The role of securitization in bank liquidity and funding management

Journal of Financial Economics 100 (2011) 663–684

Contents lists available at ScienceDirect

Journal of Financial Economics journal homepage: www.elsevier.com/locate/jfec

The role of securitization in bank liquidity and funding management$ Elena Loutskina n The Darden School of Business Administration, University of Virginia, 100 Darden Boulevard, Charlottesville, VA 22903, United States

a r t i c l e i n f o

abstract

Article history: Received 6 April 2010 Received in revised form 13 July 2010 Accepted 13 August 2010 Available online 3 March 2011

This paper studies the role of securitization in bank management. I propose a new index of ‘‘bank loan portfolio liquidity’’ which can be thought of as a weighted average of the potential to securitize loans of a given type, where the weights reflect the composition of a bank loan portfolio. I use this new index to show that by allowing banks to convert illiquid loans into liquid funds, securitization reduces banks’ holdings of liquid securities and increases their lending ability. Furthermore, securitization provides banks with an additional source of funding and makes bank lending less sensitive to cost of funds shocks. By extension, the securitization weakens the ability of the monetary authority to affect banks’ lending activity but makes banks more susceptible to liquidity and funding crisis when the securitization market is shut down. Published by Elsevier B.V.

JEL classification: G21 Keywords: Securitization Bank management Liquidity Lending channel Monetary policy

1. Introduction Since the 1970s, the market for securitized loans in the United States has grown to dominate the mortgage market and has become an increasingly important factor in lending to both consumers and businesses (Fig. 1). In

$ An earlier version of this paper was distributed under the title ‘‘Does Securitization Affect Bank Lending? Evidence from Bank Responses to Funding Shocks.’’ I am indebted to Philip Strahan and Gary Gorton for invaluable guidance and suggestions. For helpful comments and discussions, I thank Murillo Campello, Mark Carey, Thomas Chemmanur, Robert DeYoung, Edward Kane, Alan Marcus, Jeremy Stein, and seminar participants at Boston College, University of Virginia, Indiana University, University of Texas at Dallas, Bank Structure and Competition Conference (Chicago 2006), Western Finance Association meeting (2006) and European Finance Association annual meeting (2005). I gratefully acknowledge the financial support of the Foundation, Banque de France grant in the fields of money, finance, and banking. I alone am responsible for any errors or omissions. n Tel.: + 1 434 243 4031; fax: + 1 434 243 12511. E-mail address: [email protected]

0304-405X/$ - see front matter Published by Elsevier B.V. doi:10.1016/j.jfineco.2011.02.005

2007, for example, $8.1 trillion of loans outstanding were financed through securitization, or about 40% of all loans outstanding. Even in the aftermath of the 2007 financial crisis, the securitization market activity by volume exceeds the size of the corporate bond market. A rapidly evolving body of academic research aims to understand whether and how securitization changed the traditional role of banks in the economy.1 This paper contributes to

1 For analysis of the impact of securitization on loan origination decisions, see, e.g., Loutskina and Strahan (2009), Mian and Sufi (2009), Demyanyk and Van Hemert, 2009, and Keys, Mukherjee, Seru and Vig (2010). For analysis of the changing mortgage rates under the evolving asset-backed securities market, see, e.g., Black, Garbade, and Silber (1981), Kolari, Fraser, and Anari, 1998, and Heuson, Passmore, and Sparks (2000). For analysis of the role of the government-sponsored enterprises (GSEs) and the effect of government subsidies to GSEs, see Passmore (2004), Ambrose and Warga (2002), and Nothaft, Pearce, and Stevanovic (2002). For analysis of the effect of securitization on the efficacy of monetary policy in influencing real output, see Estrella (2002).

664

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

0.70 Home mortgages

Multifamily home mortgages

Commercial mortgages

Farm mortgages

Commercial and industrial loans

Consumer credit

0.60

Share securitized

0.50

0.40

0.30

0.20

0.10

0.00 1976

1978

1981

1983

1986

1988

1991

1993

1996

1998

2001

2003

2006

Fig. 1. Securitization of loans in the US economy. The figure presents the percentage of loans securitized relative to total loans outstanding for six categories of loans: (1) home mortgages, (2) multifamily residential mortgages, (3) commercial mortgages, (4) consumer credit, (5) business loans, and (6) farm mortgages. The data are from ‘‘Flow of Funds Accounts of the United States.’’ For the exact description of the methodology see Appendix A.

this strand of research by examining how securitization changed the ways individual banks manage their funding and liquidity and how these changes have in turn altered the traditional links between bank liquidity, cost of funds, and loan supply. I show, first, that securitization creates a new source of liquidity by allowing banks to convert illiquid, hard-to-sell loans into marketable securities. Second, by providing a new source of funds in the form of existing loans, securitization reduces the sensitivity of bank lending to the availability of the external sources of funds such as traditional liquid funds and deposits. As a result, securitization alleviates the impact of the local economic shocks and weakens the ability of the monetary authority to affect bank lending through open market operations. At the same time, it makes banks more vulnerable to various economic shocks when the market for securitized loans is disrupted. I propose a new bank-specific index of ‘‘bank loan portfolio liquidity’’ (Sit) that effectively captures banks’ ability to sell loans. The index is a weighted average of the potential to securitize loans of a given type (based on market-wide averages), in which the weights reflect the composition of an individual bank’s loan portfolio. Thus, market trends generate time variation in the index, whereas differences in bank loan portfolio structures generate variation across institutions. I first analyze whether securitization has reduced banks’ need to carry liquid assets to meet unexpected demands from depositors and borrowers. Using the new loan liquidity index (Sit), I show that securitization acts as a substitute for traditional liquid funds on banks’ balance sheets. Because banks choose liquidity levels and lending

jointly, I adopt two approaches to adjust for this endogeneity.2 First, I implement the instrumental variable regressions using a synthetic instrument similar to the loan liquidity index (Sit). In constructing the instrument, I use fixed bank portfolio choices as of the beginning-ofperiod values. This constant-over-time loan portfolio structure removes the effect of the managers’ discretion and ensures that the instrument varies only as a result of the deepening of the securitization market. Second, I implement a difference-in-differences analysis around two sets of regulatory interventions and market shocks that significantly changed securitization market ability or willingness to absorb new loans. The results suggest that as banks’ ability to securitize loans has increased, their holding of liquid assets on balance sheets has decreased. The magnitude of this decline is both statistically and economically significant (Fig. 2). From 1976 to 2007, the percentage of total assets held as liquid securities decreased on average by 7.33 percentage points due to the expanding secondary loan market. This decline is equivalent to roughly 69% of bank capital and cannot be explained by any time trends such as the increase in average bank size and changes in banking regulation. Because liquid funds and loans are two core components of bank assets, the decrease in the liquid funds holdings indicates a comparable increase in

2 For example, banks that prefer more liquid assets are likely to have both more liquid funds and a more securitizable loan portfolio (which can be achieved by, e.g., issuing more mortgages and fewer commercial and industrial loans), thus creating a positive bias in the relations between traditional liquidity levels and bank loan portfolio liquidity.

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

665

0.4

0.25

0.36

0.2 0.32 0.15 0.28 0.1 0.24 0.2 1976

0.05 0 1978

0.45 On-balance sheet liquidity

0.3

Loan portfolio liduidity

Loan portfolio liquidity

On-balance sheet liquidity

On-balance sheet liquidity

1981

1983

1986

Full sample

1988

1991

1993

1996

Lowest size quartile

1998

2001

2003

2006

Highest size quartile

0.4

0.35

0.3

0.25

0.2 1976

1978

1981

1983

1986

1988

1991

1993

1996

1998

2001

2003

2006

Fig. 2. Traditional liquidity, bank loan portfolio liquidity, and size relations. Panel A presents the relation over time between average level of liquidity (Bit) maintained by banks (solid line, left-hand-side scale) and the average bank loan portfolio liquidity (Sit) (dashed line, right-hand-side scale). Panel B presents the evolution of the average level of liquidity (Bit) for the full sample of bank-quarters, as well as for the lowest and highest size quartiles of the sample. See Appendix A for exact methodology for calculating Bit, Sit, and size of the banks. The sample contains bank-quarters from 1976:I through 2007:IV.

lending. Thus, securitization increased the supply of bank lending per dollar of capital in the industry. If banks can liquidate loans to finance their liquidity need, they can also do so to finance new credit, making banks less dependent on the traditional sources of funds. The increasing liquidity of bank loans ought to change the link from banks’ funding availability (e.g., deposits) to their willingness to supply credit. The existing literature shows that the availability of additional internal (Kashyap and Stein, 2000) and external (Campello, 2002; Ashcraft, 2006) sources of funds partially alleviates the effect of restrictions in availability of funds on bank loan supply. Because securitization provides banks with an additional source of both loan financing and liquidity, it should also shield banks’ willingness to supply credit from the external cost of funds shocks. To test this argument, I follow the regression framework of Kashyap and Stein (2000), which allows me to take advantage of both time series and cross-sectional variation in the loan liquidity index (Sit) and its interaction with the cost of external funds. Due to the endogeneity

problem, instead of measuring the cost of funds directly from banks’ balance sheets, I exploit the Federal Reserve’s ability to affect bank cost of funds via open market operations and construct shocks to bank funding costs that are exogenous to financial intermediaries’ decisions.3 Considering the relation between bank liquidity, lending, and loan liquidity under these exogenous shocks allows me to adjust for the endogeneity that arises due to managerial discretion. I find that securitization has made loan growth (especially growth in business loans) less sensitive to cost of funds shocks. A bank with a more liquid loan portfolio (e.g., one that holds a significant amount of mortgages) incurs a smaller on-balance sheet decrease in lending

3 The Federal Reserve’s ability to affect bank lending behavior via open market operations is called the bank lending view of monetary policy transmission. For a review of this literature, see Bernanke and Blinder (1992), Bernanke and Gertler (1995), and Kashyap and Stein (1994). The empirical evidence is shown in Kashyap and Stein (1995, 2000) and Jayaratne and Morgan (2000).

666

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

3.5

C&I loan growth (percent)

3

Monetary tightening,

Monetary loosening,

1993:IV to 1995:I

2001:II to 2003:IV

2.5 2 1.5 1 0.5 0 Banks with illiquid loan portfolios

Banks with liquid loan portfolios

Fig. 3. Bank loan portfolio liquidity and commercial and industrial loan growth. The figure presents the value-weighted C&I loan growth for two types of banks under the monetary tightening of 1993–1995 and monetary loosening of 2001–2003. I consider banks with the loan liquidity in the top 10% of Sit distribution to have liquid loan portfolios and banks with the loan liquidity in the bottom 10% of Sit distribution to have illiquid loan portfolios.

under a monetary tightening than a bank with a less liquid loan portfolio (e.g., a bank focused on business lending). Fig. 3 illustrates the result intuitively by plotting average loan growth during tight and loose monetary regimes for banks with high and low loan liquidity (Sit).4 One can see that during the period of monetary tightening, banks with more liquid loan portfolios exhibit significantly higher business loan growth than banks with illiquid loan portfolios. Given that business loans are still one of the least liquid loan categories, such funding is not mere evidence of substitutions of loans on the balance sheet. The results of more rigorous regression analysis indicate that a 100 basis point increase in the federal funds rate would reduce loan growth by 0.7% to 1.3% less at a bank with a more liquid loan portfolio (Sit at 90th percentile) compared with a bank with a less liquid loan portfolio (Sit at 10th percentile). The effect is significantly more pronounced for C&I loans, reaching 5.25% smaller decline for the first bank. The ability to securitize their existing loans insulates banks’ willingness to supply credit from a monetary policy induced cost of funds shock to the availability of external financing.5 I obtain similar evidence from a difference-in-differences analysis around exogenous regulatory changes that affected the securitization market’s ability to absorb new loans.

4 In contrast to the multivariate panel model, this simple univariate comparison does not control for loan demand. Hence, loan growth is higher during the period of tightening than during the period of loosening. 5 For comparison, two equal-sized banks with the same access to the securitization market but with levels of on-balance sheet liquidity around 10th and 90th percentiles of the level of liquidity distribution will have 0.4% to 0.7% loan growth differentials four quarters after a 100basis-points increase in the federal funds rate.

