Research Article  Open Access
Zhongyuan Geng, Xue Zhai, "Effects of the Interest Rate and Reserve Requirement Ratio on Bank Risk in China: A Panel Smooth Transition Regression Approach", Discrete Dynamics in Nature and Society, vol. 2015, Article ID 571384, 8 pages, 2015. https://doi.org/10.1155/2015/571384
Effects of the Interest Rate and Reserve Requirement Ratio on Bank Risk in China: A Panel Smooth Transition Regression Approach
Abstract
This paper applies the Panel Smooth Transition Regression (PSTR) model to simulate the effects of the interest rate and reserve requirement ratio on bank risk in China. The results reveal the nonlinearity embedded in the interest rate, reserve requirement ratio, and bank risk nexus. Both the interest rate and reserve requirement ratio exert a positive impact on bank risk for the low regime and a negative impact for the high regime. The interest rate performs a significant effect while the reserve requirement ratio shows an insignificant effect on bank risk on a statistical basis for both the high and low regimes.
1. Introduction
Loose monetary conditions, such as low interest rates, often result in excessive credit expansion, which can largely explain the financial imbalances and economic fluctuations. Following the burst of the dotcom bubble, many central banks preferred a soft monetary policy and exerted a low interest rate over an extended period to ease the potential recessions. Continuous low interest rates can boost the increase in asset prices and securitized credit and push financial entities to take more risks [1]. It seems that bank risk is therefore increased. Although it is not the time to attribute this type of monetary policy to the 2008 global financial crisis, it may have contributed to its buildup. Thus, more and more academic and practical debates are around the effect of monetary policy on bank risk, which has become a central issue off the back of 2008 global financial crisis [2, 3].
The Chinese banking system contributes the most to the financial system in China, and risks exhibited by commercial banks are the biggest threat to the nation’s financial stability. The impact of monetary policy on bank risk is thus an essential issue to which should be paid great attention when establishing its macroprudential management framework. Unlike advanced economies in which standard oneinstrument (a policy interest rate) operating procedure dominates the monetary policy tools, People’s Bank of China (PBC, China’s central bank) makes a frequent adjustment on the interest rate (a main pricebased instrument) and the reserve requirement ratio (a main quantitative instrument) simultaneously to achieve its goals. During 2007–2012, People’s Bank of China had adjusted the RMB 1year benchmark deposit rate and the reserve requirement ratio for seventeen times (six times in 2007, four times in 2008, two times in 2010, three times in 2011, and two times in 2012) and thirtyfour times (ten times in 2007, nine times in 2008, six times in 2010, seven times in 2011, and two times in 2012), respectively, which is quite uncommon in the international practice. There come some interesting questions: does interest rate and the reserve requirement ratio have different effects on the bank risk? Will it firm up the financial stability and price stability if China sticks to the frequent and simultaneous manipulation on the interest rate and the reserve requirement ratio? (About the relationship between price stability and financial stability, there are two conflicting viewpoints. One is “synergy” viewpoint that monetary policy aiming at price stability will be conducive to financial stability [4]. The other is “tradeoff” viewpoint that monetary policy aiming at price stability is not necessarily helpful for financial stability and a tradeoff relationship exists between price stability and financial stability [5, 6].) Is China’s experience in monetary policy worth learning for other economic entities in tailoring their monetary policy? The answers of the above questions require a deep dive in the empirical test of the effects of the interest rate and the reserve requirement ratio on the bank risk in China.
A linear model has been the main focus of most research on the effect of monetary policy on bank risk, while from our perspective, monetary policy instrument delivers a nonlinear effect on bank risk due to the subjective and irrational property of bank risk behavior. The bank’s risk appetite, risk perception, and risk decisionmaking behavior changes slowly, gradually, and continuously following the implementation of monetary policy. Moreover, different monetary policy instruments have various impacts on bank risk in terms of different macro environments and bank characteristics, which is subject to uncertainty thanks to the counteracting determinants as well. Therefore, the relationship between monetary policy instruments and bank risk follows a nonlinear path. It is much more reasonable to utilize a nonlinear model to analyze the effects of monetary policy instruments on bank risk in avoiding assumption errors and bias.
