THE CREDIT CYCLE and the BUSINESS CYCLE in the ECONOMY of TURKEY

Transcription

1 810 September 2014 Istanbul, Turkey 442 THE CYCLE and the BUSINESS CYCLE in the ECONOMY of TURKEY Şehnaz Bakır Yiğitbaş 1 1 Dr. Lecturer, Çanakkale Onsekiz Mart University, TURKEY, sehnazbakir@comu.edu.tr Abstract This study examines econometric relationships between bank lending and the business cycle in Turkey. Firstly the cyclical component of the real and real bank loans were determined using time series. A cointegration analysis and a vector errorcorrection model with quarterly data were used for 1987:012013:03 period. The results of cointegration analysis indicate that there is a single stable longrun equilibrium relationship between real bank loans and macroeconomic variables. The response of bank loans to shocks is positive. Bank loans have procyclical character in Turkey. Keywords: Business cycle, Credit cycle. 1. INTRODUCTION The global crisis has highlighted the connection between credit conditions and economic performance. Economists have struggled with the question of whether banks change their lending standards systematically over the business cycle. Bernanke and Gertler (1989) argue that bank lending is not only procyclical, but that the availability of bank loans may also increase the magnitude of business cycles. Gambetti and Musso (2012) argue that the impact of loan supply shocks is particularly important during slowdowns in economic activity. Banks are important institutions which finance household and business spending in Turkey. The share of bank loans in external financing of the private sector is around 82% (Central Bank, 2012). In this respect, testing the relationship between the business cycle and bank lending is important. The purpose of this study is to analyze the relationship between bank loans and business cycles in terms of Turkey s economy. The rest of this study is structured as follows: Section II explores the cyclical component of the real and real banking loans using time series in Turkey. Thus whether bank loans were procyclical, countercyclical or acyclical is determined. Section III provides the data, econometric approach and empirical results. Finally, Section IV presents the conclusions of this study. 2. THE CYCLE and the BUSINESS CYCLE in the ECONOMY of TURKEY There are a large number of statistical methods that are used to separate the cyclical component of a time series from raw data. The HodrickPrescott (HP) filter is the most widely used technique. It is used to obtain a smoothedcurve representation of a time series, one that is more sensitive to longterm than to shortterm fluctuations. The adjustment of the sensitivity of the trend to shortterm fluctuations is achieved by modifying a multiplier λ. For quarterly data Hodrick and Prescott (1997) proposed a value of q = 1/1600. Rand and Tarp (2002) observed that business cycles in developing countries are significantly shorter in duration compared to cycles in developed countries, making λ = 1600 inappropriate for developing countries. Alp et al. (2011) estimated the optimal smoothing for HodrickPrescott filter for Turkey using quarterly real data, and found that for the shorter business cycle length in Turkey the optimal λ was 98 and 17. Using our finding on the average length of the business cycle for Turkey, we estimate the HPfilter smoothing parameter for Turkey as 98 with the methodology proposed by Pedersen (2001).

2 810 September 2014 Istanbul, Turkey 443 Figure 1 shows the turning points in Turkish real series for 1987 to 2013 period. HodrickPrescott Filter (lambda=98) L_SA Trend Cycle Fig. 1. Turning Points in Turkish Real (1987Q12013Q3) Source: Central Bank of Republic of Turkey (CBRT). Real data is seasonally adjusted with Census X12 after taking natural logarithms. HodrickPrescott Filter (lambda=98) 60,000 50, ,000 30,000 20,000 10, HPTREND01 Trend Cycle Fig. 2. Business Cycles in Turkey (1987Q12013Q3) Source: Central Bank of Republic of Turkey (CBRT). Real data is seasonally adjusted with Census X12 after taking natural logarithms. The first trough point is observed in the beginning of 1989, which is followed by the one in 1994:2, which reflects repercussions of the 1994 economic crisis. The third trough appears in period reflecting the effects of Russian crisis. The trough point observed in 2001 coincides with the Turkish banking crisis. The last trough point is observed in period reflecting the effects of the global economic crisis. In the same periods, bank loans dropped to the lowest level (Figure 3).