This paper contributes to the line of research exploring how the advancements in securitization have changed the nature of banking. A number of recent studies tied securitization to excessive credit supply (Mian and Sufi, 2009; Demyanyk and Van Hemert, forthcoming; Keys, Mukherjee, Seru, and Vig, forthcoming; Loutskina and Strahan, 2009; Rajan, Seru, and Vig, 2010); lack of ex post monitoring incentives (Piskorski, Seru, and Vig, forthcoming; Parlour and Plantin, 2008); and deterioration of credit quality (Purnannandam, forthcoming; Loutskina and Strahan, forthcoming). In contrast to these studies that mostly explore the shadow banking, off-balance sheet implications of securitization, this paper illustrates that securitization leads to material changes on banks’ balance sheets and their risk exposure. Such changes are of particular importance to the regulatory authority. First, securitization has become an integral part of bank liquidity-risk management. Bank loan liquidity should now be considered along traditional balance sheet measures of liquidity, such as cash and marketable securities. Second, securitization increases banks’ credit supply across sectors. Banks’ ability to securitize liquid mortgages increases their willingness to supply illiquid business loans. This represents the beginning of the structural shift in the lending industry in which segmentation occurs with liquid loans financed through securitization and illiquid loans financed from deposits. Such a structural shift suggests changing risk profiles of financial institutions. Third, securitization weakens the link from the cost of traditional sources of funds to banks’ willingness to supply credit and, by extension, makes bank lending activity less prone to local economic shocks as well as global monetary authority interventions. With securitization, it might be necessary to make larger policy moves to achieve a significant contraction in bank lending. But because every coin has two sides, overreliance on securitization in financing liquidity and funding needs

729,011 44,628 1,135,225 192,714 2,339,017 652,555 84,861 4 109,983 0 117,497 4,618 1,673,709 30,218 2,387,841 107,572 2,993,757 80,433 1,162,807 226,420 1,724,595 580,209 2,550,586 682,110 4,614,323 1,890,898 6,932,428 3,298,285 14,284,542 7,331,595 7,450,839 2,147,536 11,044,864 3,986,066 19,946,382 8,098,756 825,087 12,509 78,880 24 1,464,743 4,267 802,207 83,251 3,846,899 1,123,994 6,113,849 1,211,512 1,377 427 0 0 421,436 421,436 553,272 103,994 1,191,754 606,799 2,423,680 4,222,233 0 885 0 0 117,260 117,260 0 2,006 0 0 31,152 31,152 162,168 50,797 409,006 204,908 802,628 1,416,542

266,231 100,097 766,107 353,902 1,486,718 2,606,727

5,301,129 2,988,224 11,135,792 6,419,087 386,091 117,347 809,733 259,953 1,806,239 40,027 3,532,530 267,921 1,083,561 27,900 2,654,591 288,341 410,641 8,991 1,554,381 212,033 110,264 6,111 977,592 142,798 27,700 1,446 489,063 100,600

Outstan ding Outstan ding Outstan ding Outstan ding Outstan ding

Securi tized

Outstan ding

Securi tized

Outstan ding

Securi tized

1991:I

Securi tized

1996:I

Securi tized

2001:I

Securi tized

2007:IV

Securi tized

667

Home mortgages Multifamily residential mortgages Commercial mortgages Farm Mortgages Loans to business (C&I loans) Consumer credit Total mortgages Total loans

6 Securitization is a process of creating new financial instruments by pooling cash flows from a number of similar assets such as mortgages or credit card accounts and putting them into a separate legal entity [or special purpose vehicle, (SPV)] often with some additional implicit or explicit guarantee or extra collateral. Creating this separate SPV isolates the cash flow generating assets or collateral so that the security is not a general claim against the issuer, just against those assets. The pooling process results in a diversified portfolio of cash flows that can be further stripped and repackaged based on various characteristics (e.g., the prepayment behavior), thereby reducing the need to monitor each underlying payment stream. For a detailed discussion of the securitization process and the role of SPVs, see Gorton and Souleles (2005).

1986:I

The US economy has seen an enormous expansion of the securitization market.6 Table 1 presents the amount of loans outstanding and loans securitized for various loan categories over the sample period 1976:I to 2007:IV. Consider, for example, home mortgages. In 1976:I, the amount of securitized home mortgages was $27.7 billion. By the end of 2007, the total amount of securitized home mortgages grew 230 times, reaching $6.42 trillion. At the same time, the amount of home mortgages outstanding grew only 22 times, from $489 billion to $11,135 billion. In 1976, neither commercial mortgages, nor C&I loans, nor consumer credit were securitizable types of loans. The securitization market outpaced loan origination by ten times. By the end of 2007, the securitized loan volume totaled $652 billion of commercial mortgages, $80 billion of C&I loans, and $682 billion of consumer credit. Through the years, C&I loans remain the least liquid loan category due to their heterogeneity, which complicates pooling and pricing processes. Fig. 1 shows how the aggregate, economy-wide share of securitized loans in total loans outstanding has been changing over the years. The share of securitized home mortgages climbed from around 5% in 1976:I to almost 60% in 2007. In 2007:IV, 28% of the commercial mortgages outstanding, 3% of C&I loans, and 27% of consumer credit were securitized. On the aggregate level, securitization has dramatically expanded from 2.2% of the total loans outstanding securitized in 1976:I to around 40% in 2007:IV.

1981:I

2. The securitization market

1976:I

made banks vulnerable to the 2007 crisis when the market for securitized loans was disrupted. The remainder of the paper is organized as follows. Section 2 describes the structure and magnitude of the market for securitized loans as well as possible channels of its influence on bank operations. Section 3 describes data and sample selection. Section 4 presents the intuition and methodology behind the bank-specific index of liquidity of a bank loan portfolio (Sit). Section 5 presents the empirical tests and results for the hypothesis of substitutability between liquid funds and securitizable loans on banks’ balance sheets. Section 6 describes empirical evidence for the argument that securitization alleviates the sensitivity of bank lending to the cost of funds shocks. Section 7 concludes the paper.

Table 1 Economy-wide loans outstanding and securitized. The table presents the aggregate, economy-wide amounts of the loans outstanding and loans securitized for six loan categories over time period 1976:I to 2007:IV. The loan categories considered are: (1) home mortgages; (2) multifamily residential mortgages; (3) commercial mortgages; (4) consumer credit; (5) business loans [commercial and industrial (C&I) loans]; and (6) farm mortgages. Total mortgages represent the aggregate of home mortgages, multifamily residential mortgages, commercial mortgages, and farm mortgages. Total loans represent the aggregate of all six loan categories. All figures are in million dollars. The data are from the ‘‘Flow of Funds Accounts of the United States.’’ For the exact description of the methodology see Appendix A.

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

668

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

2.1. Origins of securitization Securitization evolved as early as the 1970s out of concern that, at the time, the existing thrift system could not finance the growing demand for housing. While the thrifts were not in trouble, the core concern lay in their inability to attract a sufficient amount of deposits at a reasonable cost and finance the baby boomer generation reaching home-buying age in the mid to late 1970s. The major contributors to the development of bank loan securitization have been the government-sponsored enterprises (GSEs) – the Federal National Mortgage Association (Fannie Mae or FNMA) and the Federal Home Loan Mortgage Corporation (Freddie Mac) – that were created by the US Congress to provide stability and ongoing assistance to the secondary market for residential mortgages and to promote access to mortgage credit and home ownership in the US. GSEs fostered securitization by being the largest buyers of mortgages in the US. Fannie Mae and Freddie Mac combined purchase almost one-half of all conventional single-family mortgage loans originated each year (Frame and White, 2005). More important, GSEs facilitate small bank access to the securitization market by standing by to purchase individual mortgages as well as mortgage pools. They have created an environment in which the ability to securitize mortgages is similar across banks of different size. The evolution of mortgage companies and various pulling agents performing functions similar to GSEs, but on a much smaller scale, has been crucial to attain a similar equality in the nonagency securitization markets. Nevertheless, large banks continue to have economies of scale in accessing other sectors of the securitization market in which there are few or no intermediaries willing to pool and securitize loans from multiple lenders (e.g., C&I loans securitization). 2.2. Regulation of the securitization market Over four decades of continued tremendous growth, up to the 2007 Crisis, the securitization market had not experienced any dramatic speed bumps. In fact, all significant regulatory changes that have been introduced since securitization conception were nothing but friendly and market enhancing. In the early 1980s, the securitization market was facing two important problems: convoluted and unfair tax treatment and a legal investment problem that prevented regulated investors from holding securitized assets. On October 3, 1984, Congress passed the Secondary Mortgage Market Enhancement Act (SMMEA) that preempted the state laws and allowed almost any investor to hold rated mortgage backed securities. SMMEA was followed by the 1986 Tax Reform Act that created a friendlier tax environment by introducing a new tax vehicle called the Real Estate Mortgage Investment Conduits (REMIC). SMMEA and REMICS opened the door for a significant flow of capital to the securitization market and by extension dramatically enhanced loan liquidity. In fact, these two acts created an environment that had not seen any regulatory intervention (apart from minor disclosure regulations) up until the end of 2001 when Enron collapsed.

In early 2002, the role that off-balance sheet vehicles played in hiding Enron’s financial liabilities led the regulators to reexamine the treatment of the securitization conduits. The securitization market reaction was swift. For example, the asset-backed commercial paper market stalled (Acharya and Schnabl, forthcoming). In January 2003, the Financial Accounting Standards Board (FASB) issued guidance that required consolidation of norecourse conduits on banks’ balance sheets under US GAAP (FASB 46). The industry considered it a real possibility that the guidance would result in a corresponding regulatory change. Such a ruling would have had a dramatic impact on banks’ balance sheets and, given the bank capital requirements, would have rendered the securitization of loans too costly to be considered a source of liquidity or capital. In July 2004, US bank regulators issued a ruling that allowed banks to leave the conduits off balance sheets and required them to hold capital against some conduit liquidity enhancements at only a 10% conversion factor, much lower than the capital required for on-balance sheet assets. The market responded by a vivid growth in the volume of new securities issued. As a robustness test, I conduct a difference-in-differences analysis in which I exploit the 1984– 1986 regulatory changes and 2002–2004 events as exogenous shocks to the depth of the securitization market. 2.3. Implications of securitization In light of the 2007 credit crisis, it is hard to overstate the importance of securitization in shaping banks’ operations. It provides banks with a new source of financing their investment opportunities. With securitization, banks can fund new loans by securitizing them (or other outstanding loans). It changes the traditional view on the deposit institutions as liquidity providers (Diamond and Dybvig, 1983; Holmstrom and Tirole, 1998; Kashyap, Rajan, and Stein, 2002) and transforms them to pure intermediaries between borrowers and capital markets. Loan portfolios that were considered to be too cumbersome and expensive to sell 25 to 30 years ago were becoming more and more liquid. There are other channels through which securitization significantly affect the nature of banking. Securitization provides an opportunity for banks to hold more diversified loan portfolios, thus protecting them against local economic shocks while at the same time making them more prone to the global economic slowdowns (Loutskina and Strahan, forthcoming). Although deregulation has eliminated most of the legal restraints on geographic segmentation, many banks continue to originate loans in the regions or industries in which they have a superior knowledge of market conditions. With securitization, these loans can be bundled with others, bought and sold all around the country. Money can flow from the regions with excess deposits to the regions with the unsatisfied loan demand. The benefits come at a cost. Alongside eroding banks’ core role of liquidity provider, securitization jeopardizes banks’ fundamental screening and monitoring roles (Loutskina and Strahan, forthcoming; Rajan, Seru, and

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

Vig, 2010; Piskorski, Seru, and Vig, forthcoming; Parlour and Plantin, 2008). Securitization provides means to mitigate the regulatory capital requirements by allowing banks to easily move assets off their balance sheets. All these factors together have significantly reshaped the way the deposit institutions do business. In this paper, I concentrate on the on-balance sheet liquidity and funding management implications of securitization. 3. Data and sample selection In this section I provide detailed description of data and sample selection. 3.1. Bank-level data Bank-level data come from the Federal Reserve’s Report of Condition and Income (‘‘call reports’’) submitted by insured banks each quarter. I compile a data set with quarterly income statements and balance sheet information for all reporting banks over the period 1976:I through 2007:IV. Appendix A contains the detailed description of the construction of the key series. When analyzing the data set, I first exclude all the bank-quarters with missing information on total assets, total loans, and liquid funds. I exclude banks in any quarter in which they go through a merger using bank mergers data from the Federal Reserve National Information Center. Specifically, I exclude the acquiring bank in the quarters before and after a merger. To prevent the possibility of outliers driving the results, I eliminate all bank-quarters with asset growth over the last quarter in excess of 50%, those with total loan growth exceeding 100%, those with total loans-to-asset ratio below 10%, and those with a share of credit card loans in the loan portfolio above 50%.7 The final data set contains 1,537,186 bank-quarters. To analyze differences in the securitization effects across banks of different size, I separate the sample into two groups: large banks and small banks. I measure the size of a bank as a log of the real total assets in which I adjust bank size using a consumer price index equal to 100 in 1980. I assign bank-quarter to the group of small banks if its real total assets are in the bottom 75% of the size distribution and to the group of large banks if its real total assets are in the top 5% of the size distribution. The variable of interest, the on-balance sheet (or traditional) liquidity (Bit), is computed following Kashyap and Stein (2000) as a share of the marketable securities and the federal funds sold in the total assets. Cash is not included in Bit as is it likely to reflect the required reserves and hence not easy to draw down. Table 2 presents summary statistics of various balance sheet items for the obtained sample. It reports the means 7

Credit card loans have experienced a significant increase in securitization over the last ten to 15 years. I, however, can account for credit card loan securitization only as part of the consumer credit securitization. Hence, I believe that my loan liquidity index does not carefully capture the degree of loan portfolio liquidity for banks heavily involved in the credit card business. Consequently, I exclude these banks from the considered sample to avoid possible distortions due to unobservable degrees of liquidity of the credit card accounts.

669

and medians for the full sample and for the subsamples of small and large banks. It also presents how the composition of banks’ balance sheets has changed over time. When comparing small and large banks, one can see that small banks tend to hold more liquid assets (34.6% versus 25.6% of total assets) and fewer loans (55.4% versus 61.75% of total assets) in their portfolios relative to large banks. This is consistent with small banks having more trouble raising external finance and thus needing a bigger liquidity buffer as a protection against cost of funds shocks. On the liability side, small banks are mostly financed by deposits (88% of total assets) and equity (9.7%), in contrast to large banks, which use both deposits and equity to a smaller extent (78.5% and 7.9%, correspondingly). When comparing statistics for bank-quarters in 1976 and in 2007, several patterns emerge on the asset side of banks’ balance sheets. First, over time, the level of onbalance sheet liquidity fell significantly, not only for small banks (from 34.4% to 26.5%), but also for large banks (from 28.1% to 24.5%). Second, the on-balance sheet liquidity differential between large and small banks decreased dramatically. This might be attributed to increasing availability over time of the external financing to small banks and the evolution of the securitization market. Today, there is less need for small banks to maintain thick liquidity buffers if they can easily obtain funds by securitizing their loan portfolios. Third, the share of loan portfolios in total assets increased for both small and large banks. Finally, there is a decrease in the share of business loans (C&I loans) in bank loan portfolios. This decrease might be caused by the development of the commercial paper market as well as the junk bonds market.

