Credit Decomposition and Business Cycles in ... - Semantic Scholar
Credit Decomposition and Business Cycles in Emerging Market Economies Berrak Bahadiry
Inci Gumusz
University of Georgia
Sabanci University
July 24, 2014
Abstract This paper analyzes di¤erent types of private sector credit in relation to business cycles in emerging market economies. We …rst provide evidence that the results found in the literature on credit expansions being associated with economic expansions, real exchange rate appreciations and current account de…cits hold more strongly for household credit than business credit. Then, using a two-sector real business cycle model of a small open economy, we study the model dynamics generated by shocks to household credit and business credit in tradable and nontradable sectors. We show that the three types of credit shocks generate di¤erent macroeconomic dynamics in sectoral output and input levels as well as the real exchange rate. We also show that the credit shocks are important in matching key business cycle properties observed in the data, especially the moments related to credit. JEL classi…cation: E32, E44, F34, F41 Key Words: Household Credit, Business Credit, Business Cycles, Credit Shocks We would like to thank participants at the Society for Computational Economics 2012 Meeting, the SNDE 2012 Meeting, the LACEA-LAMES 2012 Meeting, the EEA-ESEM 2013 meeting, the North American Summer Meeting of the Econometric Society 2014 and seminar participants at the University of Georgia, Federal Reserve Bank of Atlanta and Central European University for valuable comments and suggestions. All remaining errors are our own. y Department of Economics, Terry College of Business, 516 Brooks Hall, Athens, GA 30602. Email: [email protected], tel: +1-706-542-3667 z Corresponding author. Sabanci University, Faculty of Arts and Social Sciences, Orhanli, Tuzla, Istanbul, 34956, Turkey. Email: [email protected], tel: +90-216-483-9318
1
1
Introduction
The literature on credit cycles, including Mendoza and Terrones (2008, 2012) and Tornell and Westermann (2002, 2003), has established that credit expansions in emerging market economies are associated with large macroeconomic expansions, widening current account de…cits and real exchange rate appreciations. The standard measure of credit used in this literature is total private credit which includes all types of credit to the private sector without di¤erentiating between lending to households and lending to …rms. These two types of credit may generate di¤erent business cycle patterns given that household credit expansions are likely to a¤ect the economy through an increase in consumption and demand for goods and services whereas business credit has the potential to increase investment and labor demand, and thereby increase output. Furthermore, the allocation of business credit between tradable and nontradable sectors is another factor that may a¤ect the business cycle dynamics. Recent developments in credit markets underline the importance of distinguishing between household and business credit in emerging markets. Figure 1 shows the evolvement of household and business credit-to-GDP ratios for nine emerging market economies and Table 1 shows the average values of these ratios for the sample period. The two types of credit exhibit di¤erent patterns: while lending to households has grown substantially over the period we analyze, the growth in business credit has been much slower in most cases. As a result, household credit has become an important component of overall private credit with potentially important consequences for business cycles. On average, household credit constitutes 33 percent of total private credit in our sample. Despite the increased importance of household credit, existing models that study emerging market business cycles abstract from household credit dynamics.
2
Business credit/GDP
Household credit/GDP 1.4
0.9 0.8
1.2
0.7 1 0.6 0.8
0.5 0.4
0.6
0.3 0.4 0.2 0.2
0.1 0 1983
1987 Brazil
1991
1995
Chile
1999
2003
2007
Czech Republic
0 1983
2011 Hungary
Korea
1987 Poland
1991
1995
1999
South Africa
2003
2007
T hailand
2011 T urkey
Figure 1. Credit-to-GDP ratios in emerging market economies
Table 1. Household and business credit Countries
HC/GDP BC/GDP HC/TC Sample Period
Brazil
11.8
19.7
37.5
1995-2012
Chile
15.4
44.4
25.8
1983-2012
Czech Republic
15.1
51.7
22.6
1995-2010
Hungary
17.5
71.4
19.7
1995-2010
Korea
54.8
92.2
37.3
1983-2012
Poland
14.5
31.6
31.5
1991-2012
South Africa
36.4
33.6
52.0
1994-2012
Thailand
47.0
70.2
40.1
1991-2012
Turkey
6.84
25.5
21.2
1993-2012
Average
24.4
48.9
33.3
-
Note: HC, BC and TC denote household, business and total credit, respectively.
In this paper, we distinguish between household and business credit, as well as the sectoral allocation of business credit, and study the macroeconomic dynamics generated by di¤erent credit types. To this end, …rst we re-examine the documented stylized facts 3
regarding credit cycles di¤erentiating between household and business credit. Using data from a group of emerging market economies, we show that the patterns documented in the literature with respect to credit and key macroeconomic variables are much stronger for household credit: household credit exhibits a strong positive comovement with output, consumption, investment and real exchange rate appreciation, and a strong negative correlation with the trade balance. Business credit, on the other hand, has weaker correlations with all of these variables. These correlations show that the empirical …ndings of the literature on credit cycles are mostly driven by household credit. We then construct a two-sector small open economy real business cycle model to assess quantitatively the role of di¤erent types of credit shocks in driving business cycles. In our model households and both the tradable and nontradable sectors are credit constrained, and …rms have a working capital requirement. The model allows us to disaggregate total credit into three subcomponents: (i) household credit, (ii) nontradable sector business credit, and (iii) tradable sector business credit. We choose to model all sectors as credit constrained since this allows us to study the model dynamics for each type of credit expansion and analyze the e¤ects in each sector separately. We analyze the model under productivity shocks in tradable and nontradable sectors and three types of credit shocks. The credit shocks are modeled as stochastic processes that a¤ect the borrowing limits of the agents. The shocks to credit are similar to the …nancial sector shocks studied in Jermann and Quadrini (2012) who analyze the e¤ects of these shocks in a closed economy model. We calibrate our model to Turkey for the period 1999Q1-2011Q4. The reason we choose to calibrate the model to Turkey is that we are able to obtain detailed sectoral data on credit variables at quarterly frequency for a relatively long time period.1 For our analysis, we not only need data on household and business credit separately, but also business credit data at the sectoral level. To the best of our knowledge, most of the emerging economies for which household credit and business credit data are separately available do not provide the sectoral credit classi…cation. 1
Turkey is a representative emerging market economy that features the standard business cycle properties observed in emerging economies documented by Aguiar and Gopinath (2007) and Neumeyer and Perri (2005): consumption is more volatile than output, trade balance is countercyclical and business cycles are very volatile (see Table 4 for more details). With respect to the correlations of di¤erent types of credit with the key macroeconomic aggregates, Turkey displays the same pattern as the average of the emerging economies in Table 2.
4
The impulse response analysis shows that the three types of credit shocks generate di¤erent macroeconomic dynamics. The starkest di¤erence appears for the real exchange rate response. We observe that an increase in the supply of credit to the nontradable sector leads to a real exchange rate depreciation whereas credit increase to households and tradable sector generates an appreciation. The credit shocks generate di¤erent responses in sectoral output and input levels as well. Nontradable output expands after each credit shock whereas tradable output only increases as a result of a positive credit shock to that sector. Investment in each sector increases as a response to the sector-speci…c credit shocks whereas consumption increases on impact after each shock. Finally, trade balance declines in all cases. The extent to which trade balance responds to credit shocks depends on the share of each type of credit in the overall economy as analyzed in Section 5.2.2. The quantitative results from the model with productivity and credit shocks can account for most of the empirical regularities of the Turkish economy. In particular, the model is successful at matching the key business cycle moments such as the relative volatilities of consumption, labor and trade balance. The model also generates a countercyclical trade balance and a strong negative correlation between real exchange rate and trade balance. More importantly, the model can account for the comovement between the key macroeconomic variables and di¤erent credit types. Household credit exhibits a strong positive correlation with output and real exchange rate, and a strong negative correlation with the trade balance whereas for business credit the correlations are weaker. In the absence of credit shocks, the model performs poorly in all aspects but especially with respect to moments related to credit. When we replace the credit shocks with an interest rate shock, the model still underperforms relative to the baseline model. Therefore we conclude that credit shocks are important in driving the business cycles in emerging market economies. The role of credit shocks for macroeconomic ‡uctuations has been recently investigated primarily in closed economy models. This literature, including Christiano, Motto and Rostagno (2010), Jermann and Quadrini (2012) and Khan and Thomas (2013), …nd that …nancial shocks are important for the dynamics of real and …nancial variables. One recent study by Perri and Quadrini (2014) studies the implications of credit shocks in a twocountry model and focuses on international comovement in real and …nancial variables. The contribution of the current paper is to show that shocks to di¤erent types of credit 5
are important to understand business cycles, as well as the real exchange rate behavior, in emerging market economies. Our paper is further related to the empirical studies on the distinction between household and business credit, such as Büyükkarabacak and Krause (2009), Büyükkarabacak and Valev (2010), and Beck et al. (2012) who underline the importance of di¤erentiating between the types of borrowers. The main conclusion of these papers is that the two types of credit serve di¤erent purposes and have distinct e¤ects on the economy.2 Our paper complements these empirical studies by providing a general equilibrium model that helps understand the transmission mechanism through which each type of credit a¤ects the business cycle properties in emerging market economies.
