! Do Credit Market Shocks affect the Real Economy ? Quasi - Experimental

This paper uses comprehensive data on bank lending and establishment-level outcomes from 1997-2011 to test whether changes in small business bank lending affect the real economy. The shift-share style research design predicts county-level lending shocks using variation in pre-existing bank market shares and estimated bank supply-shifts. Counties with negative predicted supply shocks experienced declines in small business loan originations throughout the entire period, indicating that it is costly for these businesses to find new lenders. Using confidential microdata from the Longitudinal Business Database, we find the predicted lending shocks led to statistically significant, but economically small, declines in both small firm and overall employment during the Great Recession, but did not affect employment during the 1997-2007 period. Overall, this paper’s evidence fails to support the hypothesis that the small business lending channel is an important determinant of economic activity. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!! * Michael Greenstone, University of Chicago, [email protected]; Alexandre Mas, Princeton, [email protected]; Hoai-Luu Nguyen, Haas School of Business, University of California at Berkeley, [email protected] We are grateful to Laurien Gilbert, Felipe Goncalves, Ernest Liu, and Steven Mello for excellent research assistance. We thank Daron Acemoglu, Pat Kline, Lawrence Summers, Adi Sunderam, Ivan Werning, and seminar participants at the NBER Summer Institute, Columbia, Brookings, Boston Federal Reserve, Bank of Mexico, and the AEA Meeting in Boston for helpful comments. We also thank Abigail Cooke, Javier Miranda, and Lars Vilhuber for enabling our use of the Census LBD microdata. The first version of this paper was released as a working paper in November 2012. ! ! 1 It is nearly conventional wisdom that banks play a special role in the economy. Specifically, it is widely believed that small and medium-sized businesses do not have ready substitutes for bank credit so that their influence on the economy is determined by bank finance (e.g., Brunner and Meltzer 1963; Bernanke 1983). Further, it is thought that their health can be an important determinant of macroeconomic fluctuations (Bernanke and Gertler 1995; Peek and Rosengren 2000; Ashcraft 2005). This paper gauges the credit channel’s empirical importance by separately measuring the economic consequences of shocks to small business credit during the 2007-2009 recession and during ‘normal’ economic times (i.e., 1997-2007). It is ex ante unclear whether credit channel effects should be larger in normal economic times when alternative sources of financing are likely to be more plentiful or during the Great Recession when the United States government and the Federal Reserve Board were aggressively trying to inject liquidity into financial markets. Our identification strategy leverages the substantial heterogeneity across banks in their year-toyear variation in small business lending along with geographic variation in bank market shares. Further, we isolate the portion of changes in bank lending that can be attributed to supply factors by purging each bank’s national change in lending of its exposure to local markets. In the case of the Great Recession years, we predict the change in county-level small business lending over the 2007-2009 period using interactions of banks' pre-crisis county market shares and their national change in lending. Between 2007 and 2009, for example, Citigroup reduced small business lending by 84%, while U.S. Bancorp's small business lending declined by just 3%. The essence of our approach is to ask whether counties with more Citigroup branches than U.S. Bancorp branches before the crisis experienced sharper declines in their economies over the 2007-2010 period. There is sufficient variation in banks' market shares across counties in the same state that our results are based on within-state comparisons. Using comprehensive data on both bank lending and real outcomes, the paper has three primary findings. First, counties with banks that cut lending had declines in small business loan !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!! 1 This paper is an updated version of Greenstone and Mas (2012). The substantive difference between versions is that we primarily use restricted-use LBD data, whereas Greenstone and Mas (2012) primarily relied on NETS data, which is compiled by Walls and Associates using Dun and Bradstreet’s proprietary Market Identifier files. Smaller changes include incorporating a symmetric growth measure, weighting the sample by each county’s employment count in 2006 instead of the number of establishments in 2006, and extending the sample back to 1997. ! ! 