Nonlinear theories and models have matured gradually since the 1970s, which sparks scholars in accepting the fact that nonlinear model can fit economic phenomena and economic laws in a better manner [7]. Among multiple available nonlinear models, regimeswitching models are the most popular ones. Common nonlinear regimeswitching models include the following three models: Markov regimeswitching (MRS) model, threshold regression (TR) model, and smooth transition regression (STR) model. MRS and TR models are based on the assumption that the transition from one regime to anther is discrete, which is inconsistent with the reality in many cases, and thus limits their application in practice. Hansen [8] made an initial effort on introducing threshold effects, together with a panel threshold regression (PTR) model which assumes a jumping transition through different regimes. In improving the practicability, González et al. [9] developed the Panel Smooth Transition Regression (PSTR) model, extending a smooth transition regression (STR) model to panel data heterogeneity across the panel members and over time [10]. The merged PSTR version of combining both STR and panel data enables the transition to switch between regimes over time as smooth as it can be.
This paper utilizes a PSTR model to study the nonlinearity between monetary policy instruments (i.e., the interest rate and the reserve requirement ratio) and bank risk. Our study differs from the previous literature in the following ways. First, in the PSTR model established for the analysis of effects of the interest rate and the reserve requirement ratio on bank risk, the empirical results indicate that the interest rate and the reserve requirement ratio have nonlinear impacts on bank risk. Second, the result demonstrates that the objective function of the central bank should be nonlinearoriented and bring in financial sector. The standard textbook approach adopts a linearquadratic (LQ) framework in analyzing optimal monetary policy, where the dynamic behavior of the economy is described as linear and the objective function stressing the policy goals is quadratic. Monetary policy is always keeping a balance by seeking an optimal match point at which the loss function is minimized and the squared value of the inflation gap and the squared value of the output gap are comprised at the same time [11]. Our empirical results give evidence to the fact that monetary policy instruments produce nonlinear effects on bank risk. In that case, the central bank should incorporate nonlinear elements and a financial stability variable into its objective function. Third, we try to differentiate the effects of the interest rate and the reserve requirement ratio on bank risks, thereby providing comprehensive policy guidance in China’s implementation. Additionally, some useful information can be dug up from the result as well facilitating the policymakers of other countries in designing their monetary policy.
The remainder of the paper is organized as follows: Section 2 reviews the literatures. Section 3 sketches out the general empirical model to be estimated and describes the data. Section 4 presents the empirical results and related comments. Section 5 provides the conclusive remarks.
2. Literature Review
The effects of monetary policy on bank risk are a part of, although distinguishable from, the relationship between monetary policy and financial stability which is detailed by Oosterloo and de Haan [12] and is not discussed here. So far, there is very limited theoretical support and empirical evidence about the effects of monetary policy on bank risk [1].