2
Credit and Business Cycles in Emerging Economies
This section documents business cycle regularities in relation to household and business credit for a sample of emerging market economies for which we could obtain time series data of su¢ cient length. The data for Czech Republic, Hungary, Korea, Poland, Thailand and Turkey are obtained from the Bank for International Settlements (BIS).3 For Brazil, Chile and South Africa we obtain the data from their respective central banks. Due to the lack of credit data at the sectoral level for several countries in our sample, here we analyze total business credit rather than tradable and nontradable sector credit separately. Table 2 shows the correlations between macroeconomic aggregates and the creditto-GDP ratios for household and business credit. These correlations overall show that household credit is more strongly correlated with the business cycle compared to business credit. The correlation between household credit and output is, on average, 0.55 whereas for business credit this correlation is only 0.14. The same pattern holds for consumption 2
Büyükkarabacak and Krause (2009) show that household credit leads to a deterioration in the trade balance, whereas business credit has a small but positive e¤ect. Büyükkarabacak and Valev (2010) …nd that household credit expansions have been a signi…cant predictor of banking crises. Business credit expansions are also associated with banking crises but their e¤ect is weaker. Beck et al. (2012) show that bank lending to …rms is positively associated with growth, while the relationship between household credit and growth is insigni…cant. 3 BIS also provides data for Indonesia, Mexico, Singapore and Hong Kong. For Indonesia the data starts from 2001, which does not give a su¢ ciently long time series. We opt to exclude Hong Kong, Mexico and Singapore because of the strong negative correlation between output and private credit, which contradicts with the general pattern documented in the literature.
6
and investment. The starkest di¤erence between the two types of credit is observed in their correlations with the real exchange rate. There is a very strong comovement between the real exchange rate appreciation and household credit with an average correlation coe¢ cient of 0.52. The business credit-real exchange rate correlation is lower than the correlation of household credit for all countries and negative for most of them, with an average of -0.03. In the case of the ratio of trade balance to output, household credit has a negative correlation in all countries, except for Czech Republic, with an average correlation of -0.42. The correlation between business credit and trade balance is also negative in six out of nine countries. However, the correlations are weaker and the average correlation drops to -0.19. Figures 2-4 plot the time series for three key macroeconomic variables; GDP, real exchange rate, and the ratio of trade balance to output; together with the credit-to-GDP ratios for the two types of credit. These …gures also reveal the di¤erences between the credit types in relation to the cycle. To summarize, the empirical regularities presented in this section show that the stylized facts documented in the literature with respect to private credit are to a large extent driven by household credit rather than business credit in emerging markets. Table 2. Business cycles and credit types Household Credit/GDP
Business Credit/GDP
GDP
C
I
RER
TB/GDP
GDP
C
I
RER
TB/GDP
Brazil
0.61
0.75
0.78
0.75
-0.64
0.44
0.49
0.43
0.26
-0.42
Chile
0.50
0.54
0.27
0.66
-0.43
0.35
0.38
0.23
0.26
-0.60
Korea
0.49
0.62
0.55
0.43
-0.62
-0.16
-0.15
-0.27
-0.35
0.07
S. Africa
0.80
0.83
0.79
0.50
-0.79
0.38
0.35
0.21
-0.18
-0.27
Thailand
0.47
0.41
0.30
0.54
-0.05
0.31
0.23
0.15
0.41
0.07
Turkey
0.74
0.84
0.68
0.77
-0.64
0.22
0.09
0.06
-0.09
-0.18
Czech Republic
0.62
0.40
0.44
0.68
0.09
-0.27
-0.01
-0.28
-0.18
-0.15
Poland
0.32
0.63
0.30
0.06
-0.56
0.37
0.60
0.30
-0.28
-0.59
Hungary
0.56
0.71
-0.09
0.66
-0.24
-0.15
-0.15
-0.17
-0.14
0.01
Average
0.55
0.62
0.46
0.52
-0.42
0.14
0.15
0.06
-0.03
-0.19
Notes: Annual data …ltered using Hodrick-Prescott …lter with smoothing parameter 100.
7
Brazil
Chile
Czech Republic
0.04
0.3
0.08
0.3
0.08
0.4
0.02
0.15
0.04
0.15
0.04
0.2
0
0
-0.02 -0.04 1995
0
0
-0.15 -0.04 1999
2003 Hungary
2007
2011
-0.3
0.08
0,5
0.04
0.25
0
0
-0.08 1983
0
0
-0.15 -0.04 1989
1995 2001 Korea
-0.3
2007
-0.08 1995
-0.2 1999
2003 Poland
2007
-0.4
0.1
0.14
0.04
0.3
0.05
0.07
0.02
0.15
0
0
0
0
-0.04
-0.25 -0.05
-0.07 -0.02
-0.08 1995
-0,5
-0.1 1983
-0.14 -0.04 1995
0.06
0.2
0.14
0.2
0.1
0.8
0.03
0.1
0.07
0.1
0.05
0.4
1999
2003 South Africa
2007
0
0
1989
1995 2001 Thailand
2007
0
0
-0.03
-0.1
-0.07
-0.1
-0.06 1994
-0.2
-0.14 1991
-0.2
1998
2002
2006
2010
1995
1999
Output (left axis)
2003
2007
2011
-0.15 1999
2003 Turkey
2007
0
0
-0.05 -0.1 1993
-0.3
-0,4 1997
2001
2005
2009
-0,8
Household credit/GDP (right axis)
Brazil
Chile
Czech Republic
0.04
0.2
0.08
0.2
0.08
0.2
0.02
0.1
0.04
0.1
0.04
0.1
0
0
0
0
0
-0.02
-0.1
-0.04
-0.1
-0.04
-0.04 1995
-0.2
-0.08 1983
-0.2
-0.08 1995
1999
2003 Hungary
2007
2011
0.08
0.2
0.1
0.04
0.1
0.05
0
0 -0.1
-0.08 1995
-0.2
-0.1 1983
0.06
0.14
0.03
0.07
2003 South Africa
2007
0
0
-0.03 -0.06 1994
1995 2001 Korea
2007
0
-0.04 1999
1989
2002
2006
2010
1999
2003 Poland
2007
-0.2
0.04
0.2
0.15
0.02
0.1
0
0
-0.15 -0.02
-0.1
-0.3
-0.04 1995
0.14
0.3
0.1
0.4
0.07
0.15
0.05
0.2
1989
1995 2001 Thailand
2007
0
0
-0.07 -0.07 1998
-0.1
0.3
0
-0.05
0
-0.14 -0.14 1991
1999
2003 Turkey
2007
0
0
-0.15 -0.05 1995
1999
Output (left axis)
2003
2007
2011
-0.3
-0.1 1993
Business credit/GDP (right axis)
Figure 2. Output and credit-to-GDP ratios
8
-0.2
-0.2 1997
2001
2005
2009
-0.4
Brazil
Chile
0.3 0.15 0
0.2
0.3
0.2
0.4
0.15
0.1
0.15
0.1
0.2
0
-0.15
Czech Republic
0.3
0
0
0
0
-0.15
-0.1
-0.15
-0.1
-0.3
-0.2 1983
-0.3
-0.2 1995
0.2
0.5
0.3
0.14
0.2
0.3
0.1
0.25
0.15
0.07
0.1
0.15
-0.3 1995
1999
2003 Hungary
2007
2011
0
0
1989
1995 2001 Korea
2007
0
0
-0.2 1999
2003 Poland
2007
0
-0.4
0
-0.1
-0.25 -0.15
-0.07
-0.1
-0.2 1995
-0,5
-0.3 1983
-0.14
-0.2 1995
0.4
0.2
0.2
0.2
0.3
0.8
0.2
0.1
0.1
0.1
0.15
0.4
1999