3.2. Monetary policy proxies To proxy the cost of external financing for banks, I use three different monetary policy indicators: the federal funds rate (Fed Funds); the difference between the rates paid on six-month prime-rated commercial papers and 180-day Treasury bills (Paper-bill); and the Strongin measure of monetary policy (Strongin).8 These indicators of monetary policy are constructed using time series data available from the Federal Reserve and are described in detail in Appendix A. Kashyap and Stein (2000) and Bernanke and Blinder (1992) provide a detailed discussion of these three proxies. All policy measures are transformed so that increases in their levels represent Fed tightening. They are also normalized to have the same standard deviation.

8 The paper-bill measure is based on Bernanke (1990) argument that increases in paper-bill index capture Fed tightening because banks cut loans and corporations are forced to substitute commercial paper for bank loans. Strongin measures the portion of nonborrowed reserve growth that is orthogonal to total reserve growth. Strongin (1995) argues that the Fed is constrained to meet total reserve demands in the short run (failure to do so would lead to wild swings in the funds rate), but it can effectively tighten policy by reducing nonborrowed reserves and forcing banks to borrow more from the discount window.

670

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

Table 2 Summary statistics. The table presents the means and medians for with-in category distribution of various balance sheet items (see definition in Appendix A) and the total number of bank-quarters in the full sample and in the subsamples of large (top 5% in size) and small (bottom 75% in size) banks. The data is from the Federal Reserve’s Report of Condition and Income for the period from 1976:I to 2003:IV and Flow of Funds Accounts of the United States database. I eliminate all bank-quarters with asset growth over the last quarter in excess of 50%, those with total loan growth exceeding 100%, those with total loans-to-asset ratio below 10%, and those with the share of credit card loans in the loan portfolio above 50%. 1976Q1–2007Q4 Full sample Mean Total assets (millions of 1983 dollars) Liquid assets (percent of total assets, Bit) Total loans (percent of total assets) Commerical and industrial loans (percent of total loans) Home mortgages (percent of total loans) Multifamily mortgages (percent of total loans) Farm mortgages (percent of total loans) Commercial mortgages (percent of total loans) Consumer loans (percent of total loans) Total deposits (percent of total assets) Net income (percent of total assets) Total equity (percent of total assets) Securitizability of loan portfolio (percent, Sit) Number of observations

Small banks

Median Mean Median

1976Q1–1980Q4 Large banks Mean

Median Mean Median

Median Mean Median

Large banks Mean

Median

3995.04 933.59

28.17 23.37

2921.47 857.76

36.51 33.41

6307.16 887.90

33.21

31.78

34.62 33.33

25.60

24.09

34.40 33.00

26.57

26.23

28.16 26.07

24.58

22.10

56.80

58.05

55.41 56.64

61.75

62.86

55.97 57.54

55.21

56.31

62.08 63.76

65.40

67.63

30.93

27.13

33.33 29.60

24.92

23.55

36.52 32.13

32.37

32.57

26.31 22.89

18.14

15.66

25.11

21.92

24.04 21.10

25.45

20.75

19.09 15.85

17.91

14.60

27.83 24.90

28.11

24.74

1.10

0.05

0.80

0.00

2.66

1.00

0.48

0.00

1.80

0.51

1.32

0.18

3.31

1.64

4.75

1.56

5.77

2.69

0.56

0.08

4.93

2.44

0.35

0.07

8.27

4.73

1.06

0.15

12.24

9.29

10.49 7.62

15.49

13.36

6.64

5.27

9.42

8.38

17.61 14.46

23.93

23.22

19.51

16.65

20.38 17.31

16.55

14.40

28.05 25.85

24.09

24.20

11.12 8.56

9.98

4.73

86.91

88.68

87.77 89.08

78.50

80.87

89.33 90.20

82.44

83.79

84.07 85.53

73.83

76.64

0.60

0.59

0.59

0.60

0.59

0.54

0.79

0.78

0.49

0.44

0.58

0.78

0.66

9.39

8.60

9.71

8.86

7.89

7.27

8.95

8.28

6.51

6.32

11.34 10.19

10.13

9.04

12.09

8.45

11.01 7.15

14.48

12.53

1.79

1.50

1.70

1.28

24.03 23.36

25.45

24.44

1,461,539

6  X



Mean

Small banks

31.18 26.72

1,096,155

73,076

In this paper I propose a new index of liquidity of a bank’s loan portfolio. The individual bank portfolio structure and economy-wide securitization are crucial factors to be considered in constructing the index. Consider two banks: Bank A and Bank B. Assume that Bank A holds 80% of its loan portfolio in home mortgages and 20% in C&I loans and that Bank B holds 20% of its loans in home mortgages and 80% in C&I loans. Because home mortgages have been more liquid than C&I loans over the years, Bank A faces fewer frictions in liquidating its loans than Bank B. Following this intuition, I construct an index to proxy each bank’s potential to securitize (sell) its loans (Sit) in a way that captures both the composition of a bank’s loan portfolio and the growth in the depth of the securitization market over time. The proposed measure is computed as

j¼1

Large banks

255.20 37.36

4. Measuring bank-level loan portfolio liquidity (Sit)

Sit ¼

Small banks

2001Q1–2007Q4

Economy-wide Securitized Loans of Type j at Time t Economy-wide Total Loans Outstanding of Type j at Time t Share of Type j Loans

in Bank i Portfolio at Time t



! :

ð1Þ

234,325

9,889

0.53

135,042

16,150

The index can be thought of as a weighted average of the potential to securitize loans of a given type (based on marketwide averages), in which the weights reflect the composition of an individual bank’s loan portfolio. Thus, market trends generate time variation in the index, and the differences in bank loan portfolios generate variation across institutions. I construct this measure by breaking down a bank loan portfolio into six categories: (1) home mortgages, (2) multifamily residential mortgages, (3) commercial mortgages, (4) consumer credit, (5) business loans not secured by real estate (commercial and industrial loans), and (6) farm mortgages. The index can be computed using market-level data from the US Flow of Funds and individual bank-level data on loans from the Report of Condition and Income. Appendix A discusses in detail the construction of the economy-wide time series components of Eq. (1), and provides exact definitions of bank loan portfolio components. Table 2 presents average Sit for the full sample of bankquarters, large and small bank subsamples, and the beginning and ending points of the considered sample. The average bank loan liquidity index in my sample is 12.09%. Over the sample period of 1976 to 2007, the average Sit has increased from about 1.8% to roughly 24%. Without a doubt, the increase in the loan liquidity is

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

economically significant and calls for careful analysis of the securitization effects on banks’ operations.

671

predictable. Finally, large banks are more likely to have access to significant internal capital markets (e.g., Campello, 2002; Houston, James, and Marcus, 1997).

5. Securitization and on-balance sheet liquidity 5.1. Univariate tests Liquidity creation is a core role of banking institutions and is a cornerstone of the majority of the financial intermediation theories. By definition, securitization is a process of creating liquid financial instruments out of assets that could be too cumbersome or expensive to sell individually. Such a transformation calls for an evaluation of liquid loans’ role in financing the liquidity needs of banks. If a bank can securitize the existing loans as easily as it can convert liquid assets into cash, such a bank is unlikely to hold liquid assets (assuming that liquid securities continue to offer less return than loan origination). When facing a new lending opportunity or deposit withdrawal, this bank converts a necessary amount of the existing loans into cash. In the absence of the market for securitized loans, a bank needs to hold enough liquid funds to provide liquidity to borrowers and depositors on demand. This argument suggests that banks with liquid loan portfolios are likely to contract their liquid securities holding (i.e., consider liquid securities and liquid loans as substitutes). Panel A of Fig. 2 graphically presents the relation between average loan liquidity Sit and average on-balance sheet liquidity Bit over time. The evolution of the securitization market coincides with decreasing amounts of liquid funds on bank balance sheets. The time series correlation between average Sit and average Bit in my sample is 0.61, significant at the 1% level, indicating substitutability between the liquidity sources. What about large banks and small banks? Does the expanding securitization market affect those two groups differently? There exists a dynamic trade-off. The process of securitization involves pooling a diversified loan portfolio and packaging securities and, hence, requires significant services from GSEs or other pooling agencies (e.g., investment banks). Large banks tend to have much tighter relations with investment banks, which provide them a direct route to the derivatives markets. These relations are likely to decrease the costs of securitization as well as the time it takes to securitize loans for large banks relative to small banks. While small banks are forced to attract other banks or pooling agencies to the securitization process, large banks can issue securities backed by a pool of their own loans. Even though large banks have significant advantages accessing the securitization market, the marginal benefit of securitization as a liquidity substitute might be smaller for large banks because they tend to be more efficient in managing their liquid funds even in the absence of securitization. Large banks can afford to maintain less liquid assets on their balance sheet for a number of reasons.9 First, they usually face less severe principal-agent problems while trying to raise the uninsured funds [e.g., certificates of deposits (CDs)] compared with small banks. Second, large deposit institutions have a more diversified depositor base that makes the deposit withdrawals less volatile and more 9 In Kashyap and Stein (2000) and Jayaratne and Morgan (2000) similar arguments are empirically tested in.

Table 3 presents a cross-sectional analysis of bank loan portfolio liquidity (Sit), levels of on-balance sheet liquidity (Bit), and bank size for various subperiods of the sample. In Panel A of Table 3, I separate the sample of bank-quarters into four quartiles based on the distribution of the loan liquidity measure Sit and compute the average on-balance sheet liquidity Bit in each quartile. The results suggest that banks with lower liquidity of loan portfolios have more liquid funds on their balance sheets. The difference in bank liquidity Bit between banks in the least liquid loans quartile and banks in the most liquid loans quartile is significant at the 1% level for the full sample as well as its various subsamples. Panel B of Table 3 evaluates the relation between the banks’ size and levels of on-balance sheet liquidity (loan liquidity) in the cross-sectional framework. It presents the average liquidity measures across size quartiles for the full sample of bank-quarters as well as for various subsamples. I find that, over the years, large banks tend to maintain lower levels of on-balance sheet liquidity than small banks. Panel B of Fig. 2 graphically illustrates this relation. Banks in the highest size quartile have around 6.6% less total assets held as liquid securities compared with lowest size quartile banks. This is equivalent to roughly 18% less liquid funds in total assets for banks in the largest size quartile relative to the banks in the lowest size quartile. The results are economically and statistically significant at the 1% level. The evidence is consistent with the argument that large banks are more efficient in managing their liquid funds. 5.2. Instrumental variable analysis The level of liquidity maintained by banks can be affected by numerous economy-wide factors such as deregulation, consolidation in the banking industry, and technological advancements. I next conduct a more rigorous regression analysis looking to evaluate the relation between bank liquid fund holdings (Bit) and loan portfolio liquidity (Sit) and how the strength of this relation changes with bank size (the log of real total assets) or reputation (ratio of letters of credit to total assets). Additional control variables include access to additional internal sources of funds (net income as percentage of total assets, affiliation with a bank holding company, cost of deposit), ability to raise funds externally [bank size or reputation (ratio of letters of credit to total assets)], the level of capitalization (equity capital as a percentage of total assets, share of deposit financing), and the loan portfolio performance (share of nonperforming loans in the total loan portfolio).10 I also include time dummies for 10 The share of nonperforming loans is not available for about 200,000 observations in my sample from 1976 to 1983. The results hold, irrespective of controlling for the nonperforming loans, and produce coefficients of similar magnitudes.