3 1987Q1 1987Q3 1988Q1 1988Q3 1989Q1 1989Q3 1990Q1 1990Q3 1991Q1 1991Q3 1992Q1 1992Q3 1993Q1 1993Q3 1994Q1 810 September 2014 Istanbul, Turkey 444 Figure 2 shows business cycles in Turkey for 1987:012013:03 period. Real narrowed substantially during 1994, and crises. HodrickPrescott Filter (lambda=98) L Trend Cycle Fig. 3. Turning Points in Turkish Real Bank Loans (1987Q12013Q3) Source: Central Bank of Republic of Turkey (CBRT). Loan data were converted to real terms using 1987 = 100 index. To capture financial fluctuations and real fluctuations, loan growth rate and growth rate are used (Egert and Sutherland, 2012). Figure 4, 5, 6, 7 and 8 show the relationship between loan growth and growth in Turkey. In this study, five different subperiods are determined for the period These periods are as follows: 1987Q11994:01, 1994:021997:04, 1998:012001:04, 2002:012006:04, and 2007:12013:03. Table 1 shows the descriptive statistics of real growth rate and real loan growth rate for each period. The real credit growth and real gross domestic product coincide in Turkey. Therefore, it is possible that to say the bank loans in Turkey have procyclical character Fig :011994:1 Period

4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 810 September 2014 Istanbul, Turkey , , ,0000 5, ,0000 Fig :21997:04 Period Fig :012001:04 Period 25, , ,0000 5, ,0000 Fig :012006:04 Period

5 810 September 2014 Istanbul, Turkey , , ,0000 5, ,0000 Figure 8: 2007:012013:01 Period Table 1. Real Growth Rate and Real Credit Growth Rate (Descriptive Statistics) 1987: :03 RGSYH RKREDİ Mean Median Maximum Minimum Std. Dev Skewness Kurtosis JarqueBera Probability : :04 RGSYH RKREDİ 1987: :01 RGSYH RKREDİ Mean Median Maximum Minimum Std. Dev Skewness Kurtosis JarqueBera Probability : :04 RGSYH RKREDİ 1994: :04 RGSYH RKREDİ Mean Median Maximum Minimum Std. Dev Skewness Kurtosis JarqueBera Probability : :03 RGSYH RKREDİ Mean Mean Mean Median Median Median Maximum Maximum Maximum Minimum Minimum Minimum Std. Dev Std. Dev Std. Dev Skewness Skewness Skewness Kurtosis Kurtosis Kurtosis JarqueBera JarqueBera JarqueBera

6 810 September 2014 Istanbul, Turkey 447 Probability Probability Probability EMPIRICAL RESULTS We begin with a standard vector autoregression (VAR), which recognizes the endogeneity of macroeconomic variables and is written as follows: yt A0 A y... A y 1 t 1 p, t t p t (1) where y is a vector of observable endogenous variables. The core of our VAR comprises just four variables: log real loan volumes (crd), log real (y), the general price level (cpi), and log real monetary supply (m2). Also crisis variables are added to the model as dummy variables. These variables are 1994 crisis, 2001 crisis and 2008 crisis. All variables were transformed from nominal values to real values by using the CPI (1987=100). The model was estimated with quarterly data for period 1987Q12013Q3. All data were taken from the Central Bank of the Turkish Republic Electronic Data Distribution System. Table II. Results of the Unit Root Test Variables (Level) (1stDif.) intercept intercept and trend intercept intercept and trend crd 1.18(1) 1.24(1) 7.52(0)* 7.93(0)* y 0.45(0) 3.11(0) 10.42(0)* 10.38(0)* cpi 1.10(3) 2.24(3) 13.61(2)* 13.52(2)* m2 0.76(0) 1.97(0) 9.29(0)* 9.49(0)* Note: *The critical values at a significance level of 1% in the ADF test, for intercept model and intercept and trend model is 3.49 and 4.05 respectively. As the first differences of the series are stationary (Table II), the Johansen cointegration test is used to identify the longterm relationship between variables. Criteria LR (Likelihood Ratio) is used to determine the lag length of the VAR model, and it is essential that the VAR model with the selected lag length has no autocorrelation or heteroscedasticity problems. According to the LR test, the determined lag length is five (k=5). Table III. Johansen Cointegration Rank Test Results Null Hypothesis Eigenvalue Trace Statistic 5% Critical value H 0: r = * H 0: r * H 0: r Note: * It shows that the hypothesis is rejected at a level of significance of 5%.