Advanced economies typically use a policy interest rate as the monetary policy instrument. So, the studies on the effects of monetary policy on bank risk concentrate on the effects of interest rate on bank risk. The theoretical research in the literature suggests several channels that interest rate affects bank risk [13, 14]: (1) “Asset valuation” channel: A reduction in the interest rate boosts asset and collateral values, which in turn can modify banks’ estimation of probabilities of default, loss given default, and volatilities, and it incents banks to take on risk. (2) “Search for yield” channel: Low interest rates cause banks’ target revenue to decline, which provokes banks to invest in highmargin and highrisk areas or financial instruments. (3) “Asset substitution” channel: The decline in interest rates will lead to a low proportion of safe assets in the bank assets portfolio. Riskneutral banks will increase the demand for risky assets until a new equilibrium arises in the ratio of safe assets and risky assets. (4) “Constant leverage” channel: Commercial banks target a constant leverage ratio. Low interest rates will boost the assets prices. Bank equity will increase and banks will respond to the fall in leverage by increasing their demand for risky assets. This reaction reinforces the initial boost to asset values, and so on. The result is a more fragile banking system that is more exposed to negative shocks to asset values and thus riskier. (5) “Central bank communication” channel: If the central bank has transparent policy and credible commitment, low interest rate is an implicit commitment that will induce collective moral hazard. Low interest rates mean loose monetary and regulatory environment, which stimulates banks to take on more risk. (6) “Assetliability mismatch” channel: When interest rates are low, banks can only absorb shortterm deposits. The mismatch between shortterm deposits and longterm project finance tends to high leverage. The more leveraged the banks are, the higher the risk of failure is. (7) “Habit formation” channel: If the interest rate is low, investors tend to consume more and the expected credit spread is high. Thus, investors are willing and able to get more loans from the bank or invest in highrisk financial instruments, which results in higher bank risk. In addition, some studies suggest that interest rates have an uncertain effect on bank risk, which depends on many factors that affect the mutually countervailing forces [1]. The effect of changes in interest rates on bank risk may change over time, along with a change in the banking system or a change in the characteristics of the bank itself [15].
Empirical research shows some conflicting findings. One such finding claims that low interest rates lead to an increase in bank risk and high interest rates can prevent its accumulation [16–19], while others [20, 21] claim that the reverse is true. Interestingly, Thakor [22], Jiménez et al. [23], and Martha López et al. [24] document an uncertain effect of interest rates on bank risk. Interest rates have a smaller impact on the risky assets of banks with higher capital, but a larger effect on the banks with more offbalance business. Certain banks can react heterogeneously to interest rate changes. Banks with a high capital adequacy rate and income diversification perform more radically in their risktaking.
Some observation can be noted from the literature cited above. First, most of the previous studies have employed the ordinary least squares and generalized least squares methodology to establish a linear model to study the impact of monetary policy on bank risk in the context of crosssectional or time series. Yener et al. [25] studied the nonlinear effects of monetary policy on bank risk by simply incorporating the quadratic term of an explanatory variablecredit expansion into the linear regression equation, without establishing a cuttingedge nonlinear model. Second, few studies compare the different effects of the interest rate and the reserve requirement ratio on bank risk. In advanced economies, interest rate is the major monetary policy instrument. Thus, they focus primarily on the effect of interest rates on bank risk. In China, some scholars concentrate on the effects of the interest rate and the reserve requirement ratio on bank risk, but they all draw the same conclusion that both the interest rate and the reserve requirement ratio have a negative effect on bank risk [26, 27].
3. Model Specification and Data
When the sample size is not sufficiently large, the introduction of too many explanatory variables will result in the decline in the degrees of freedom and multicollinearity. Therefore, this study will only concentrate on the impact of macroeconomic factors on bank risk and does not consider effects of the banklevel micro factors on bank risk. From the angle of monetary policy instruments, we construct the following PSTR model to study the effects of interest rate and reserve requirement ratio on the bank risk (about the detailed methodology of PSTR model, see Granger and Teräsvirta [28], Teräsvirta [29], Eitrheim and Teräsvirta [30], Hansen [8], and González et al. [9]). Considerwhere , , and and denote the cross section and timedimension of the panel, respectively. = expected default frequency, which is the dependent variable, represents the fixed effects, = interest rate; = reserve requirement ratio; = real estate price index; = purchasing managers’ index, a threshold variable. The is the transition functions, normalized to be bounded between 0 and 1. (When the transition function equals 0 or 1, the corresponding model is, resp., called low regime or high regime. The values of transition function transit between 0 and 1 smoothly, which makes the model transit between low regime and high regime smoothly.) slope parameter denotes the speed of transition from one regime to the other, the threshold parameters, the residual term, and the regression coefficients.
