Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Aairs Federal Reserve Board,
Washington, D.C.
Bank Lending Channels during the Great Recession
Samuel Haltenhof, Seung Jung Lee, and Viktors Stebunovs

2014-06
NOTE: Sta working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research sta or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.
Bank Lending Channels during the Great Recession

Samuel Haltenhof University of Michigan Seung Jung Lee Viktors Stebunovs Board of Governors of the Federal Reserve System October 29, 2013
Abstract We study the existence and economic signicance of bank lending channels that aect employment in U.S. manufacturing industries. In particular, we address the question of how a dramatic worsening of rm and consumer access to bank credit, such as the one observed over the Great Recession, translates into job losses in these industries. To identify these channels, we rely on dierences in the degree of external nance dependence and of asset tangibility across manufacturing industries and in the sensitivity of these industries output to changes in the supply of consumer credit. We show that household access to bank loans matters more for employment than rm access to local bank loans. Our results suggest that, over the recent nancial crisis, tightening access to commercial and industrial loans and consumer installment loans explains jointly about a quarter of the drop in employment in the manufacturing sector. In addition, a decrease in the availability of home equity loans explains an extra one-tenth of the drop. JEL Classification: G21, G28, G30, J20, L25 Keywords: bank credit channels, bank lending standards, home equity extraction, credit crunch, employment, job losses, Great Recession.
The views expressed in this paper are solely the responsibility of the authors and should not be interpreted as reecting the views of the Board of Governors of the Federal Reserve System or the Federal Reserve System. All errors and omissions are our own responsibility alone. University of Michigan, Department of Economics; emal: shalten@umich.edu Board of Governors of the Federal Reserve System, 20th Street and Constitution Avenue, NW, Washington, DC 20551; email: seung.j.lee@frb.gov and viktors.stebunovs@frb.gov, respectively.
Figure 1: Four credit supply channels

Firm depends on external sources of funds
C&I loans Bank Tangible coll. Firm Funds
Small business owner
Home equity Bank HELOCs
Firm produces durable goods Cons. loans Money Firm Durable goods HELOCs Household Home equity Bank
Introduction
We study the existence and economic signicance of several bank lending channels.
In particular, we address the question of how a dramatic worsening of rm and consumer access to bank credit, such as the one observed during the Great Recession, translates into job losses in U.S. manufacturing industries. The focus of the paper is on the economic signicance of bank lending channels, rather than credit channels more broadly, and the key for identifying the economic eects of bank lending channels is that bank loans are not perfectly substitutable with other types of external nance. To x the ideas further, Figure 1 shows four separate channels through which bank credit might aect employment and industry dynamics in the manufacturing sector: (1) through the supply of commercial and industrial (C&I) loans directly to rms, (2) through the accessibility of home equity lines of credit (HELOCs) to small business owners to prop up their businesses, (3) through the supply of consumer installment loans to households, and (4) through the accessibility of HELOCs to households for consumption purposes. In this paper, we examine these four channels using data for U.S. manufacturing industries and large commercial banks over the 1991-2011 period. There are three main reasons behind our choice of studying specically the linkages between access to bank credit and U.S. manufacturing employment. First, by studying the real eects of changes in the supply of bank credit, we account for the possible substitution of funding sources at the rm and household levels. One might imagine that rms and households will substitute away from more limited bank credit to more easily available alternatives, perhaps mitigating the eect of a decline in supply of bank credit on manu-
facturing employment. Second, over the past few decades, manufacturing industries have had relatively stable establishment structures and banks seem to have continued to supply a signicant share of their C&I loans to the manufacturing sector. In contrast, other industries, such as retail trade, have experienced a shift over time toward multi-unit rms, often with access to national capital markets. Hence, this shift likely has weakened these non-manufacturing industries reliance on local bank credit. Third, manufacturing industries outputin particular, durable goods outputis sensitive to changes in the supply of consumer credit. Indeed, most purchases of household durable goods, such as cars or large appliances, tend to be nanced. This economic feature allows us to judge the importance of consumer access to bank credit on employment in manufacturing. Our explained variablestotal employment, number of establishments, and average establishment sizeare from the Quarterly Census of Employment and Wages (QCEW). To our knowledge, this is a novel application of the QCEW data. One of the advantages of this data set is that it does not contain a structural break in the classication of industries due to the transition from the Standard Industrial Classication (SIC) to the North American Industry Classication System (NAICS) in the late 1990s. Hence, the data set covers a few recessions, including the Great Recession, on a consistent basis. We focus on employment in manufacturing rather than output because employment at the industry-state level can be more precisely measured. While the U.S. Census makes industry-state level output data available, the data are noisy by the Census own admission, and the data compilation approaches alternate between census and non-census years. Our explanatory variables come from a few sources. We associate changes in rm and household access to bank credit with changes in commercial banks C&I lending standards and in their willingness to originate consumer installment loans based on bank-specic responses to questions in the Senior Loan Ocer Opinion Survey on Bank Lending Practices (SLOOS). Indeed, the data on the net fraction of banks indicating a tightening of C&I lending standards is highly negatively correlated with the growth in employment in the manufacturing sector, as shown in Figure 2. We also associate changes in household access to home equity loans with growth in home equity. We use the state- and national-level house price indices compiled by CoreLogic, state- and national-level mortgage debt per borrower taken from TransUnions Trend Data, and the national level household balance sheet data from the Federal Reserves statistical release Z.1 to construct proxies for growth in home equity. To better isolate the dierent bank lending channels, we minimize the risk that our results will be driven by either reverse causality or by omitted variables. Our identication assumption takes advantage of the dierences in national bank presence across U.S. states and is somewhat similar to that is regularly used in the literature.
Figure 2: Changes in C&I lending standards and growth in employment in the U.S. manufacturing sector
Percent 80 Tightening 60 40 20 0 -20 Easing -40 -60 -80 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 -10 -15 5 0 -5 Percent 15
Changes in standards (left axis) Growth in employment (right axis)
10
*Net percentage of respondent banks reporting tightening of standards on C&I loans, weighted by total loans. Source: Federal Reserve Board, Senior Loan Officer Opinions Survey on Bank Lending Practices.
Note: Shaded areas are NBER-dened recessions.
We posit that changes in major banks C&I lending standards, apportioned to a particular state, and households growth in home equity are exogenous to developments in a given manufacturing industry in a given state and at a given point in time. That is, in accordance with the questions in the SLOOS, we postulate that national banks tighten (or ease) C&I lending standards for small rms broadly across the country rather than targeting a particular state or a particular industry in a given state. Variation in geographical presence of banks and in their timing of tightening (or easing) generates variation of our C&I lending standards measure across states. Similarly, we postulate that changes in commercial banks willingness to originate consumer installment loans is exogenous to developments in a given manufacturing industry in a given state and at a given point in time. We control for omitted variable bias (1) by comparing manufacturing industries that do or do not depend on external sources of nance and do or do not have a high degree of pledgeable tangible assets and (2) by comparing industries that produce durable vs. non-durable goods. The rst comparison is an improvement on the setup that has been widely used in the literature to tease out a dierential impact of credit supply changes on industries dependent on external nance, for example, as in Cetorelli and Strahan (2006). Its novelty is that we take into account not only the need to borrow to nance physical capital investment (captured by the Rajan and Zingales (1998) measure of external nance dependence) but also the ability to access C&I loans by manufacturing rms (captured by the tangible asset measure as in Braun (2002) and Claessens and Laeven (2003)). We interact these
industry-specic indicators with measures of changes in state-level commercial banks C&I loan lending standards to isolate the local bank lending channel to rms. The second comparison is quite intuitive, too.1 Consumption of durable goods is more likely to be nanced rather than paid for outright, and hence changes in consumer access to credit, though aecting both consumption of durable and non-durable goods, are more likely to aect the consumption of the former. We recognize that the location of production and the location of consumption of durable goods are usually not the same. After all, if households in one state have diculties obtaining consumer installment or home equity loans, then the production of durable goods in other states should be aected. Hence, we interact national-level measures of household access to bank loans to identify the bank lending channel to households. Our results show that measures of changes in lending standards on C&I loans and changes in willingness to originate consumer installment loans by major commercial banks, and the availability of home equity lines of credit aect notably the manufacturing employment over the sample period.2 We show that household access to loans matters more for employment than rm access to local loans and that access to bank credit aects employment mostly through changes in the size of rms rather than through changes in their numbers. The latter nding is consistent with a couple of literature strands. First, for larger rms, consistent with the exporter hysteresis international trade literature, it may be that, following a tightening of access to credit, the sunk cost aspect of the rm entry decision in the presence of xed per-period costs results in these rms continuing to serve the market despite unfavorable economic or nancial conditions, but perhaps at a smaller scale requiring fewer employees. Second, at the other extreme, for smaller rms, consistent with the literature on lending relationships, as in Berger and Udell (1994) and Petersen and Rajan (1994), smaller (nascent) rms may depend less on bank loans than larger (older) ones, and, hence, their entry decisions might depend less on bank credit availability. Indeed, in our data, while the smallest establishments employ relatively a small fraction of employees in aggregate, they are numerous and introduce some noise into the aggregate series for the number of establishments. Our results highlight the adverse eects that tightening access to credit, especially for households, over the Great Recession and the subsequent slow recovery had on employment in manufacturing industries. Structural break tests support the idea that a signicant porThis breakdown of industries into industries producing durable vs. non-durable goods has been used in the growth and nancial development literature. 2 Although we do not consider a short-term credit supply channel, which includes trade credit or shortterm bank loans, we believe that disruptions in the supply of loans of these types over the Great Recession also possibly had a signicant impact on employment in manufacturing industries.
1
tion of such employment losses were the manifestation of an unusually large tightening of credit availability to the economy rather than a structural change in the bank lending channels. Indeed, our model-based back-of-the-envelope calculations suggest that, between 2007 and 2010, the tightening of lending standards and a decrease in willingness to orginate loans caused a 4.4 percent decline in employment in the manufacturing sector. The explanatory power of the eects of these bank lending channel based on our approach is notable, as the actual drop in employment was 17 percent. Our exercise suggests that, over the recent nancial crisis, tightening access to C&I loans and and a decrease in willingness to originate consumer installment loans explain jointly about a quarter of the drop in employment in the manufacturing sector. Our further back-of-the envelope calculations suggest that, over the same period, the decline in home equity loan accessibility explains an extra tenth of the manufacturing employment decline. Our estimate of the impact of changes in C&I lending standards, changes in willingness to originate consumer loans, and the growth in the availability of home equity on the manufacturing industries have implications for the recovery in the labor market going forward. To some extent, the tightening of lending standards and the decrease in willingness to originate loans reect commercial banks eorts to deleverage. The greater anticipated regulatory burden faced by commercial banks may have temporarily held back employment growth in manufacturing, thus contributing to weak overall labor market conditions. Moreover, the sluggish housing market improvement may have been a further drag on manufacturing employment, for example, because of limited access to HELOCs. Although in the longer term, the displaced manufacturing workers might be absorbed by other sectors in the economy. When dealing with the unusually low levels of employment in manufacturing in an environment like the Great Recession, the policy prescription that follows from our back-ofthe-envelope exercise is that policymakers should pay attention to restoring the functioning of the bank credit supply channel to households relatively more than that to rms. Our results should not be taken though as advocating an implementation of Keynesian policies in a recession. Note the focus of the paper is on employment losses in the manufacturing industries attributable to cutbacks in credit supply. As long as the supply of credit is constrained by bank-related factors, ceteris paribus, policymakers might be urged to step in to restore credit ow, which implies a dierent policy response than simply trying to boost aggregate demand. The outline of the paper is as follows. After a short literature review, the following section provides a description of our data sources and the ways we transformed the raw data. The fourth section goes over our empirical strategy and econometric specication. The fth section presents the estimation results. We then detail the economic signicance
of the bank lending channels by estimating employment losses in manufacturing industries attributable to tightening access to C&I loans or decreased willingness to originate consumer installment loans over the Great Recession. We end with some concluding remarks.
Literature
As mentioned earlier, our identication assumption takes advantage of the dierences
in bank presence across U.S. states and is somewhat similar to that in Peek and Rosengren (2000), Garmaise and Moskowitz (2006), and Lee and Stebunovs (2012). For example, Peek and Rosengren (2000) use the Japanese banking crisis to test whether a loan supply shock to branches and agencies of Japanese banks aected construction activity in the U.S. commercial real estate market. Similarly, Garmaise and Moskowitz (2006) study the eects of changes in large bank mergers on changes in crime at the MSA level, arguing that such merger activity instruments for changes in bank competition at the local level. Lee and Stebunovs (2012) use a similar setup to study the eects of bank balance sheet pressure manifested through state-level bank capital ratios on the employment and net rm dynamics in dierent manufacturing industries within a given state. Our paper also contributes to the nascent literature that tackle various bank lending channels and their real eects over the Great Recession. In this literature strand, using employee-specic data, Duygan-Bump, Levkov, and Montroiol-Garriga (2010) nd that workers in small rms are more likely to become unemployed during the 2007-2009 nancial crisis if they work in industries with high external nancing needs. From a more international perspective, Bijlsma, Dubovik, and Straathof (2010) nd evidence that the credit crunch in 2008 and 2009 resulted in lower industrial growth in industries that are more dependent on external nance in the OECD countries. Bentolila, Jansen, Jim enez, and Ruano (2013) provide evidence from Spain on how employment at rms funded by weak banks fell considerably more than employment at rms funded by healthier banks. Finally, Fort, Haltiwanger, Jarmin, and Miranda (2013) suggest that the collapse in house prices accounts for a signicant part of the large decline in young/small businesses during the Great Recession. However, because of their VAR approach, their nancing channel admittedly reects both the associated credit demand and supply factors, whereas we attempt to identify the economic impact of disruptions in bank credit supply only.
Description of the data