2003 South Africa
2007
0
0
-0.2 -0.4 1994
1998
2002
2006
2010
1989
1995 2001 Thailand
2007
0
-0.1
-0.1
-0.2
-0.2 1991
0 -0.1 1995
1999
2003
Real exchange rate (left axis)
Brazil
2007
2011
-0.2
-0.15 1999
2003 Turkey
2007
0
0
-0.15 -0.3 1993
-0.3
-0.4 1997
2001
2005
2009
-0.8
Household credit/GDP (right axis)
Chile
Czech Republic
0.3
0.2
0.2
0.2
0.2
0.2
0.15
0.1
0.1
0.1
0.1
0.1
0
0 -0.15 -0.3 1995
1999
2003 Hungary
2007
0
-0.1
-0.1
-0.2
-0.2 1983
2011
0.2
0.2
0.3
0.1
0.1
0.15
0
0 -0.1
-0.2 1995
-0.2
-0.3 1983
0.4
0.14
0.2
0.07
2003 South Africa
2007
0
0
1989
1995 2001 Korea
2007
0
-0.1 1999
0
-0.2
-0.2 1995
0.3
0.2
0.2
0.15
0.1
0.1
2007
-0.2
0
-0.1 -0.2 1995
0.2
0.3
0.3
0.4
0.1
0.15
0.15
0.2
1989
1995 2001 Thailand
2007
0
0
-0.1
-0.15 -0.15
-0.2 1991
-0.3
1995
1999
Real exchange rate (left axis)
2003
2007
2011
-0.1 1999
2007
-0.3 1993
-0.2
0 -0.2 1997
Business credit/GDP (right axis)
Figure 3. Real exchange rate and credit-to-GDP ratios
9
2003 Turkey
0
-0.14
2010
2003 Poland
-0.3
-0.07 2006
1999
-0.15
-0.4 1994
2002
-0.1
0
-0.2 1998
0
-0.1
0
-0.15
0
-0.1
2001
2005
2009
-0.4
Brazil
Chile
Czech Republic
2.8
0.3
8
0.3
1.4
0.15
4
0.15
0
0
0
0 -1.4 -2.8 1995
1999
2003 Hungary
2007
-0.15
-4
-0.3
-8 1983
2011
4
0,5
2
0.25
0
0
-2
-0.25
-4 1995
-0,5
1999
2003 South Africa
2007
-0.15
3
0.4
1.5
0.2
0
0
-1.5
-0.2
-0.3
-3 1995
12
0.14
4
0.3
6
0.07
2
0.15
0
0
0
0
1989
1995 2001 Korea
2007
-6 -12 1983
1989
1995 2001 Thailand
2007
-0.07
-2
-0.14
-4 1995
1999
2003 Poland
2007
-0.4
-0.15 1999
2003 Turkey
2007
-0.3
4
0.2
14
0.2
6
0.8
2
0.1
7
0.1
3
0.4
0
0
0
0
0
0
-2
-0.1
-4 1994
-0.2
1998
2002
2006
2010
-7 -14 1991
1995
1999
2003
2011
-3
-0.2
-6 1993
-0.4 1997
2001
2005
2009
-0.8
Household credit/GDP (right axis)
Trade balance/GDP (left axis)
Brazil
2007
-0.1
Chile
Czech Republic
2.8
0.2
8
0.2
3
0.2
1.4
0.1
4
0.1
1.5
0.1
0
0
0
0
0
0
-1.4
-0.1
-4
-0.1
-2.8 1995
-0.2
-8 1983
-0.2
-3 1995
4
0.2
12
0.3
4
0.2
2
0.1
6
0.15
2
0.1
0
0
0
0
0
0
1999
2003 Hungary
2007
2011
-2
-0.1
-4 1995
-0.2
1999
2003 South Africa
2007
1989
1995 2001 Korea
2007
-6 -12 1983
1989
1995 2001 Thailand
2007
-1.5
-0.1 1999
2003 Poland
2007
-0.2
-0.15
-2
-0.3
-4 1995
-0.1
0.3
6
0.4
1999
2003 Turkey
2007
-0.2
4
0.14
14
2
0.07
7
0.15
3
0.2
0
0
0
0
0
0
-2 -4 1994
-0,07 1998
2002
2006
2010
-0,14
-7 -14 1991
1995
1999
Trade balance/GDP (left axis)
2003
2007
2011
-0.15
-3
-0.3
-6 1993
-0.2 1997
2001
Business credit/GDP (right axis)
Figure 4. Trade balance-to-GDP and credit-to-GDP ratios
10
2005
2009
-0.4
3
The Model
3.1
Households
Households choose consumption and labor to maximize their expected lifetime utility given by E0
1 X t=0
where labor,
h
(
cht (cht;N ; cht;T ) ) 1
lt
h t
1
;
> 1;
> 0;
(1)
2 (0; 1) is the discount factor, cht is the consumption aggregator, lt represents
is the risk aversion parameter,
is the parameter that governs the intertemporal
elasticity of substitution in labor supply, and
is the measure of disutility from work-
ing. Consumption is an aggregate of the consumption of nontradable goods, cht;N ; and the consumption of tradable goods, cht;T . The budget constraint of households is given by cht;T + pt;N cht;N + Rbht
1
= bht + wt;T lt;T + pt;N wt;N lt;N ;
(2)
where bht denotes the amount borrowed at time t, R = (1 + r) is the gross interest rate and r is the net real interest rate, which is taken as constant. The variables lt;T and lt;N denote labor supplied to tradable and nontradable sectors, respectively, wt;T and wt;N denote the wage rates in the two sectors and pt;N is the relative price of the nontradable good, where the price of the tradable good is normalized to one. Households face a credit constraint in every period. The total value of their debt including both interest and principal cannot exceed a fraction of their expected income in the next period. As in Ludvigson (1999), we choose to tie borrowing to income because many banks require income statements before they provide funds to the borrowers since income is associated with some observable measure of the borrower’s …nancial health. The credit constraint of households is of the form Rbht
mht Et (wt+1;T lt+1;T + pt+1;N wt+1;N lt+1;N ) :
11
(3)
In the calibration of the model,
h
is chosen such that
h
< 1=R. This condition
guarantees that the credit constraint is binding in and around the steady state. The loanto-income (LTI) ratio, denoted by mht , is modeled as a stochastic process.
3.2
Entrepreneurs
3.2.1
Tradable sector
Entrepreneurs who produce tradable goods combine households’labor services with capital, kt
1;T .
Output is produced by a Cobb-Douglas technology: yt;T = eAt;T kt
1 1;T lt;T
;
(4)
where At;T is an exogenous stochastic productivity shock. The capital accumulation decision is made by the entrepreneurs and the capital accumulation equation for the tradable sector is given by it;T = kt;T
(1
)kt
1;T ;
(5)
where it;T denotes investment in the tradable sector. The investment good used in both sectors is assumed to be tradable and
is the common depreciation rate.
Firms in both sectors have to pay a fraction
of the wages before output becomes
available and they need working capital loans from foreign lenders. Thus, tradable sector …rms borrow wt;T lt;T at the beginning of period t and repay R wt;T lt;T at the end of the period as in Neumeyer and Perri (2005). As households, entrepreneurs are also restricted in their borrowing due to enforceability problems. Following Mendoza (2010), we assume that the entrepreneur’s total debt, which includes intertemporal debt, beT t ; and within-period working capital loans, cannot exceed a fraction of the collateral assets, which are capital holdings in our model. In the case of the tradable sector, the credit constraint takes the form RbeT t + R wt;T lt;T
k meT t Et qt+1;T kt;T :
12
(6)
The loan-to-capital (LTC) ratio, denoted by meT t ; is modeled as a stochastic process, and k qt+1;T is the price of capital at time t + 1: We use adjustment costs for capital accumulation
to reduce the volatility of investment. Therefore, the price of capital in terms of tradable consumption di¤ers from one and is given by k =1+ qt;T
where
(kt
1;T ; it;T )
@ (kt 1;T ; it;T ) ; @it;T
(7)
is the capital adjustment cost function.
