Maxim Nikitin International College of Economics and Finance Higher School of Economics Moscow, Russia and R. Todd Smith University of Alberta Department of Economics Edmonton, AB, Canada Bank Spreads and Business Cycles We analyze the business-cycle dynamics of commercial bank spreads and their components for Canada, Italy, Germany, the Netherlands, Switzerland, the United Kingdom, and the United States. In particular, we focus on loan spread, i.e. the spread between the loan rate and the interbank market rate, as well as the deposit spread, i.e. the gap between the interbank rate and the deposit rate. We analyze spreads for 3 kinds of deposits, savings deposits, time deposits and eurodeposits (the deposits nominated in the foreign currency). Motivation for this analysis is twofold. First, the amplitude of business cycle fluctuations depends on spreads. Output dynamics likely depends heavily on banks’ pricing of loans (relative to other interest rates) at different stages of the business cycle. Second, estimates of the behavior of bank loan and deposit pricing over business cycle may provide insight into whether banks exploit market power in certain deposit market segments (e.g., ordinary savings deposits) either to buffer adverse shocks to other (more competitive) deposit market segments (e.g., time deposits or eurodeposits) or to their loan customers. Previous empirical studies of bank spreads had several limitations: 1) they focused almost exclusively on the US; 2) they analyzed the total spreads between the loan rates and deposit rates, but not their components; 3) they identified various factors that affect bank spreads, including the volatility of short-term interest rates, bank market power in loan and deposit markets, bank risk aversion, and default risk on bank loans. However, they typically made no attempt to link these factors to the business cycle fluctuations. The paper tests several hypotheses. 1) The baseline hypothesis is that controlling for various other determinants of bank spreads, the loan-deposit spreads are countercyclical—they increase during recessions and decrease during economic expansions—and the reason for this is that loan markups (over some benchmark interest rate) have a marked countercyclical component. The baseline hypothesis is based on the financial accelerator theory developed by Bernanke et al. (1996) and Kiyotaki and Moore (1997), who show that business cycles are exacerbated by the cyclical behavior of bank lending rates because borrowers’ net worth (or collateral) deteriorates during recessions. 2)Relationship hypothesis: banks use their market power in the deposit markets to smooth the loan spreads over the business cycle: deposit spreads are expected to be countercyclical as well. 3) Cyclical competitiveness hypothesis: savings deposit spreads are more countercyclical than time deposit spreads, and, especially, than eurodeposit spreads. This might be true because banks enjoy more market power in the market for savings deposits, than in the market for time deposits, and, especially, in the market for eurodeposits. Therefore, banks exploit their market power in the markets for saving deposists to 1 buffer adverse shocks to banks’ cost of funds in more competitive segments of the deposit market. 4) Countercyclicality of deposit spreads is more pronounced in continental Europe than in Anglo-Saxon countries. This is related to the fact that capital markets are more developed in Anglo-Saxon countries, and hence commercial banks there face stiffer competition in the deposit and loan markets. On the other hand, the relationship banking is much less developed there. Therefore the ability of commercial banks in the Anglo-Saxon countries to exploit their market power in the deposit market, as well as the desire to do it, are much smaller than in continental Europe. Data Description The analysis uses quarterly data. Sample periods begin in the late 1960s or early 1970s and end in the late 1990s. Exceptions are eurodeposit rates for Canada and Italy which are not available until the mid to late 1970s. Unfortunately, the data on ordinary savings deposits rates is unavailable for Germany and for the United States. The date comes from the following databases: Datastream, OECD Financial Statistics Monthly Databank, International Financial Statistics, OECD Analytical Database. Econometric Specification The main estimated equation is as follows: SPREADi,t = Et A(L)Yi,t βy + Xi,t βx + εi,t (1) Here SPREADi,t - is a spread in country i in quarter t, Yi,t - GDP growth rate in country i in quarter t, Xi,t – row vector of control variables, A(L) - polynomial in the lag/lead