In this section, we justify our focus on manufacturing industries and review our data
sources and the ways we transformed the raw data. Our explained variables come from the
QCEW. We derive our explanatory variables from the SLOOS, TransUnions Trend Data, CoreLogics data, Federal Reserves statistical release Z.1, and other sources.
3.1
Focus on manufacturing industries at the state level
Our analysis focuses on manufacturing industries. These industries are often studied in the nance and banking literaturefor example, Cetorelli and Strahan (2006) and Kerr and Nanda (2009). In contrast to some other industries that have experienced a shift over time toward multi-unit rms, the manufacturing industries have had relatively stable structures and many smaller manufacturing rms continue to rely on local bank loans.3
3.2
The Quarterly Census of Employment and Wages

The QCEW program publishes quarterly employment and wages data reported by
employers accounting for 98 percent of U.S. jobs, available at the county, MSA, state, and national levels by industry.4 The programs primary outlet is the tabulation of the employment and wages of establishments that report to the Unemployment Insurance (UI) programs of the United States. Employment covered by these UI programs represents about 99.7 percent of all wage and salary civilian employment in the country. Ultimately, the QCEW data have broad economic signicance for the evaluation of labor market trends and major industry developments, for time-series analyses, and for interindustry comparisons. The QCEW data are collected on an establishment basis. An establishment is an economic unit, such as a farm, mine, factory, or store that produces goods or provides services. It is typically at a single physical location and engaged in one, or predominantly one, type of economic activity for which a single industrial classication may be applied. According the Bureau of Labor Statistics (BLS), most employers have only one establishment; thus, the establishment is the predominant reporting unit or statistical entity for reporting employment data. Admittedly, if someone is interested in the number of rms rather than the number of establishments in a given industry, then there might be some measurement error in our dependent variable induced by the fact that large rms often operate multiple establishments. Nevertheless, the number of establishments from the QCEW is highly correlated with the economic quantity, such as rms or net rm entry, that we are trying to observe, for at least two reasons. First, the analysis of the U.S. Census Bureaus Longitudinal Business
For example, the retail trade sector has undergone a pronounced shift away from single-unit rms to national chains with access to national capital markets. In fact, Jarmin, Klimek, and Miranda (2009) report that the share of U.S. retail activity accounted for by single-establishment rms fell from 60 percent in 1967 to just 39 percent in 1997. 4 We draw on the Bureau of Labor Statistics materials to write parts of this section.
3
Database suggests that most U.S. rms have only one establishment.5 For example, Davis, Haltiwanger, Jarmin, and Miranda (2006) report that on average, each publicly traded rm operates about 90 establishments, while each privately held rm only 1.16 establishments. Note that in 1990, there were over 4.2 million privately held rms and less than 6 thousand publicly traded rms. In 2000, there were about 4.7 milion privately hed rms and less than 7.4 thousand publicly traded rms.6 Second, earlier researchby, for example Black and Strahan (2002)has shown that the rate of the creation of new businesses is correlated with the share of new establishments in a local economy. In accordance with BLS policy, data reported under a promise of condentiality are not published in an identiable way and are used only for specied statistical purposes. The BLS withholds the publication of UI-covered employment and wage data for any industry level when necessary to protect the identity of cooperating employers. Totals at the industry level for the states and the nation include the undisclosed data suppressed within the detailed tables. Although the QCEW data provide a wealth of disaggregate information, we limit ourselves to studying total employment, the number of establishments, and the average establishment size measured in employees over the 1991-2011 period at the state level.7
3.3
External nance dependence for rms and households

We consider several channels through which bank credit aects manufacturing employ-
ment: the supply of C&I loans directly to rms, the supply of home equity loans to small business owners to prop up their businesses, the supply of consumer installment loans to households, and the supply of home equity loans to households for consumption purposes. To examine the bank lending channels to rms or small business owners, we follow the literature in constructing a measure of dependence on external sources of nance. One measure for external nance dependence based on Rajan and Zingales (1998) is calculated as the fraction of total capital expenditure not nanced by internal cash ows from operations and reects rms requirements for outside capital.8 This measure is viewed as a techno5 The Longitudinal Business Database (LBD) covers all business establishments in the U.S. private nonfarm economy that le payroll taxes with the IRS. As such, it covers all establishments in the U.S. nonfarm business sector with at least one paid employee. In some industries, the share of multi-establishment rms is higher than in others. For example, the retail trade sector has undergone a pronounced shift away from single-unit rms to national chains. 6 Similarly, Haltiwanger, Jarmin, and Miranda (2009) say that the LBD covers about 6 million rms and 7 million establishments in a typical year that have at least one paid employee, which implies that, on average, each rm operates 1.17 establishments. 7 More precisely, average establishment size calculated as the average number of employees per establishment. 8 We calculate the Rajan-Zingales measures for each manufacturing industry at the three digit NAICS level for manufacturing industries using COMPUSTAT. This measure is dened as capital expenditures (CAPX) minus cash ows from operations divided by capital expenditures. Cash ows are calculated by
logically determined characteristic of a sector that is innate to the manufacturing process and exogenous from the perspective of an individual rm. Following Cetorelli and Strahan (2006), we calculate a Rajan-Zingales measure for each NAICS three-digit manufacturing industry by summing both capital expenditures and cash ows over the sample period for each U.S.-based rm that has been in existence in the COMPUSTAT database for more than ten years and by taking the median value of rms in each manufacturing industry. We then categorize industries that lie above the median value of the Rajan-Zingales measures as industries dependent on external nance (EF = 1) and those that lie below as industries not dependent on external nance (EF = 0). Although rms dependence on bank loans may be somewhat limited in the U.S., this dependence is larger for some rms (such as relatively small privately held rms) than for others (such as large publicly traded rms). Moreover, as shown in Colla, Ippolito, and Li (2013), even larger rms may specialize in a certain debt type and may not be able to substitute away from scarce bank loans quickly and, hence, may be forced to downsize or shut down. In other words, if small and large rms alike were able to substitute away from scarce bank loans completely, we would not be able to identify the banking lending channels using our regression models. Moreover, using a more specic measure capturing dependence on bank loans subjects us to endogeneity concerns as low dependence may simply indicate nancing constraints. Likewise, Cetorelli and Strahan (2006) argue that the EF measures computed for only mature rms provide a powerful instrument for small rms demand for bank credit, but a direct measure of bank credit dependence, for example, based on bank-loans-to-assets ratios of small businesses from the 1998 Survey of Small Business Finance (SSBF) does not.9 To sharpen our identication approach, we consider rms ability to pledge collateral in securing bank loans. As suggested by the Survey of Terms of Bank Lending (FRB: E.2 Release), C&I loans tend to be secured by collateral, such as equipment and machinery. To reect this particular feature of C&I loans, we consider asset tangibility by industry. We reason that rms in manufacturing industries with a high share of tangible assets relative to total book-value assets should have easier access to external nance. Following the guidelines in Braun (2002) and Claessens and Laeven (2003), we compute the tangibility ratios at the three-digit NAICS level using the data for large U.S.-based rms over the sample period. Again, we are not interested in the exact value of the asset tangibility metric for each industry as such; we sort industries into the industries with a low tangible
summing up the following items in Compustat: IBC, DPC, TXDC, ESUBC, SPPIV, and FOPO. 9 Still, we re-estimate the benchmark model with control and treatment groups based on the 1998 SSBF data on bank loans, with the estimation results shown in the second column in Table 10 in the Appendix. Not surprisingly, the results are somewhat dierent and are explained in our robustness checks.
asset share (indicating a low ability to pledge collateral for C&I loans, T A = 0) and a high tangible asset share (indicating a high ability to pledge collateral for C&I loans, T A = 1) based on the whether an asset tangibility ratio for a given industry is below or above the median.10 Finally, we can dene one of our treatment groups: these are the industries that are dependent on external sources of funding in the Rajan and Zingales (1998) sense and have ahigh ability to pledge collateral to secure access to C&I loans in the sense of Braun (2002) and Claessens and Laeven (2003). The rst indicator tells us the need to borrow in a given industry, and the second indicator, the ability to do so. In addition, we also consider the EF = 1 and T A = 1 treatment groups separately to stay consistent with the literature. To examine the bank lending channel to households-consumers, we recognize that the degree of consumer reliance on bank credit for consumption of non-durable goods is dierent from that for consumption of durable goods. Consumption of durable goods is more likely to be nanced with consumer installment or home equity loans (rather than paid for outright or nanced directly in capital markets) than consumption of non-durable goods. Hence, to a large extent, the producers of durable goods are at the mercy of lenders to consumers. We follow the U.S. Census Bureaus breakdown of manufacturing industries into either durable goods (DG = 1) or non-durable goods producers (DG = 0). Table 1 shows the breakdown of three-digit NAICS manufacturing industries into industries dependent on external nance (EF = 1), industries with high tangible asset shares (T A = 1), and industries producing durable goods (DG = 1). Some industries do not have any of these characteristics, yet others have one, two, or all three of these characteristics, which helps with our identication. Having dened the control and treatment groups, we look into the growth in total employment, the number of establishments, and the average establishment size in each of the groups. Figures 3 to 5 plot these measures for the entire economy. The gures suggest that employment and average establishment size in the treatment group are more procyclical than that in the control group. However, for the number of establishments, the business cycle pattern for the treatment group relative to the control group is less clear.11
We calculate the asset tangibility measures based on all rms as used in the literature. Although the Rajan-Zingales measure is sensitive to whether we use only mature rms or all rms, the tangibility measure is not, implying that nancially constrained rms may be more constrained in the total size of the balance sheet, and not necessarily in the composition of their assets. 11 We discuss later in the paper why this might be the case.
10
10
Table 1: Characteristics of manufacturing industries