The entrepreneur’s problem is to maximize her expected utility E0
1 X
(
eT t
)
t=0
eT eT 1 (ceT t (ct;N ; ct;T )) 1
(8)
subject to technology and borrowing constraints, and the following ‡ow of funds constraint eT ceT t;T + pt;N ct;N + wt;T lt;T + it;T + (kt
1;T ; it;T )
+ RbeT t 1 + (R
1) wt;T lt;T = yt;T + beT t : (9)
As in the case of households, consumption of the tradable sector entrepreneur, ceT t ; is an eT aggregate of the consumption of nontradable and tradable goods, ceT t;N and ct;T ; respectively.
Similar to household’s discount factor, we assume that
eT
< 1=R so that the credit
constraint is binding in and around the steady state. 3.2.2
Nontradable sector
Entrepreneurs in the nontradable sector also produce output with a Cobb-Douglas technology: yt;N = eAt;N kt
1 1;N lt;N
(10)
;
where At;N is an exogenous stochastic productivity shock and kt
1;N
denotes capital used
in the production of the nontradable good. Capital is accumulated by the entrepreneur and the equation for capital accumulation is given by it;N = kt;N
(1
)kt
1;N ;
where it;N denotes investment in the nontradable sector. 13
(11)
Similar to the tradable sector, …rms in the nontradable sector also have a working capital requirement and face a credit constraint. The entrepreneur’s total value of debt including the interest payments cannot exceed a fraction of the expected value of the capital holdings: RbeN t + R wt;N lt;N
k meN t Et qt+1;N kt;N ;
(12)
denotes the intertemporal debt issued at time t by the nontradable sector entrewhere beN t preneur. The loan-to-capital ratio, denoted by meN t ; is modeled as a stochastic process. We use adjustment costs for capital accumulation in the nontradable sector as well, in order to reduce the volatility of investment. The price of capital in terms of tradable consumption, k ; is given by qt;N k qt;N =1+
where
(kt
1;N ; it;N )
@ (kt 1;N ; it;N ) ; @it;N
(13)
is the capital adjustment cost function.
The entrepreneur in the nontradable sector maximizes her expected utility E0
1 X
(
eN t
t=0
)
eN 1 (ceN ceN t t;N ; ct;T ) 1
(14)
subject to technology and borrowing constraints, as well as the following ‡ow of funds constraint eN ceN t;T +pt;N ct;N +pt;N wt;N lt;N +it;N + (kt
eN 1;N ; it;N )+Rbt 1 +(R
1) wt;N lt;N = pt;N yt:N +beN t : (15)
Consumption of the nontradable sector entrepreneur is also an aggregate of the coneN sumption of nontradable and tradable goods, ceN t;N and ct;T , respectively.
We also assume for the nontradable sector entrepreneur that
eN
< 1=R to make sure
that the credit constraint is binding in and around the steady state.
3.3
Equilibrium
eN Given initial conditions bh0 ; beT 0 ; b0 ; k0;T ; k0;N ; the constant interest rate r; the sequence of
shocks to sectoral productivity levels, the loan-to-income ratio of the household and the 14
loan-to-capital ratios of the entrepreneurs, the competitive equilibrium is de…ned as a set of allocations and prices
eT eT eN yt;T ; yt;N ; lt;T ; lt;N ; kt;T ; kt;N ; it;T ; it;N ; cht ; cht;T ; cht;N ; ceT t ; ct;T ; ct;N ; ct ;
eN h eT eN ceN t;T ; ct;N ; bt ; bt ; bt ; pt;N ; wt;T ; wt;N
such that (i) the allocations solve the problems of
households, and entrepreneurs in the tradable and nontradable sectors at the equilibrium prices, (ii) factor markets clear, and (iii) the resource constraints for the tradable and nontradable sectors hold: eN cht;T + ceT t;T + ct;T + it;T + it;N + (kt
1;T ; it;T )
+ tbt = yt;T
(16)
eN cht;N + ceT t;N + ct;N = yt;N
(17)
+ (kt
1;N ; it;N )
where the trade balance is de…ned as tbt = R bht
4
1
eN + beT t 1 + bt 1 + (R
1) wt;T lt;T + (R
1) wt;N lt;N
eN bht + beT : (18) t + bt
Calibration
The model is solved using quarterly Turkish data for the period 1999Q1-2011Q4. The construction of the series used in the model solution is explained in detail in the Appendix. The parameter values of the model are summarized in Table 3. We take the real interest rate as constant and set it equal to the average real interest rate in Turkey, which equals 0.0117. We set the discount factors such that the credit constraints bind in and around the steady state. The values for value for
eT
h
and
eN
are set to 0.94, and the
is set to 0.97, which are the highest possible values that guarantee binding
credit constraints in the solution of the model. The value of ; which determines the intertemporal elasticity of substitution in labor supply, is set to 1.7 following Correia et al. (1995). The coe¢ cient of relative risk aversion is set to 1, which corresponds to log-utility. The annual depreciation rate is set to 0.08 following Meza and Quintin (2007).
15
Table 3. Parameter values of the benchmark model Parameter Value
Description
h
0.94
Discount factor of households
eN
0.94
Discount factor of nontradable sector entrepreneurs
eT
0.97
Discount factor of tradable sector entrepreneurs
1
Relative risk aversion coe¢ cient
1.7
Labor curvature
1.589
Labor weight in utility
0.54
Nontradable weight in the consumption aggregator
0.35
Capital exponent in the tradable sector
0.25
Capital exponent in the nontradable sector
0.08
Annual depreciation rate
r
0.0117 Real interest rate
'T
5.21
Capital adjustment cost coe¢ cient in the tradable sector
'N
18.30
Capital adjustment cost coe¢ cient in the nontradable sector
0.25
Working capital coe¢ cient
mh
0.423
Loan-to-income ratio
meN
0.266
Loan-to-capital ratio in the nontradable sector
meT
0.102
Loan-to-capital ratio in the tradable sector
Stochastic processes AT
0.662
h T
0.697
("AT )
AN
0.777
h N
0.062
("AN ) 0.0148
h
0.817
eT T
0.153
("h )
0.0350
eT
0.656
eN N
0.362
("eT )
0.0287
eN
0.803
("eN )
0.0270
0.0283
We cannot calibrate the capital share parameters in the tradable and nontradable sectors for Turkey due to unavailability of data. Di¤erent values have been used in the literature for these parameters and the general consensus is that the tradable sector is more capital intensive than the nontradable sector. Therefore, we set the capital’s share of income equal
16
to 0.35 in the tradable sector and 0.25 in the nontradable sector, which are close to the values used in the literature.4 The value of
is set to 1.589 so that the steady state labor supply equals 0.17, which
is the average value of time spent working as a percentage of total discretionary time in Turkey. The share of nontradable goods in the consumption aggregator, ; is set equal to the average share of nontradable consumption in total consumption in Turkey. The steadystate value of the loan-to-capital ratio in the nontradable and tradable sectors, meN and meT ; are set to match the average value of business credit in each sector as a ratio of GDP for the sample period, which are 11.2% and 9.6%, respectively. Likewise, the steady-state value of the loan-to-income ratio, mh ; is set to match the average value of the ratio of household credit to GDP in the data, which is 7.2%. For the calibration of the parameter ; we use data on short-term bank loans from the Company Accounts database of the Central Bank of Turkey. Total liabilities of …rms are composed of intertemporal loans and working capital loans in our model, and the loans for working capital have a shorter duration compared to the other loans. Therefore, we choose to approximate the working capital loans with short-term bank loans. We calibrate
by
taking the average of the ratio of short-term loans to the compensation of employees, which is equal to 0.25. Since the data on short-term bank loans are not available at the sectoral level, we use the same value for both tradable and nontradable sectors. The consumption aggregator is assumed to be of the following form for all agents: cjt cjt;N ; cjt;T = (cjt;N ) (cjt;T )1 ;
0
Inci Gumusz
University of Georgia
Sabanci University
July 24, 2014
Abstract This paper analyzes di¤erent types of private sector credit in relation to business cycles in emerging market economies. We …rst provide evidence that the results found in the literature on credit expansions being associated with economic expansions, real exchange rate appreciations and current account de…cits hold more strongly for household credit than business credit. Then, using a two-sector real business cycle model of a small open economy, we study the model dynamics generated by shocks to household credit and business credit in tradable and nontradable sectors. We show that the three types of credit shocks generate di¤erent macroeconomic dynamics in sectoral output and input levels as well as the real exchange rate. We also show that the credit shocks are important in matching key business cycle properties observed in the data, especially the moments related to credit. JEL classi…cation: E32, E44, F34, F41 Key Words: Household Credit, Business Credit, Business Cycles, Credit Shocks We would like to thank participants at the Society for Computational Economics 2012 Meeting, the SNDE 2012 Meeting, the LACEA-LAMES 2012 Meeting, the EEA-ESEM 2013 meeting, the North American Summer Meeting of the Econometric Society 2014 and seminar participants at the University of Georgia, Federal Reserve Bank of Atlanta and Central European University for valuable comments and suggestions. All remaining errors are our own. y Department of Economics, Terry College of Business, 516 Brooks Hall, Athens, GA 30602. Email: [email protected], tel: +1-706-542-3667 z Corresponding author. Sabanci University, Faculty of Arts and Social Sciences, Orhanli, Tuzla, Istanbul, 34956, Turkey. Email: [email protected], tel: +90-216-483-9318