672

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

Table 3 Traditional liquidity, loan portfolio liquidity, and bank’s size: univariate analysis. The table presents the univariate analysis of the relations between banks’ size (measured by log real total assets), level of on-balance sheet liquidity (Bit), and the bank-specific loan portfolio liquidity (Sit). Panel A presents the average loan portfolio liquidity for four liquidity quartiles and the loan portfolio securitizability index differential between more liquid and less liquid banks. Panel B presents the average on-balance sheet liquidity for four size quartiles as well as the liquidity differential between banks in the largest and smallest size quartiles. The averages are computed for the full sample 1976:I to 2007:IV, as well as for various sub-periods of the sample. The standard errors for the estimates of the averages for each group are reported. The t-statistics for differences in means are reported in brackets. *, **, and *** correspond to below 10%, 5%, and 1% significance of t-statistics, respectively. Panel A: Average liquidity Bit across loan liquidity Sit quartiles (%) Loan liquidity quartiles 1976:I–2007:IV 1980:IV

1985:IV

1990:IV

1995:IV

2000:IV

2005:IV

35.1 (0.130) 35.0 (0.161) 31.3 (0.956) 29.1 (0.897) 5.96 [6.58]***

34.0 (0.246) 34.1 (0.172) 36.2 (0.334) 27.4 (1.850) 6.60 [3.54]***

38.9 (0.735) 37.9 (0.275) 35.0 (0.188) 31.1 (0.382) 7.80 [9.42]***

34.6 (0.205) 36.4 (0.458) 35.6 (0.216) 33.4 (0.213) 1.20 [4.06]***

30.6 (0.639) 31.6 (0.857) 29.2 (0.291) 28.6 (0.189) 2.00 [3.00]***

29.9 (0.460) 29.2 (1.060) 25.5 (0.321) 26.5 (0.214) 3.40 [6.70]***

Panel B: Average liquidity measure Bit across size quartiles (%) Size quartiles 1976:I–2007:IV 1980:IV

1985:IV

1990:IV

1995:IV

2000:IV

2005:IV

36.1 (0.243) 35.4 (0.252) 35.4 (0.258) 29.8 (0.276) 6.30 [13.32]*** 282,808

38.5 (0.271) 37.1 (0.274) 36.5 (0.283) 29.2 (0.282) 9.30 [21.84]*** 263,784

37.6 (0.309) 35.8 (0.285) 34.9 (0.275) 31.3 (0.274) 6.30 [18.00]*** 224,195

32.6 (0.366) 30.1 (0.317) 28.0 (0.295) 26.5 (0.275) 6.10 [14.16]*** 186,715

31.5 (0.462) 29.0 (0.391) 25.8 (0.328) 23.3 (0.275) 8.20 [3.40]*** 99,861

Illiquid Q1 loans Q2 Q3 Liquid Q4 loans Q1–Q4

Small Q1 Q2 Q3 Large Q4 Q1–Q4 Observations

34.2 (0.026) 35.1 (0.026) 33.4 (0.026) 30.3 (0.026) 3.90 [107.30]***

36.0 (0.026) 34.5 (0.026) 33.4 (0.026) 29.1 (0.026) 6.90 [191.33]*** 1,344,696

37.4 (0.176) 35.3 (0.198) 34.3 (0.213) 29.8 (0.234) 7.60 [30.26]*** 287,333

each quarter to account for changes in the regulation, business cycle effects, and other trends. Under the substitutability hypotheses, the coefficient of loan liquidity is expected to be negative. Because large banks and more reputable banks experience fewer information frictions in accessing capital markets and hence can maintain less liquid funds, I anticipate the coefficients of bank size and reputation also to be negative. If the marginal benefit of the loan liquidity on traditional liquidity level decreases with increasing bank size or reputation, I anticipate the coefficient on Sit  Size (or Sit  Letters of Credit) to be positive. Finally, given that the securitization originated out of the concern for the deposit base to be insufficient or too costly to fund housing demands, I evaluate whether banks with higher costs of deposits or more unstable deposit base enjoy larger marginal benefit from potential loan securitization in managing liquidity. I expect the coefficient on Sit  Cost of Deposits (Sit  Share of Core Deposits) to be negative (positive).11 A potential endogeneity exists between on-balance sheet liquidity and banks’ loan portfolio liquidity due to the ability of banks’ management to choose the onbalance sheet liquidity level and the structure of a bank loan portfolio simultaneously. Specifically, there might be

11 Core (or demand) deposits are sticky even in the face of competition, thus leading to more predictable and smaller in magnitude liquidity demands from depositors. As a result, securitization might be less important for high-demand-deposit banking institutions as a liquidity management tool.

a positive bias in the relation between Bit and Sit because banks that prefer more liquid assets are likely to have both more liquid funds and more liquid loan portfolio (which can be achieved by, e.g., issuing more mortgages and fewer C&I loans). To adjust for this endogeneity due to managerial discretion, I adopt three approaches: a lagged independent variable and instrumental variable analysis (presented in this section), and a difference-indifferences analysis around regulatory discontinuities (in Section 5.3). Table 4 presents the ordinary regression analysis with lagged independent bank-specific variables (Panel A) and the instrumental variable approach (Panel B) in which the instrumental for Sit equals: Instrumentit ¼ 6  X j¼1

Avgi

Economy-wide Securitized Loans of Type j at Time t Economy-wide Total Loans Outstanding of Type j at Time t Share of Type j Loans in Bank i Portfolio at Time t



!

ð2Þ

In constructing this instrument I use a fixed portfolio structure computed for each bank as the average portfolio structure over the first four quarters available in my sample.12 This fixed loan portfolio structure captures an 12 Similarly, I instrument the interaction terms between loan liquidity(Sit) and Size(Sit  Letters of Creditit, Sit  Cost of Deposits(Sit  Share of Core Deposits)) using instruments from Eq. (2) and its interaction with Size and other variables correspondingly.

Table 4 Traditional liquidity, loan securitization, and bank’s size: multivariate analysis. The table reports the results of the regression analysis in which the dependent variable is the level of on-balance sheet liquidity (Bit). The independent variables are loan liquidity Sit  1, the log of real total assets, letters of credit to total assets, bank holding company dummy, equity capital to total assets, share and cost of deposit financing, net income to total assets, and unused loan commitments to total assets. The time-specific fixed effects accommodate the effects of changes in the regulation and other economic conditions. Specification 1–5 present the results of ordinary regressions specifications 5–10 present the results of the instrumental variable analysis. The instrumental variable for Sit  1 is constructed using Eq. (2) where instead of variable over time bank loan portfolio structure I use the average loan portfolio structure of an individual bank over the first four quarters available in my sample. Sit  1  Sizeit  1 and Sit  1  Letters of Creditit  1 are also instrumented. Robust t-statistics are reported in parentheses. *, **, and *** correspond to below 10%, 5%, and 1% significance of t-statistics, respectively.

(1) Liquidity of loan portfolio: Sit  1

 0.0524*** (3.997)

Sizeit  1  Sit  1

(2)  0.082 (1.345) 0.00269 (0.50)

Letters of Creditit  1  Sit  1

(3)  0.054*** (4.034)

Instrumental variable regressions (4)

 0.0245 (1.51)

Bank holding company dummy Equity Capitalit  1 (percent in total assets) Share of Deposit Financingit  1 (percent in total assets) Cost of Depositsit  1 Net Incomeit  1 (percent in total assets) Time dummies Number of observations R-squared

 0.249*** (13.12)

(7)  17.14*** (16.23) 1.507*** (16.20)

(8)  0.257*** (13.17)

 0.386** (2.30)

 0.005*** (6.466)  2.368*** (22.36)  0.0163*** (9.875) 0.995*** (27.47) 0.0709*** (3.73)  0.022*** (3.503) 4.436*** (23.07) + 1,309,131 0.152

 0.0053*** (5.619)  2.362*** (22.19)  0.0162*** (9.836) 0.995*** (27.44) 0.0715*** (3.75)  0.022*** (3.504) 4.432*** (23.04) + 1,309,131 0.152

 0.005*** (6.465)  2.436*** (17.09)  0.0163*** (9.882) 0.995*** (27.46) 0.0703*** (3.70)  0.022*** (3.504) 4.425*** (22.98) + 1,309,131 0.152

(9) 0.0681 (0.47)

(10)  0.336*** (16.53)

3.551*** (3.74)

Transaction Depositsit  1  Sit  1

Letters of Creditit  1

 0.199*** (12.07)

(6)

0.766 (0.82)

Cost of Depositsit  1  Sit  1

Sizeit  1 (log of real total assets)

(5)

 0.00505*** (6.595)  2.347*** (22.16)  0.0166*** (10.06) 0.969*** (25.74) 0.0781*** (4.16)  0.00591*** (2.840) 0.530*** (24.57) + 1,309,131 0.159

 5.208** (2.158) 0.845*** (16.46)  0.00368*** (4.05)  2.409*** (20.13)  0.0129*** (6.03) 1.080*** (23.80)  0.0051 (0.24)  0.0118*** (3.73) 0.346*** (13.04) + 923,104 0.166

 0.003*** (3.817)  2.578*** (24.17)  0.0209*** (12.61) 0.984*** (27.36) 0.0778*** (4.13)  0.0214*** (3.894) 4.327*** (22.69) + 1,309,131 0.157

 0.184*** (16.66) 0.106 (0.57)  0.00705*** (4.077) 0.848*** (22.89) 0.409*** (14.38)  0.00508 (0.918) 1.608*** (6.36) + 1,309,131 0.159

 0.003*** (3.810)  2.894*** (19.30)  0.0210*** (12.64) 0.983*** (27.33) 0.0747*** (3.97)  0.0214*** (3.895) 4.274*** (22.38) + 1,309,131 0.157

 0.00304*** (3.930)  2.623*** (23.79)  0.0220*** (13.11) 0.979*** (25.47) 0.0758*** (4.03) 0.0386* (1.92) 0.492*** (20.65) + 1,309,131 0.166

0.727*** (12.92)  0.00196** (2.13)  2.562*** (21.17)  0.0166*** (7.74) 1.070*** (23.78) 0.00756 (0.36)  0.0117*** (3.52) 0.341*** (12.85) + 923,104 0.166

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

Ordinary regression

673

674

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

individual bank’s loan specialization and at the same time eliminates the source of endogeneity (managerial discretion) as well as the effect of securitization on the composition of a bank’s loan portfolio. The constructed instrumental variable captures the changes in the loan liquidity index for a bank that does not change its loan portfolio structure in response to the changing depth of the securitization market.13 Because I include time fixed effects in the instrumental variable regressions, the coefficient of Sit is driven by within a quarter variation in the instrumental variable. The results are consistent across Panel A and Panel B of Table 4. I find that the level of on-balance sheet liquidity is negatively correlated with the widening of the securitization market. The coefficients of Sit have significantly higher magnitude in Panel B than in Panel A, which is consistent with the initial expectation of a positive bias due to endogeneity. The evidence suggests that, as bank loan portfolio liquidity increases by 1%, the level of on-balance sheet liquidity maintained by a bank decreases on average by around 28 basis points. As I have shown in Table 2, the average loan portfolio liquidity index increased from around 1.5% in the 1970s to around 25% over the 2000– 2007 period. This corresponds to around a 6.6% decrease in the total assets held as liquid securities, which is in turn equivalent to roughly a 68% decrease in the amount of liquid funds per dollar of equity held by banks. The average level of on-balance sheet liquidity held by banks decreased by 12% from around 36% in 1976 to around 24% in 2007. Thus, securitization is responsible for roughly one-half of this decline in banks’ liquid fund holdings. One can look at these results from a different perspective. Because the liquid funds and loans are two dominant components of the asset side of a bank balance sheet, a 6.1% decrease in the share of liquid funds in total assets is likely to lead to an increase in the share of the loan portfolio of a similar magnitude. The interaction coefficients in Table 4 also indicate that banks with easier access to external funds (larger, more reputable, or those enjoying cheaper deposits) and banks with more stable and predictable liquidity demands (high core deposits banks) draw smaller marginal benefits of securitization in managing liquidity. Securitization is likely to be responsible for a larger decline in liquid securities held by small banks (from 33% in the 1970s to 26% in 2000–2007) relative to the change observable for large banks (from 27% in the 1970s to 24% in 2000–2003). Securitization, thus, bridges the gap in liquidity levels between large and small banks and

potentially eliminates the competitive advantage of banks with less expensive deposits. 5.3. Regulatory discontinuity analysis The second identification strategy is centered around the set of exogenous regulatory and macroeconomic shocks altering the ability of the securitization market to absorb new loans. Specifically I evaluate whether banks with more liquid loan portfolios behaved differently around two sets of exogenous regulatory events discussed in detail in Section 2. First, I evaluate the changes in banks’ liquid fund holdings around the exogenous SMMEA (October 1984) and REMICS (1986) that dramatically enhanced the ability of institutional investors to acquire securitized loans. I exploit the regression specification similar to one used in Table 4 and use 1983:I to 1988:II subsamples surrounding the event dates: Bit ¼ j þ aSit1 þ a1984 Sit1 D1984 þ a1986 Sit1 D1986 þ Bank Controlsit1 þ gt þ eit

Bit ¼ j þ aSit1 þ a2002 Sit1 D2002 þ a2004 Sit1 D2004 þ Bank Controlsit1 þ gt þ eit , 13 The average loan portfolio structure for the first four bankquarters available for each bank in my sample alleviates the effect of securitization on a bank loan portfolio composition. A bank loan portfolio structure is less likely to be affected by securitization in early bank-quarters than in recent years. For robustness, I also construct the instrumental variable using the average bank portfolio composition over all available quarters for each individual bank. The results of the instrumental variable regressions in this case are similar to those presented above.

ð3Þ

where D1984 (D1986) is a dummy variable equal to one for every quarter after and including 1984:IV (1986:I) and zero otherwise. The coefficients a1984 and a1986 capture the increase in the marginal benefit of securitization in allowing banks to decrease their liquid fund holdings. Both a1984 and a1986 are expected to be negative. I allow for the impact of other bank characteristics (bank size, income, deposit base and cost, letters of credit, etc.) to change upon the change in regulation by interacting all the bank-specific control variables with D1984 and D1986. Table 5, Panel A, presents only the coefficients of interest: a, a1984, and a1986. Given the endogeneity between traditional and loan liquidity, only the coefficient on the interaction term with the exogenous event dummies (D1984 and D1986) are identified. The negative values of both coefficients indicate that both regulatory acts enhanced the role of the securitization market in bank liquidity management. The magnitudes of the coefficients are comparable or even larger than those reported in Table 4, leading me to conclude that the change was not only statistically but also economically significant. Finally, the results cannot be explained by any bank’s local economic conditions captured by the bank-specific fixed effects accounted for in column 2 of Table 5. Panel B of Table 5 presents a similar difference-indifferences analysis centered around the Enron collapse in late 2001 and subsequent regulation of the off-balance-sheet vehicles. Using the 2000:I to 2006:IV subsample, I evaluate the results of the following regression specification: ð4Þ

where D2002 (D2004) is a dummy variable equal to one for every quarter after and including 2002:I (2004:III) and zero otherwise. The collapse of Enron led the banking industry to expect regulation requiring additional (and potentially significant) capital requirements to be posted even for loans sold without recourse, thus stalling the market (see Fig. 1). I expect a2002 to be positive, indicating the decline in the role of securitization as a liquidity management tool.