NAICS 311 312 313 314 315 316 321 322 323 324 325 326 327 331 332 333 334 335 336 337 339 31-33 Description Food Manufacturing Beverage and Tobacco Product Manufacturing Textile Mills Textile Product Mills Apparel Manufacturing Leather and Allied Product Manufacturing Wood Product Manufacturing Paper Manufacturing Printing and Related Support Activities Petroleum and Coal Products Manufacturing Chemical Manufacturing Plastics and Rubber Products Manufacturing Nonmetallic Mineral Product Manufacturing Primary Metal Manufacturing Fabricated Metal Product Manufacturing Machinery Manufacturing Computer and Electronic Product Manufacturing Electrical Equipment, Appliance, and Component Transportation Equipment Manufacturing Furniture and Related Product Manufacturing Miscellaneous Manufacturing Total Manufacturing EF TA DG Empl. Share (Percent) 1.2 0.0 0.2 0.0 0.2 0.0 0.4 0.3 0.4 0.1 0.6 0.5 0.4 0.3 1.1 0.8 0.9 0.3 1.2 0.4 0.5 9.7 Output Share (Percent) 2.8 0.0 0.2 0.0 0.1 0.0 0.4 0.7 0.4 2.3 2.6 0.8 0.5 1.0 1.3 1.3 1.6 0.5 2.9 0.3 0.6 20.3
Note: Employment and output shares are relative to the entire economy as of 2007.
3.4
3.4.1
Denitions of loan types and the Senior Loan Ocer Opinion Survey
Denitions of loan types
To identify dierent bank lending channels, we focus on three types of loans: C&I loans, consumer installment loans, and HELOCs. C&I loans include loans for commercial and industrial purposes to sole proprietorships, partnerships, corporations, and other business enterprises, whether secured or unsecured, single payment, or installment. C&I loans exclude the following: loans secured by real estate; loans to nancial institutions; loans to nance agricultural production and other loans to farmers; loans to individuals for household, family, and other personal expenditures; and other miscellaneous loan categories. Typically, the interest rate for C&I loans is set as a spread over the prime rate or Libor and adjusts with movement in the benchmark rate over the loan term. Consumer installment loans are loans to individualsfor household, family, and other personal expendituresthat are not secured by real estate, such as auto loans. Typically, the interest rate for new consumer installment loans is set as a spread over the prime rate or Libor and remains xed over the full loan term. In recent years, the popularity of HELOCsrevolving, open-ended lines of credit secured by residential propertieshas overshadowed the use of non-collateralized consumer 11
Figure 3: Growth in employment in manufacturing industries in the U.S.

Percent 10
EFxTA=0, DG=0 EFxTA=1 DG=1
-5
-10
-15
-20 1992
Source: QCEW
1994
1996
1998
2000
2002
2004
2006
2008
2010
Note: EF T A = 0 are manufacturing industries that either do not dependent on external sources of funding or do not have the ability to pledge collateral to secure access to C&I loans. DG = 0 are manufacturing industries that produce nondurable goods. EF T A10 are manufacturing industries that depend on external sources of funding and have the ability to pledge collateral to secure access to C&I loans. DG = 1 are manufacturing industries that produce durable goods. Shaded areas are NBER-dened recessions.
installment loans. In part due to the popularity of home equity loans, the growth rate of consumer installment loans has been one-half of the growth rates of other types of core loans prior to the Great Recession. HELOCs are typically secured by junior liens and are usually accessible by check or credit card. The rate on new home equity loans is often set as a spread to the prime rate or Libor. Lenders typically oer home equity loans only up to 100 percent of the appraised property value, less the amount of any rst mortgage lien. 3.4.2 The Senior Loan Ocer Opinion Survey
Changes in C&I lending standards and willingess to originate consumer installment loans are based on bank-specic responses to questions about changes in lending standards and terms from the Federal Reserves SLOOS.12 The survey is usually conducted four times per year by the Federal Reserve Board, and up to 60 banks participate in each survey. The survey is voluntary and typically includes
Individual bank survey responses are condential. For more details, see Bassett, Chosak, Driscoll, and Zakraj sek (forthcoming).
12
12
Figure 4: Growth in the number of establishments in manufacturing industries in the U.S.

