SFAS 115, Bank Balance Sheet Liquidity and Loan Growth DISSERTATION

Transcription

1 SFAS 115, Bank Balance Sheet Liquidity and Loan Growth DISSERTATION Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University By Anthony A. Meder Graduate Program in Accounting and MIS The Ohio State University 2011 Dissertation Committee: Professor Anne Beatty, Adviser Professor Darren Roulstone Professor Douglas Schroeder Professor Jennifer Altamuro

2 Copyrighted by Anthony Alan Meder 2011

3 Abstract I examine the effect of marketable security holdings on monetary policy when those securities are classified under SFAS 115. Prior research has shown that loan growth declines in monetary contractions and that marketable security holdings mitigate that decline. Those studies consider the securities in aggregate; I am the first to consider the securities classification in conjunction with monetary policy. I ask whether held-tomaturity securities, which are less liquid than non-held-to-maturity securities due to their held-to-maturity classification, are negatively related to loan growth during monetary contractions, relative to non-held-to-maturity securities. I find that held-to-maturity securities are more negatively related to loan growth, relative to non held-to-maturity securities, during monetary contractions. I also find that held-to-maturity securities are incrementally more negatively related to loan growth during monetary contractions, relative to non-tightening times. Finally, I find that both of these effects are stronger for small banks, relative to large banks. Given these findings, I conclude that held-to-maturity securities actually enhance, not mitigate, the effect of monetary tightening on bank lending. ii

4 Dedication Without the support and patience of my wife, Debra, this dissertation could not have happened. Her faith and confidence in me was the well from which I drank when I felt overwhelmed. She may not have a co-author credit on this document, but she is my life's co-author and I am indebted to her for her support in all I have done and will do. I dedicate this dissertation to you, Debbie. iii

5 Acknowledgments This dissertation would not exist without the support, guidance and patience of my committee members and I am very thankful for their commitment to me and my education. I also want to thank the Accounting Department of Fisher College for supporting me over the past five years. I particularly thank Anne Beatty for being my adviser, for her support and for her availability as I worked my way through this dissertation. I also thank Anne for her honesty in evaluating my progress throughout the program and her obvious dedication to my success. I am grateful to Doug Schroeder for his patience during our econometrics, model specification and causality talks and I thank him for improving my understanding of these issues. Darren Roulstone and Jennifer Altamuro also provided important advice and critiques regarding my dissertation; I thank both of them for suggestions that focused and clarified the exposition and direction of my dissertation. iv

6 Vita B.S. Business Admin., University of Michigan--Flint M. Acc. Accounting, Oakland University Ph.D. Accounting and MIS, The Ohio State University Publications Structured Finance and Mark-to-Model Accounting: A Few Simple Illustrations. with S.Schwartz, R. Spires and R. Young, Accounting Horizons, September Fields of Study Major Field: Accounting and MIS v

7 Table of Contents Abstract... ii Dedication... iii Acknowledgments... iv Vita... v Publications... v Fields of Study... v Table of Contents... vi List of Tables... viii List of Figures... ix Chapter 1: Introduction... 1 Chapter 2: Background and Literature Review Federal Reserve and Monetary Policy Monetary Policy and the Bank Lending Channel SFAS115 and Balance Sheet Liquidity Chapter 3: Hypothesis Development vi

8 Chapter 4: Sample and Research Design Sample Selection Research Design Additional Analysis: Further Disaggregation Chapter 5: Results and Analysis Descriptive Statistics Multivariable Analysis Main Results Large/small Bank Results Disaggregated Analysis Results Sensitivity Analysis Using Alternative Tightening Measures Sensitivity Analysis on Alternative Fed Funds Measures of Tightening Additional Sensitivity Analysis Chapter 6: Summary and Conclusions References APPENDIX A: BALANCE SHEET (Commercial Bankshares, Inc., AS OF DEC. 31, 2006) vii

9 List of Tables Table 1: Variable definitions Table 2: Descriptive statistics Table 3: Pearson correlations Table 4: Regression analysis using Fed Funds rate as proxy for monetary policy Table 5: Regression analysis of small vs. large bank comparison Table 6: Descriptive statistics and correlations for disaggregated securities analysis Table 7: Regression analysis using the disaggregated securities Table 8: Regression analysis with six-month commercial paper premium as monetary policy proxy Table 9: Regression analysis using three-month commercial paper premium as monetary policy proxy viii

10 List of Figures Figure 1: Average Ratio of HTM sold or transferred annually from 1996 to Figure 2: Average ratio of HTM securities sold or transferred vs. TED spread Figure 3: Balance sheet for Commercial Bankshares, Inc ix

