A TIME-SERIES ANALYSIS OF US SAVINGS AND LOAN PERFORMANCE: MAJOR TRENDS AND POLICY ISSUES AFTER THE HOUSING CRISIS Mine Aysen Doyran, Lehman College—CUNY1 ABSTRACT The savings and loan (S&L) debacle of the 1980s produced one of the largest financial crises since the Great Depression. Since then scholarly research has devoted substantial attention to profitability and loan performance in US commercial banking. Yet, the implications of these studies for research on non-bank institutions are limited. This paper investigates the determinants of S&L asset quality and loan performance over the period 1978-2009. The aim of this paper is to establish which of these potential determinants of loan performance prevail in the US S&L industry. An equation of loan performance ratio is submitted using first-order differences of variables and least squares trend fitting. Our analysis highlights that higher return on assets (ROA) and ratio of net worth to total assets lead to a lower ratio of poor loan performance. Higher the ratio of consumers to default on their loans, the lower the return on assets and hence less bank profits. In addition, the leverage ratio has a significant negative coefficient on loan performance. By and large, our analysis is consistent with earlier studies that diminishing net worth and under-performing loans have the potential to render S&Ls vulnerable to financial shocks, thus contributing to financial instability. KEY WORDS: S&L Industry, Banking Regulation, Asset Quality, Subprime Mortgage Crisis, Loan Performance

INTRODUCTION Like the S&L crisis of the 1980s and early 1990s, the current financial crisis exposed weaknesses in the U.S. financial system not seen since the Great Depression. During this downturn, financial losses have been largely concentrated in the housing market; mortgage market participants have been hard hit by declines in profitability, increases in bad loans, and customer defaults. Since then the S&L industry has also realized sizable

1

I am are grateful to the Center of Excellence in Teaching and the Liberal Arts Department of the Fashion Institute of Technology of the State University of New York’s for their outstanding and generous support, both financial and moral. I am alone responsible for errors.

1

losses on loans and investments, thus resulting in a further rise in non-performing loans and credit-related write-offs in mortgage lending related sectors.

The S&L industry in general responded to the crisis by using loan loss reserves—an expense set aside as a cushion against customer defaults and bad loans. Thrifts were reported as having set aside a record of $39.3 billion in loss reserves and an additional $5.8 billion in the first quarter of 2009. Yet, reflecting the present housing market downturn and increase in unemployment rates, asset quality and problem loans (delinquent loans plus ―loans in nonaccrual status‖ and ―repossessed assets‖) have increased to a degree not seen since the early 1990s (US Treasury Department, 2010). Institutions of all types and business models have experienced an increase in problem loans especially institutions offering subprime and other types of high-risk mortgages. For example, the FDIC took control of the California based savings and loan association IndyMac in what regulators called the second-largest bank failure in U.S. history.

Asset quality is one of the key indicators of financial viability and overall performance of banking institutions. While S&L associations continue to expand their repertory of savings deposits, mortgage loans and other financial products, the loan performance is typically the principal indicator of their asset quality. Accordingly, loan quality remains a key indicator of their overall condition and earnings. Since the banking crises of the early 1980s, scholarly research has devoted substantial attention to profitability and loan performance in US commercial banking. Yet, the implications of these studies for research on non-bank institutions are quite limited.

2

Accordingly, this paper examines the determinants of asset quality and loan performance in the S&L industry over the period 1978-2009. An equation of asset quality ratio using first-order differences of time-series values is estimated. Using ADF as a statistical test by estimation of least squares trend fitting, the study takes into consideration industry specific as well as macroeconomic indicator as control variables. The ADF (Augmented Dickey Fuller) test is used for detecting the existence of a unit root in autoregressive model and stationary trend fitting. Overall, our analysis highlights that higher return on assets (ROA) and ratio of net worth to total assets leads to a lower ratio of poor loan performance. Higher the ratio of consumers to default on their loans, the lower the return on assets and hence less bank profits. In addition, the leverage ratio has a significant negative coefficient on loan performance. By and large, our analysis is consistent with the view that decreasing net worth and under-performing loans have the potential to render S&Ls vulnerable to financial shocks, thus contributing to financial instability.

The rest of the article is organized as follows. The second section examines historical developments in the S&L industry asset structure and underwriting standards since the beginning of the current downturn. It discuses the significance of recent market and financial innovations and assesses the role of regulation in light of historical trends. The third section reviews the literature on asset quality and mortgage loan performance. The fourth section introduces the data and illustrates the methodology and hypothesis; the fifth section presents the empirical results. The sixth section briefly discusses the implications of the federal loan modification programs for the prudential regulation of S&Ls, which have sought to help households in foreclosure and prevent delinquency. The

3

final section draws strategic lessons from this experience for future researchers and practitioners in the field of risk management.