2 originations over the 2008-2009 period. For example, a one standard deviation reduction in predicted lending in 2009 is associated with a 17% reduction in total county-level small business loan originations from 2009 through 2010. In the short term, at least, it appears that the costs to switching lenders are meaningful for small firms. Second, this same predicted negative shock in lending depresses 2009-2010 employment growth rates for small standalone firms (single-unit establishments with fewer than 20 employees) by 0.4 percentage points. Under a series of polar assumptions described below, our estimates suggest these lending shocks can account for just 3 percent of the overall decline in small business employment in this period and 5 percent of the total decline in employment. Of course, we cannot reject that the employment effects would have been larger in the absence of the extraordinary interventions undertaken by the Federal Reserve and the U.S. Government to aid banks. Third, there is a significant relationship between predicted lending shocks and bank loans to small businesses during the 1997-2007 period, but these predicted shocks are not associated with changes in economic activity. This finding runs counter to much of the previous literature, because it suggests that, at least with this identification strategy, the credit channel is not empirically important in normal times. While small businesses appear to have access to alternatives sources of credit in these periods, we cannot directly observe whether these alternatives are more costly. This paper adds to the literature in several ways. First, our study is nationally representative, which allows us to consider the aggregate implications of our estimates without external validity concerns. Previous papers studying the U.S. have focused on subsamples defined by particular sets of firms, episodes, or regions (e.g., Chodorow-Reich 2014; Peek and Rosengren 2000; Ashcraft 2005). Second, while other papers have also focused on small firms, which are more likely to be affected by bank supply decisions (e.g., Duygan-Bump et al. 2014), our paper !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!! !Chodorow-Reich examines disruptions in the syndicated loan market following the collapse of Lehman Brothers in 2008, and finds that borrowers of weaker banks faced restrictions in credit supply, which translated into greater cuts in employment at these firms. Its sample is comprised of 2,040 firms that have an average employment size in 2008 of 2985 employees (median 620). Ashcraft examines the closing of healthy banks by the FDIC in some Texas counties in the late 1980s and early 1990s. Peek and Rosengren find that shocks to the balance sheets of Japanese banks operating in the U.S. from economic conditions in Japan affected construction activity in U.S. markets where the Japanese banks operated. !! 3 Outside of the U.S., Bentolila et al. (2015) and Hochfellner et al. (2015) use administrative data to estimate the impact of credit supply shocks on employment in Spain and Germany, respectively ! ! 3 additionally measures the impacts on overall county-level employment. Consequently, our estimates incorporate establishment entry, exit, and expansion/shrinkage, as well as any multiplier-style effects or indirect effects via competitor responses without relying on assumption-dependent theoretical models. We believe this is unique for a national study of the U.S. Third, we utilize a research design that allows us to control for confounding demand factors that may have affected employment growth. Our paper also contributes to the literature on the causes of the Great Recession and the subsequent slow recovery. The range of explanations for this deep decline and slow pace of recovery include reduced aggregate demand (Mian and Sufi 2014), uncertainty (Baker, Bloom, and Davis 2015; Bloom et al. 2014), and structural factors (Charles, Hurst, and Notowidigdo 2012). The list of explanations certainly also includes the tightening of bank lending standards and, at a high level, this theory is supported by the disproportionate employment losses incurred by small firms that are more reliant on bank lending than other firms (Charnes and Krueger 2011; CBO 2012; Fort et al. 2013). Based on this observation, some policymakers (e.g., Bernanke 2010; Krueger 2010) suggested that fractured credit markets played a major role in overall employment declines. Indeed, restoring access to credit was a key feature of the policy response following the financial crisis. The remainder of the paper is organized as follows. Section II provides some brief background on the financial crisis and the role of small businesses in the U.S. economy. Section III describes the data sources. Section IV explains the research design and how it is implemented. Section V outlines the econometric models and presents the results. Section VI interprets the findings, and Section VII concludes. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!! 