11 Chapter 1: Introduction Prior monetary policy research has shown theoretically (Stein, 1998) and empirically (Bernanke & Blinder, 1992; Kashyap & Stein 1995, 2000) that bank loan growth declines with monetary tightening and that marketable securities held by banks affect that association. Specifically, the literature shows that a lending decline following a monetary tightening by the Federal Reserve is mitigated by marketable securities held by banks (Bernanke & Blinder, 1992; Kashyap & Stein, 2000). Prior monetary policy research has not considered that accounting standards may play a role in the trading of these marketable securities. Monetary policy is affected by balance sheet liquidity and Statement of Financial Accounting Standards 115: Accounting for Certain Investments in Debt and Equity Securities (SFAS 115) is associated with balance sheet liquidity; therefore, I consider the interactive effect of SFAS 115 and monetary policy. A monetary tightening by the Federal Reserve is meant to reduce cash in the economy. If the economy is growing too rapidly, inflation becomes a concern and the Federal Reserve will reduce cash in the economy to slow the growth and prevent inflation. To enact a tightening, the Federal Reserve increases the targeted Fed Funds rate and sells treasuries to banks to meet that target and to reduce cash reserves in the banking system. Monetary policy literature shows that one channel for this monetary 1

12 tightening is a reduction in bank lending. However, that literature also shows marketable securities held by banks will mitigate the reduction in lending and lessen the effectiveness of the lending channel as a transmission method of monetary tightening policy. The monetary policy literature tests the effects of marketable securities without considering the classification of those securities under SFAS 115. Under that policy, securities that would be otherwise liquid may become effectively illiquid when classified as held-to-maturity. Securities classified as held-to-maturity can be sold prior to maturity only under very specific circumstances, per the standard. While this standard-induced illiquidity could reduce loan growth at any time, it may be more problematic during monetary tightening. This study's innovation is that it examines the interactive effect of SFAS 115 and monetary tightening. A tightening is meant to slow the economy and reduced lending may be part of the Federal Reserve's plan for reducing cash in the system and slowing the economy. However, if the loan reduction is greater than expected due to standard-induced illiquidity of held-to-maturity securities, a tightening meant to slow the rate of economic growth could become an over-tightening and cause economic contraction instead. If the Federal Reserve considers the documented effects of marketable securities on loan growth without also considering the effect of held-tomaturity classification under SFAS 115, this over-tightening scenario may be even more likely to occur. 2

13 Stein (1998) first models the loan portfolio reaction to a decrease in deposits and shows a decrease in loans following a decrease in deposits. He also shows that marketable securities mitigate the decrease in lending. Stein (1998) does not require a specific deposit-reducing event. To address the interactive effects of monetary policy and SFAS 115, I consider a particular deposit reducing-event: an increase in the fed funds rate. Federal open market security sales are designed to reduce cash reserves in the banking system and to raise the federal funds rate. The higher fed funds rate translates into higher interest rates through the financial markets, which makes finding new depositors to replace the lost reserves more difficult for banks. Potential depositors (whose deposits are insured by the FDIC) have interest-bearing options and, with an increase in interest rates, may choose the higher return rather than the safety of an insured deposit. When banks experience a decrease in insured deposits (e.g. transaction accounts, passbook savings accounts) they may not be able to replace those lost funds with uninsured funds (such as large CDs or equity capital). When trying to attract investors with non FDIC-insured products, banks face more competition for the investors' funds. Empirical monetary policy transmission research supports the link between balance sheet liquidity and loan volume. The bank lending channel theory predicts that bank lending is one conduit of monetary policy; when the Fed acts to tighten the money supply, banks reduce lending in reaction to the loss of liquidity. This is the liquidity effect of the bank lending channel and is supported by research on both the banks and the banking system as a whole. Deposits, loan balances, and security 3

14 holdings decrease after a monetary tightening (Bernanke & Blinder, 1992). Kashyap & Stein (2000) find a similar loan reduction for commercial banks after the money supply is tightened; however, that reduction is mitigated by marketable securities. Balance sheet liquidity offsets, at least partially, the reduction in loans after monetary tightening. SFAS 115 requires that firms classify their marketable securities as trading (TRADE), available-for-sale (AFS) or held-to-maturity (HTM). Securities obtained to provide short-term profits are trading securities; debt securities for which the entity has the intent and ability to retain until maturity are classified as HTM. All other securities are classified as AFS. The HTM classification for debt securities constrains otherwise liquid securities (e.g. U.S. treasuries, securities supported by governmental agencies, state government securities). Once a bank classifies a security as HTM, the bank cannot liquidate that security without violating SFAS 115 and potentially facing penalties (FASB 1993). Past literature regarding SFAS 115 concentrates on the adoption period and finds that the SFAS 115 classifications are important for bank decision-making; banks classify securities for a variety of reasons, including capital concerns and, in the case of classification at adoption, whether the securities had accumulated holding gains or losses (Beatty, 1995; Hodder et al. 2002). Future liquidity concerns affect HTM classification decisions (Hodder et al. 2002); limiting the sale of HTM securities may reduce the ability to manage liquidity and to fulfill loan demand (Beatty, 1995). While SFAS 115 requires that marketable securities are separated by classification, 4