Before carrying out the empirical analysis, we should discuss the variables used and the dataset. In view of the availability of the data, we use the quarterly (from 04/2007 to 03/2012) data of thirteen Chinese listed banks. (The data of listed banks is from the share market rather than share market. We exclude Agricultural Bank of China, China Everbright Bank, and China Construction Bank. The reasons are as follows. Agricultural Bank of China and China Everbright Bank became a listed bank in 2010 and there is not much data available. In December 2011, the number of total shares, shares, and shares of China Construction Bank is, resp., 250.01 billion, 9.593 billion, and 214.83 billion. The ratio of shares to total shares is only 3.84%, producing the mismatch between share market value (too less) and liabilities (too more). This mismatch will result in the fact that calculated value of EDF cannot objectively reflect the expected default frequency of China Construction Bank.) Considering that the calculation of EDF requires bank’s stock returns and market value data which can be present only after the bank lists and that the Bank of Communications, the Industrial Bank, the CITIC Bank, the Bank of Ningbo, the Bank of Nanjing, and the Bank of Beijing became listed banks in 2007, our sample starts from the fourth quarter of 2007 so as to obtain sample data as much as possible. The sample contains three large commercial banks, seven jointstock commercial banks, and three city commercial banks. Among those, the large commercial banks include the Industrial and Commercial Bank of China (ICBC), the Bank of China (BOC), and the Bank of Communications (BOCOM); the jointstock commercial banks include CITIC Bank, Huaxia Bank (HB), Pingan Bank (PB), China Merchants Bank (CMB), Shanghai Pudong Development Bank (SPDB), Industrial Bank (IB), and China Minsheng Banking Co. (CMSB); the city commercial banks include Beijing Bank (BB), Nanjing Bank (NJB), and Ningbo Bank (NBB).
Table 1 provides descriptive statistics for the variables used in the empirical analysis. Table 2 reports correlation coefficients between these variables. According to Gujarati Damodar [31], if the zeroorder correlation coefficient of two regressors is over 0.8, the multicollinearity problem will be severe. Correlations in our study are at acceptable levels as shown in Table 2.


In what follows, we analyze the choice of the dependent, explanatory, transition, and control variables.
3.1. Bank Risk
As for the dependent variable measuring bank risk, we chose the expected default frequency (EDF). This indicator has become a popular measure of bank soundness in related empirical work on financial stability. Theoretically, according to the nature of the risk, EDF that utilizes stock price and earnings volatility to characterize the risk behaviour of the bank is undoubtedly the ideal choice [25]. The reasons are listed below. First, EDF is relatively objective because it is calculated on the basis of stock transaction data and financial data found in the financial statements of the listed banks. Second, EDF is a dynamic index and can be updated based on changes of stock transaction data and regularly published financial statements of listed banks. Therefore, EDF can reflect changes of bank risk over time. Third, EDF overcomes the bias caused by applying historical data to represent future trends. EDF is calculated on the basis of realtime situations of the stock market. Changes of the yields and market value in the stock market can reflect the bank’s performance, market expectations, and future trends.
EDFs are the outcome of Moody’s KMV model which establishes a functional relationship between distance to default and the probability of default. The EDF of a company varies over time, reflecting the changing economic prosperity of the firm or its industry sector. A detailed description of the mapping between the distance to default and the EDF measure can also refer to Crouhy et al. [32].
We use the method of Brandimarte [33] to calculate the EDF. Results of EDF calculations by Matlab7.0 software are reported in Table 3. The riskfree interest rate needs to be used to calculate EDF and it is based on the daily weighted average of the RMB 1year benchmark deposit rate. (The following data is from RESSET Financial Research Database and RESSET is China’s leading provider of financial databases: thirteen listed banks’ daily yield, daily total market value, quarterly longterm liabilities, and quarterly shortterm liabilities that are used to calculate EDF; broad money; quarterly real estate price index; RMB 1year benchmark deposit rate. The data of purchasing manager’s index (PMI) is derived from CEInet statistics database (CEI: China Economic Information).)