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

675

Table 5 Traditional liquidity and loan securitization: evidence from regulatory changes. The table reports the results of the regression analysis in which the dependent variable is the level of on-balance sheet liquidity (Bit) and the core variable of interest (loan liquidity Sit  1) is interacted with a set of dummies corresponding to a set of exogenous events. D1984 (D1986) is a dummy variable equal to one for every quarter after and including 1984:IV (1986:I) and zero otherwise. D2002(D2004) is defined similarly. The table reports only the coefficients of interest and omits an extensive set of control variables. Bank specific controls, namely, the log of real total assets, letters of credit to total assets, bank holding company dummy, equity capital to total assets, share and cost of deposit financing, net income to total assets, nonperforming loans to total loans, and unused loan commitments to total assets, are interacted with the dummies. I further control for the time- and bank-specific fixed effects. Panel A presents the impact of the Secondary Mortgage Market Enhancement Act and The Real Estate Mortgage Investment Conduits on the role of securitization in bank liquidity management. Panel B analyses the impact of the Enron collapse and the subsequent 2004 regulation. Robust t-statistics are reported in parentheses. *, **, and *** correspond to below 10%, 5%, and 1% significance of t-statistics, respectively. Panel A: Subsample period 1983:I to 1988:II Liquidity of loan portfolio Sit  1 Liquidity of loan portfolio Sit  1  D1984 Liquidity of loan portfolio Sit  1  D1986 Number of observations Adjusted R2 Panel B: Subsample period 2000:I to 2006:IV Liquidity of loan portfolio Sit  1 Liquidity of loan portfolio Sit  1  D2002 Liquidity of loan portfolio Sit  1  D2004 Number of observations Adjusted R2 Other bank controls interacted with event dummies Quarter dummies Bank dummies

The subsequent regulation of 2004 not only eased the uncertainty but also did not introduce any dramatic changes in capital requirements for sold loans, thus reviving the securitization market (a2004 o0). The results presented in Panel B of Table 5 confirm the expectations. The direct coefficient on Sit 1 is still not identified. a2002 is negative and a2004 is positive, indicating that the Enron collapse led the banking industry to be skeptical that the securitization market could be considered a source of liquidity, but the 2004 regulation restored their confidence.14 Overall, the results support the substitutability hypothesis between liquid securities and liquid loans held on banks’ balance sheets.

6. The effect of securitization on banks lending under funding shocks The presence of a large buffer of liquid funds (Kashyap and Stein, 2000) or other external sources of capital (Campello, 2002; Ashcraft, 2006) insulates banks from the cost of funds shocks. Securitization not only acts as a substitute for liquidity on banks’ balance sheets but also offers an additional mechanism to finance loans in the face 14 One might argue that all three regulatory acts of 1984, 1986, and 2004 were somewhat anticipated. Such potential anticipation is unlikely to be relevant given the relatively short-term nature of banks liquidity management or relative immediacy of the banks’ lending decisions. If relevant, it would bias the results against me. In addition, the Enron events are truly exogenous to both the banking and financial system and unrelated to any macroeconomic events in general and securitization market in particular.

0.515*** (12.91)  0.413*** (13.27)  0.352*** (13.96) 245,258 0.166

0.300*** (19.79)  0.180*** (16.19)  0.224*** (29.48) 245,258 0.861

 0.211*** (9.43) 0.116*** (6.264)  0.140*** (7.98) 208261 0.173

 0.0367*** (6.17) 0.108*** (26.93)  0.0718*** (15.85) 208261 0.869

Yes Yes No

Yes Yes Yes

of restricted availability of external financing. Apart from cutting back on lending and draining down liquid funds in the face of financial constraints, a bank can securitize existing loans, thus obtaining funds for new lending opportunities. It can also finance new loans by tapping into the securities market (securitizing issued loans immediately). In this section I empirically evaluate this issue exploiting the Federal Reserve’s ability to affect bank costs of funds via open market operations (the bank lending channel). The main argument behind the lending channel of monetary policy transmission is that, by selling bonds in the open market, the Federal Reserve drains the reserves of the depositary institutions, thus causing a reduction in the availability of insured deposits—the cheapest source of the loanable funds for banks. It is not optimal for banks to completely offset this decline in deposits by borrowing directly from economic agents using the uninsured financing instruments.15 Consequently, in the past, a bank facing tightened monetary policy would reduce lending in response to an increase in the marginal costs of raising deposits. With securitization enhancing banks’ liquidity and financing sources, a bank with a more liquid loan portfolio should

15 The uninsured financing instruments are not free from the traditional principal-agent problem and, hence, require the additional risk premium. CDs in excess of $100,000, for example, are not protected by the deposit insurance and, therefore, carry more risk as well as the necessity for monitoring by lenders. For detailed discussions of the sufficient conditions for the existence of the bank lending channel of the monetary policy see Bernanke and Blinder (1992), Bernanke and Gertler (1995), and Kashyap and Stein (1994).

676

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

experience a smaller contraction in its lending activity under a restricted availability of external finance than a bank with lower loan liquidity.16 Furthermore, because large banks have a competitive advantage in their ability to access the securitization market (see discussion in Section 5.1), the effect of securitization on bank lending activity should be more pronounced for large banks than for small banks. My strategy of analyzing the relation between the amount of liquid funds maintained by banks, their lending behavior, and loan liquidity in the framework of bank lending responses to the external funding shock has a number of advantages. First, it allows to adjust for potential endogeneity. Because the Federal Reserve induced funding shocks are exogenous to financial intermediaries’ decisions, the cross-sectional differences in the relation between financial institutions’ investments (loans) and their ability to securitize them (sell them) can be isolated using monetary policy as a state variable. Second, it allows to test whether securitization offers an additional mechanism to finance loans in the face of central bank tightening, thereby potentially weakening the link from monetary policy to loan supply. 6.1. Econometric specifications In choosing the regression specification, I start from socalled univariate one-step regression specification similar to Kashyap and Stein (2000) and Ehrmann, Gambacorta, Martinez-Pages, Sevestre, and Worms (2001). I regress the log real loan growth (DlogðLit Þ) against: four lags of itself; five lags of changes in a monetary policy indicator (DMt); linear time trend; bank-specific fixed effects; liquidity of a bank’s balance sheet (Bit 1), as well as cross effect of liquidity and changes in the monetary policy indicator17

DlogðLit Þ ¼ gi þ

4 X

Following Ehrmann, Gambacorta, Martinez-Pages, Sevestre, and Worms (2001), I adopt the bank-specific fixed effects since the variation in the level of liquidity, as well as the composition of a banks’ balance sheet, growth of the loan portfolio, etc., might be affected by the banks’ internal characteristics such as clientele base, management team, and availability of the lending opportunities (e.g., individual home mortgages versus business loans). The control variables include quarter dummies; bank holding company dummy; size of the bank (log real total assets); bank capitalization as percentage of total assets; profitability as percentage of total assets; share of deposit in bank financing; cost of deposits; and the interaction of these variables with monetary policy indicator Mt. An increase in the level of Mt corresponds to an increase in the costs of banks’ traditional financing sources (a monetary tightening). The traditional theory of the lending channel of monetary policy transmission argues that a contractionary monetary shock drains banks’ insured deposits and, hence, causes a decrease in banks’ lending volumes. Thus, the sum of m’s should be negative. Following Kashyap and Stein (2000), the availability of liquid funds reduces banks’ loan growth sensitivity toward a positive cost of funds shock. Thus, the sum of b’s should be positive. To capture the securitization market influence on a bank’s loan growth, I augment the set of the independent variables in the basic univariate regression specification Eq. (5) with the bank-specific index of a loan portfolio liquidity Sit proposed in this paper and five lags of Sit  1DMt  j ðj ¼ 0,4Þ.

DlogðLit Þ ¼ gi þ

4 X

aj DlogðLitj Þ þ Y0 Timet

j¼1

0

aj DlogðLitj Þ þ Y0 Timet

j¼1

0

þ Bit1 @b þ

4 X

þ Bit1 @b þ

1

0

bj DMtj A þ Other Controls þ eit

j¼0

þ Sit1 @c þ

4 X j¼0 4 X

1

bj DMtj A 1

xj DMtj A þOther Controls þ eit

j¼0

ð5Þ

16 The existing literature shows that the availability of additional internal and external sources of funds partially alleviates the effect of funding shocks on the supply of loans. Kashyap and Stein (2000) find that more liquid banks are less susceptible to monetary authority moves than less liquid ones. Campello (2002) shows that internal capital markets also help shield banks from the impact of increase in the costs of funds. Because securitization provides banks with an additional source of financing and liquidity, my argument is in tune with this literature. 17 In the dynamic models for panel data which contain the individual specific fixed effects lagged dependent variables become nonexogeneous if the sample has small time dimension (T). Arellano and Bond (1991) and Anderson and Hsiao (1982) propose the solution for this problem by using generalized method of moments estimation procedures. However, the literature considers this problem to be present only in samples with time-series number of periods below 15. Because my sample contains bank-quarters from 112 periods and I restrict each deposit institution to have at least 20 quarters of data present (84% of the deposit institutions in my sample are present for more than 40 quarters), I do not use the GMM procedures proposed by Arellano and Bond (1991) and Andersen and Hsiao (1982).

ð6Þ A bank with a higher level of loan liquidity should have smaller contraction in lending activity under a positive cost of funds shock than a bank with lower loan portfolio liquidity. Hence, I expect the sum of x’s to be positive. While I focus on the question of how a bank’s decisions on the asset side of the balance sheet affect its loan growth, other mechanisms can generate a similar effect on bank lending. Specifically, a similar effect might be generated by a bank’s inability to fulfill capital adequacy requirements.18 To disentangle this inadequate capitalization

18 A number of papers [see, e.g., Diamond and Rajan (2000), Hubbard, Kuttner, and Palia (2002), and Sharpe (1995)] argue that insufficient bank equity capital can be one of the restricting forces behind banks’ lending activity. According to this story, monetary tightening simply raises rates and suppresses economic activity, thus causing banks to experience loan losses, and, hence, reduction in capital. This, in turn, forces weaker, more capital constraint banks to cut back on new lending.

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

alternative, I implement a so-called bivariate regression analysis in which I add the log real gross domestic product (GDP) growth to the set of the independent variables (Kashyap and Stein, 2000). Adjusting for the GDP growth also allows me to control for loan demand differences across quarters

DlogðLit Þ ¼ gi þ

4 X

aj DlogðLitj Þ þ Y0 Timet

j¼1

0

þBit1 @b þ 0 þSit1 @c þ

4 X

1

bj DMtj A

j¼0 4 X

1

xj DMtj A

j¼0

þ

4 X

DlogGDPtj ðy1j þ y2j Bit1 þ y3j Sit1 Þ

j¼0

þOther Controls þ eit

ð7Þ

Finally, to analyze whether the groups of large and small banks exhibit different relation between loan growth and on-balance sheet liquidity, and between loan growth and loan liquidity sensitivity under a cost of funds shock. I estimate Eqs. (6) and (7) for large and small bank subsamples. The standard errors for each set of the ‘‘difference coefficients’’ are estimated via SUR system using either univariate [Eq. (6)] or bivariate [Eq. (7)] specifications. 6.2. On-balance sheet loans versus bank lending activity The empirical predictions relate the loan liquidity and on-balance sheet liquidity to banks’ lending activity or, in other words, banks loan origination. I cannot observe the loan origination fully. Some loans are being sold immediately upon issuance and find no reflection on banks’ balance sheets. The balance sheet loan volumes could be significantly smaller than the actual loan volumes extended to the economic agents by a bank. As a result, using the on-balance sheet loan growth as a proxy for bank lending presents a problem. To alleviate this problem, I consider the balance sheet loan growth not only for total loans but also for C&I loans. Because C&I loans are still difficult to securitize and they are last to be sold by a depositary institution when it faces a funding constraint, the on-balance sheet C&I loan growth better reflects the actual bank lending activity. I anticipate that the ability of securitization to alleviate the sensitivity of the loan portfolios to the availability of the traditional sources of financing should be more pronounced for least liquid C&I balance sheet loan volumes (that have smaller measurement error) and less pronounced for balance sheet volumes of total loans that contain first-to-be-sold mortgages. The magnitudes of the securitization effect on C&I loans are likely to reveal the actual effect securitization has on banks’ ability to supply credit. 6.3. Empirical tests and results Table 6 presents regression analysis of the total loan growth. The table gives a compact overview of the various regression estimations. Panel A presents the estimates of