15 monetary policy literature considers marketable securities in the aggregate. I considered the interactive effect of SFAS 115 classification and monetary policy. I first consider the differential effect of HTM securities during periods of monetary tightening. I hypothesize that during monetary tightening, HTM securities, relative to non-htm securities, are negatively related to loan growth. In addition, I consider the effect of tightening relative to non-tightening on this differential effect of HTM securities. I hypothesize that the negative differential effect of HTM securities relative to non-htm securities is stronger during monetary tightening relative to nontightening periods. Monetary policy literature has shown that small banks are more intensely affected by monetary policy (Kashyap & Stein, 2000; Kishan & Opiela, 2000 and 2006). The monetary policy literature uses size as a natural identification technique to address the potentially endogenous nature of loan growth. Small banks have more loans on their balance sheet than large banks (Kashyap & Stein 1995, 2000; Kishan & Opiela 2000, 2006). I follow that literature and also examine the difference in loan growth effects of HTM securities using size as a natural identifier. I hypothesize that the predicted HTM securities effects are stronger for small banks, relative to large banks. To test my hypotheses, I use overall liquidity (ratio of total securities to assets) and the ratio of HTM securities to total assets to capture the differential effect of HTM securities. Total securities is the sum of HTM, AFS, trade and fed funds sold (FFS). My proxy for monetary policy is the changes in the fed funds rate (Bernanke & Blinder, 1992; Kashyap & Stein 1995, 2000; Kishan & Opiela 2000, 2006). I also use 5

16 the three- and six-month commercial paper premiums as measures of monetary tightening in sensitivity analyses. The premiums are the difference between a risk free rate (the rate for treasury securities of the same maturity) and the financial commercial paper rate. Increases in the fed funds rate (or higher commercial paper premiums) indicate a tighter monetary environment. For testing the difference in effects between small and large banks, I follow the monetary policy literature and classify banks based upon total assets; those banks in the 98 th percentile are classified as large banks while those banks no higher than the 90 th percentile are classified as small banks (Kashyap & Stein 1995, 2000; Kishan & Opiela 2000, 2006). I use quarterly CALL report data from the first quarter of 1996 through fourth quarter of Regulated U.S. financial institutions are required to file a CALL report on a quarterly basis. CALL reports are essentially balance sheets, income statements and approximately 20 additional schedules that provide information about the institution s income and financial condition for the quarter. My sample consists of 386,257 quarterly CALL reports representing 11,355 commercial banks. I find support for the hypothesis that loan growth is decreasing in HTM securities, compared to non-htm securities, during monetary contractions relative to non-contraction periods. I also find that HTM securities are more negatively related to loan growth during contractions, relative to non-contraction periods. Lastly, I show that the effects are more pronounced for small banks relative to large banks during contractions. 6

17 Prior monetary policy literature finds that marketable securities mitigate the bank lending channel transmission mechanism of monetary policy; I contribute to this literature by showing that accounting policy for marketable securities alters that mitigation effect. HTM securities enhance, not mitigate, the bank lending channel's transmission of monetary policy into the economy during monetary tightening. I am the first to examine this interactive effect. This paper contributes to the bank lending and accounting literature. The Fed needs to consider both the amount of marketable securities held by banks and the composition of that portfolio of marketable securities when setting monetary policy. As the ratio of HTM securities increases, the Fed should consider less of a rate increase during tightening actions. Otherwise, a move meant to slow an overheating economy could stop the economic growth and perhaps trigger an economic contraction. Small businesses rely on small banks for capital to continue in business and to expand. Therefore, a tightening that does not consider that interactive effect may not only result in an over-tightening but also may restrict job growth through less lending by small banks to small businesses. The rest of the paper is organized as follows. Section 2 provides background and section 3 is the hypothesis development. Section 4 presents the sample and research design; empirical results are discussed in section 5. Section 6 concludes. 7

18 Chapter 2: Background and Literature Review 2.1 Federal Reserve and Monetary Policy The Federal Open Market Committee (FOMC) is the component of the Federal Reserve System (Fed) that conducts open market operations. An open market operation is the sale (purchase) of treasury securities to implement a monetary contraction (expansion). The Fed decreases (increases) reserve account balances after the sale (purchase) of securities. The Fed uses open market operations as its primary method of implementing monetary policy (Edwards 1997). The open market operation decisions are made to target a fed funds rate. 1 Since the open market committee targets a fed funds rate, the fed funds rate has been used to proxy for monetary policy by the Fed (Bernanke & Blinder, 1992; Kashyap & Stein, 2000). The commercial paper premium is also a good proxy for monetary policy (Bernanke & Blinder, 1992; Becher et al. 2008). The commercial paper premium is the interest rate on commercial paper less a risk free rate, typically the rate on a treasury security of the same maturity; the typical maturity used is six months. 1 While the goal of the operations could be a reserve level or a price level (fed funds rate), the FOMC has been targeting a funds rate since the 1980s, started announcing policy changes in 1994 and then started announcing specific fed funds target rates in