1
1
Introduction
The literature on credit cycles, including Mendoza and Terrones (2008, 2012) and Tornell and Westermann (2002, 2003), has established that credit expansions in emerging market economies are associated with large macroeconomic expansions, widening current account de…cits and real exchange rate appreciations. The standard measure of credit used in this literature is total private credit which includes all types of credit to the private sector without di¤erentiating between lending to households and lending to …rms. These two types of credit may generate di¤erent business cycle patterns given that household credit expansions are likely to a¤ect the economy through an increase in consumption and demand for goods and services whereas business credit has the potential to increase investment and labor demand, and thereby increase output. Furthermore, the allocation of business credit between tradable and nontradable sectors is another factor that may a¤ect the business cycle dynamics. Recent developments in credit markets underline the importance of distinguishing between household and business credit in emerging markets. Figure 1 shows the evolvement of household and business credit-to-GDP ratios for nine emerging market economies and Table 1 shows the average values of these ratios for the sample period. The two types of credit exhibit di¤erent patterns: while lending to households has grown substantially over the period we analyze, the growth in business credit has been much slower in most cases. As a result, household credit has become an important component of overall private credit with potentially important consequences for business cycles. On average, household credit constitutes 33 percent of total private credit in our sample. Despite the increased importance of household credit, existing models that study emerging market business cycles abstract from household credit dynamics.
2
Business credit/GDP
Household credit/GDP 1.4
0.9 0.8
1.2
0.7 1 0.6 0.8
0.5 0.4
0.6
0.3 0.4 0.2 0.2
0.1 0 1983
1987 Brazil
1991
1995
Chile
1999
2003
2007
Czech Republic
0 1983
2011 Hungary
Korea
1987 Poland
1991
1995
1999
South Africa
2003
2007
T hailand
2011 T urkey
Figure 1. Credit-to-GDP ratios in emerging market economies
Table 1. Household and business credit Countries
HC/GDP BC/GDP HC/TC Sample Period
Brazil
11.8
19.7
37.5
1995-2012
Chile
15.4
44.4
25.8
1983-2012
Czech Republic
15.1
51.7
22.6
1995-2010
Hungary
17.5
71.4
19.7
1995-2010
Korea
54.8
92.2
37.3
1983-2012
Poland
14.5
31.6
31.5
1991-2012
South Africa
36.4
33.6
52.0
1994-2012
Thailand
47.0
70.2
40.1
1991-2012
Turkey
6.84
25.5
21.2
1993-2012
Average
24.4
48.9
33.3
-
Note: HC, BC and TC denote household, business and total credit, respectively.
In this paper, we distinguish between household and business credit, as well as the sectoral allocation of business credit, and study the macroeconomic dynamics generated by di¤erent credit types. To this end, …rst we re-examine the documented stylized facts 3
regarding credit cycles di¤erentiating between household and business credit. Using data from a group of emerging market economies, we show that the patterns documented in the literature with respect to credit and key macroeconomic variables are much stronger for household credit: household credit exhibits a strong positive comovement with output, consumption, investment and real exchange rate appreciation, and a strong negative correlation with the trade balance. Business credit, on the other hand, has weaker correlations with all of these variables. These correlations show that the empirical …ndings of the literature on credit cycles are mostly driven by household credit. We then construct a two-sector small open economy real business cycle model to assess quantitatively the role of di¤erent types of credit shocks in driving business cycles. In our model households and both the tradable and nontradable sectors are credit constrained, and …rms have a working capital requirement. The model allows us to disaggregate total credit into three subcomponents: (i) household credit, (ii) nontradable sector business credit, and (iii) tradable sector business credit. We choose to model all sectors as credit constrained since this allows us to study the model dynamics for each type of credit expansion and analyze the e¤ects in each sector separately. We analyze the model under productivity shocks in tradable and nontradable sectors and three types of credit shocks. The credit shocks are modeled as stochastic processes that a¤ect the borrowing limits of the agents. The shocks to credit are similar to the …nancial sector shocks studied in Jermann and Quadrini (2012) who analyze the e¤ects of these shocks in a closed economy model. We calibrate our model to Turkey for the period 1999Q1-2011Q4. The reason we choose to calibrate the model to Turkey is that we are able to obtain detailed sectoral data on credit variables at quarterly frequency for a relatively long time period.1 For our analysis, we not only need data on household and business credit separately, but also business credit data at the sectoral level. To the best of our knowledge, most of the emerging economies for which household credit and business credit data are separately available do not provide the sectoral credit classi…cation. 1
Turkey is a representative emerging market economy that features the standard business cycle properties observed in emerging economies documented by Aguiar and Gopinath (2007) and Neumeyer and Perri (2005): consumption is more volatile than output, trade balance is countercyclical and business cycles are very volatile (see Table 4 for more details). With respect to the correlations of di¤erent types of credit with the key macroeconomic aggregates, Turkey displays the same pattern as the average of the emerging economies in Table 2.
4
The impulse response analysis shows that the three types of credit shocks generate di¤erent macroeconomic dynamics. The starkest di¤erence appears for the real exchange rate response. We observe that an increase in the supply of credit to the nontradable sector leads to a real exchange rate depreciation whereas credit increase to households and tradable sector generates an appreciation. The credit shocks generate di¤erent responses in sectoral output and input levels as well. Nontradable output expands after each credit shock whereas tradable output only increases as a result of a positive credit shock to that sector. Investment in each sector increases as a response to the sector-speci…c credit shocks whereas consumption increases on impact after each shock. Finally, trade balance declines in all cases. The extent to which trade balance responds to credit shocks depends on the share of each type of credit in the overall economy as analyzed in Section 5.2.2. The quantitative results from the model with productivity and credit shocks can account for most of the empirical regularities of the Turkish economy. In particular, the model is successful at matching the key business cycle moments such as the relative volatilities of consumption, labor and trade balance. The model also generates a countercyclical trade balance and a strong negative correlation between real exchange rate and trade balance. More importantly, the model can account for the comovement between the key macroeconomic variables and di¤erent credit types. Household credit exhibits a strong positive correlation with output and real exchange rate, and a strong negative correlation with the trade balance whereas for business credit the correlations are weaker. In the absence of credit shocks, the model performs poorly in all aspects but especially with respect to moments related to credit. When we replace the credit shocks with an interest rate shock, the model still underperforms relative to the baseline model. Therefore we conclude that credit shocks are important in driving the business cycles in emerging market economies. The role of credit shocks for macroeconomic ‡uctuations has been recently investigated primarily in closed economy models. This literature, including Christiano, Motto and Rostagno (2010), Jermann and Quadrini (2012) and Khan and Thomas (2013), …nd that …nancial shocks are important for the dynamics of real and …nancial variables. One recent study by Perri and Quadrini (2014) studies the implications of credit shocks in a twocountry model and focuses on international comovement in real and …nancial variables. The contribution of the current paper is to show that shocks to di¤erent types of credit 5
are important to understand business cycles, as well as the real exchange rate behavior, in emerging market economies. Our paper is further related to the empirical studies on the distinction between household and business credit, such as Büyükkarabacak and Krause (2009), Büyükkarabacak and Valev (2010), and Beck et al. (2012) who underline the importance of di¤erentiating between the types of borrowers. The main conclusion of these papers is that the two types of credit serve di¤erent purposes and have distinct e¤ects on the economy.2 Our paper complements these empirical studies by providing a general equilibrium model that helps understand the transmission mechanism through which each type of credit a¤ects the business cycle properties in emerging market economies.