Percent 10
8 6 4 2 0 -2 -4 -6 -8 -10
1992
Source: QCEW
1994
1996
1998
2000
2002
2004
2006
2008
2010
the largest banks in each Federal Reserve district and is roughly nationally representative. The surveyed banks are all considered large in the sense that no bank has assets of less than $3 billion in the survey. The top 5 banks (by total assets) had deposit-taking branches in more than 20 states in the beginning of our sample period and have had branches in all 29 states in our sample since 2004. On average, only 7 banks have bank branches in only one state in our survey per year. Banks are asked to report whether they have changed their credit standards over the past three months on six categories of core loans including C&I loans. Data measuring changes in credit standards on C&I loans are available beginning with the May 1990 survey. Questions regarding changes in standards on credit card loans and other consumer loans were added to the survey in February 1996 and May 1996, respectively. However, a series indicating changes in banks willingness to originate consumer loans is available over the entire sample period. The SLOOS surveys follow a somewhat irregular schedule. The SLOOS asks banks to report changes in their lending practices over the previous three months, and the survey is
13
Figure 5: Growth in the average size of establishments in manufacturing industries in the U.S.
Percent 15
10
-5
-10
-15 1992
Source: QCEW
1994
1996
1998
2000
2002
2004
2006
2008
2010
conducted so that it coincides with regular meetings of the Federal Open Market Committee. Hence, the January SLOOS refers to the period from October to December of the prior year. To make individual bank responses operational, we transform them into an aggregate measure in two steps. First, we map individual bank responses into indicator variables. The question about changes in C&I lending standards reads, Over the past three months, how have your banks credit standards for approving applications for C&I loans or credit linesother than those to be used to nance mergers and acquisitionsto large and middle-market rms and to small rms changed?Banks respond to that question using a categorical scale from 1 to 5: 1 = eased considerably, 2 = eased somewhat, 3 = remained about unchanged, 4 = tightened somewhat, and 5 = tightened considerably. We use the answers based on banks responses with respect to small rms because the QCEW data is predominantly composed of small businesses and we attempt to capture the local bank lending channel to rms.13 Although
However, changes in C&I lending standards for large and middle-market rms is highly correlated to those for small rms.
13
14
we do not have a breakdown of changes in lending standards in the SLOOS across dierent types of industries, available data sources suggest that banks have continued to provide a signicant share of C&I loans to the manufacturing sector over our sample period.14 In addition, though banks were extremely unlikely to characterize their changes in lending standards as eased considerably or tightened considerably, we depart from Bassett, Chosak, Driscoll, and Zakraj sek (forthcoming) in that we use all of the ve classications available to survey respondents. Letting j index the respondent banks and t index time, we dene an indicator variable Tj,t as follows: Tj,t = 2 if bank j reported considerable easing of standards at time t, Tj,t = 1 if bank j reported somewhat easing, Tj,t = 0 if bank j reported no change in standards, Tj,t = 1 if bank j reported somewhat tightening, and Tj,t = 2 if bank j reported considerable tightening. Second, we aggregate individual bank responses across banks for each U.S. state and convert those from quarterly to annual frequency. Using the indicator variables, we construct a composite of changes in lending standards for a particular state s, weighted by total business loans (C&I loans plus commercial real estate loans) for each year t, in part to strengthen the exogeneity assumption that banks with exposure to a broader economy aect industry-state employment dynamics. In other words, the largest of banks with branches in multiple states get weighted the mostthis is ideal because a large fraction of small business loans are originated by the largest banks.15 More specically, the tightening measure we use is calculated as follows: Ts,t =
4 q =1 J j =1 (business loans)j,q,t Tj,q,t , 4 J ( business loans ) j,q,t q =1 j =1
where q denotes a quarter of the year. Out of all the banks that participate in the SLOOS, we select only those that have deposit taking branches in a state s according to the Summary of Deposits. Hence, the total number of banks, J , in a given state may be below 60 for a particular state. We limit the coverage to 29 states (including the District of Columbia) where the J selected banks have at least a 15 percent cumulative share of deposits in every year of our sample.
14
16
We believe that these lters ensure that our state-level tightness
According to the G.27 Federal Reserve statistical release, about 28 percent of C&I loans outstanding at large commercial banks (with assets of more than $1 billion) were to the manufacturing sector as of December 1982. According to the FR Y-14 (a supervisory data collection covering bank holding companies with assets over $50 billion), as of the third quarter of 2012, 23 percent of all C&I loans outstanding were still to the manufacturing sector. In contrast, employment at manufacturing rms represented only about 9 percent of total nonfarm employment as of August, 2012. 15 According to regulatory Call Reports, as of the second quarter of 2013, about one-third of all loans with original amounts of less than $1 million, a proxy for small business loans, were booked in the top ve banks in terms of total assets. 16 The 29 states, which include the three largest economies in the country, are: Arizona, California, Colorado, Connecticut, Delaware, District of Columbia, Florida, Georgia, Illinois, Indiana, Kentucky, Maine,
15
measure is in fact representative for a given state. Questions regarding changes in standards on credit card loans and other consumer loans were added to the SLOOS only in 1996. However, we associate these changes with the changes in banks willingness to originate consumer installment loans, which are available over the entire sample period. The question about changes in consumer installment loans reads, Please indicate your banks willingness to make consumer installment loans now as opposed to three months ago. By analogy with the C&I index, we construct a national composite index of changes in willingness to make consumer installment loans, Wt , weighted by total consumer loans (excluding residential real-estate loans). We consider Wt as a proxy for changes in the tightness of lending standards for consumer loans. Figure 6 shows our measure of the changes in the net fraction of banks reporting a tightening of C&I lending standards for three states: New York, Texas, and California. It shows a drastic tightening of C&I lending standards around the past three recessions as well as a notable loosening of the standards in the mid-2000s. Across the 29 states in our sample, the variation is particularly high around the recessions, which provides us with adequate cross-sectional variation for our identication. Figure 7 plots our national measure of the net share of banks reporting an increased willingess to originate consumer installment loans. Notice that credit cycles vary by loan category in terms of timing and magnitude, which can be quite dierent from the macroeconomic business cycle. According to the Federal Reserve H.8 statistical release, C&I loans outstanding at commercial banks stopped contracting in the fourth quarter of 2004, whereas consumer loans were minimally aected by the 2001 recession. In terms of the credit markets, if the consumer loan market was relatively more impaired than the C&I loan market in the 2007 recession, it was the opposite for the 2001 recession; hence, it took the C&I loan market longer to recover after the 2001 recession.
3.5
State and national variables
We use the real GDP series from the Bureau of Economic Analysis and the state; the national level house price indices compiled by CoreLogic; and mortgages, HELOCs, home equity loans secured by junior liens, and net worth from the Federal Reserves statistical release Z.1. We also use state-level mortgage debt per borrower from TransUnions Trend Data.17 There is notable heterogeneity across states in the timing and magnitudes of house prices changes. Some areas experienced strong decreases in home values over the recent crisis, while other areas avoided the housing boom and experienced no signicant house
Massachusetts, Michigan, Minnesota, Missouri, Nevada, New York, North Carolina, Ohio, Oregon, Pennsylvania, Rhode Island, South Carolina, South Dakota, Tennessee, Texas, Virginia, and Washington. 17 Trend Data is an aggregated consumer credit database that oers quarterly snapshots of randomly sampled consumers.
16
Figure 6: Net fraction of banks reporting a tightening of C&I Lending Standards for small rms
Percent 80
New York Texas California
60
40 Tightening
20
Easing
-20
-40
-60
-80 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011
Source: Federal Reserve Board, Senior Loan Officer Opinions Survey on Bank Lending Practices.
price depreciation. We use a measure of growth in unemcumbered home equity as a proxy for changes in lending standards for HELOCs. Indeed, though we only have data going back to 2007, the correlation between the home equity growth we estimate on a national level and an aggregate measure of tightening of credit standards for HELOCs constructed based on the SLOOS is -0.90 on an annual basis and -0.77 on a quarterly basis, implying that unemcumbered home equity growth is a good indicator for changes in credit standards as well. We construct our measure of growth in home equity at the state or national level as follows. We start with the premise of Avery, Brevoort, and Samolyk (2011) that the dierence between house prices and outstanding mortgage debt should approximate home equity. Since we cast our regression models in growth rates, we construct a proxy for the growth rate of the equity ratio (the inverse of the loan-to-value ratio): HEs,t = HPs,t M Ds,t , where HEs,t is the growth rate of home equity in state s at time t, HPs,t is the growth rate of the house price (value) index in state s at time t, and M Ds,t is the growth rate of 17
Figure 7: Net fraction of banks reporting an increase in their willingness to originate consumer installment loans
Percent 80
U.S.
60
More Willing
40
20
Less Willing
-20
-40
-60
-80 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011
Source: Federal Reserve Board, Senior Loan Officer Opinions Survey on Bank Lending Practices.
mortgage debt in state s at time t.18 This admittedly might be a noisy proxy for growth in the equity ratio, but we believe it is the best available state-level measure. At the national level, there is an alternative reported in the Z.1 statistical release, which we denote as HEt .19 To make sure that growth in home equity is indeed an indicator of home equity loans availability, rather than a proxy for household wealth-driven demand, we include into our regression models growth in aggregate household net worth (N Wt ), constructed using the Federal Reserve statistical release Z.1.
Dene the equity ratio as HEs,t = HPs,t /M Ds,t and, after taking logs and dierentiating, obtain the expression in the text. 19 Comparing the national measure of the home equity growth rate computed from TransUnion with the national ocial level reported in the Z.1 Federal Reserve statistical release, we nd a correlation of 0.74. Therefore, we are comfortable using the state-level measure as an adequate proxy for growth in home equity. CoreLogic produces another variable of interestthe share of negative home equity measures at the state level; however, the reporting began only in 2009.
18
18
Variation, identication, and the empirical model

We choose our unit of observation to be a NAICS three-digit manufacturing industry
in a given state and a given year. To ensure even more robust identication, we could have worked with county- or MSA-level data, but, at such a low level of aggregation, there would have been too many missing observations due to condentiality and non-disclosure issues. In contrast, the QCEW industry data at the state level are available over a long period and include the undisclosed data suppressed within the detailed disaggregated tables. Hence, working with state-level data appears to strike a balance between exogeneity concerns and data availability. Although the QCEW is a quarterly frequency data set, we choose to work with annual averages for a few reasons. We are interested neither in immediate responses of employment to changes in access in credit, which later might be reversed, nor in the seasonality of manufacturing employment.20 We examine how credit supply conditions for both rms and households aect manufacturing industry dynamics. To isolate these eects and control for omitted variable bias, we exploit the variation in rm external nance dependence and asset tangibility across the manufacturing industries and the variation in consumer loan dependence across non-durable and durable goods manufacturing industries. Specically, we examine whether changes in C&I lending standards, the ability for home equity extraction by small business owners, banks willingness to originate consumer installment loans, and the ability for home equity extraction by households for consumption purposes matter for employment in manufacturing industries. Given a high degree of persistence in the number of manufacturing establishments and their average size and other variables over the sample period, as well as the nature of our measure of changes in C&I lending standards and the willingness to originate consumer installment loans, we work with an empirical model cast in growth rates; this model is stationary and allows us to control for aggregate trends in levels and growth rates (because of included xed eects). Growth rates of the explained variables are not persistent, with autoregressive coecients below 0.20; so, to avoid potential endogeneity, we omit lagged dependent variables from our regression models. Our identication assumption is that changes in major banks lending standards, apportioned to a particular state, and growth in home equity are exogenous to developments in a given manufacturing industry in a given state and at a given point in time. In accordance with the questions in the SLOOS survey, we postulate that banks tighten C&I lending standards broadly across the country rather than targeting a particular state and/or a particular industry. Variation in the geographical presence of banks and in the timing of tightening
In a quarterly model, a set of lagged explanatory variables would weaken identication because of collinearity.
20
19
generates variation in the changes our measure of changes in C&I lending standards across states.21 Besides omitted variables, we control for aggregate credit, state, and national economic conditions. Aggregate credit conditions are proxied by the change in the realized real interest rate. As a proxy for national economic conditions, we include the growth rate of U.S. real GDP. To address the potential endogeneity of industry location choices and industry-statespecic trends, we include industry-state xed eects in the benchmark model. To check the robustness of our results, we estimate additional models with various error clustering assumptions. Putting all the pieces together, we estimate the following specication:
Yi,s,t = T EFi T Ai Ts,t + H EFi HEs,t

factors for supply of credit to rms
+W DGi Wt + H DGi HEt

factors for supply of credit to households
+N DGi N Wt + S SCs,t + N N Ct + i,s + i,s,t , where Yi,s,t is the growth rate of employment (or the growth rate of the number of establishments or the growth rate of the average establishment size) in industry i and state s at time t; T is the coecient of the interaction term between the indicator for external nance dependence of industry i, EFi , asset tangibility, T Ai , and the net percentage of banks tightening standards for C&I loans in state s at time t, Ts,t ; H is the coecient of the interaction term between the indicator for the external nance dependence of industry i, EFi , and the growth rate of house equity in state s at time t, HEs,t ; W is the coecient of the interaction term between the indicator for durable goods industry i, DGi , and the net percentage of banks reporting increased willingness to make consumer installment loans, Wt ; H is the coecient of the interaction term between the indicator for durable goods industry i, DGi , and the growth rate of aggregate home equity at time t, HEt ;
21 Moreover, the SLOOS data suggest that C&I lending standards tightening for large and small rms is highly correlated, and so many banks change their standards on C&I loans in general rather than targeting a subset of borrowers.
20
N is the coecient of the interaction term between the indicator for durable goods industry i, DGi , and the growth rate of aggregate household net worth at time t, N Wt ;22 S and N are the coecients of the state and national economic conditions variables, captured by SCs,t and N Cs,t , respectively, which include Ts,t , Wt , HEt , N Wt , U.S. real GDP growth, and the rst dierence of the Libor-based realized real interest rate; i,s is the coecient for the industry-state xed eect, which should take into account any trends in the level of employment in the manufacturing sector by industry and state;23 and nally i,s,t is the error term robust to heteroskedasticity. We compute errors clustered separately in several ways: clustering by industry state, clustering by year, and double clustering by industry state and year. The multiple clustered errors are calculated using the Cameron, Gelbach, and Miller (2011) code.
4.1
Sample selection and representativeness