19 2.2 Monetary Policy and the Bank Lending Channel The monetary policy literature provides evidence that the bank lending channel (BLC) is a transmission conduit for monetary policy into the economy. According to the BLC theory, monetary policy by the Fed can change loan supply (Bernanke & Blinder, 1992). BLC theory says a monetary contraction will reduce loan supply; for that reduction to affect the economy, there must exist loan-dependent firms that cannot wholly replace debt with other funding (Hubbard 1995). 2 Van den Heuvel (2007) models the BLC through the capital effect. He develops a dynamic model that predicts a monetary contraction will reduce loans supplied by banks with regulatory-constrained equity capital. The reduction is caused by an increase in short-term rates that lowers future bank capital. Stein (1998) models the BLC transmission via the liquidity effect. Faced with a decrease of insured deposits, banks replace those deposits with uninsured liabilities. 3 Investors who do not desire the protection of FDIC-insured deposits have many options for their investments; banks face more competition for those funds and will not be able to fully replace lost deposits (Stein 1998). In a simplified model excluding marketable securities, Stein shows that banks reduce lending to provide liquidity when uninsured liabilities do not fully replace insured deposits. He then introduces marketable securities and shows that banks will, at least partly, offset lending reductions with security sales. 2 Cecchetti (1995) and Bernanke & Gertler (1995) also detail the specifics of the BLC theory. 3 Stein identifies equity as an uninsured source of funding, but banks could also use large (greater than $100,000) time deposits to replace insured deposits. 9

20 Bernanke and Blinder (1992) examine the liquidity channel of monetary policy transmission. They use the fed funds rate to proxy for monetary policy and their results show that after a 31-basis point increase in the fed funds rate (one standard deviation), aggregated loans, deposits and securities all decline while the unemployment rate rises. The proxy seems reasonable since the Fed considers a fund rate target when enacting monetary policy. Bernanke and Blinder (1992) further show that the fed funds rate is systematically related to economic activity 4 and that the rate is informative of supply shocks rather than demand shocks; a regression of the fed funds rate on a set of demand variables produces insignificant coefficients (Bernanke & Blinder, 1992). They also find that the commercial paper premium (CPP) is a strong proxy for monetary policy. 5 However, they use the fed funds rate since the CPP is informative because is it indicative of the stance of monetary policy and the fed funds rate is a more direct measure (Bernanke & Blinder, 1992; Bernanke & Mihov, 1998). 6 Having shown the appropriateness of the fed funds rate as a proxy for monetary policy, Bernanke and Blinder (1992) then examine the effect of monetary policy on aggregate bank loans, securities and deposits as well as on the unemployment rate. Deposits and securities decrease almost immediately while loans begin to decline after 4-6 months and unemployment rates begin climbing after 4-5 months (Bernanke and Blinder 1992). The bank balance sheet effects and the unemployment change are 4 The fed funds rate decreases (increases) in response to unemployment (inflation) shocks. 5 See Friedman & Kuttner (1992). More recently, Becher et al. (2008) supports the informativeness of the CPP. 6 For a similar reason, the fed funds rate was chosen over narrative based proxies such as the Boschen-Mills Index (Boschen & Mills, 1995). 10

21 evidence that monetary tightening has an economic effect. Overall, Bernanke & Blinder (1992) provide evidence that the Fed conducts monetary policy with a fed funds rate target in mind, that the fed funds rate can proxy for monetary policy and that a liquiditybased BLC does exist. Kashyap and Stein (1995) extend Bernanke and Blinder (1992) by disaggregating the banking information into 5 bank-size categories (as measured by total assets). They hypothesize that banks reactions to a decrease in liquidity are differentiated by bank size; larger banks should have an easier time replacing the lost deposits and, therefore, monetary policy would have less effect on the lending policies of those larger banks. They find that for all banks, changes in deposits are negatively related to changes in the fed funds rate. For small banks, loan volume and marketable securities are more negatively related to changes in the fed funds rate (Kashyap & Stein, 1995). Overall, Kashyap and Stein (1995) provide support for the liquidity BLC transmitting monetary policy. Kashyap and Stein (2000) further extend the work of Bernanke and Blinder (1992) by examining the effect of monetary policy on individual banks. Kashyap and Stein (2000) test the liquidity BLC using quarterly financial information over the period of 1976 through 1993 from the CALL reports that banks are required to file with the Federal Reserve. They find support for the liquidity BLC; monetary tightening reduces bank lending but the effect is decreasing in balance sheet liquidity. They further split their sample by size (measured by average assets) and find that the change in lending is strongest for small banks with illiquid balance sheets. 11

22 2.3 SFAS115 and Balance Sheet Liquidity The monetary policy literature shows that bank balance sheet liquidity contributes to loan growth and may insulate banks, at least partly, from monetary tightening. The sum of securities and fed funds sold, scaled by total assets, is the typical liquidity measure. For this to be an appropriate liquidity measure, securities and fed funds sold need to be liquid assets. 7 The fed funds market is an overnight interbank market and should pose little liquidity concern. 8 However, the extant literature does not appear to address the effect of SFAS 115 accounting on liquidity of securities and, subsequently, on loan growth. HTM securities are reported at amortized cost and are not to be sold or transferred into another category unless specific, significant circumstances apply (none of which include liquidity needs) (FASB 1993): 1. Evidence of deterioration of the issuer s creditworthiness 2. Tax law change regarding the tax exempt status of the interest 3. Change in regulatory or statutory environment regarding securities portfolio composition 4. Increase in required regulatory capital 5. Increase in risk weights on debt securities used in risk-based capital 6. If the security is within 90 days of maturity 7. If the bank has already collected 85% of the principal In the Bank Accounting Series, the Chief Accountant of the Office of the Comptroller of Currency (OCC) provides interpretations of US GAAP and guidance for national banks accounting practices. The Chief Accountant s position is that, transferring securities out of HTM (by sale or re-classification) for other than the safe 7 Bernanke & Blinder (1992), Kashyap & Stein (2000), Kishan & Opiela (2000, 2006). 8 Fed funds transactions with stated maturities > 1 day are called term fed funds. They are not reported as fed funds sold. 12