2
Credit and Business Cycles in Emerging Economies
This section documents business cycle regularities in relation to household and business credit for a sample of emerging market economies for which we could obtain time series data of su¢ cient length. The data for Czech Republic, Hungary, Korea, Poland, Thailand and Turkey are obtained from the Bank for International Settlements (BIS).3 For Brazil, Chile and South Africa we obtain the data from their respective central banks. Due to the lack of credit data at the sectoral level for several countries in our sample, here we analyze total business credit rather than tradable and nontradable sector credit separately. Table 2 shows the correlations between macroeconomic aggregates and the creditto-GDP ratios for household and business credit. These correlations overall show that household credit is more strongly correlated with the business cycle compared to business credit. The correlation between household credit and output is, on average, 0.55 whereas for business credit this correlation is only 0.14. The same pattern holds for consumption 2
Büyükkarabacak and Krause (2009) show that household credit leads to a deterioration in the trade balance, whereas business credit has a small but positive e¤ect. Büyükkarabacak and Valev (2010) …nd that household credit expansions have been a signi…cant predictor of banking crises. Business credit expansions are also associated with banking crises but their e¤ect is weaker. Beck et al. (2012) show that bank lending to …rms is positively associated with growth, while the relationship between household credit and growth is insigni…cant. 3 BIS also provides data for Indonesia, Mexico, Singapore and Hong Kong. For Indonesia the data starts from 2001, which does not give a su¢ ciently long time series. We opt to exclude Hong Kong, Mexico and Singapore because of the strong negative correlation between output and private credit, which contradicts with the general pattern documented in the literature.
6
and investment. The starkest di¤erence between the two types of credit is observed in their correlations with the real exchange rate. There is a very strong comovement between the real exchange rate appreciation and household credit with an average correlation coe¢ cient of 0.52. The business credit-real exchange rate correlation is lower than the correlation of household credit for all countries and negative for most of them, with an average of -0.03. In the case of the ratio of trade balance to output, household credit has a negative correlation in all countries, except for Czech Republic, with an average correlation of -0.42. The correlation between business credit and trade balance is also negative in six out of nine countries. However, the correlations are weaker and the average correlation drops to -0.19. Figures 2-4 plot the time series for three key macroeconomic variables; GDP, real exchange rate, and the ratio of trade balance to output; together with the credit-to-GDP ratios for the two types of credit. These …gures also reveal the di¤erences between the credit types in relation to the cycle. To summarize, the empirical regularities presented in this section show that the stylized facts documented in the literature with respect to private credit are to a large extent driven by household credit rather than business credit in emerging markets. Table 2. Business cycles and credit types Household Credit/GDP
Business Credit/GDP
GDP
C
I
RER
TB/GDP
GDP
C
I
RER
TB/GDP
Brazil
0.61
0.75
0.78
0.75
-0.64
0.44
0.49
0.43
0.26
-0.42
Chile
0.50
0.54
0.27
0.66
-0.43
0.35
0.38
0.23
0.26
-0.60
Korea
0.49
0.62
0.55
0.43
-0.62
-0.16
-0.15
-0.27
-0.35
0.07
S. Africa
0.80
0.83
0.79
0.50
-0.79
0.38
0.35
0.21
-0.18
-0.27
Thailand
0.47
0.41
0.30
0.54
-0.05
0.31
0.23
0.15
0.41
0.07
Turkey
0.74
0.84
0.68
0.77
-0.64
0.22
0.09
0.06
-0.09
-0.18
Czech Republic
0.62
0.40
0.44
0.68
0.09
-0.27
-0.01
-0.28
-0.18
-0.15
Poland
0.32
0.63
0.30
0.06
-0.56
0.37
0.60
0.30
-0.28
-0.59
Hungary
0.56
0.71
-0.09
0.66
-0.24
-0.15
-0.15
-0.17
-0.14
0.01
Average
0.55
0.62
0.46
0.52
-0.42
0.14
0.15
0.06
-0.03
-0.19
Notes: Annual data …ltered using Hodrick-Prescott …lter with smoothing parameter 100.
7
Brazil
Chile
Czech Republic
0.04
0.3
0.08
0.3
0.08
0.4
0.02
0.15
0.04
0.15
0.04
0.2
0
0
-0.02 -0.04 1995
0
0
-0.15 -0.04 1999
2003 Hungary
2007
2011
-0.3
0.08
0,5
0.04
0.25
0
0
-0.08 1983
0
0
-0.15 -0.04 1989
1995 2001 Korea
-0.3
2007
-0.08 1995
-0.2 1999
2003 Poland
2007
-0.4
0.1
0.14
0.04
0.3
0.05
0.07
0.02
0.15
0
0
0
0
-0.04
-0.25 -0.05
-0.07 -0.02
-0.08 1995
-0,5
-0.1 1983
-0.14 -0.04 1995
0.06
0.2
0.14
0.2
0.1
0.8
0.03
0.1
0.07
0.1
0.05
0.4
1999
2003 South Africa
2007
0
0
1989
1995 2001 Thailand
2007
0
0
-0.03
-0.1
-0.07
-0.1
-0.06 1994
-0.2
-0.14 1991
-0.2
1998
2002
2006
2010
1995
1999
Output (left axis)
2003
2007
2011
-0.15 1999
2003 Turkey
2007
0
0
-0.05 -0.1 1993
-0.3
-0,4 1997
2001
2005
2009
-0,8
Household credit/GDP (right axis)
Brazil
Chile
Czech Republic
0.04
0.2
0.08
0.2
0.08
0.2
0.02
0.1
0.04
0.1
0.04
0.1
0
0
0
0
0
-0.02
-0.1
-0.04
-0.1
-0.04
-0.04 1995
-0.2
-0.08 1983
-0.2
-0.08 1995
1999
2003 Hungary
2007
2011
0.08
0.2
0.1
0.04
0.1
0.05
0
0 -0.1
-0.08 1995
-0.2
-0.1 1983
0.06
0.14
0.03
0.07
2003 South Africa
2007
0
0
-0.03 -0.06 1994
1995 2001 Korea
2007
0
-0.04 1999
1989
2002
2006
2010
1999
2003 Poland
2007
-0.2
0.04
0.2
0.15
0.02
0.1
0
0
-0.15 -0.02
-0.1
-0.3
-0.04 1995
0.14
0.3
0.1
0.4
0.07
0.15
0.05
0.2
1989
1995 2001 Thailand
2007
0
0
-0.07 -0.07 1998
-0.1
0.3
0
-0.05
0
-0.14 -0.14 1991
1999
2003 Turkey
2007
0
0
-0.15 -0.05 1995
1999
Output (left axis)
2003
2007
2011
-0.3
-0.1 1993
Business credit/GDP (right axis)
Figure 2. Output and credit-to-GDP ratios
8
-0.2
-0.2 1997
2001
2005
2009
-0.4
Brazil
Chile
0.3 0.15 0
0.2
0.3
0.2
0.4
0.15
0.1
0.15
0.1
0.2
0
-0.15
Czech Republic
0.3
0
0
0
0
-0.15
-0.1
-0.15
-0.1
-0.3
-0.2 1983
-0.3
-0.2 1995
0.2
0.5
0.3
0.14
0.2
0.3
0.1
0.25
0.15
0.07
0.1
0.15
-0.3 1995
1999
2003 Hungary
2007
2011
0
0
1989
1995 2001 Korea
2007
0
0
-0.2 1999
2003 Poland
2007
0
-0.4
0
-0.1
-0.25 -0.15
-0.07
-0.1
-0.2 1995
-0,5
-0.3 1983
-0.14
-0.2 1995
0.4
0.2
0.2
0.2
0.3
0.8
0.2
0.1
0.1
0.1
0.15
0.4
1999