Our restricted sample for 29 states appears to be representative of the population of
manufacturing industries in the entire country. We checked the data breakdown by employment, number of establishments, and average establishment size for two years, 2007 and 2010. The population measures are shown in Table 2. In percentage terms, the breakdown of employment and number of establishments in our sample is very similar to that in the population, and the average establishment size in the sample is nearly identical to that in the population.24
5
5.1
Results
Four bank lending channels
First, we see if each credit channel exists separately from the others in Table 3. We
exploit the variation in rm external nance dependence across all manufacturing industries (to identify the eects of the supply of credit to rms) and the sensitivity of the manufacturing industries output to changes in consumer credit (to identify the eect of the supply
Unfortunately, we do not have an acceptable estimate for state-level net worth. However, because we do not nd statistically signicant eects from the home equity channel from the business side, we believe that this is less critical to have in our regressions. 23 The industry-state xed eects should also control for self-selected locations across the states. 24 We also check whether balancing the panel introduces any selection biases by estimating the regression model on an unbalanced panel. The results are largely unaected.
22
21
Table 2: Breakdown of employment in manufacturing industrie

Employment in manufacturing in 2007 EF T A = 0 EF T A = 1 DG = 0 3,514,947 1,537,086 DG = 1 7,311,679 1,469,312 Column total 10,826,626 3,006,398 Employment in manufacturing in 2010 EF T A = 0 EF T A = 1 DG = 0 3,193,000 1,254,642 DG = 1 5,971,005 1,068,850 Column total 9,164,005 2,323,492 Row total 5,052,033 8,780,991 13,833,024
Row total 4,447,642 7,039,855 11,487,497
Percentage change in employment in manufacturing (2007-10) EF T A = 0 EF T A = 1 Row total DG = 0 -9.2 -18.4 -12.0 DG = 1 -18.3 -27.3 -19.8 Column total -15.4 -22.7 -17.0
of credit to households). Specically, we nd that employment in industries dependent on external nance and with high asset tangibility is aected by the net share of commercial banks reporting changes in C&I loan lending standards, in the rst column. In the second column, we also nd that the availability for home equity extraction by households-business owners also matters for those industries dependent on external nance. In addition, employment in durable goods manufacturing industries falls with a decline in the net share of banks reporting a decreased willingness to originate consumer installment loans, in the third column. Finally, in the fourth column, the availability of home equity for extraction has a positive eect on employment in durable goods industries as well, even controlling for the growth in household net worth. All coecients are statistically signicant at the 1 percent level. Next, we present our empirical results with all of the four bank credit channels combined in Tables 4 to 6. We are also interested in examining whether credit supply conditions aect employment in manufacturing on the extensive or intensive margin. Hence, we estimate two additional sets of models: one for the number of establishments and another for the average establishment size. Given that the growth rate of employment is just a sum of the growth rates of the number of establishments and the average establishment size, the regression coecients in the employment regression are nearly exact sums of the corresponding coecients in Tables 5 and 6. Table 4 shows regression results for the models for total employment in manufacturing industries. While all the three models in the table have common industry-state xed eects, 22
Table 3: Separate lending channel regression results for employment

EF T A Ts EF HEs DG W (1) 0.040 (5.232) (2) (3) (4)
0.068 (2.115) 0.059 (7.739) 0.098 (3.122) 0.071 (3.479) 0.028 (4.304) 0.010 (0.546) 0.040 (7.402) 0.004 (0.229) 0.023 (2.206) 0.938 (13.385) 0.091 (2.311) 9500 500 0.14 I S I S 0.042 (7.506) 0.024 (0.924) 0.041 (7.423) 0.004 (0.239) 0.023 (2.190) 0.939 (13.439) 0.092 (2.325) 9500 500 0.14 I S I S 0.043 (7.622) 0.010 (0.580) 0.012 (1.929) 0.004 (0.242) 0.023 (2.208) 0.937 (13.432) 0.092 (2.318) 9500 500 0.15 I S I S 0.042 (7.557) 0.010 (0.570) 0.041 (7.435) 0.043 (1.590) 0.057 (3.369) 0.939 (13.466) 0.092 (2.320) 9500 500 0.14 I S I S
DG HE
DG N W
Ts HEs W
HE
N W Real GDP Real interest rate Num. of observations Num. of clusters R-sq. overall Fixed eects Error clustering
Note: EF denotes a dummy for industries dependent on external nance, T Aa dummy for industries with relatively high asset tangibility, and DGa dummy for industries producing durable goods. Ts is the net fraction of banks reporting a tightening of C&I lending standards for small rms at the state level. HEs is growth in home equity at the state level. W is the net fraction of banks reporting an increase in willingness to originate consumer installment loans at the national level. HE is growth in home equity at the national level. N W is the growth of net worth at the national level. Standard errors are clustered by I S (industry state). Coecients are reported with * if signicant at the 10% level, ** at the 5% level, and *** at the 1% level. t-statistics are reported below the coecients.
each model has a dierent specication of error clustering: in the rst column, errors are clustered by industry-state; in the second, by year; and, in the third, double-clustered by industry-state and year. With respect to the credit supply to rms, the results show that a percentage point increase in the net share of banks reporting a tightening of C&I lending standards reduces 23
the growth rate in employment in manufacturing industries with a high share of tangible assets and which are dependent on external nance by 0.04 percentage point. Note that this regression coecient is statistically signicant at conventional levels in the regression with double-clustered errors. The economic signicance of this and other coecients will be explored later in a separate section. The regression coecient capturing the impact of the ability for home equity extraction to nance rms dependent on external nance is positive, but no longer robustly statistically signicant. As for the credit supply to households, the results show that a percentage point increase in the net share of banks reporting an increased willingness to originate consumer installment loans increases the growth rate of employment in the manufacturing industries producing durable goods by 0.07 percentage point. In addition, a 1 percentage point increase in the growth in home equitya measure of potential home equity extraction by householdsboosts the growth rate in employment in the manufacturing industries producing durable goods by 0.15 percentage point. Table 5 shows regression results for the models for the number of establishments. For credit supply to rms, the results show that our measure for tightening credit standards does not appear to have a statistically signicant eect for two out of the three model specications. As for credit supply to households, the results also show that an increase in the net share of banks reporting an increased willingness to originate consumer installment loans increases the growth rate of the number of establishments in industries producing durable goods in a statistically signicant manner for only one out of the three specications. The ability for home equity extraction appears to have no statistically signicant eect on either the growth rate of establishments in industries dependent on external nance or that in industries which produce durable goods all three specications. The lack of statistically robust evidence that the supply of bank credit to rms has little eect on the number of establishments appears to be consistent with a few literature strands. First, for the the larger and more mature rms, consistent with the exporter hysterisis international trade literature, it may be that, following a tightening of access to credit, the sunk cost aspect of the rm entry decision in the presence of xed per-period costs results in these rm continuing to serve the market despite unfavorable economic or nancial conditions, but perhaps at a smaller scale requiring fewer employees, as noted in Alessandria and Choi (2007) and other works. Second, at the other extreme, for smaller rms, consistent with the literature on lending relationships as in Berger and Udell (1994) and Petersen and Rajan (1994), nascent rms may depend less on bank loans than older rms. In addition, setting up a rm may not be that costly. For example, according to Djankov, Porta, LopezDe-Silanes, and Shleifer (2002), entrepreneurs average cost of starting a rm (including the time to start up a rm) was 1.7 percent of per-capita income in the United States in 1999, or $520. Likewise, layos by rms that are induced by stricter lending standards may spur
24
Table 4: Employment regression results

(1) EF T A Ts EF HEs DG W 0.043 (6.113) 0.042 (1.345) 0.066 (8.864) 0.154 (4.816) 0.046 (2.356) 0.028 (4.319) 0.010 (0.393) 0.008 (1.338) 0.071 (2.544) 0.046 (2.716) 0.935 (13.391) 0.091 (2.300) 9500 500 0.16 I S I S (2) 0.043 (2.304) 0.042 (1.842) 0.066 (2.760) 0.154 (2.569) 0.046 (0.907) 0.028 (1.529) 0.010 (0.283) 0.008 (0.360) 0.071 (0.908) 0.046 (0.720) 0.935 (3.286) 0.091 (0.453) 9500 19 0.16 I S Y (3) 0.043 (2.051) 0.043 (1.196) 0.066 (2.506) 0.154 (2.232) 0.046 (0.813) 0.028 (1.338) 0.012 (0.272) 0.008 (0.320) 0.070 (0.776) 0.046 (0.625) 0.937 (2.917) 0.092 (0.399) 9500 500 19 0.16 I S I SY
DG HE
DG N W
Ts HEs W
HE
Note: EF denotes a dummy for industries dependent on external nance, T Aa dummy for industries with relatively high asset tangibility, and DGa dummy for industries producing durable goods. Ts is the net fraction of banks reporting a tightening of C&I lending standards for small rms at the state level. HEs is growth in home equity at the state level. W is the net fraction of banks reporting an increase in willingness to originate consumer installment loans at the national level. HE is growth in home equity at the national level. N W is the growth of net worth at the national level. Standard errors are clustered by I S (industry state) in column (1), by Y (year ) in column (2), and double-clustered by I S Y (industry state and year ) in column (3). Coecients are reported with * if signicant at the 10% level, ** at the 5% level, and *** at the 1% level. t-statistics are reported below the coecients.
some creation of establishments, which may boost the number of establishments in times of distress. Aaronson, Rissman, and Sullivan (2004), for example, document the increase in the number of rms, which was accompanied by a fall in employment at the aggregate level.
25
Table 5: Number of establishments regression results