23 harbor reasons given in SFAS 115 taints the entire HTM portfolio and all HTM assets will have to be re-classified as AFS. The response continues that consistent with the Securities and Exchange Commission the bank will be prohibited from using the HTM classification for two years. 9 The SEC s Corporate Finance Division posts frequently requested accounting and reporting interpretations and guidance. 10 The Corporate Finance Division monitors the reporting of banks. When a bank sells HTM securities for reasons other than those allowed under SFAS 115, the Finance Division presumes that the remaining balance of HTM securities should be transferred out of HTM classification and the bank may be forced to justify the remaining HTM classifications. According to the guidance, Finance division staff may then challenge the original classification of the transferred securities, the classification of securities still carried as HTM and the future classification of securities as HTM. These challenges can lead to restating past financial reports (if prior classification is determined to be invalid), tainting the remaining balance of the HTM portfolio and prohibiting the use of HTM going forward. In two separate civil cases, the SEC obtained judgments against Fannie Mae (in 2006) and against Freddie Mac (in 2007) 11. In the Fannie Mae case, the SEC alleged that Fannie Mae would temporarily classify all debt securities as HTM upon acquisition and then assign permanent classifications at the end of the acquisition month. This practice was in violation of SFAS 115 and also, according to the SEC, highlighted a lack Both in US District Court, District of Columbus. The Freddie Mac case number is 1:07-CV41728RCL while the Fannie Mae case number is 1:06CV

24 of internal controls. Fannie Mae could strategically re-classify HTM securities based upon market value changes since the acquisition. For these, and other alleged violations of the Securities Act of 1933, Fannie Mae agreed to a judgment without admitting guilt and was ordered not to violate the Act and to pay $1 of disgorgement and $350,000,000 (three hundred fifty million) as a civil penalty. In the Freddie Mac case, the SEC alleged that Freddie Mac initiated many securities transactions at the end of 2000 in order to generate securities losses that would offset an anticipated gain from the transition to SFAS 133. A number of other improper transactions and re-classifications during November and December of 2000 were designed to classify or to reclassify securities either to AFS (so that value changes would flow to other comprehensive income rather than net income) or to HTM for amortized cost accounting. Freddie Mac was alleged to have violated the Securities Act of 1933 and the Exchange Act of 1934, including having defrauded investors. Freddie Mac agreed to a judgment with no admission of guilt and was ordered not to violate the Acts and to pay $1 disgorgement and $50,000,000 (fifty million) in civil penalties. Neither of these cases was strictly about SFAS 115 violations; Fannie and Freddie were charged with numerous violations and allegedly fraudulent acts. The settlements do not assign a penalty amount for each alleged violation so a specific amount cannot be associated with SFAS 115 violations. It may be that the bulk of the fines were not related to the SFAS 115 violations; however, these cases do show that the SFAS 115 violations may be included in court actions. 14

25 Trading and AFS securities are reported at fair value and the unrealized gains and losses of trading (AFS) securities are included in earnings (other comprehensive income). At adoption (fiscal years beginning after Dec. 15, 1993), bank regulators had announced that regulatory capital calculations would include the unrealized holding gains and losses resulting from SFAS 115 rules. In late 1994, bank regulators reversed their position and announced that SFAS 115-induced equity changes would be excluded from regulatory capital calculations. The FASB instituted an amnesty period from mid-november through December of During that period, banks had the option to re-allocate securities among the SFAS 115 categories without penalty (Hodder et al. 2002). The SFAS 115 literature primarily examines reactions around the adoption period. Hodder et al. (2002) examines security classifications for both the initial adoption period ( ) and the post-amnesty period ( ). Hodder et al. (2002) find that banks faced a trade-off between liquidity and equity capital stability at the inception of SFAS 115. Banks and regulators opposed SFAS 115 claiming that the standard would artificially increase capital equity volatility (Beatty 1995). Beatty (1995) finds that banks change their investment portfolio management strategy to reduce that volatility as well as to inflate earnings by classifying securities with holding gains as AFS securities. She finds that the proportion of securities held by banks declined, as did the maturity of those securities. Shortened maturities could reduce the bank s interest income and the illiquidity of held-to-maturity securities could restrict the bank s ability to lend (Beatty 1995). 15