operator. We take three lags, three leads and the contemporaneous value of the GDP growth rate. Et is the expectation operator at time t, βx βy are the column vectors of estimated coefficients, and εi,t are the regression residuals. Equation (1) is estimated separately for each kind of loan and deposit spread. A(L) is the main focus of the paper. Real GDP growth rate is normalized by its sample standard deviation, so the coefficients on Yi,t can be interpreted as the effects of a one-standard deviation change in real GDP growth on the spread. Lagged and contemporaneous GDP growth may matter for spreads because borrower creditworthiness is countercyclical; this is the main point of Bernanke et al. (1996) and Kiyotaki and Moore (1997). Expected future GDP may matter for similar reasons: an anticipated deterioration in borrower creditworthiness could be reflected in current loan rates. The list of control variables includes variables that showed their significance in the earlier studies of bank spreads, namely, short-term interbank interest rate, volatility of short-term interest rate (measured by the average over the quarter of squared monthly deviations of the short-term interest rate from the trend of this interest rate), ratio of bank reserves to total assets, inflation rate, slope of the yield curve (defined as the spread between a three-month interbank rate and the overnight rate), lagged ratio of savings deposits to demand deposits, lagged ratio of loans to total assets and the time trend. Evaluating equation (1) involves several econometric problems. First, there is an endogeneity problem, because the GDP growth rate in the current quarter and short-term interest rate are determined simultaneously with spreads. This problem can be resolved using the 2 instrumental variables method. The instrument set includes lags of right-hand side variables (beginning with period t-1) as well as the contemporaneous values of inflation and bank reserves. Second, the regression residuals are autocorrelated and heteroskedastic. To address these problems we use heteroskedasticity and autocorrelation consistent estimator (HAC) developed by Newey and West (1987). This is a generalization of the White’s heteroskedacticity-robust estimator for the case when both both autocorrelation and heteroskedasticity have unknown form. Estimation Results The main estimation results are summarized in the table below. Country Loan spread Savingsdeposit spread Time-deposit spread Euro-deposit spread Canada 0.062 0.333*** 0.043 Germany 0.373 -1.591*** -0.150 Italy -0.277 -1.664*** -1.089*** 1.019*** Netherlands -0.692*** -1.346*** -1.350*** 0.174** Switzerland -0.468*** -0.312*** -0.121 -0.032 UK -0.319*** -0.252* 0.413*** 0.697*** US -0.569*** - 0.165*** 0.062 -0.389*** - The coefficients reported in the table are the effects of a one-standard-deviation increase in GDP growth rate in 3 preceding quarters, the current quarter and 3 subsequent quarters. As usual, three stars denote coefficients significant at 1 % significance level, and two stars denote coefficients significant at 5 % significance level. The main conclusions from the table above are as follows: 1) Loan spreads are countercyclical in all countries except Canada and Germany. The result is highly significant and quantitatively important in the US, UK, the Netherlands and Switzerland. Therefore, the baseline hypothesis is largely supported by the data. 2) Savings-deposit spreads are countercyclical in all countries, but they are more countercyclical in continental Europe. The coefficients are below 0.4 by absolute value in the UK and Canada, while they are greater than one in Italy and the Netherlands. 3) Time-deposit spreads in most of the continental European countries display marked countercyclicality. 3 4) Time-deposit spreads in Anglo-Saxon countries are modestly procyclical. 5) In all countries eurodeposit spreads are either acyclical (Canada, Germany, Switzerland, US) or procyclical (Italy, the Netherlands, UK). Therefore all theoretically motivated hypotheses are supported by the data. However, the difference between continental Europe and Anglo-Saxon countries is significant only for the time deposits. _____________________________________________________________________________ Bernanke, Ben, Gertler, Mark, and Simon Gilchrist (1996) The Financial Accelerator and the Flight to Quality. Review of Economics and Statistics 78, 1–15. Kiyotaki, Nobuhiro and John Moore (1997) Credit Chains. Journal of Political Economy 105, 211–248. Newey, Whitney K. and Kenneth D. West (1987) A Simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix. Econometrica 55, 703– 708. 4