DG HE
DG N W
Ts HEs W
HE
Indeed, in our data, while the smallest establishments employ relatively a small fraction of employees in aggregate, they are numerous and introduce some noise into the aggregate number of establishment series.
26
Although conducting our analysis separately for small and larger rms would be ideal to analyze this further, our data do not have this categorization at the industry-state level that we require to ensure the exogeneity setup. However, we can infer from a separate data set, the Census Bureaus County Business Patterns (CBP) data, that, indeed, for the entire country, changes in the number of establishments in the smallest establishment size classes introduce noise to the aggregate number of establishments.25 However, because they account in aggregate for a small fraction of the employment, changes in employment at these establishments have little impact in the aggregate. Breaking down these establishments by industry types reveals, at times, perverse behavior of metrics for the treatment group industries relative to the control groups, introducing further noise into the estimation. Table 6 shows regression results for the models for the average establishment size. For credit supply to rms, the results show that a percentage point increase in the net fraction of banks reporting a tightening of credit standards reduces the growth rate of the average size of the establishments with a high share of tangible assets and dependent on external nance by 0.03 percentage point. The results also show that an improvement in access to HELOCs for rms does not have a statistically robust impact on the average establishment size for two of the three specications. As for credit supply to households, the results show that a percentage point increase in the net share of banks reporting an increased willingness to originate consumer installment loans increases the growth rate of the average establishment size in industries producing durable goods by 0.06 percentage point. In addition, a 1 percentage point increase in home equitya measure of potential home equity extraction by householdsboosts the growth rate of the average size by 0.14 percentage point.
5.2
Breakdown of the EF T A channel for employment

In this section, we check whether dependence on external nance or pledgable asset
availability matters equally for employment in manufacturing industries. To accomplish this, we estimate separately the impact of access to external funding on EF = 1 and T A = 1 industries. As Table 7 shows, it is the former that is statistically signicant. However, we still believe it is correct to study the intersection of EF = 1 and T A = 1 industries for at least two reasons. First, the C&I loan denition and other survey results suggest the importance of pledgable assets for rms access to bank credit. Indeed, it might be the case that the literature denes pledgable assets too narrowly. For example, accounts receivable may serve as collateral for C&I loans as well. Second, the consideration
The reason we do not use the CBP data for our analysis is that the CBP time coverage is too short and the industry-state-size-class coverage is not comprehensive for our purposes. The CBP data have some breaks in industry classication, making it impossible to merge the data into a long sample. In addition, for some industries in some states, coverage is quite limited due to condentiality concerns. Finally, the CBP data are compiled as of the rst quarter for each year, rather than on a year-average basis.
25
27
Table 6: Average size of establishments regression results

DG HE
DG N W
Ts HEs W
HE
of EF = 1 and T A = 1 industries jointly, as a comparison of Tables 4 and 7 reveals, strengthens the impact of the eects of tightening of C&I lending standards on employment in manufacturing industries. More specically, the coecient on the triple-interacted term
28
Table 7: Separation of EF and T A channels for employment

(1) EF Ts T A Ts EF HEs DG W 0.025 (3.200) 0.004 (0.507) 0.045 (1.423) 0.064 (7.970) 0.155 (4.867) 0.044 (2.209) 0.028 (3.520) 0.011 (0.448) 0.010 (1.490) 0.071 (2.561) 0.045 (2.621) 0.936 (13.430) 0.092 (2.310) 9500 500 0.15 I S I S (2) 0.025 (2.975) 0.004 (0.271) 0.045 (2.081) 0.064 (2.548) 0.155 (2.636) 0.044 (0.885) 0.028 (1.321) 0.011 (0.328) 0.010 (0.401) 0.071 (0.922) 0.045 (0.702) 0.936 (3.289) 0.092 (0.456) 9500 19 0.15 I S Y (3) 0.025 (2.257) 0.004 (0.230) 0.046 (1.273) 0.064 (2.317) 0.155 (2.289) 0.044 (0.792) 0.028 (1.165) 0.014 (0.304) 0.010 (0.355) 0.071 (0.788) 0.045 (0.609) 0.938 (2.915) 0.092 (0.402) 9500 500 19 0.15 I S I SY
DG HE
DG N W
Ts HEs W
HE
we use in our baseline specication is almost as twice as large as the coecient on the changes in C&I credit standards interacted with EF only. 29
5.3
Structural breaks in the bank lending channels for employment

In this section, we check whether our results are driven by the developments over the
Great Recession and the subsequent recovery. To do so, we estimate a regression model that allows a change in the coecients on the measures of access to rm-oriented nance (the change in C&I lending standards and the state-level home equity growth measure) and to consumption-oriented nance (the change in the willingness to originate consumer installment loans and the country-level home equity growth measure). The estimation results shown in Table 8 on the various interactions with the Crisis indicator, denoting the period from 2007 to 2011, suggest no statistically robust evidence of structural breaks in these channels. That said, there is some indication that the change in home equity extraction ability for rm funding may have become more important and changes in willingness to originate consumer installment loans and home equity extraction ability for durable goods consumption may have become less important over the crisis.
Economic signicance and the real macro eects of bank lending channels
6.1
The real eects of bank lending channels during the Great Recession
With the estimates of the marginal eects in hand, we now perform back-of-the-
envelope calculations of the eects of bank lending channels on employment during the Great Recession based on the estimated dierential eects. In our calculations, we assume that the eect of tightening of C&I lending standards or a decline in the willingness to originate consumer loans does not aect the manufacturing industries in the control groupthat is, the non-durables industries with a low degree of external nance dependence and asset tangibility.26 The economic signicance of our results can be quantied by looking at the combined marginal eect of a tightening in C&I lending standards and a decrease in the willingness to originate consumer installment loans over the Great Recession and the subsequent slow recovery. The marginal eect of a 1 percentage point increase (decrease) in the net fraction of banks reporting a tightening in C&I lending standards or a decrease in the willingness to originate consumer installment loans are directly inferred from Table 4. We base these calculatesion from the changes in the net share of banks reporting a change in C&I lending standards and a willingness to originate consumer installment loans, which are inferred from Figures 6 and 7. These gures suggest that about 33 percent of banks tightened C&I lending
26 This approach provides a conservative estimate of the impact of credit-supply disruptions on employment in the manufacturing industries.
30
Table 8: Structural breaks in bank lending channels for employment

(1) EF T A Ts 0.039 (4.614)
(2) 0.039 (3.046)
(3) 0.039 (2.616) 0.003 (0.040) 0.004 (0.149) 0.097 (0.956) 0.112 (2.093) 0.206# (1.533) 0.057 (0.640) 0.069 (1.008) 0.114 (0.652) 0.025 (0.189) 0.028 (1.552) 0.001 (0.024) 0.008 (0.300) 0.082 (0.868) 0.048 (0.685) 0.930 (3.017) 0.128 (0.574) 9500 500x19 0.16 IS ISY
EF HEs
0.005 (0.149) 0.004 (0.344) 0.098 (2.013) 0.112 (9.682) 0.207 (5.719) 0.057 (2.548) 0.069 (4.982) 0.115 (3.057) 0.025 (1.021) 0.028 (3.989) 0.001 (0.052) 0.008 (1.165) 0.083 (2.904) 0.048 (2.813) 0.928 (11.678) 0.127 (2.576) 9500 500 0.16 IS IS
0.005 (0.094) 0.004 (0.178) 0.098 (1.185) 0.112 (2.307) 0.207# (1.725) 0.057 (0.708) 0.069 (1.120) 0.115 (0.719) 0.025 (0.210) 0.028 (1.808) 0.001 (0.039) 0.008 (0.334) 0.083 (1.038) 0.048 (0.789) 0.928 (3.404) 0.127 (0.651) 9500 19 0.16 IS Y
EF T A Ts Crisis
EF HEs Crisis
DG W
DG HE
DG N W
DG W Crisis
DG HE Crisis
DG N W Crisis
Ts
HEs
HE
N W
Real GDP
Real interest rate Num. of observations Num. of clusters R-sq. overall Fixed eects Error clustering
Note: EF denotes a dummy for industries dependent on external nance, T Aa dummy for industries with relatively high asset tangibility, and DGa dummy for industries producing durable goods. Ts is the net fraction of banks reporting a tightening of C&I lending standards for small rms at the state level. HEs is growth in home equity at the state level. W is the net fraction of banks reporting an increase in willingness to originate consumer installment loans at the national level. HE is growth in home equity at the national level. N W is the growth of net worth at the national level. Crisis is a dummy variable indicating the period from the beginning of the most recent recession to the subsequent recovery from 2007 to 2011. Standard errors are clustered by I S (industry state) in column (1), by Y (year ) in column (2), and double-clustered by I S Y (industry state and year ) in column (3). Coecients are reported with * if signicant at the 10% level, ** at the 5% level, *** at the 1% level, and # at the under 12% level. t-statistics are reported below the coecients.
31
Table 9: Back-of-the envelope macro eects in growth rates