26 Chapter 3: Hypothesis Development Monetary policy transmission theory suggest that higher balance sheet liquidity promotes stronger loan growth and mitigates declines in lending due to monetary contractions (Bernanke & Blinder, 1992; Kashyap & Stein 1995, 2000). Accounting rules effectively alter liquidity characteristics for securities based upon the SFAS 115 classifications. Specifically, marketable securities that are classified as HTM become effectively illiquid due to that classification while non-htm securities maintain their liquidity. Banks face liquidity and other balance sheet management consequences from their classification choices (Beatty, 1995; Hodder et al. 2002). Furthermore, non-htm security value changes are treated as other comprehensive income and there exists a perception that this type of income is overly volatile and detrimental to equity value (Chambers et al., 2007; Bamber et al., 2010). Banks are expected to abide by the intent and ability stated in SFAS 115 and not sell or reclassify HTM securities until they mature, unless very specific conditions are met. None of those conditions include liquidity needs. Inherently liquid securities are rendered effectively illiquid by HTM classification. Selling, or reclassifying, HTM securities for reasons other than those prescribed in SFAS 115 taints the entire balance of HTM securities and the remaining HTM balance must be reclassified as AFS and the 16

27 gains and losses of the reclassified securities included in reported capital. 12 Additionally, selling HTM securities for non-recognized reasons indicates the securities were misreported on prior CALL reports; banks can be fined for misreporting. Stein (1998) showed that marketable securities allow banks to sell or call fewer loans when faced with a liquidity shortfall. An implicit assumption is that the securities are liquid and able to be sold. I examine this assumption for HTM securities; HTM securities should be rendered effectively illiquid due to their SFAS 115 classification. Of course, none of the above consequences or liquidity concerns are relevant if the bank s intent was to hold the HTM securities to their maturity and if business conditions have not altered that intent and ability. However, there is some evidence that banks will reclassify when they can do so without penalty. Accounting Statement SFAS 133 allowed banks to reclassify their securities upon adoption (FASB1998). Figure 1 shows a spike in HTM reclassifications and sales when the standard went into effect. Recently, concerns have been raised that banks opportunistically reclassified out of the HTM classification, specifically when SFAS 159 provided the opportunity (Leone 2007; Taub 2007). In Figure 2, I add the average to my Figure 1 graph. The TED spread is the difference between interbank loans (measured as the London Interbank Offered Rate, LIBOR) and short-term U.S. government debt (measured as three-month U.S. Treasuries). An increasing TED spread is indicative of liquidity concerns in the banking sector (Brunnermeier & Pederson, 2009). Figure 2 supports the Leone (2007) and Taub (2007) articles since the TED spread was declining during the period with the highest

28 ratio of HTM sales or reclassifications. Also during 2001, interest rates began declining. After hovering around the 5% to 6% range from 1996 to 2000, rates for U.S. treasuries in the 3-year constant maturity series dropped to lows of around 2% from 2000 through Dropping interest rates lead to higher bond prices; banks may have taken advantage of this change by selling out of the HTM classification. SFAS 115 accounting is meant to capture the intent and ability of the bank regarding its HTM securities. If the accounting was correct and the intent does not change, then there should be no association between loan growth and HTM securities. However, if intent is not constant or if future business conditions (e.g. monetary tightening) were not accurately anticipated, the illiquidity of HTM securities should have an effect on loan growth and that effect is a reduction in the mitigation of loan growth by marketable securities. Therefore, I hypothesize: H1a: During monetary tightening periods, the balance sheet liquidity effect on loan growth is decreasing in HTM securities, relative to non-htm securities. My argument is that there is an interactive effect between the SFAS 115 accounting standard and monetary tightening as specified in H1a. Stein (1998) shows that banks reduce lending when faced with a decline in deposits. After a tightening, banks' loans, securities and deposits decline (Bernanke & Blinder 1992). The loan decline during tightening periods is lower for banks holding greater amounts of marketable securities; banks with more liquid balance sheets (those with more marketable 13 Treasuries of other maturities experienced similar interest rate declines. 18

29 securities) experience less negative impact to their loan portfolio during monetary tightening. (Kashyap & Stein, 2000; Kishan & Opiela, 2000, 2006). HTM securities are effectively illiquid due to their classification and the mitigation of negative loan growth by marketable securities relies upon the liquidity of the securities. While monetary tightening does not increase the illiquidity of HTM securities, the constraint that the illiquidity of HTM securities imposes on a bank's ability to replace lost deposits should be more severe during monetary tightening than non-tightening periods. If my tests support H1a then I will have shown that HTM securities, relative to non-htm securities, are more negatively related to loan growth during monetary tightening periods. However, it may not be entirely clear whether the differential negative effect is due to the HTM classification, the monetary tightening policy, or an interaction of the two (my hypothesis). The main focus of this study is that there is an interactive effect and not just independent effects of tightening or HTM classification on loan growth. To address this concern, I compare the magnitudes of the differential effects of HTM, relative to non-htm, securities, between the tightening and nontightening periods. If there is an interactive effect, then I would expect that the negative differential effect of HTM compared to non-htm securities will be more negative during monetary tightening relative to non-tightening periods. Therefore, I further hypothesize: H1b: The differential effect of HTM securities compared to non-htm securities is more negative during monetary tightening relative to non-tightening periods. 19