2003 South Africa
2007
0
0
-0.2 -0.4 1994
1998
2002
2006
2010
1989
1995 2001 Thailand
2007
0
-0.1
-0.1
-0.2
-0.2 1991
0 -0.1 1995
1999
2003
Real exchange rate (left axis)
Brazil
2007
2011
-0.2
-0.15 1999
2003 Turkey
2007
0
0
-0.15 -0.3 1993
-0.3
-0.4 1997
2001
2005
2009
-0.8
Household credit/GDP (right axis)
Chile
Czech Republic
0.3
0.2
0.2
0.2
0.2
0.2
0.15
0.1
0.1
0.1
0.1
0.1
0
0 -0.15 -0.3 1995
1999
2003 Hungary
2007
0
-0.1
-0.1
-0.2
-0.2 1983
2011
0.2
0.2
0.3
0.1
0.1
0.15
0
0 -0.1
-0.2 1995
-0.2
-0.3 1983
0.4
0.14
0.2
0.07
2003 South Africa
2007
0
0
1989
1995 2001 Korea
2007
0
-0.1 1999
0
-0.2
-0.2 1995
0.3
0.2
0.2
0.15
0.1
0.1
2007
-0.2
0
-0.1 -0.2 1995
0.2
0.3
0.3
0.4
0.1
0.15
0.15
0.2
1989
1995 2001 Thailand
2007
0
0
-0.1
-0.15 -0.15
-0.2 1991
-0.3
1995
1999
Real exchange rate (left axis)
2003
2007
2011
-0.1 1999
2007
-0.3 1993
-0.2
0 -0.2 1997
Business credit/GDP (right axis)
Figure 3. Real exchange rate and credit-to-GDP ratios
9
2003 Turkey
0
-0.14
2010
2003 Poland
-0.3
-0.07 2006
1999
-0.15
-0.4 1994
2002
-0.1
0
-0.2 1998
0
-0.1
0
-0.15
0
-0.1
2001
2005
2009
-0.4
Brazil
Chile
Czech Republic
2.8
0.3
8
0.3
1.4
0.15
4
0.15
0
0
0
0 -1.4 -2.8 1995
1999
2003 Hungary
2007
-0.15
-4
-0.3
-8 1983
2011
4
0,5
2
0.25
0
0
-2
-0.25
-4 1995
-0,5
1999
2003 South Africa
2007
-0.15
3
0.4
1.5
0.2
0
0
-1.5
-0.2
-0.3
-3 1995
12
0.14
4
0.3
6
0.07
2
0.15
0
0
0
0
1989
1995 2001 Korea
2007
-6 -12 1983
1989
1995 2001 Thailand
2007
-0.07
-2
-0.14
-4 1995
1999
2003 Poland
2007
-0.4
-0.15 1999
2003 Turkey
2007
-0.3
4
0.2
14
0.2
6
0.8
2
0.1
7
0.1
3
0.4
0
0
0
0
0
0
-2
-0.1
-4 1994
-0.2
1998
2002
2006
2010
-7 -14 1991
1995
1999
2003
2011
-3
-0.2
-6 1993
-0.4 1997
2001
2005
2009
-0.8
Household credit/GDP (right axis)
Trade balance/GDP (left axis)
Brazil
2007
-0.1
Chile
Czech Republic
2.8
0.2
8
0.2
3
0.2
1.4
0.1
4
0.1
1.5
0.1
0
0
0
0
0
0
-1.4
-0.1
-4
-0.1
-2.8 1995
-0.2
-8 1983
-0.2
-3 1995
4
0.2
12
0.3
4
0.2
2
0.1
6
0.15
2
0.1
0
0
0
0
0
0
1999
2003 Hungary
2007
2011
-2
-0.1
-4 1995
-0.2
1999
2003 South Africa
2007
1989
1995 2001 Korea
2007
-6 -12 1983
1989
1995 2001 Thailand
2007
-1.5
-0.1 1999
2003 Poland
2007
-0.2
-0.15
-2
-0.3
-4 1995
-0.1
0.3
6
0.4
1999
2003 Turkey
2007
-0.2
4
0.14
14
2
0.07
7
0.15
3
0.2
0
0
0
0
0
0
-2 -4 1994
-0,07 1998
2002
2006
2010
-0,14
-7 -14 1991
1995
1999
Trade balance/GDP (left axis)
2003
2007
2011
-0.15
-3
-0.3
-6 1993
-0.2 1997
2001
Business credit/GDP (right axis)
Figure 4. Trade balance-to-GDP and credit-to-GDP ratios
10
2005
2009
-0.4
3
The Model
3.1
Households
Households choose consumption and labor to maximize their expected lifetime utility given by E0
1 X t=0
where labor,
h
(
cht (cht;N ; cht;T ) ) 1
lt
h t
1
;
> 1;
> 0;
(1)
2 (0; 1) is the discount factor, cht is the consumption aggregator, lt represents
is the risk aversion parameter,
is the parameter that governs the intertemporal
elasticity of substitution in labor supply, and
is the measure of disutility from work-
ing. Consumption is an aggregate of the consumption of nontradable goods, cht;N ; and the consumption of tradable goods, cht;T . The budget constraint of households is given by cht;T + pt;N cht;N + Rbht
1
= bht + wt;T lt;T + pt;N wt;N lt;N ;
(2)
where bht denotes the amount borrowed at time t, R = (1 + r) is the gross interest rate and r is the net real interest rate, which is taken as constant. The variables lt;T and lt;N denote labor supplied to tradable and nontradable sectors, respectively, wt;T and wt;N denote the wage rates in the two sectors and pt;N is the relative price of the nontradable good, where the price of the tradable good is normalized to one. Households face a credit constraint in every period. The total value of their debt including both interest and principal cannot exceed a fraction of their expected income in the next period. As in Ludvigson (1999), we choose to tie borrowing to income because many banks require income statements before they provide funds to the borrowers since income is associated with some observable measure of the borrower’s …nancial health. The credit constraint of households is of the form Rbht
mht Et (wt+1;T lt+1;T + pt+1;N wt+1;N lt+1;N ) :
11
(3)
In the calibration of the model,
h
is chosen such that
h
< 1=R. This condition
guarantees that the credit constraint is binding in and around the steady state. The loanto-income (LTI) ratio, denoted by mht , is modeled as a stochastic process.
3.2
Entrepreneurs
3.2.1
Tradable sector
Entrepreneurs who produce tradable goods combine households’labor services with capital, kt
1;T .
Output is produced by a Cobb-Douglas technology: yt;T = eAt;T kt
1 1;T lt;T
;
(4)
where At;T is an exogenous stochastic productivity shock. The capital accumulation decision is made by the entrepreneurs and the capital accumulation equation for the tradable sector is given by it;T = kt;T
(1
)kt
1;T ;
(5)
where it;T denotes investment in the tradable sector. The investment good used in both sectors is assumed to be tradable and
is the common depreciation rate.
Firms in both sectors have to pay a fraction
of the wages before output becomes
available and they need working capital loans from foreign lenders. Thus, tradable sector …rms borrow wt;T lt;T at the beginning of period t and repay R wt;T lt;T at the end of the period as in Neumeyer and Perri (2005). As households, entrepreneurs are also restricted in their borrowing due to enforceability problems. Following Mendoza (2010), we assume that the entrepreneur’s total debt, which includes intertemporal debt, beT t ; and within-period working capital loans, cannot exceed a fraction of the collateral assets, which are capital holdings in our model. In the case of the tradable sector, the credit constraint takes the form RbeT t + R wt;T lt;T
k meT t Et qt+1;T kt;T :
12
(6)
The loan-to-capital (LTC) ratio, denoted by meT t ; is modeled as a stochastic process, and k qt+1;T is the price of capital at time t + 1: We use adjustment costs for capital accumulation
to reduce the volatility of investment. Therefore, the price of capital in terms of tradable consumption di¤ers from one and is given by k =1+ qt;T
where
(kt
1;T ; it;T )
@ (kt 1;T ; it;T ) ; @it;T
(7)
is the capital adjustment cost function.