4 Speaking in July 2010 at the Federal Reserve Meeting Series, “Addressing the Financing Needs of Small Businesses,” Chairman Ben Bernanke stated that "making credit accessible to sound small businesses is crucial to our economic recovery and so should be front and center among our current policy challenges,” and that "the formation and growth of small businesses depends critically on access to credit, unfortunately, those businesses report that credit conditions remain very difficult.” ! ! 4 II. Background The heart of the theory that banks are critical suppliers of credit for small businesses is the idea that it is costly for lenders to obtain information about these firms. Direct measurement of these costs cannot be measured with available data sets, but the existing evidence is often supportive of this possibility. For example using data from the Survey of Small Business Finances through 2003, Brevoort, Holmes, and Wolken (2010) estimate the median distance between firms and their suppliers of credit to be just 3 miles. Further, they find that only 14.5 percent of small firms borrowed from an institution that was more than 30 miles from their headquarters. A number of empirical studies have investigated the benefits of long-term lending relationships as a way to overcome information asymmetries in the lending market (e.g., Cole 1998, Berger and Udell 1995, Hoshi et al. 1990, Petersen and Rajan 1994). Berger et al. (2002) argue that firms that borrow from large banks tend to be more credit rationed, suggesting that firms that are cut off from credit from larger banks (as in our study) may not be able to obtain credit elsewhere. Nguyen (2015) provides direct evidence of this by showing that branch closings that follow mergers between large banks lead to prolonged declines in local small business lending, indicating that borrowers who lose access to credit have difficulty obtaining credit from other (bank) lenders. In the macroeconomics literature, credit market frictions have been suggested as a channel for the transmission of monetary policy, specifically through the effect of interest rates on the external finance premium, which arises through imperfections in credit markets (Bernanke and Gertler 1995). There is evidence that the lending channel for small businesses may have been compromised during the Great Recession. Most relevant to our context is that the liquidity crisis translated into less available credit across the economy, and the decline in commercial bank lending was especially severe for small business loans. According to data from the Federal Reserve Survey of Senior Loan Officers, the net percentage of loan officers reporting tightening standards for !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!!!!! 5 Amel and Brevoort (2005) use survey data from the National Federation of Independent Business Research Foundation to show the median distance over which small firms search for credit is only 4.3 miles. Using data from a single large commercial bank, Agarwal and Hauswald (2010) find a similar median distance between the lending branch and the firm (2.6 miles), and argue this is because geographic proximity facilitates the acquisition of “soft” information. Using data from the Community Reinvestment Act, Laderman (2008) finds that only about 10 percent of small business lending is from banks with no branch in the local market.! 6 The mechanisms behind the unraveling of the financial system in 2008 are complex, and have been analyzed in depth by Brunnermeier (2009) and Shleifer and Vishny (2011) among others.! ! ! 5 medium and large firms was 64 percent in the first quarter of 2009 as compared to zero percent in the first quarter of 2007.!!Data from banks reporting under the Community Reinvestment Act show loan originations to small businesses fell by 52% between 2007 and 2010. A similar pattern is seen in the survey of members of the National Federation of Independent Business: loan availability began to decline in the beginning of 2007, did not reach its nadir until 2009, and has been on a slow recovery since then (Dunkelberg and Wade 2012). A number of papers explore the underlying mechanisms for this decline in lending and conclude that it was, in large part, “supply-driven.” Ivashina and Scharfstein (2010), for example, document that new loans to large borrowers fell by 79% between the second quarter of 2007 and the fourth quarter of 2008. They argue that an important mechanism behind this decline was banks’ reduced access to short-term debt following the failure of Lehman, coupled with a drawdown of credit lines by their borrowers. This severe contraction in small business lending may have led to significant real economic effects given both the importance of small businesses in the U.S. economy (in 2007, firms with less than 100 employees represented approximately 36% of employment and 20% of net job creation in the U.S.) and their dependence on local bank credit.