30 Large banks tend to have more funding options; they are not as dependent on depositors as smaller banks (Kashyap & Stein, 2000). Since Fed actions typically lead to lower deposits, banks that are less dependent on deposits should be less affected by those Fed actions. In effect, monetary policy research uses size classifications as a natural identification technique to control for the potentially endogenous nature of loan growth. Small banks are less able to obtain external financing to replace the deposits lost after a monetary tightening(kashyap & Stein 1995, 2000; Kishan & Opiela 2000, 2006). Typical size splits are based upon regulatory cutoffs by asset size (Beatty & Laio, 2009) or by size percentiles (Kashyap & Stein 1995, 2000; Kishan & Opiela, 2000, 2006). An additional concern that makes the small versus large comparison important is the effect of lending on job growth. Small business is often called the engine of the economy, and small businesses borrow from small banks. Reductions in loans by small banks hamper the growth of small business (Cowley and Maltby, 2008). If small banks already face greater decline in loan growth, the incremental negative effect due to HTM securities could further restrict economic growth. I follow past monetary policy literature and consider large banks as those in the 98 th percentile by total assets and small banks are those below the 90 th percentile. I hypothesize: H2a: The negative effect of HTM securities, relative to non-htm securities, on loan growth during monetary tightening is more negative for small banks than large banks. H2b: The more negative differential effect of HTM securities compared to non-htm securities, during monetary tightening relative to non-tightening periods is stronger in small banks than in large banks. 20

31 Ratio of HTM sold / (HTM sold + HTM) Figure 1: Average Ratio of HTM sold or transferred annually from 1996 to 2008 Average ratio of HTM sold per bank during years Year HTMsold / HTMsold+HTM The largest spike in HTM sold ratio occurs when SFAS133 goes into effect, In the following figure, I include a proxy for bank liquidity concerns. 21

32 Ratio of HTM sold / (HTM sold + HTM) TED Spread in % Figure 2: Average ratio of HTM securities sold or transferred vs. TED spread Average ratio of HTM sold or transferred by each bank vs. TED Spread 1996 to Year HTM SOLD TED The TED Spread is the difference in interest rate as 3-month LIBOR less 3-month U.S. T-Bills. A higher TED Spread is indicative of liquidity concerns in the banking industry (Brunnermeier & Pederson 2009). The TED spread was declining during the 2001 spike in HTM sold and transferred. The graph shows that, even though there were not encountering liquidity concerns, banks, on average, sold or transferred a large portion of their HTM securities. 22

33 Chapter 4: Sample and Research Design 4.1 Sample Selection For hypothesis testing, I use quarterly bank level data reported in CALL reports from 1996:Q1 through 2008:Q4. The CALL reports provide balance sheet, income statement and schedules for commercial banks. I start my sample with 1996:Q1 because two important things happened between adoption and December First, when SFAS115 went into effect, bank regulators had decided that the unrealized gains and losses created by SFAS 115 classification would be included in the calculating regulatory capital ratios. However, that decision was rescinded in October 1994 and banks no longer had to consider the regulatory capital effects of unrealized gains and losses due to classification (Hodder et al. 2002). Second, in November 1995 the FASB announced an amnesty period. Between mid-november and the end of December in 1995, banks could freely rebalance their securities among the three SFAS 115 classifications. My initial sample is 472,404 bank quarters during the sample period. Restricting to commercial banks 14 reduces the sample to 420,925 observations. As suggested by Holod and Peek (2007), I further exclude commercial banks that are essentially credit 14 Charter code 200 in the CALL reports indicates commercial banks, depository trust companies, credit card companies with commercial bank charters, private banks, development banks limited charter banks and foreign banks. Institutions excluded include central banks, government agencies, regulatory bodies, credit unions, securities brokers, and other non-depository institutions 23

34 card banks (those banks where at least 50% of outstanding loans are credit card balances)which reduces the sample to 418,684 observations. Finally, excluding observations that do not have the financial data for my tests leaves 386,257 bank-quarter observations for 11,355 commercial banks. 4.2 Research Design To examine how HTM securities change the balance sheet liquidity effect on loan growth, relative to non-htm securities, I test the association between changes in bank lending and balance sheet liquidity and the held-to-maturity securities. I use OLS to estimate the following equation to test H1; the model variables are interacted with the Tighten variables. Continuous variables are winsorized at 1 st and 99 th percentile. Standard errors are clustered by bank and by quarter using the Peterson (2009) method based upon Cameron et al. (2006). 15 ChgLoans t = β 0 + β 1 * ChgLoans t-1 +β 2 * ChgLoans t-1 *Tighten +β 3 * ChgLoans t-2 + β 4 *Tighten*ChgLoans t-2 + β 5 *Liq + β 6 * Liq*Tighten + β 7 *HTMA + β 8 * HTMA*Tighten + β 9 *Size+ β 10 * Size*Tighten + β 11 *Tier1Ratio + β 12 *Tier1Ratio*Tighten +β 13 *ChgDeposits + β 14 *ChgDeposits*Tighten + β 15 *Delinquency + β 16 *Delinquency*Tighten + β 17 *ChgUnemp + β 18 * ChgUnemp*Tighten+ β 19 *Tighten+ ε (1) where 15 With a panel of bank-quarters, I expect that residuals of the OLS estimation will not be independent. I could have used firm dummy variables and clustered by quarter; however, using firm dummies in the place of clustering by firm requires that I assume the firm effect is fixed. If the firm effect decays over time, then the firm dummy method will not fully correct for the dependence among residuals. Using clustering when the effect is fixed, however, needlessly reduces sample size since observations are based on clusters rather than individual observations (Peterson 2009). I have no reason to assume a constant firm effect and therefore use clustering rather than firm dummies. 24