Predicted percentage changes in employment in manufacturing (2007-10) EF T A = 0 EF T A = 1 Row total DG = 0 0.0 -4.2 -1.3 DG = 1 -5.4 -9.6 -6.1 Column total -3.7 -6.9 -4.4 Predicted changes in employment in manufacturing (2007-10) EF T A = 0 EF T A = 1 Row total DG = 0 0 -64,773 -64,773 DG = 1 -395,708 -141,436 -537,144 Column total -395,708 -206,209 -601,917 Goodness of t (predicted percentage change/actual percentage change) EF T A = 0 EF T A = 1 Row total DG = 0 0.0 22.9 10.7 DG = 1 29.5 35.3 30.9 Column total 23.8 30.2 25.7
standards on net per year, and that about 27 percent of banks decreased their willingness to originate consumer installment loans on net per year from 2007 to 2010. Bassett, Chosak, Driscoll, and Zakraj sek (forthcoming) and Lown and Morgan (2006) also use similar logic in their VAR framework in describing changes in credit standards and how they aect lending and output.27 For the entire U.S. manufacturing sector (rather than the sample used in estimation), the pre- and post-crisis breakdown in employment are shown in Table 2. Over the Great Recession total employment in the manufacturing sector declined, with the treatment groups experiencing notably larger declines than the control group (EF T A = 0 or DG = 0). And among the treatment groups, the manufacturing industries depending on external nance, with high asset tangibility, and producing durable goods were aected the most. Table 9 summarizes the impact estimates in growth rates. The model predicts a 4.4 percent decline in overall employment, while the actual decline was about 17 percent. For the industries dependent on external sources of funding with high tangible asset shares and produce durable goods (EF T A = 1 and DG = 1), the predicted percentage reduction in employment is the largest, followed by the industries that do not depend on external sources of funding or have relatively little tangible assets, but produce durable goods (EF T A = 0 and DG = 1). The impact on employment in the industries dependent on external sources of funding with a high share of tangible assets and produce non-durable goods (EF T A = 1 and DG = 0) was relatively small.28
These percentages are cardinal numbers and have a specic interpretation, though the underlying data on whether a bank reports a somewhat tightening of lending standards, for example, are subjective. 28 Although we could have evaluated a direct eect, we believe that such calculations may reect factors not directly related to credit supply, including demand-side factors.
27
32
These knock-o growth eects can be easily mapped into level eects on employment in manufacturing industries. The results in Table 9 suggest that the model predicts about a 602,000 decline in employment, while the actual decline was about 2.3 million employees. The number of workers displaced in the manufacturing industries dependent on external nance with high tangible asset shares, which also produce durable goods (EF T A = 1 and DG = 1), is predicted to be around 141,000 or about 10 percent of the employment in these industries in the base year. The number of workers displaced in industries not dependent on external nance or having a small share of tangible assets, but producing durable goods(EF T A = 0 and DG = 1) is about 396,000 (about 5.4 percent of the corresponding employment), while that number in industries dependent on external nance with a high degree of tangible assets, but producing non-durable goods (EF T A = 1 and DG = 0) is about 65,000 (about 4 percent of the corresponding employment in 2007). Again, our identication scheme does not inform on the impact of the tightening of lending standards or decreased willingness to originate loans on the industries not dependent on external nance and producing non-durable goods. Given the estimation results shown in Tables 5 and 6, it is clear that the employment declines were driven primarily by a shrinkage in the establishment sizes in the intensive margin rather than the number of establishments at the extensive margin. We are able to explain the larget share of actual declines in employment in manufacturing industries dependent on external nance with a high share of tangible assets, and at the same time, which produce durable goods. As Table 9 shows, we have the highest goodness of t for that categoryabout 35 percent. Generally, though, our calculations for other types of manufacturing industries tend to have lower goodness of t.
6.2
The real eects of the availability of home equity extraction

It should be noted that we have used only two variables so farthe changes in the net
fraction of banks reporting a tightening of C&I lending standards and a decreased willingness to originate consumer installment loansto explain changes in employment. When it comes to the industries producing durable goods, our regression results suggest that there is an additional statistically signicant bank lending channel at workthe availability of home equity extraction by households for consuming durable goods. Our proxy for the growth in home equity declined, on average, about 6 percent per year during the 2007-2010 period. Performing similar calculations to that above, we estimate that the reduction in home equity extraction predicts an additional 1.8 percent decline in manufacturing employment. Overall, our back-of-the envelope exercise suggests that cutbacks in bank credit availability account for about 35 percent of the decline in employment in the manufacturing sector. This leaves a sizeable remainder unexplained, most of which is presumably largely 33
related to a fall in demand. In fact, Mian and Su (2012) point to a large negative demand shock, in part, attributable to household overindebtedness, underlying employment losses in the non-traded goods sector over the Great Recession.
6.3
Generalization of the back-of-the-envelope exercise to the entire economy

We prefer to err on the conservative side in generalizing the back-of-the-envelope exer-
cises. In particular, we are cautious about equating job losses in a given industry to those in the entire economy. By concentrating on the manufacturing sector alone, we do not consider how many of the displaced manufacturing workers might have been absorbed by other sectors aected less by adverse economic and nancial conditions. However, we still believe that our ndings are indicative of developments in other sectors in the economy reliant on bank lending.
6.4
Prospects of the economic recovery and policy prescriptions

