شناسایی ترازنامه و کانال های وام دهی انتقال پولی : تجزیه و تحلیل در سطح وام
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|20498||2013||11 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Banking & Finance, Volume 37, Issue 8, August 2013, Pages 2812–2822
We make a novel attempt at comparing the strength of the lending and balance sheet channels of monetary transmission. To make this comparison, we use loan-level data to determine how borrower balance sheets and bank liquidity are related to bank lending decisions and how monetary policy can affect these relationships. The key innovation in this paper is the use of loan-level data. This enables us to measure the independent effects of the two channels and directly account for borrower balance sheets and lender liquidity instead of using proxies. Our results show that the balance sheet channel is the main mechanism through which monetary policy shocks are transmitted to the economy and that the lending channel does not play a significant role.
The consensus in the monetary economics literature is that monetary policy has a non-negligible short-run effect (for at least 2 years) on the real economy. A large number of empirical studies (Romer and Romer, 1989, Bernanke and Blinder, 1992 and Christiano et al., 1994) reach this conclusion by using various strategies to identify monetary policy shocks in Structural Vector Auto-regressive (SVAR) models. The standard cost of capital (or interest rate) channel falls short of explaining this short-run effectiveness of monetary policy. Research in the past 20 years has, hence, sought alternative explanations. The credit channel theory, or the effect of monetary policy on the level of financial frictions and thus the amount of bank lending, is currently the most prominent explanation and it is supported by empirical evidence (e.g. Bernanke and Gertler, 1995). Studies that investigate how this channel operates by analyzing bank behavior, however, overwhelmingly produce the conflicting result that the response of total loan supply (an indicator of economic activity) to monetary policy shocks is small and has gradually declined over the past 30 years. The decline in this component of the credit channel (more commonly known as the lending channel of monetary transmission) is mostly explained by the easier access to liquidity that banks – especially larger banks – have gained together with deeper and more developed financial markets (e.g. Cetorelli and Goldberg, 2011; Kashyap and Stein, 2000).2,3 So how exactly does monetary policy affect the economy? Credit channel theory offers a well-documented, alternative explanation for the non-negligible effect of monetary policy. According to this theory, monetary policy can also affect the amount of lending by having an impact on borrower balance sheets (the value of collateral) and on lender sensitivity to these balance sheets. This channel of monetary transmission (more commonly known as the balance sheet channel), therefore, explains the effectiveness of monetary policy by focusing on the borrower side of financial contracts. Most economists believe that the balance sheet channel is operative, and the mechanism has thus become a common component in macroeconomic models.4 Empirical research that measures the strength of this channel, however, is very scarce and to the best of our knowledge, the strength of the lending and the balance sheet channels have not been compared before. This is not surprising. To fully measure the balance sheet channel, one needs loan-level data to identify borrowers and lenders, and link borrower balance sheets and lender liquidity with the terms of the loan contract; these data are not as readily available as bank-level data that are used to measure the lending channel of monetary transmission. Bernanke and Gertler (1995) summarize the difficulty in identifying the independent effects of these channels as follows: It is extremely difficult to carry out an empirical test that would conclusively separate the bank lending channel from the balance sheet channel. For this reason, we are more confident in the existence of a credit channel in general than we are in our ability to distinguish sharply between the two mechanisms of the credit channel. In this paper, we use a unique loan-level (commercial and industrial loans) dataset and identify the independent effects of borrower balance sheets and bank liquidity on the lending decisions of banks and investigate how monetary policy affects these decisions. To construct our dataset we first obtain loan-level data such as the spread between the interest rate on the loan and a risk free rate (hereafter, the lending spread), the amount and the maturity of the loan from the Thomson Reuters DealScan database. We then combine these data with borrower and lender specific variables that we obtain from the Capital IQ Compustat database. Our dataset covers 15,794 loan deals from 1995 to 2009. By using this large, loan-level dataset, our paper makes a first attempt at comparing the strength of the lending and balance sheet channels of monetary transmission. To clarify our contribution, we should point out that this is certainly not the first empirical investigation of the balance sheet channel. For example, studies such as de Bondt (2004) and Ashcraft and Campello (2007) find that this channel is operative by using proxies for the strength of balance sheets such as average corporate bonds spreads or state GDP gaps. Although these studies usually find that central banks exert a significant effect on the economy through borrower balance sheets, using aggregate variables as proxies for balance sheets does not allow them to fully assess the strength of this channel and to make a comparison with the lending channel. We are also not the first to use loan level data.5 To the best of our knowledge, this paper is the first attempt, however, at measuring the strength of the different channels of monetary transmission using loan-level data. Our results show that balance sheet strength, captured by the borrower leverage ratio, is a significant and important determinant of lending spreads. We find, for example, that lending spreads increase by 0.85 basis points on average when the leverage ratio increases by 1% point. In contrast, we find that lender liquidity is not significantly related to lending spreads. Our more central result is that the balance sheet channel of monetary transmission is significant and large in magnitude. A comparison with the lending channel reveals that the balance sheet channel is also the main way in which monetary policy affects bank lending. For example, we find that a 1% orthogonal shock to