7 810 September 2014 Istanbul, Turkey 448 The longterm equilibrium model containing the variables is as follows: crd 4.15 y 0.14 m cpi 0.03dm(1994) 0.08dm(2001) 0.118dm(2008) (2) (0.307) (0.072) (0.050) (0.023) (0.018) (0.026) This result indicates that there is a longterm equilibrium relationship between all variables. As the Johansen cointegration test proved the presence of a longterm equilibrium relationship between macroeconomic variables and financial variables, one may look at the causality relationship between the variables. An error correction model is used to determine the direction of causality between cointegrated series. The lag length in the error correction model is four (k=4). Table IV. Vector Error Correction Model Variables Coefficient Standard Error t statistic Probability ΔCRD (1) ** ΔCRD(2) ** ΔCRD(3) ** ΔCRD(4) ** ΔM2(2) ** ΔDM(1994) * ΔDM(2008) *** ECM coefficient ** Rsquared Mean dependent var Adjusted Rsquared S.D. dependent var S.E. of regression Akaike info criterion Sum squared resid Schwarz criterion Log likelihood HannanQuinn criter Fstatistic DurbinWatson stat Prob (Fstatistic) In the estimated error correction model (Table IV), the coefficients of the lagged variables represent the shortterm dynamics of the dependent variable. The error correction term in the model is negative and statistically significant. The negative coefficient of the error correction term indicates that the dependent variable meets the shortterm adaptation rate for longterm equilibrium. The coefficient of the money supply variable is positive and significant; the coefficients of dummy variables (1994, 2008) are negative and significant. This finding indicates that there is no causal relationship between real and inflation and real loans. Figure 9 shows the impulse response function of the variables. According Figure.., and money supply shocks have a positive impact on real loan volumes. The real loan volumes exhibit a positive response to two quarters and turn negative soon after. The response of loans remains positive after about eight quarters. The response of the loan volumes to inflation rate is negative.

8 810 September 2014 Istanbul, Turkey 449 Response of Credit to Response of Credit to M Response of Credit to CPI Figure 9: Impulse Responses for VECM 4. CONCLUSIONS An economic analysis which determined the cyclical component of the real and real bank loans using time series indicated that the real loan growth and real gross domestic product coincide in Turkey. The variables used in the econometric model are analyzed for both shortterm and longterm relationships. The main results of the empirical analysis are as follows. The results of cointegration analysis indicate that there is a single stable longrun equilibrium relationship between real loan volumes, real, real money supply, inflation rate and crisis variables. Money supply in the 1994 and 2008 crises are found to influence real loan volumes in the short term. The 2001 crisis affects real loan volumes in the long term. Another important finding is that the effect of real on real bank loans is lost in the short term. On the other hand, impulse response functions show that the real shocks appear to have a significant effect on credit markets. 5. REFERENCES LIST Alp, H., Başkaya, Y. S., Kılınç, M., and Yüksel, C. (2011). Estimating Optimal HodrickPrescott Filter Smoothing Parameter for Turkey: Anadolu International Conference in Economics II), June 1717, 2011, Eskişehir, Turkey. Bernanke, B. and Gertler, M. (1989). Agency Costs, Net Worth and Business Fluctuations: American Economic Review, vol. 79, Gambetti, L. and Musso, A. (2012). Loan Supply Shocks and the Business Cycle: Working Paper Series, No: Hodrick, R. J. and Prescott, E. C. (1997). Postwar U. S. Business Cycles: An Empirical Investigation: Journal of Money, Credit and Banking, vol. 29, 116. Pedersen, T. M. (2001). The HodrickPrescott Filter, the Slutzky Effect, and the Distortionary Effect of Filters: Journal of Economic Dynamics and Control, vol. 25, Rand, J. and Tarp, F. (2002). Business Cycles in Developing Countries: Are They Different?: World Development, vol. 30,