677

the univariate specification Eq. (6), and Panel B presents the results for the bivariate specification Eq. (7). Because both specifications contain bank-specific fixed effects, I require that each bank has at least 20 observations in my sample to ensure correct estimation of these parameters.19 For each regression only the sums of the coefficients bj and xj are reported. Along with the different regression specifications, I consider three monetary policy indicators: Fed funds rate, Paper-bill, and Strongin. In addition, I test whether the considered sets of coefficients are different across the subsamples of small and large banks. Both Panel A and Panel B show that the total loan growth is positively affected by the loan liquidity and the availability of liquid assets in the presence of a cost of funds shock. All specifications (univariate and bivariate) combined with various monetary indicators unanimously agree on the directions of the effects of the considered independent variables: DMt jBit 1 and DMt jSit 1. The sum of coefficients on DMtSit 1 is positive in all specifications consistent with securitization providing an additional source of funding for banks facing a funding constraint and allowing lending to expand.20 Consistent with Kashyap and Stein (2000), I also find that a thicker liquidity buffer allows banks to maintain more loans on their balance sheet upon a monetary contraction: The sum of coefficients of DMt jBit 1 is positive. These results are statistically significant at below a 1% level in most regression specifications. Table 6 shows that the full sample results are mostly driven by the group of small banks. Consistent with large banks being able to access other sources of external finance, the sum of coefficients of DMt  jBit  1 is insignificant for the subsample of large banks in most regression specifications indicating a negligible role that the onbalance sheet liquidity plays as a source of funds for large banks. Securitization has a significant positive effect on the lending activity for both small and large banks. Furthermore, most of the specifications show that securitization affects large banks to a greater extent than small ones. Overall, I find statistically and economically strong evidence that banks’ loan liquidity shields their lending from restricted availability of external financing. 6.4. C&I loans growth and securitization Table 7 presents regression analysis of C&I lending. The structure of the table is similar to Table 6. Panel A presents the estimates of the univariate specification Eq. (6), and Panel B presents the results for the bivariate specification Eq. (7). For each regression, only the sums of 19 For robustness, I implement similar regression analysis without imposing 20-quarter restriction on my sample, without including the bank-specific fixed effects in the regression specifications, and using time fixed effects instead of the bank-specific fixed effects. I find the results to be qualitatively similar. 20 In addition to the statistical tests presented in Tables 5 and 6, I implement two tests of joint hypotheses. First, I test if the coefficient of all lagged DMt  jBit  1, (DMt  jSit  1) are jointly equal to zero. Second, I test if all x and r coefficients are jointly equal to zero. Since, the obtained results showed rejection of these hypotheses at least at the 2% level for all regression specifications, I do not present these tests in Tables 5 and 6.

678

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

Table 6 Analysis of securitization effect on total loans. The table presents the results of the regression analysis of real total loans growth for individual banks. Panel A presents the results of the univariate regression. Panel B presents the results of bivariate regressions where five lags of the log of real gross domestic product growth are added. The dependent variable is the log of real loan growth. The independent variables are bank specific fixed effects, four lags of real loan growth, five lags of change in a measure of monetary policy (DMt  j, jA{0,y,4}), lagged measure of liquidity Bit  1, lagged measure of loan liquidity Sit  1, and the cross effects for these variables. The monetary policy indicators are federal funds rate, paper-bill, and Strongin measure. All policy indicators are transformed so that increases in their levels represent Fed tightening they are also normalized to have the same standard deviation. Only the sum of the coefficients for DMt  1, DMtSit  1, and the cross-effect DMtSit  1Bit  1 are shown. Full sample stands for all banks-quarters available in the sample period 1976:I-2007:IV. Small banks represents only bank-quarters in the bottom 75th percentile of the size distribution, where size is measured as the log of real total assets. Large banks represent only bank-quarters in the top 5th percentile of the size distribution. Large–small statistics are estimated via SUR system. The robust standard errors are reported in parentheses. *, **, and *** correspond to below 10%, 5%, and 1% significance of t-statistics, respectively. Sample

Federal funds rate (1)

Panel A: Univariate analysis Full sample DMtBit  1

Small banks

DMtBit  1

DMtBit  1

DMtBit  1

DMtBit  1

DMtBit  1

DMtBit  1 DMtSit  1

 0.036 (3.98)***

 0.022 (3.08)***

 0.043 (6.58)***  0.012 (1.70)

0.021 (10.73)***

0.038 (18.47)***

0.018 (13.71)*** 0.023 (11.16)***

0.022 (9.43)***

0.048 (17.45)***

0.018 (11.02)*** 0.029 (10.46)***

 0.008 (0.92)

0.032 (4.77)***

 0.019 (2.87)*** 0.009 (1.37)

 0.031 (3.31)***

 0.015 (2.18)***

 0.037 (5.48)***  0.019 (2.66)***

 0.016 (2.44)**

DMtSit  1 Large–small

 0.009 (1.00)

0.049 (7.38)***

 0.019 (2.94)*** 0.042 (6.31)***

0.017 (10.5)***

DMtSit  1 Large banks

0.027 (12.73)***

0.071 (29.96)***

0.025 (16.93)*** 0.054 (22.70)***

0.018 (13.4)***

DMtSit  1 Small banks

0.025 (14.09)***

0.067 (37.25)***

0.026 (21.22)*** 0.054 (29.92)***

 0.039 (5.86)***

DMtSit  1 Panel B: Bivariate analysis Full sample DMtBit  1

(4)

 0.016 (2.54)***

DMtSit  1 Large–small

(3)

0.022 (15.5)***

DMtSit  1 Large banks

(2)

0.023 (18.8)***

DMtSit  1

Paper-bill

 0.033 (4.93)***

the coefficients for DMtBit  1 and DMtSit  1 are presented. In the regression analysis of the real C&I loan growth I impose an additional sample restriction. Specifically, I omit all banks that have less than 5% of their loan portfolio in C&I loans to avoid any distortion due to banks that do only negligible amount of C&I lending. This restriction eliminates around 6.4% of my sample. The regression results for the C&I loans are statistically significant and similar across univariate and bivariate specifications. The loan liquidity plays a positive role in maintaining the C&I loan growth under a monetary tightening when the cost of fund goes up (the sum of the coefficients of DMtSit  1 is positive and statistically

Strongin measure

(5)

(6)

(7)

0.020 (10.23)***

0.084 (34.04)***

0.030 (16.80)*** 0.074 (29.72)***

0.018 (7.60)***

0.075 (22.70)***

0.030 (13.75)*** 0.057 (17.25)***

0.044 (3.94)***

0.072 (8.05)***

 0.009 (1.02) 0.071 (8.01)***

0.026 (2.29)**

 0.003 (0.32)

 0.039 (4.26)*** 0.014 (1.51)

0.014 (5.67)***

0.032 (11.52)***

0.021 (10.56)*** 0.020 (7.33)***

0.013 (4.68)***

0.026 (6.94)***

0.021 (8.86)*** 0.008 (2.24)**

0.001 (0.08)

0.044 (4.89)***

 0.009 (1.03) 0.033 (3.56)***

0.014 (1.21)

0.019 (1.95)***

 0.030 (3.28)*** 0.024 (2.53)***

(8)

(9)

0.071 (9.63)***

0.015 (7.55)*** 0.093 (12.71)***

0.075 (3.56)***

0.015 (6.29)*** 0.046 (4.86)***

0.023 (9.00)***

0.041 (3.72)*** 0.021 (8.26)***

 0.052 (7.96)***

0.026 (2.34)**  0.025 (6.11)***

0.060 (7.92)***

0.007 (3.05)*** 0.085 (11.31)***

0.071 (5.73)***

0.008 (2.93)*** 0.056 (2.65)**

0.022 (8.53)***

0.001 (0.07) 0.020 (7.78)***

 0.049 (8.36)***

0.008 (0.64)  0.036 (6.43)***

significant at the 1% level). As anticipated, the magnitude of these regression coefficients for C&I loans is four to ten times larger than that for total loans. The hypotheses that the liquidity buffer helps protect bank loan growth under a monetary contraction is strongly supported for C&I loans (the sum of coefficients of DMtBit  1 is positive and statistically significant at the 1% level). Small banks behave similarly to the full sample case. The coefficients of DMtBit  1 and DMtSit  1 are positive and have magnitudes similar to the full sample estimates. The case of large banks, meanwhile, is much more interesting. First, large banks show considerable ability to exploit the securitization market in protecting their business loans.

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

679

Table 7 Analysis of securitization effect on commercial and industrial loans. The table presents the results of the regression analysis of real C&I loans growth. Panel A presents the results of the univariate regression. Panel B presents the results of bivariate regressions where five lags of the log of real gross domestic product growth are added. The dependent variable is the log of real loan growth. The independent variables are bank specific fixed effects, four lags of real loan growth, five lags of change in a measure of monetary policy (DMt  j, jA{0,y,4}), lagged measure of liquidity Bit  1, lagged measure of loan liquidity Sit  1, and the cross effects for these variables. The monetary policy indicators are federal funds rate, paper-bill, and Strongin measure. All policy indicators are transformed so that increases in their levels represent Fed tightening they are also normalized to have the same standard deviation. Only the sum of the coefficients for DMtBit  1, DMtSit  1, and the cross-effect DMtSit  1Bit  1 are shown. Full sample stands for all banks-quarters available in the sample period 1976:I-2007:IV. Small banks represents only bankquarters in the bottom 75th percentile of the size distribution, where size is measured as the log of real total assets. Large banks represents only bankquarters in the top 5th percentile of the size distribution. Large–small statistics are estimated via SUR system. The robust standard errors are reported in parentheses. *, **, and *** correspond to below 10%, 5%, and 1% significance of t-statistics, respectively. Sample

Federal funds rate (1)

Panel A: Univariate analysis Full sample DMtBit  1

Small banks

DMtBit  1

DMtBit  1

DMtBit  1

DMtBit  1

DMtBit  1

DMtBit  1 DMtSit  1

 0.046 (1.63)

 0.013 (0.57)

 0.049 (2.38)***  0.008 (0.35)

0.029 (4.75)***

0.100 (15.19)***

0.032 (7.73)*** 0.094 (14.06)***

0.028 (3.90)***

0.127 (15.04)***

0.032 (6.53)*** 0.118 (13.83)***

 0.023 (0.79)

0.123 (5.71)***

 0.022 (1.09) 0.101 (4.57)***

 0.051 (1.74)***

 0.004 (0.19)***

 0.054 (2.62)***  0.017 (0.75)***

 0.012 (0.57)

DMtSit  1 Large –small

 0.009 (0.32)

0.155 (7.30)***

 0.012 (0.58) 0.150 (7.07)***

0.029 (5.82)***

DMtSit  1 Large banks

0.037 (5.75)***

0.168 (22.86)***

0.037 (8.45)*** 0.158 (21.44)***

0.029 (7.15)***

DMtSit  1 Small banks

0.039 (7.09)***

0.158 (27.66)***

0.041 (10.82)*** 0.153 (26.59)***

 0.036 (2.75)***

DMtSit  1 Panel B: Bivariate analysis Full sample DMtBit  1

(4)

 0.003 (0.13)

DMtSit  1 Large –small

(3)

0.033 (7.56)***

DMtSit  1 Large banks

(2)

0.036 (9.60)***

DMtSit  1

Paper-bill

 0.040 (1.93)***

The sum of the coefficients of DMtSit  1 for the subsample of large banks is not only positive but it is also significantly larger than that for the subsample of small banks. Second, similar to the case of the total loans, large banks do not consistently exhibit any dependency on the liquid funds in maintaining their lending activity under a monetary tightening. To summarize, the analysis of C&I loans shows that a bank with higher loan liquidity tends to have larger C&I loan volume growth under a constrained availability of external sources of funds than a bank with lower loan liquidity. As expected for difficult-to-securitize loans, the magnitude of the positive securitization effect on bank

Strongin measure

(5)

(6)

(7)

0.051 (8.34)***

0.159 (20.18)***

0.045 (8.08)*** 0.156 (19.63)***

0.046 (6.50)***

0.151 (14.73)***

0.041 (6.24)*** 0.140 (13.50)***

0.080 (2.39)**

0.252 (8.86)***

 0.009 (0.34) 0.251 (8.82)***

0.034 (0.99)

0.100 (3.34)***

 0.050 (1.77)* 0.111 (3.70)***

0.047 (6.23)***

0.099 (11.11)***

0.029 (4.74)*** 0.094 (10.49)***

0.045 (5.19)***

0.102 (8.81)***

0.029 (4.02)*** 0.091 (7.80)***

0.032 (0.92)

0.226 (7.85)***

 0.032 (1.12) 0.221 (7.48)***

 0.012 (0.35)***

0.124 (4.10)***

 0.061 (2.10)*** 0.129 (4.17)***

(8)

(9)

0.338 (2.64)**

0.042 (6.77)*** 0.355 (2.38)***

0.488 (3.66)***

0.041 (5.71)*** 0.258 (3.88)***

0.405 (4.99)***

0.078 (2.32)** 0.382 (4.70)***

 0.083 (5.29)***

0.037 (1.08) 0.124 (4.75)***

0.331 (4.28)***

0.033 (4.43)*** 0.312 (3.49)***

0.439 (4.58)***

0.036 (4.13)*** 0.414 (3.76)***

0.327 (4.00)***

0.033 (0.93) 0.301 (3.68)***

 0.112 (5.43)***

 0.003 (0.08)***  0.113 (4.83)***

lending under a funding shock is significantly larger for balance sheet volumes of C&I loans than for balance sheet total loans. In addition, I find that securitization has a significantly greater impact on C&I loans for large banks when compared with small ones. This supports the argument that large banks are able to exploit the benefits of the securitization to a greater extent. 6.5. Economic significance of the results In the previous subsections I discuss the effect of securitization on bank lending, concentrating on the sign and the statistical significance of the obtained estimates.