non-borrowed reserves growth (captured by the Strongin (1995) index) increases the sensitivity to balance sheets by approximately 0.34%. The corresponding effect of monetary policy on the sensitivity to lender liquidity is only 0.007% and is insignificant. These results confirm the usual finding in the literature that the lending channel of monetary transmission is insignificant; but, more importantly, they show that monetary policy changes the course of the economy mainly through its effect on borrower balance sheets. To obtain these results we follow a two-step approach. First, we use quarterly cross section data to measure the sensitivity of lending spreads to the indicators of borrower leverage and lender liquidity. We measure these sensitivities for each quarter in our dataset. Second, we stack the sensitivity coefficients obtained in the first stage to form a time series. We do this separately for the lender liquidity and borrower leverage coefficients. We then measure the effect of monetary policy on these sensitivities in a second stage regression. We also follow this two-step estimation strategy by including different indicators of monetary policy stance and by accounting for first stage measurement errors. Finally, we check the robustness of our benchmark results to using a single stage estimation methodology. The results are similar. The main advantage of using a loan-level dataset is that it allows us to control for lender and borrower specific variables when identifying the independent effects of the balance sheet channel and the lending channel, respectively. As explained later in the paper, however, our baseline methodology restricts our dataset by excluding banks for which financial information could not be obtained from the Capital IQ Compustat database (e.g. foreign banks). To overcome this shortcoming, we follow an alternative methodology to measure the balance sheet channel. Specifically, we use data from banks that lend more than once in each quarter and compare the lending spreads and borrower leverage on these loans. This is similar, in spirit, to the methodology in Ashcraft and Campello (2007). This identification strategy allows us to fully control for lender liquidity, obtain a more pure measure of the balance sheet channel and use more observations, but it does not allow us to make a comparison with the lending channel. We do, however, find that the balance sheet channel is similarly significant and large in magnitude. The rest of the paper is organized as follows: Section 2 presents the theoretical framework that illustrates the balance sheet and lending channels of monetary transmission. Section 3 discusses our two-step empirical methodology. Section 4 describes the dataset and Section 5 presents our baseline results. Section 6 checks the robustness of our results and compares the strength of the two channels. Section 7 concludes.
نتیجه گیری انگلیسی
In this paper, we use a large number of loan-level observations to estimate the importance of the balance sheet and lending channels in monetary policy transmission. The unique feature of our analysis is that by using a loan-level dataset, we are able to identify the independent effects of borrower balance sheets and lender liquidity on monetary transmission. Consistent with theory, we find that monetary policy affects the sensitivity of bank lending to borrower balance sheets such that a weak balance sheet position of a borrower increases the lending spread on a bank loan more under tight than under loose monetary policy. In other words, banks’ lending decisions are amplified by monetary policy as suggested in the financial accelerator literature (e.g. Bernanke et al., 1999). Furthermore, our finding of a diminished role for the lending channel confirms the results of earlier studies that suggest that financial innovation makes banks less dependent on deposits as a means of raising loanable funds. These results suggest that if monetary policy effectiveness is measured more broadly by using the balance sheet channel in addition to the more common lending channel measures, the usual conclusions in the monetary literature can be reversed: Monetary policy that operates through the broad credit channel is still effective. The behavior of the economy before, after and during the 2008– 2009 credit market crisis reinforces this prediction and supports our main finding – the balance sheet channel is the more powerful of the two credit channel mechanisms of monetary transmission. Before the crisis, both borrower balance sheets and bank loans were very sensitive to monetary policy and a relatively small degree of monetary contraction triggered a severe financial crisis mainly through its effects on borrower balance sheets. The lack of liquidity on the supply side of financial markets has also not been the main culprit for the slow economic recovery and for the low amount of lending following the crisis since banks have been holding substantial amounts of liquid reserves. The slow pace of recovery at the early stages was, conversely, blamed on economic uncertainty and weak borrower balance sheets (for example caused by a depressed real estate market). Finally there is evidence showing that during the crisis, monetary policy played a stabilizing role by preventing adverse feedback loops between borrower balance sheets and the cost of external finance (see, Mishkin, 2009, 2011). Cecchetti et al. (2011) find evidence showing a similar stabilizing role of monetary policy in Korea and emerging market economies. The main crisis management and general policy implication of our results is that central banks should be more cognizant about the effects that they have on economic activity through borrower balance sheets. They should develop methods, similar to stress-testing for banks that quantify these effects and help prevent potential crisis episodes. Despite the advantage of using a large number of loan-level data, our analysis is not without drawbacks. For example, the reduced form equations that we estimate to draw conclusions for monetary transmission do not allow us to account for the direct effects of monetary policy on borrower balance sheets and lender liquidity. It would be interesting for future research to embed a partial equilibrium framework where it is costly to raise loanable funds into a model that includes borrower balance sheets and effective monetary policy to compare the strength of the different channels. This comparison can be made more meaningful by calibrating the non-standard parameters in the model such as sensitivity to balance sheets and lender liquidity by using the results from a loan-level data analysis such as ours.