The entrepreneur’s problem is to maximize her expected utility E0
1 X
(
eT t
)
t=0
eT eT 1 (ceT t (ct;N ; ct;T )) 1
(8)
subject to technology and borrowing constraints, and the following ‡ow of funds constraint eT ceT t;T + pt;N ct;N + wt;T lt;T + it;T + (kt
1;T ; it;T )
+ RbeT t 1 + (R
1) wt;T lt;T = yt;T + beT t : (9)
As in the case of households, consumption of the tradable sector entrepreneur, ceT t ; is an eT aggregate of the consumption of nontradable and tradable goods, ceT t;N and ct;T ; respectively.
Similar to household’s discount factor, we assume that
eT
< 1=R so that the credit
constraint is binding in and around the steady state. 3.2.2
Nontradable sector
Entrepreneurs in the nontradable sector also produce output with a Cobb-Douglas technology: yt;N = eAt;N kt
1 1;N lt;N
(10)
;
where At;N is an exogenous stochastic productivity shock and kt
1;N
denotes capital used
in the production of the nontradable good. Capital is accumulated by the entrepreneur and the equation for capital accumulation is given by it;N = kt;N
(1
)kt
1;N ;
where it;N denotes investment in the nontradable sector. 13
(11)
Similar to the tradable sector, …rms in the nontradable sector also have a working capital requirement and face a credit constraint. The entrepreneur’s total value of debt including the interest payments cannot exceed a fraction of the expected value of the capital holdings: RbeN t + R wt;N lt;N
k meN t Et qt+1;N kt;N ;
(12)
denotes the intertemporal debt issued at time t by the nontradable sector entrewhere beN t preneur. The loan-to-capital ratio, denoted by meN t ; is modeled as a stochastic process. We use adjustment costs for capital accumulation in the nontradable sector as well, in order to reduce the volatility of investment. The price of capital in terms of tradable consumption, k ; is given by qt;N k qt;N =1+
where
(kt
1;N ; it;N )
@ (kt 1;N ; it;N ) ; @it;N
(13)
is the capital adjustment cost function.
The entrepreneur in the nontradable sector maximizes her expected utility E0
1 X
(
eN t
t=0
)
eN 1 (ceN ceN t t;N ; ct;T ) 1
(14)
subject to technology and borrowing constraints, as well as the following ‡ow of funds constraint eN ceN t;T +pt;N ct;N +pt;N wt;N lt;N +it;N + (kt
eN 1;N ; it;N )+Rbt 1 +(R
1) wt;N lt;N = pt;N yt:N +beN t : (15)
Consumption of the nontradable sector entrepreneur is also an aggregate of the coneN sumption of nontradable and tradable goods, ceN t;N and ct;T , respectively.
We also assume for the nontradable sector entrepreneur that
eN
< 1=R to make sure
that the credit constraint is binding in and around the steady state.
3.3
Equilibrium
eN Given initial conditions bh0 ; beT 0 ; b0 ; k0;T ; k0;N ; the constant interest rate r; the sequence of
shocks to sectoral productivity levels, the loan-to-income ratio of the household and the 14
loan-to-capital ratios of the entrepreneurs, the competitive equilibrium is de…ned as a set of allocations and prices
eT eT eN yt;T ; yt;N ; lt;T ; lt;N ; kt;T ; kt;N ; it;T ; it;N ; cht ; cht;T ; cht;N ; ceT t ; ct;T ; ct;N ; ct ;
eN h eT eN ceN t;T ; ct;N ; bt ; bt ; bt ; pt;N ; wt;T ; wt;N
such that (i) the allocations solve the problems of
households, and entrepreneurs in the tradable and nontradable sectors at the equilibrium prices, (ii) factor markets clear, and (iii) the resource constraints for the tradable and nontradable sectors hold: eN cht;T + ceT t;T + ct;T + it;T + it;N + (kt
1;T ; it;T )
+ tbt = yt;T
(16)
eN cht;N + ceT t;N + ct;N = yt;N
(17)
+ (kt
1;N ; it;N )
where the trade balance is de…ned as tbt = R bht
4
1
eN + beT t 1 + bt 1 + (R
1) wt;T lt;T + (R
1) wt;N lt;N
eN bht + beT : (18) t + bt
Calibration
The model is solved using quarterly Turkish data for the period 1999Q1-2011Q4. The construction of the series used in the model solution is explained in detail in the Appendix. The parameter values of the model are summarized in Table 3. We take the real interest rate as constant and set it equal to the average real interest rate in Turkey, which equals 0.0117. We set the discount factors such that the credit constraints bind in and around the steady state. The values for value for
eT
h
and
eN
are set to 0.94, and the
is set to 0.97, which are the highest possible values that guarantee binding
credit constraints in the solution of the model. The value of ; which determines the intertemporal elasticity of substitution in labor supply, is set to 1.7 following Correia et al. (1995). The coe¢ cient of relative risk aversion is set to 1, which corresponds to log-utility. The annual depreciation rate is set to 0.08 following Meza and Quintin (2007).
15
Table 3. Parameter values of the benchmark model Parameter Value
Description
h
0.94
Discount factor of households
eN
0.94
Discount factor of nontradable sector entrepreneurs
eT
0.97
Discount factor of tradable sector entrepreneurs
1
Relative risk aversion coe¢ cient
1.7
Labor curvature
1.589
Labor weight in utility
0.54
Nontradable weight in the consumption aggregator
0.35
Capital exponent in the tradable sector
0.25
Capital exponent in the nontradable sector
0.08
Annual depreciation rate
r
0.0117 Real interest rate
'T
5.21
Capital adjustment cost coe¢ cient in the tradable sector
'N
18.30
Capital adjustment cost coe¢ cient in the nontradable sector
0.25
Working capital coe¢ cient
mh
0.423
Loan-to-income ratio
meN
0.266
Loan-to-capital ratio in the nontradable sector
meT
0.102
Loan-to-capital ratio in the tradable sector
Stochastic processes AT
0.662
h T
0.697
("AT )
AN
0.777
h N
0.062
("AN ) 0.0148
h
0.817
eT T
0.153
("h )
0.0350
eT
0.656
eN N
0.362
("eT )
0.0287
eN
0.803
("eN )
0.0270
0.0283
We cannot calibrate the capital share parameters in the tradable and nontradable sectors for Turkey due to unavailability of data. Di¤erent values have been used in the literature for these parameters and the general consensus is that the tradable sector is more capital intensive than the nontradable sector. Therefore, we set the capital’s share of income equal
16
to 0.35 in the tradable sector and 0.25 in the nontradable sector, which are close to the values used in the literature.4 The value of
is set to 1.589 so that the steady state labor supply equals 0.17, which
is the average value of time spent working as a percentage of total discretionary time in Turkey. The share of nontradable goods in the consumption aggregator, ; is set equal to the average share of nontradable consumption in total consumption in Turkey. The steadystate value of the loan-to-capital ratio in the nontradable and tradable sectors, meN and meT ; are set to match the average value of business credit in each sector as a ratio of GDP for the sample period, which are 11.2% and 9.6%, respectively. Likewise, the steady-state value of the loan-to-income ratio, mh ; is set to match the average value of the ratio of household credit to GDP in the data, which is 7.2%. For the calibration of the parameter ; we use data on short-term bank loans from the Company Accounts database of the Central Bank of Turkey. Total liabilities of …rms are composed of intertemporal loans and working capital loans in our model, and the loans for working capital have a shorter duration compared to the other loans. Therefore, we choose to approximate the working capital loans with short-term bank loans. We calibrate
by
taking the average of the ratio of short-term loans to the compensation of employees, which is equal to 0.25. Since the data on short-term bank loans are not available at the sectoral level, we use the same value for both tradable and nontradable sectors. The consumption aggregator is assumed to be of the following form for all agents: cjt cjt;N ; cjt;T = (cjt;N ) (cjt;T )1 ;
0