680

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

Table 8 Implied effect of the liquidity and securitization on the aggregate lending four quarters after a federal funds rate shock of 100 basis points. The table shows how two equal-size banks with different levels of liquidity Bit and the loan portfolio liquidity Sit would respond to a monetary tightening. I present the estimates of the combined liquidity-securitization effect on loan growth under a 100 basis points hike in the federal funds rate implied by the estimates of b, x, and r presented in Tables 5 and 6 for various combinations of Bit and Sit. Panel A shows the results for total loans and Panel B, for commercial and industrial (C&I) loans. The computed magnitudes are drawn on the parameter estimates from the univariate and bivariate regression for the full sample, where Bit  1, Sit  1, and the cross-effect Bit  1Sit  1 are included in the set of the independent variables. The percentage of the total effect attributed to the presence of securitization is reported in parentheses. Univariate regression Percentiles of Bit Percentiles of Sit

10th (15.37%)

35th (26.86%)

Bivariate regression Percentiles of Bit

65th (37.93%)

90th (53.44%)

90th–10th

10th (15.37%)

35th (26.86%)

65th (37.93%)

90th (53.44%)

90th–10th

Panel A: Total loans growth (percent) 10th 0.2207 0.3738 (0.76%) (9.46) (6.57) 35th 0.2935 0.4595 (3.43%) (31.94) (24.00) 65th 0.5384 0.7476 (12.38%) (62.89) (53.28) 90th 0.9422 1.2227 (27.14%) (78.79) (71.44) 90th–10th 0.7215 0.8489

0.5212 (5.39) 0.6193 (20.38) 0.9490 (48.04) 1.4927 (66.96) 0.9715

0.7278 (4.54) 0.8433 (17.61) 1.2313 (43.57) 1.8712 (62.87) 1.1433

0.5072

0.1082 (16.22) 0.1695 (46.51) 0.3754 (75.85) 0.7150 (87.32) 0.6068

0.1822 (13.02) 0.2650 (40.19) 0.5432 (70.82) 1.0021 (84.18) 0.8199

0.2535 (11.70) 0.3570 (37.31) 0.7049 (68.25) 1.2786 (82.50) 1.0252

0.3533 (10.75) 0.4859 (35.10) 0.9313 (66.14) 1.6661 (81.07) 1.3128

0.2451

Panel B: C&I loans growth (percent) 10th 0.4638 0.6963 (0.76%) (32.74) (21.68) 35th 0.9941 1.2235 (3.43%) (68.62) (55.43) 65th 2.7756 2.9946 (12.38%) (88.76) (81.79) 90th 5.7140 5.9158 (27.14%) (94.54) (90.78) 90th–10th 5.25 5.22

0.9202 (16.32) 1.4443 (46.68) 3.2054 (75.98) 6.1102 (87.40) 5.19

1.2339 (12.07) 1.7538 (38.14) 3.5009 (69.01) 6.3825 (83.00) 5.15

0.3411 (39.17) 0.8075 (74.31) 2.3746 (91.26) 4.9595 (95.82) 4.62

0.5020 (27.76) 0.9886 (63.31) 2.6233 (86.18) 5.3196 (93.18) 4.82

0.6570 (22.06) 1.1629 (55.96) 2.8628 (82.11) 5.6664 (90.96) 5.01

0.8742 (17.47) 1.4072 (48.73) 3.1983 (77.44) 6.1524 (88.27) 5.28

Another question needs to be addressed. Specifically, it is important to evaluate whether the estimated coefficients imply economically significant magnitudes. I address this issue by quantifying how two equal-size banks with different levels of on-balance sheet liquidity (Bit) and different degrees of loan liquidity (Sit) would respond to a cost of funds shock. In Table 8, I present the estimates of the combined liquidity-securitization effect on loan growth under a 100 basis points hike in the federal funds rate implied by the estimates of the coefficients of DMtBit 1 and DMtSit 1 presented in Tables 6 and 7 for various combinations of Bit and Sit.21 Panel A shows the results for total loans and Panel B shows the results for C&I loans. In an attempt to disentangle the magnitudes of liquidity and securitization effects without crowding the table any further, I compute the percentage of the total effect attributed to the presence of securitization and report it in parentheses. In constructing the table, I consider liquid and illiquid banks as having Bit of 52.4% and 15.2%, respectively, which corresponds to the 90th and 10th percentile of the Bit distributions in my sample. I also consider high loan liquidity banks and low loan liquidity

21 Table 7 is drawn on the parameter estimates from the univariate and bivariate regression for the full sample (because they are most conservative) where Bit  1, Sit  1, and the cross effect Bit  1Sit  1 are included in the set of the independent variables. The magnitudes of the effects are similar whether the coefficient estimates from univariate or bivariate specifications are used.

0.5498 0.6929 0.9290 0.4218 0.7700 0.7598 0.7253 0.6686 (0.10)

0.3164 0.5559 0.9511 0.7060 0.5331 0.5997 0.8236 1.1929 0.66

banks as having Sit of 29.2% and 0.8%, respectively, which corresponds to 90th and 10th percentile of the Sit distribution. The magnitudes presented in Table 8 are positive because they do not contain the direct negative effect of a monetary tightening on bank lending. Consequently, in this table only the magnitudes of the differences in the lending activity between liquid and illiquid banks as well as between banks with high and low loan liquidity (when this direct effect cancels out) are correctly identified. Panel A of Table 8 shows that, four quarters after a 100 basis points increase in the federal funds rate, a high loan liquidity bank would on average exhibit 0.7% to 1.14% higher growth in total loans than a bank with a low loan liquidity (depending on the amount of the traditional onbalance sheet liquidity). This indicates that if both banks start with $1,000 in their total loan portfolios, then, after the tightening, the high loan liquidity bank will have roughly $7 to $11.4 more total loans than a low loan liquidity bank, purely based on the differences in the loan portfolio liquidity. Similarly, four quarters after a 100 basis points hike in the federal funds rate, a liquid bank would on average exhibit 0.51% to 0.92% higher growth in total loans than an illiquid bank. The share of securitization related total loan growth in the combined positive liquidity-securitization effect starts from 4.5% for a liquid bank with low loan liquidity and reaches around 78.8% for an illiquid bank with high loan liquidity. The liquidity effect for C&I loans is similar in magnitude to the total loans case. A liquid bank would have on

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

average 0.5% to 0.92% higher C&I loan growth relative to an illiquid bank four quarters after a 100 basis points increase in the federal funds rate solely due to the onbalance sheet liquidity differential. In contrast, the magnitude of the securitization effect for C&I loans is roughly six times larger than that for total loans. Under a monetary contraction of 100 basis points in the federal funds rate, a high loan liquidity bank would on average exhibit 5.15% to 5.25% more C&I loan growth than a low loan liquidity bank based purely on the difference in the loan liquidity. The liquidity of a bank loan portfolio accounts for 12% to 94% of the combined liquidity-securitization effect and clearly dominates. Because the analysis of C&I loans is likely to reveal the actual effect securitization has on banks’ ability to supply credit, I argue that the magnitude of the securitization effect introduced in this study is seven to ten times larger than the magnitude of the pure liquidity effect shown by Kashyap and Stein (2000). It is hard to deny that securitization has not only statistically but also economically significant effect on C&I lending under a restricted availability of external financing. 6.6. Regulatory discontinuity analysis As a final robustness test, I evaluate how the ability of the existing loan liquidity to affect bank lending activity changes around the regulation of 1984–1986 and Enron related events of 2002–2004. These exogenous events have a direct impact on the ability of the securitization

681

market to absorb new loans on demand but have no impact on the cost of other external financing for banks. I estimate the following regressions using 1983:I to 1988:II and 2000:I to 2006:IV subsamples, correspondingly

DlogðLit Þ ¼ bSit1 þ b1984 Sit1 D1984 þ b1986 Sit1 D1986 þBank Controlsit1 þ ji þ gt þ eit

ð8Þ

and

DlogðLit Þ ¼ bSit1 þ b2002 Sit1 D2002 þ b2004 Sit1 D2002 þBank Controlsit1 þ ji þ gt þ eit :

ð9Þ

Here, bank fixed effects absorb bank local economic conditions and, given a much shorter sample span, effectively capture the past lags of loan growth. The quarter fixed effects capture any direct impact of any other macroeconomic factors. I use the growth in C&I lending as a dependent variable because on-balance sheet volumes of C&I loans better capture the actual banks’ lending activity. In addition, following the arguments of Ivashina and Scharfstein (2010), I evaluate the growth in the extended C&I credit measured as a sum of outstanding C&I loans and C&I loan commitments. The results are presented in Table 9. The results in Panel A confirm that, by enhancing investors’ access to securitization market, both SMMEA and REMICS increased the marginal benefit of securitization as an additional source of funding for banks. Two positive coefficients suggest that by increasing the securitization market ability to absorb (at the time mostly

Table 9 Securitization as a source of loan funding: evidence from regulatory changes. The table reports the results of the regression analysis of commercial and industrial (C&I) loan growth captured by two variables: C&I loans on banks’ balance sheets and C&I credit (C&I loans plus C&I loan commitments). The core variable of interest (loan liquidity Sit  1) is interacted with a set of dummies corresponding to a set of exogenous events. D1984 (D1986) is a dummy variable equal to one for every quarter after and including 1984:IV (1986:I) and zero otherwise. D2002andD2004 are defined similarly. The table reports only the coefficients of interest and omits an extensive set of control variables. Bank-specific controls, namely, the log of real total assets, letters of credit to total assets, bank holding company dummy, equity capital to total assets, share and cost of deposit financing, net income to total assets, nonperforming loans to total loans, and unused loan commitments to total assets, are interacted with the dummies. I further control for the time- and bank-specific fixed effects. Panel A presents the impact of the Secondary Mortgage Market Enhancement Act and The Real Estate Mortgage Investment Conduits on the role of securitization in funding new lending. Panel B presents the analysis of the Enron collapse and the subsequent 2004 regulation impact. Robust t-statistics are reported in parentheses. *, **, and *** correspond to below 10%, 5%, and 1% significance of t-statistics, respectively. Dependent variable Panel A: Subsample period 1983:I to 1988:II Liquidity of loan portfolio Sit  1 Liquidity of loan portfolio Sit  1  D1984 Liquidity of loan portfolio Sit  1  D1986 Number of observations Adjusted R2 Panel B: Subsample period 2000:I to 2006:IV Liquidity of loan portfolio Sit  1 Liquidity of loan portfolio Sit  1  D2002 Liquidity of loan portfolio Sit  1  D2004 Number of observations Adjusted R2 Bank-specific control variables Quarter dummies Bank dummies

C&I loan growth

C&I credit growth

 1.127*** (  32.34) 0.150*** (6.00) 0.302*** (9.40) 198,198 0.83

C&I loan growth

C&I credit growth

 0.798*** (41.78) 0.154*** (9.41) 0.315*** (23.52) 198,198 0.92

 1.199*** (  84.89)  0.877*** (64.86) 0.0939*** (13.10) 196,926 0.79

 0.387*** (  31.33)  0.0460*** (  4.026) 0.120*** (12.21) 196,926 0.83 Yes Yes No

 0.157*** (24.87)  0.0950*** (20.60) 0.0440*** (9.57) 196,926 0.92

 0.258*** (38.05)  0.0536*** (10.83) 0.0500*** (10.11) 196,926 0.92 Yes Yes Yes

682

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

mortgage) loans, the regulators increased the amount of funds available to C&I borrowers and allowed banks with more liquid loans portfolios to lend more. Similarly, the uncertainty about capital requirements after Enron collapse led to a decline in use of securitization as a source of funding (Panel B). The 2004 regulation brought the financial system back on track. In addition to supporting the core hypothesis of this paper, namely, the securitizable loans act as an additional source of funds for banks, the results emphasize an important spillover effect: the ability to securitize mortgages enhances the financial institutions willingness to finance business loans. 7. Conclusion and discussion This paper analyzes the role of securitization in affecting the nature of banking. I propose a new bank-level index of bank loan portfolio liquidity and use it to evaluate the impact of securitization on bank liquidity and lending management. First, I find that securitization acts as a substitute for banks’ on-balance sheet liquidity as it provides deposit institutions with an effective channel to convert illiquid loans into liquid securities. It alleviates the advantages of large banks in terms of liquidity management allowing the liquidity levels of small and large banks to converge. Banks’ ability to securitize has become an integral part of their liquidityrisk management. Overlooked earlier, loan liquidity should now be considered along with the traditional liquidity measures. Second, securitization increases credit availability across sectors as it reduces the sensitivity of bank loan portfolios toward availability of the traditional sources of financing (e.g., deposits). The credit supply of banks with more liquid (securitizable) loan portfolios is less susceptible toward shocks in the costs of external financing than the credit supply of banks with less liquid loan portfolio. As a result banks now seem to hold more of their assets in loans than in the past. In addition, large banks are able to exploit the benefits of securitization for loan origination to a greater extent relative to small banks. This paper adds to a rapidly evolving line of papers that contribute to a better understanding of how securitization reshaped the banking industry and how such changes contributed to the 2007 credit crisis. In contrast to the majority securitization studies that mostly explore the shadow banking, off-balance sheet, implications of securitization, this paper illustrates that securitization led to a material changes on banks’ balance sheets and in banks’ risk exposure. Securitization now holds a very important role in banks liquidity and funding management. Securitization pushes banks to deviate from being traditional liquidity providers to becoming pure intermediaries. The transition should motivate the researches to evaluate how securitization reshapes existing banking theories. The monetary authority should also pay a closer look at the impact this market has on the efficacy of its policy. The evidence indicates that access to the securitization market can potentially offset the impact of

Fed policies on banks’ loan supply. With securitization, it might be necessary to make a larger policy moves to achieve a significant contraction in banks’ lending. The federal government now can affect bank lending by tightening banks’ access to the secondary loan market or introducing regulation that might render securitization too costly. This new channel was effectively demonstrated during the 2002–2004 Enron-related events when a mere expectation of tightening regulation led to a significant decline in lending and increase in banks’ liquidity levels. Similarly, one can argue that significant federal capital infusion and changes in GSEs’ charter during the 2007 crisis supported the mortgage market and by extension purchasing power of the consumers and banks willingness to lend to businesses. The 2007 crisis demonstrated how collapse of investor confidence in securities backed by loans can infect other markets. The credit crisis was a perfect storm of events starting with a freeze of the subprime mortgage market and going to a full collapse of the short-term lending market (Greenspan, 2010). A plentitude of the shadow banking activities found their way back to banks’ balancesheets calling for funding and adequate capital. It would be interesting to explore whether and to what extent the quick response of the Fed in supporting GSEs alleviated the impact that the crisis had on banks. The events of the 2007 crisis call on the Federal Reserve to reevaluate the regulatory environment governing the securitization market. The results in this paper suggest that it would need to carefully consider and balance economic stability and consumers’ and businesses’ access to finance.