35 ChgLoans t : Liq: HTMA: Size: Tier1Ratio: ChgDeposits: Delinquency: ChgUnemp: Tighten: Interactions: Change in loans outstanding as (Loans t -Loans t-1 )/Loans t-1 for all Sum of held-to-maturity, available-for-sale and trading securities and fed funds sold scaled by total assets, all at start of the period. Ratio of held-to-maturity securities / total assets; both at beginning of period. Ln (Total Assets) at start of period. Tier 1 Capital scaled by total assets; both at start of period Change in total deposits during the period (Total Deposits t Total Deposits t-1 ) scaled by total assets at the start of the period. Ratio of non-performing loans / total assets loans at the start of the period. Change in quarterly unemployment during the prior period, from Is one of three proxies for tight monetary conditions: (1) an indicator=1 for an aggregate increase in the fed funds rate over the 2 nd and 3 rd prior periods, 0 otherwise; (2) the six month commercial paper premium (CPP6) and (3) the three month commercial paper premium (CPP3). Interactions of all variables with Tighten measures. Under H1a, I expect (β 7 + β 8 ) to be negative to indicate that during monetary tightening, HTM securities are more negatively related to loan growth, relative to non- HTM securities. H1b predicts that, the differential effect of HTM securities compared to non-htm securities is more negative during monetary tightening relative to nontightening periods. Accordingly, I expect that β 8 will be negative. Besides the variables of interest, I include the changes in total loans for each of the two prior quarters ChgLoans t-1 and ChgLoans t-2, which follows prior loan growth research (Kashyap & Stein 2000; Kishan & Opiela 2000). I include these prior period loan growth measures to control for banks loan growth trends. The Tighten variable is my proxy for monetary policy. I use the changes in fed funds rate for the second and third prior quarters since there is a lag between the change in policy and its impact on bank lending (Bernanke & Blinder 1992). 25

36 In sensitivity analysis, I also test my hypotheses using the six-month commercial paper premium (CPP6) since that measure has also been shown to be a good proxy for monetary conditions (Bernanke & Blinder, 1992; Becher et al. 2008). The CPP6 is measured as the rate on six-month commercial paper less the rate for six-month treasury bills at the end of each quarter. The six-month commercial paper series was discontinued by the Federal Reserve in August, From January 1997 until August 1997, both the new three month series and the new six month series co-existed. I follow Becher et al. (2008) and calculate the average difference in the premium between the six-month measures and the three-month measures during the eight months that both series existed. During that time, the six-month premium was, on average, 2.5 basis points greater than the three-month premium. I then calculated the three-month premium and added 2.5 basis points to derive the six-month premium. I also use the three-month premium (CPP3) in additional sensitivity tests. The three month premium is the difference between the three-month commercial paper rate and the three-month treasury rate. I include the Tier1Ratio to control for capital effects on loan growth; Kishan and Opiela (2000, 2006) show that the bank lending channel is sensitive to capital ratios. Available deposits are another liquid source of funds; therefore, I include ChgDep to control for this source of liquidity (Bernanke & Blinder 1992; Kishan & Opiela 2000). I include Size to ensure I am not capturing size effects. Delinquency is included as a control for the condition of the loan portfolio (Kishan & Opiela 2000). ChgUnemp is my proxy for loan demand. Loan demand decreases during a slow economy and the unemployment rate is an indicator of a slowing economy (Bernanke & Lown, 1991). 26

37 When the Open Market Committee initiates a securities sale, the resulting decrease in reserves leads to an increase in the fed funds rate. Subsequently, other interest rates will also increase. Therefore, I expect the Tighten coefficient to be negative. I further expect that the coefficients on Tier1Ratio, and ChgDep will be positive. Banks that are less concerned about violating regulatory capital requirements will lend more, as will banks with sufficient deposits. I do not predict a direction for the ChgLoans t-1, ChgLoans t-2 and Delinquency. Current loan growth could be positively associated with prior loan growth if the bank is growing its overall loan portfolio. However, the bank could be trying to slow prior growth or reverse a prior contraction in its loan portfolio; under those circumstances, the association between current and prior growth could be negative. The association between Delinquency and loan growth is similarly ambiguous. A negative association could be described as a numerator effect; if the value of non-performing loans is increasing then the bank could perceive a problem in their loan portfolio and reduce future lending However, a positive association could be a denominator effect; the value of total loans is decreasing so the bank increases new lending. For my second hypothesis, I estimate a stacked regression by bank size. I use an indicator Small for banks with total assets below the 90 th percentile threshold. I use another indicator Large for banks in the 98 th percentile of total assets. To test the second hypothesis, I compare the differences between small and large banks of the variables of interest. H2a is that the negative effect of HTM securities, relative to non-htm securities, during monetary tightening will be stronger in small banks than in large banks. 27