Our estimate of the impact of changes in C&I lending standards, changes in willing-
ness to originate consumer loans, and the growth in the availability of home equity on the manufacturing industries have implications for the recovery in the labor market going forward. To some extent, the tightening of lending standards and the decrease in willingness to originate loans reect commercial banks eorts to deleverage. By changing the composition of their balance sheets from loans toward U.S. Treasury securities, commercial banks may improve their risk-weighted capital ratios. By slowing the growth in the size of their overall balance sheets, banks can also improve their regulatory leverage ratios. The greater anticipated regulatory burden faced by commercial banks may temporarily hold back employment growth in manufacturing industries dependent on external nance (and/or producing durable goods), thus contributing to continued weak labor market conditions. Moreover, the sluggish housing market improvement may have been a further drag on manufacturing employment. In the longer term, though, the displaced workers in these industries may be absorbed by other sectors in the economy, as our results do not suggest permanent impediments to growth. When dealing with the unusually low levels of employment in manufacturing in an environment like the Great Recession, the policy prescription that follows from our back-ofthe-envelope exercise is that policymakers should pay attention to restoring the functioning of the bank credit supply channels to households relatively more than those to rms. Indeed, our result that household access to bank credit seems to matter more for employment in the manufacturing sector than rms access to bank loans appears to be consistent with some stylized facts about the U.S. nancial system. While households rely more on banks for 34
credit, many large rms have the ability to access capital markets directly, implying at least some substitutability between bank and capital markets funding. In addition, our results are consistent with the fact that the share of consumption in GDP, at about 70 percent, by far dominates the share of physical capital investment. However, our results should not be taken as advocating an implementation of Keynesian policies in a recession. In this paper, we focus on employment losses in the manufacturing industries attributable to cutbacks in credit supply only. As long as supply of credit is constrained by bank-related factors, ceteris paribus, policymakers might be urged to step in to restore credit ow, which implies dierent policies from simply trying to boost aggregate demand.
Conclusion
We examine how bank credit supply conditions for both rms and households aect
employment in the U.S. manufacturing industries. To isolate these eects, we exploit variation in changes in C&I lending standards, changes in willingness to originate consumer installment loans, and growth in home equity across U.S. states and time. To control for omitted variable bias, we rely on dierences in the degree of external nance dependence across manufacturing industries and in the sensitivity of these industries output to changes in consumer credit. We show that changes in C&I lending standards and the willingness to originate consumer installment loans by major commercial banks and the availability of home equity extraction by households aect notably employment in manufacturing industries over the 1991-2011 period. In particular, our results highlight the adverse eects that disruptions in the supply of bank credit have had on employment in manufacturing industries over the Great Recession.
35
Acknowledgments
We thank Bill Bassett, Nick Bloom, Christopher Carroll, John Driscoll, Lucca Guerrieri, John Haltiwanger, and Jonathan Rose as well as seminar participants at the Federal Reserve Board for helpful comments and suggestions. In addition, we thank discussants Adonis Antoniades, Chiara Banti, Jin Cao, Enrique Martinez-Garcia, Anthony Murphy, Leonard Nakamura, Francisco Rodr guez-Fern andez, and Pei Shao and participants at numerous conference and meetings for valuable comments and suggestions. Finally, we are grateful to two anonymous referees and the editor for helpful suggestions. We are also grateful to Robert Avery and Neil Bhutta for their help with accessing TransUnions Trend Data. We thank Shaily Patel for excellent research assistance. Viktors Stebunovs also thanks the Oce of Financial Stability Policy and Research for its hospitality.
36
References
Aaronson, D., E. R. Rissman, and D. G. Sullivan (2004): Assessing the Jobless Recovery, Federal Reserve Bank of Chicago Economic Perspectives, 28(2), 220. Alessandria, G., and H. Choi (2007): Do Sunk Costs of Exporting Matter for Net Export Dynamics?, Quarterly Journal of Economics, 122(1), 289336. Avery, R., K. P. Brevoort, and K. Samolyk (2011): Housing Cycles, Household Credit Performance, and Economic Activity, Working Paper. Bassett, W. F., M. B. Chosak, J. C. Driscoll, and E. Zakraj sek (forthcoming): Changes in Bank Lending Standards and the Macroeconomy, Journal of Monetary Economics. nez, and S. Ruano (2013): When Credit Dries Bentolila, S., M. Jansen, G. Jime Up: Job Losses in the Great Recession, Working Paper. Berger, A. N., and G. F. Udell (1994): Did Risk-Based Capital Allocate Bank Credit and Cause a Credit Crunch in the U.S.?, Journal of Money, Credit and Banking, 26(3), 585628. Bijlsma, M., A. Dubovik, and B. Straathof (2010): How Large was the Credit Crunch in the OECD?, CPB Discussion Paper 232, CPB Netherlands Bureau for Economic Policy Analysis. Black, S. E., and P. E. Strahan (2002): Entrepreneurship and Bank Credit Availability, Journal of Finance, 57(6), 28072833. Braun, M. (2002): Financial Contractibility and Assets Hardness: Industrial Composition and Growth, Working Paper. Cameron, A. C., J. B. Gelbach, and D. L. Miller (2011): Robust Inference With Multiway Clustering, Journal of Business and Economic Statistics, 29(2), 238249. Cetorelli, N., and P. E. Strahan (2006): Finance as a Barrier to Entry: Bank Competition and Industry Structure in Local U.S. Markets, Journal of Finance, 61(1), 437461. Chor, D., and K. Manova (2012): O the Cli and Back? Credit Conditions and International Trade during the Global Financial Crisis, Journal of International Economics, 87(1), 117133.
37
Claessens, S., and L. Laeven (2003): Financial Development, Property Rights, and Growth, Journal of Finance, 58(6), 24012436. Colla, P., F. Ippolito, and K. Li (2013): Debt Specialization, Journal of Finance, 68(5), 21172141. Davis, S. J., J. Haltiwanger, R. Jarmin, and J. Miranda (2006): Volatility and Dispersion in Business Growth Rates: Publicly Traded versus Privately Held Firms, NBER Working Paper No. 12354. Djankov, S., R. L. Porta, F. Lopez-De-Silanes, and A. Shleifer (2002): The Regulation of Entry, Quarterly Journal of Economics, 117(1), 137. Duygan-Bump, B., A. Levkov, and J. Montroiol-Garriga (2010): Financing Constraints and Unemployment: Evidence from the Great Recession, Quantitative Analysis Unit Series Paper QUA10-6, Federal Reserve Bank of Boston. Fort, T. C., J. Haltiwanger, R. Jarmin, and J. Miranda (2013): How Firms Respond to Business Cycles: The Role of Firm Age and Firm Size, NBER Working Paper No. 19134. Garmaise, M. J., and T. J. Moskowitz (2006): Bank Mergers and Crime: The Real and Social Eects of Credit Market Competition, Journal of Finance, 61(2), 495539. Haltiwanger, J., R. Jarmin, and J. Miranda (2009): Business Dynamics Statistics: An Overview, Marion Ewing Kauman Foundation BDS Briefs. Jarmin, R. S., S. D. Klimek, and J. Miranda (2009): The Role of Retail Chains: National, Regional and Industry Results, in Producer Dynamics: New Evidence from Micro Data, ed. by T. Dunne, J. B. Jensen, and M. J. Roberts, pp. 237262. University of Chicago Press. Kerr, W. R., and R. Nanda (2009): Democratizing Entry: Banking Deregulations, Financing Constraints, and Entrepreneurship, Journal of Financial Economics, 94(1), 124149. Lee, S. J., and V. Stebunovs (2012): Bank Capital Ratios and the Structure of Nonnancial Industries, Finance and Economics Discussion Series 2012-53, Board of Governors of the Federal Reserve System (U.S.). Lown, C. S., and D. P. Morgan (2006): The Credit Cycle and the Business Cycle: New Findings Using the Loan Ocer Opinion Survey, Journal of Money, Credit, and Banking, 38(6), 15751597. 38
Mian, A. R., and A. Sufi (2012): What Explains High Unemployment? The Aggregate Demand Channel, NBER Working Paper No. 17830. Peek, J., and E. S. Rosengren (2000): Collateral Damage: Eects of the Japanese Bank Crisis on Real Activity in the United States, American Economic Review, 90(1), 3045. Petersen, M. A., and R. G. Rajan (1994): The Benets of Lending Relationships: Evidence from Small Business Data, Journal of Finance, 49(1), 337. Rajan, R., and L. Zingales (1998): Financial Dependence and Growth, American Economic Review, 88(3), 559586.
39
Appendices
Robustness checks
In this appendix, we describe our additional robustness checks in further detail. First, we check whether are results are robust to using scale measures of external nance dependence and asset tangibility, rather than the categorical measures indicating external nance dependence and high asset tangibility used in the benchmark models. Both the Rajan-Zingales measure and the net fraction of banks tightening C&I credit standards can be either positive or negative, which introduces ambiguity in interpretation of the triple interaction term, EF T AT . Note that, the reason why other papers, such as Cetorelli and Strahan (2006), do not run into this issue is that they interact the Rajan-Zingales measure with a categorical variablefor example, with the interstate bank deregulation dummymaking interpretation of the respective regression coecient straightforward. In contrast, in more recent work, such as Chor and Manova (2012), in which interaction terms included scale variables, the authors rely on the median-based approach to separate industries into control and treatment groups. Nevertheless, we consider a model under the full variation approach, with the estimation results shown in the rst column of Table 10. To estimate such a model, we performed an ane transformation of the Rajan-Zingales measure that preserved the ordinal ranking of industries in terms of their dependence on external sources of nance. More specically, we added a suciently large positive integer to each measure, so that all these measures became positive. The estimation results show that the sign on the triple interaction term is still negative, indicating that, with a tightening in C&I credit standards, employment falls more in industries with higher dependence on external nance and higher asset tangibility. We shy away from using the full variation approach in the benchmark regressions because of issues with economic interpretation of the regression coecients. Second, at a risk of introducing endogeneity, we check whether our results carry over when we replace the Rajan-Zingales nancial dependence measure with a ratio of bank loans to assets for small rms based on the 1998 Survey of Small Business Finance (SSBF).29 The SSBF-based regression results, shown in the second column in Table 10, suggest that the home equity extraction channel for rm funding is both statistically and economically signicant. As for using a measure capturing dependence on bank loans in the benchmark regressions, we are cautious about subjecting the model to endogeneity concerns. In contrast to the Rajan and Zingales measure, which reect deep technological parameters computed
29 We use the ratios from Cetorelli and Strahan (2006) computed using the 1998 SSBF at a two-digit SIC level. We map these measures to a three-digit NAICS level. This mapping is necessarily rough as the two classications are not fully compatible even at low levels of aggregation.
40
for nancially-unconstrained rms, the bank-loan-to-assets ratio, for example, estimated using the SSBF, is necessarily subject to endogeneity, where low dependence may simply indicate nancing constraints.30 Third, we re-estimate our benchmark model with separate time trends for each of our 21 manufacturing industries included, with the estimation results shown in the third column of Table 10. As the table suggests, our results are robust to accounting for even these trends in growth rates for each industry. However, we believe the inclusion of these additional regressors in the model is redundant; moreover, only a few industries have statistically signicant time trends. We emphasize that our benchmark regressions are for growth rates in total employment, the number of establishment, and their average size measured in employees, which should take into account any trends in the level of total employment in the manufacturing sector. In addition, we have industry-state xed eects that control for (a) the trend in growth rates that dier by industry and by state and (b) industries self-selected locations across the states. Finally, we conduct several additional robustness checks. Our results are robust to the exclusion of the growth rate of real GDP, inclusion of state-level GDP, dierent home equity ratio denitions, and the inclusion of lagged dependent variables. Furthermore, we also use alternative measures of changes in C&I lending standards and the willingness to originate consumer installment loans and nd no material change in our results: we experiment with dierent weights in construction of the tightness and willingness measures, with classication of responses from the SLOOS, and with the threshold for cumulative shares of deposits in every year of our sample for a given state. Our results also still hold when we exclude bankfriendly states, such as Delaware and South Dakota, and large states, such as California, New York, and Texas, from our sample. We nd some evidence that the trade credit channel was operational. Finally, we investigate whether an alternative, equity-requirement-based approach to identifying home equity extraction for rm funding was preferable, and it was not.
30 In the Rajan-Zingales context, whether physical capital nancing gaps are met with bank or market funding is irrelevant. Moreover, for most large rms, these cannot be reliably estimated using publicly available data.
41
Table 10: Regression results for employment with (1) full variation in EF and T A Measures, (2) SSBF-based EF measure, and (3) industry-specic trends
(1) EF T A Ts EF HEs DG W 0.024 (1.928) 0.027 (1.145) 0.067 (8.402) 0.157 (4.868) 0.047 (2.360) 0.026 (2.329) 0.057 (0.850) 0.008 (1.220) 0.072 (2.580) 0.046 (2.716) 0.936 (13.435) 0.091 (2.297) 9500 500 0.15 I S I S (2) 0.008 (1.011) 0.146 (4.586) 0.066 (8.548) 0.174 (5.313) 0.045 (2.248) 0.039 (6.129) 0.064 (2.383) 0.008 (1.295) 0.081 (2.882) 0.045 (2.632) 0.936 (13.449) 0.092 (2.316) 9500 19 0.15 I S I S (3) 0.032 (4.425) 0.021 (0.662) 0.068 (8.033) 0.162 (4.838) 0.048 (2.366) 0.038 (5.687) 0.001 (0.030) 0.001 (0.150) 0.080 (2.786) 0.050 (2.880) 0.846 (11.899) 0.105 (2.624) 9500 500 0.17 I S I S
DG HE
DG N W
Ts HEs W
HE
Note: EF denotes a dummy for industries dependent on external nance, T Aa dummy for industries with relatively high asset tangibility, and DGa dummy for industries producing durable goods. Coecients are reported with * if signicant at the 10% level, ** at the 5% level, and *** at the 1% level. Ts is the net fraction of banks reporting a tightening of C&I lending standards for small rms at the state level. HEs is growth in home equity at the state level. W is the net fraction of banks reporting an increase in willingness to originate consumer installment loans at the national level. HE is growth in home equity at the national level. N W is the growth of net worth at the national level. Column (1) uses an ane transformation of the EF measure to incorporate the full variation in EF and T A Measures, column (2) uses an EF measure based on the bank-loans-to-assets ratio based on the 1998 SSBF, and column (3) includes industry-specic time trends. Standard errors are clustered by I S (industry state) t-statistics are reported below the coecients. Industry-specic trend coecients are not shown.
42

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C

Transféré par

Droits d'auteur :

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Aairs Federal Reserve Board,

Bank Lending Channels during the Great Recession

Samuel Haltenhof, Seung Jung Lee, and Viktors Stebunovs

Bank Lending Channels during the Great Recession

Figure 1: Four credit supply channels

C&I loans Bank Tangible coll. Firm Funds

Small business owner

Home equity Bank HELOCs

Changes in standards (left axis) Growth in employment (right axis)

Note: Shaded areas are NBER-dened recessions.

Description of the data

Focus on manufacturing industries at the state level

The Quarterly Census of Employment and Wages

External nance dependence for rms and households

Table 1: Characteristics of manufacturing industries

Figure 3: Growth in employment in manufacturing industries in the U.S.

EFxTA=0, DG=0 EFxTA=1 DG=1

Figure 4: Growth in the number of establishments in manufacturing industries in the U.S.

EFxTA=0, DG=0 EFxTA=1 DG=1

EFxTA=0, DG=0 EFxTA=1 DG=1

We believe that these lters ensure that our state-level tightness

State and national variables

New York Texas California

Note: Shaded areas are NBER-dened recessions.

Note: Shaded areas are NBER-dened recessions.

Variation, identication, and the empirical model

Yi,s,t = T EFi T Ai Ts,t + H EFi HEs,t

+W DGi Wt + H DGi HEt

Sample selection and representativeness

Table 2: Breakdown of employment in manufacturing industrie

Row total 4,447,642 7,039,855 11,487,497

Table 3: Separate lending channel regression results for employment

Table 4: Employment regression results

Table 5: Number of establishments regression results

Breakdown of the EF T A channel for employment

Table 6: Average size of establishments regression results

Table 7: Separation of EF and T A channels for employment

Structural breaks in the bank lending channels for employment

Table 8: Structural breaks in bank lending channels for employment

(2) 0.039 (3.046)

Table 9: Back-of-the envelope macro eects in growth rates

The real eects of the availability of home equity extraction

Generalization of the back-of-the-envelope exercise to the entire economy

Prospects of the economic recovery and policy prescriptions

Vous aimerez peut-être aussi