Appendix A A.1. Bank-level variables Whenever possible, in defining a bank-level variable, I follow the series definitions in the Federal Reserve notes on forming consistent time series. Federal reserve physical district code: The district code is taken from item RSSD9170. Total loans: Total loans are reported in the call report item RCFD1400. Starting in 1984:I, this item also includes lease-financing receivables. To ensure continuity, total loans must be computed as the sum of RCFD1400 and RCFD2165 (lease-financing receivables) for the period prior to 1984:I. Total mortgages: Total mortgages are taken from item RCFD1410. Home mortgages: Home mortgages are computed as RCON1430. Multifamily residential mortgages: Multifamily residential mortgages are computed as RCON1460. Commercial mortgages: Commercial mortgages are taken from item RCON1480. Farm mortgages: Farm mortgages are taken from item RCON1420. Assets: Total assets are taken from item RCFD2170. C&I loans: Commercial and industrial loans are computed as the sum of RCFD1600 and RCFD1590.

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

Consumer loans: Consumer loans (include car loans and student loans) are computed as RCFD1975. Liquidity: The measure of a bank liquidity is computed as the sum of RCFD0400 (US treasury securities), RCFD0600 (US government agency and corporate obligations), RCFD0900 (obligations of states and political subdivisions), RCFD0380 (all other bonds, stocks, and securities), and RCFD1350 (Fed funds sold and securities purchased under agreements to resell) for the period up to 1983:IV. For the 1984:I–1993:IV period, liquidity is the sum of RCFD0390 (total investment securities), RCFD1350, and RCFD2146 (assets held in trading account). Finally, for the 1994:I– 2003:IV period, it equals the sum of RCFD1350, RCFD1754 (securities held to maturity), and RCFD1773 (available for sale securities). I eliminate cash in vaults because a greater portion of banks’ cash is stored for purposes of reserve requirements. Following Kashyap and Stein (2000), I compute the measure of bank’s on-balance sheet liquidity Bit as a percentage of the liquid funds in the total assets. Deposits: Total deposits are computed as RCFD2200. Total equity: Total equity is computed as RCFD3210. Income: Net income comes from item RIAD4340. Nonperforming loans: Following Campello (2002), I use the measure of loan performance that avoids managerial discretion in reporting losses. It equals the sum of loans over 90 days late (RCFD1407), plus loans not accruing (RCFD1403). Credit card loans: The amount of credit card loans comes from item RCFD2008. Standby letters of credit: The standby letters of credit are computed as either item RCFD3375 or as the sum of items RCFD3376 and RCFD3377 when the data on item RCFD3375 are missing. A.2. Measures of monetary policy Fed funds: I use the monthly series of the effective annualized federal funds rates provided by the Board of Governors Release H.15. This proxy is advocated by Bernanke and Blinder (1992), who show that the federal funds rate captures the stance of monetary policy well because it is sensitive to shocks to the supply of bank reserves. The fed funds rate is the prevalent measure of monetary policy in empirical work. However, the adequacy of this proxy has been questioned for periods when the Fed’s operating procedures were modified (e.g., the Volcker period). For robustness, I employ alternative measures of the monetary policy. Paper-bill: This proxy is computed as the difference between the rates paid on six-month prime rated commercial papers and 180-day Treasury bills. These series are available from the Board of Governors’. Bernanke (1990) argues that increases in paper-bill index capture Fed tightening since banks will cut loans and corporations are forced to substitute commercial paper for bank loans. Strongin: Strongin (1995) argues that previous studies that attempted to identify the stance of monetary policy fail to properly address the Fed’s strategy of accommodating reserve demand shocks. Strongin measures the portion of nonborrowed reserves growth that is orthogonal to total reserve growth. It equals the residual of the

683

linear regression of total reserves on nonborrowed reserves, in which both series are normalized by a 24-month moving average of total reserves prior to the estimation. Strongin argues that the Fed is constrained to meet total reserve demands in the short run (failure to do so would lead to wild swings in the funds rate), but it can effectively tighten policy by reducing non-borrowed reserves and forcing banks to borrow more from the discount window. As suggested by Bernanke and Mihov (1998), I estimate the regression parameters based on the monthly data separately for the subperiod January 1964 to October 1, 1979 and for subperiod November 1979 to January 2004. I perform this computation using data from the Federal Reserve Economic Data (FRED) bank. All indicators of the monetary policy are converted so that an increase in the indicator represents a monetary tightening.

A.3. Bank loan portfolio liquidity I compute the degree of loan liquidity for six loan categories as the ratio of loans securitized to total loans outstanding. All the data are taken from the ‘‘Flow of Funds Accounts of the United States.’’ Home mortgages: Home mortgages outstanding are taken from Table L.2 item FL193165105. Home mortgages securitized are computed as the sum of item FL413065105 from Table L.126 and item FL673065105 from Table L.127. Multifamily residential mortgages: Multifamily residential mortgages outstanding are taken from Table L.2 item FL123165405. Multifamily residential mortgages securitized are computed as thesum of item FL413065405 from Table L.126 and item FL673065405 from Table L.127. Commercial mortgages: Commercial mortgages outstanding are taken from Table L.2 item FL193165505. Commercial mortgages securitized are computed as the sum of item FL413065505 from Table L.126 and item FL673065505 from Table L.127. Farm mortgages: Farm mortgages outstanding are taken from Table L.2 item FL893065605. Farm mortgages securitized are taken from Table L.126 item FL413065605. C&I loans: Loans to business outstanding are computed as the sum of item FL253169255, item FL193168005, item FL263168005, item FL263169255 from Table L.2. Loans to business securitized are taken from Table L.127 item FL673069505. Consumer credit: Consumer credit outstanding is taken from Table L.2 item FL153166000. Consumer credit securitized is computed as the sum of items FL673066000 and item FL673069153 from Table L.127. References Acharya, V.V., Schnabl, P. Do global banks spread global imbalances? The case of asset-backed commercial paper during the financial crisis of 2007–09. International Monetary Fund, Economic Review, forthcoming. Ambrose, B.W., Warga, A., 2002. Measuring potential GSE funding advantages. Journal of Real Estate Finance and Economics 25, 129–150. Anderson, T.W., Hsiao, C., 1982. Formulation and estimation of dynamic models using panel data. Journal of Econometrics 18, 47–82.

684

E. Loutskina / Journal of Financial Economics 100 (2011) 663–684

Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies 58, 277–297. Ashcraft, A.B., 2006. New evidence on the lending channel. Journal of Money, Credit, and Banking 38 (3), 751–776. Bernanke, B.S., 1990. On the predictive power of interest rates and interest rate spreads. New England Economic Review (November– December), Federal Reserve Bank of Boston, 51–68. Bernanke, B.S., Blinder, Alan S., 1992. The federal funds rate and the channels of monetary transmission. American Economic Review 82, 901–921. Bernanke, B.S., Gertler, M., 1995. Inside the black box: the credit channel of monetary policy transmission. Journal of Economic Perspectives 9, 27–48. Bernanke, B.S., Mihov, I., 1998. Measuring monetary policy. Quarterly Journal of Economics 9, 869–902. Black, D.G., Garbade, K.D., Silber, W.L., 1981. The impact of the GNMA pass-through program on FHA mortgage costs. Journal of Finance 36, 457–469. Campello, M., 2002. Internal capital markets in financial conglomerates: evidence from small bank responses to monetary policy. Journal of Finance 57, 2773–2805. Demyanyk, Y., Van Hemert, O. Understanding the subprime mortgage crisis. Review of Financial Studies, forthcoming. Diamond, D., Dybvig, P., 1983. Bank runs, deposit insurance, and liquidity. Journal of Political Economy 91, 401–419. Diamond, D.W., Rajan, R.G., 2000. A theory of bank capital. Journal of Finance 55, 2431–2465. Ehrmann, M., Gambacorta, L., Martinez-Pages, J., Sevestre, P., Worms, A., 2001. Financial systems and the role of banks in monetary policy transmission in the euro area. Working Paper Series No. 105, European Central Bank, Frankfurt, Germany. Estrella, Arthuro, 2002. Securitization and efficacy of monetary policy. FRBNY Economic Policy Review (May), 243–255. Frame, W.S., White, L.J., 2005. Fussing and fuming about Fannie and Freddie: how much smoke, how much fire? Journal of Economic Perspectives (Spring), 159–184 Gorton, G.B., Souleles, N.S., 2005. Special purpose vehicles and securitization. NBER Working Paper Series, vol. w11190. Greenspan, A., 2010. The crisis. Unpublished Working Paper. Heuson, A., Passmore, W., Sparks, R., 2000. Credit scoring and mortgage securitization: implications for mortgage rates and credit availability. Finance and Economics Discussion Series No. 2000-44, Board of Governors of the Federal Reserve System, Washington, DC. Holmstrom, B., Tirole, J., 1998. Private and public supply of liquidity. Journal of Political Economy 106, 1–40. Houston, J., James, C., Marcus, D., 1997. Capital market frictions and the role of internal capital markets in banking. Journal of Financial Economics 46, 135–164. Hubbard, R.G., Kuttner, K.N., Palia, D.N., 2002. Are there bank effects in borrowers’ costs of funds? Evidence from a matched sample of borrowers and banks. Journal of Business 75, 559–583.

Ivashina, V., Scharfstein, D.S., 2010. Bank lending during the financial crisis of 2008. Journal of Financial Economics 97 (3), 319–338. Jayaratne, J., Morgan, D.P., 2000. Capital market frictions and deposit constraints at banks. Journal of Money Credit and Banking 32, 74–92. Kashyap, A.K., Stein, J.C., 1994. Monetary policy and bank lending. In: Mankiw, G.N. (Ed.), Monetary Policy. University of Chicago Press, Chicago IL, pp. 221–256. Kashyap, A.K., Stein, J.C., 1995. The impact of monetary policy on banks’ balance sheets. Carnegie-Rochester Conference Series on Public Policy 42, 151–195. Kashyap, A.K., Stein, J.C., 2000. What do a million observations on banks say about the transmission of monetary policy? American Economic Review 90, 407–428 Kashyap, A.K., Rajan, U.A., Stein, J.C., 2002. Banks as liquidity providers: an explanation for the coexistence of lending and deposit-taking. Journal of Finance 57, 33–73. Keys, B.J., Mukherjee, T.K., Seru, A., Vig, V. Did securitization lead to lax screening? Evidence from subprime loans. Quarterly Journal of Economics, forthcoming. Kolari, J.W., Fraiser, D.R., Anari, A., 1998. The effect of securitization on the mortgage market yield: a cointegration analysis. Real Estate Economics 26, 677–693. Loutskina, E., Strahan, P.E., 2009. Securitization and the declining impact of bank finance on loan supply: evidence from mortgage acceptance rates. Journal of Finance 64, 861–889. Loutskina, E., Strahan, P.E. Informed and uninformed investment in housing: the downside of diversification. Review of Financial Studies, forthcoming. Mian, A., Sufi, A., 2009. The consequences of mortgage credit expansion: evidence from the 2007 mortgage default crisis. Quarterly Journal of Economics 124 (4), 1449–1496. Nothaft, F., Pearce, J., Stevanovic, S., 2002. Debt spreads between GSEs and other corporations. Journal of Real Estate Finance and Economics 25, 151–172. Passmore, Wayne, 2004. The GSE implicit subsidy and value of government ambiguity. Unpublished Working Paper, Board of Governors of the Federal Reserve System, Washington, DC. Parlour, C., Plantin, G., 2008. Loan sales and relation banking. Journal of Finance 63, 1291–1314. Piskorski, T., Seru, A., Vig, V. Securitization and distressed loan negotiation: evidence from the subprime mortgage crisis. Journal of Financial Economics, forthcoming. Purnannandam, A. Originate-to-distribute model and the subprime mortgage crisis. Review of Financial Studies, forthcoming. Rajan, U., Seru, A., Vig, V., 2010, The failure of models that predict failure: distance, incentives and default. Unpublished Working Paper, University of Chicago, Chicago IL. Sharpe, S.A., 1995. Bank capitalization, regulation, and credit crunch: a critical review of the research findings. Working Paper No. 95-20, Federal Reserve Board, Washington, DC. Strongin, S., 1995. The identification of monetary policy disturbances: explaining the liquidity puzzle. Journal of Monetary Economics 35, 463–497.