Kiyotaki, N. and Moore, J. (1997). Credit Cycles, Joum_al of Political Economy.
105(2), 211-248

[25] Kydland, F. and Zarazaga, C. (2002). Argentina’s Lost Decade, Review of
Economic Dmmics. 5(1), 152-165

[26] Li, N. (2007). Cyclical Wage Movements in Emerging Markets Compared to
Developed Economics: A Contractual Approach, Dissertation, Stanford University

90

 

[27] Maia, J. and Kweitel, M. (2003). Argentina: Sustainable Output Growth After
the Collapse, Workding paper

[28] Mauro, P., Sussman, N. and Yafeh, Y. (2002) Emerging Market Spreads: Then
versus Now. The Quarterly Journal of Economics. 695-733

[29] Mendoza, E. (1991). Real Business Cycles in a Small-Open Economy.
American Economic Review 81(4), 797-818.

[30] Mendoza, E. and Terrones, M. (2008) An Anatomy of Credit Booms: Evidence
from Macro Aggregates and Micro Data, NBER Working Pagr, No. 14049.

[31] Minetti, R. (2007). Bank Capital, Firm Liquidity and Project Quality, Journal of
Monetarv Economics. forthcoming

[32] Neumeyer, PA. and Perri, F. (2005) Business Cycles in Emerging Markets:
The Role of Interest Rates, Journal of Monetary Economics, 52(2), 345-380.

[33] Nogues, J. and Grandes, M. (2001) Country Risk, Economic Policy, Contagion
Effect or Political Noise? Journal of Applied Economics. IV (1), 125-162.

[34] Noy, I. (2003) Do IMF Bailouts Result in Moral Hazard? An Events-Study
Approach, Working Papgr, University of Hawaii at Manoa.

[35] Oviedo, PM. (2005). World Interest Rates, Business Cycles, and Financial
Intermediation in Small Open Economies, Working Paper, Iowa State University

[36] Schmitt-Grohe, S. and Uribe, M. (2003) Clossing Small Open Economy
Models, Joumjfl of International Economics 61(1), 163-185.

 

[37] Uhlig, H. (1997) A Toolkit for Analyzing Nonlinear Dynamic Stochastic
Models Easily, Working papar

[38] Uribe, M. and Yue, V.Z. (2006) Country Spreads and Emerging Countries: Who
Drives Whom?, Journal of International Economics 69(1), 6-39.

[39] Van den Heuvel, S. (2007) The Bank Capital Channel of Monetary Policy,
Working Paper, The Whaton School, University of Pennsylvania.

[40] Wooldridge, J. (2006) Introductory Econometrics: A Modern Approach. Third
Edition, Thomson, South-Westem Press.

91

 

   

lillllglllllljll(litI