BACKGROUND: THE S&L INDUSTRY PERFORMANCE DURING THE FINANCIAL DOWNTURN Like commercial banks, the S&L industry went through radical changes during the late 1980s and early 1990s. Because of rapid market innovations and deregulatory measures instituted by the US government, S&Ls began to offer a more diversified range of services than ever before. With their new freedoms, they were able to enter a wider range of businesses that they might have otherwise sought, such as commercial lending, trust services and nonresidential consumer lending. In 1980, Depository Institutions Deregulation and Monetary Control Act was passed, increasing the deposit insurance coverage for S&Ls from $40,000 to $100,000. During this period, many S&Ls failed due to engagement in large-scale speculation, especially in real estate (Balderston, 1985). However, a more unstable period ensued with the closing of the Resolution Trust Corporation (RTC) in mid-1995—a government agency responsible for resolving 747 thrifts with total assets of $394 billion. From January 1986, to December 1995, the US S&L industry lost nearly 50% of its total institutions—a period during which number of federally insured S&Ls declined from 3,234 to 1,645 (Curry and Shibut, 2000)

During the 1990s, the S&L industry was shaped by the rapid growth of subprime mortgage lending. Various federal government actions as well as market forces fostered this development. The later revisions to the Community Reinvestment Act (CRA) of

4

1977 chiefly offered banking institutions deregulatory incentives to provide loans to lowand moderate-income borrowers, mainly those classified as subprime borrowers. Furthermore, the ―Federal Housing Administration, which guarantees mortgage loans of many first-time borrowers, liberalized its rules for guaranteeing mortgages, increasing competition in the market and lowering interest rates faced by some subprime mortgage borrowers‖ (Gramlich, 2004). This group of borrowers included customers who were previously denied credit for real estate and other consumer loans. While subprime lending has provided new opportunities for increased homeownership among high risk borrowers, it was also associated with poor loan performance related to higher levels of delinquency, foreclosure, and in some cases predatory and discriminatory lending practices.

While high interest rates and inflation during the early 1980s hampered financial growth, mortgage loan volume has shown a dramatic increase since 2000. The recent financial expansion illustrates the ―confluence of rising borrower demand, historically low interest rates, intense lender competition, innovations in the structure and marketing of mortgages, and an abundance of capital from lenders and mortgage securities investors‖ (Angell and Rowley, 2007). However, a booming housing market came at a cost of deteriorating credit quality in light of lending practices. While a lack of oversight over mortgage lending and deregulation were in part to blame for risky loans to homebuyers, the securitization model generally involved larger risks. This was particularly true for S&Ls which traditionally specialized in processing conventional home mortgages rather than securitizing subprime loans with adjustable rates. The elimination of Regulation Q

5

in 1980 shifted the mortgage market dramatically from savings institutions to commercial banks and to government-sponsored enterprises (GSEs). GSEs led the expansion of secondary mortgage market by securitizing mortgages in the form of mortgage-backed securities (MBS). This was the market for the sale of securities/bonds backed or collateralized by the value of mortgage loans. By 2005, nearly 68% of home mortgages were securitized. In 2005, total private-label MBS accounted for 29% of gross outstanding MBS in 2005, doubling its share from 2003. During the same period, twothirds of private-label securitizations included nonprime loans, increasing from 46% in 2003 (Angel and Rowley, 2007). With the increased exposure to competition from other lenders, S&Ls appeared willing to assume greater risk in their search for high profits. Chart 1: S&L Share of Mortgage Originations, $ in Millions 900000 800000 700000 600000 500000 400000

MORIG

300000

Source: OTS, 2010

200000

100000

2008

2006

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

1978

0

Mortgage securitization enables lenders to shift credit risk and interest rate risk to investors who have greater risk forbearance. As a recent by product of secondary 6

mortgage markets, ―private-label‖ MBS’s are securitized by institutions other than government-sponsored enterprises and therefore lack any explicit or implicit backing. Since subprime loans lack an explicit guarantee, they elevate credit risks. Lenders naturally take into account a borrower’s credit history when determining loan eligibility. When ―compared with prime loans, subprime loans typically have higher loan-to-value ratio, reflecting that subprime borrowers face in making down payments and the tendency of these borrowers to seek equity during refinancing‖. Because of their tendency to go into delinquency or default more often, subprime loans also carry higher interest rates than prime loans. While subprime lenders are subject to certain lending standards, fraud, abuse, and predatory lending problems are also problematic features of the subprime market (Gramlich, 2004).

Yet none of these problems prompted subprime regulation. Over the period 1994-2003, subprime mortgage loan originations increased by a rate of 25% per year, nearly a tenfold increase in nine years. Similarly, subprime originations as a share of total mortgage originations grew from 4.5% in 1994 to 13.2% in 2000 and 8.8% in 2003. In 2002, only 5 commercial banks and 11 S&Ls offered subprime loans though these five banks were large enough to represent 27% of the mortgage loans of the subprime lenders. Likewise, 35 affiliates of financial holding companies (for example, City Financial) constituted only 19% of the subprime lenders but accounted for 43% of the subprime loans. 11 S&Ls constituted only 6% of subprime lenders but accounted for 13.8% of total subprime mortgage loans. On the other hand, there were 135 independent mortgage companies, but they only accounted for 12%, the relatively small share of subprime mortgage loans

7

(Gramlich, 2004). Chart 1 plots the S&L share of mortgage originations during the period 1978-2009. As can be seen, the mortgage originations illustrate some cyclicality but no apparent trend in the 1980s and 1990s. It is important to note, however, that mortgage originations increased sharply during the housing boom and then decreased during the cyclical downturn of the early 2007 to levels statically below that of the S&L crisis period.

Because of their larger share of the mortgage securities market, U.S. banks and securities trading firms suffered the greatest losses when the market collapsed in 2008. Federally insured S&Ls specialized in nonprime mortgages were also impacted. American Home Mortgage, which operated as a real estate investment trust, collapsed and filed for bankruptcy on August 6, 2007. In January 2008, Bank of America acquired Countrywide Financial, the largest U.S. mortgage lender, for $4 billion after its stock prices had dropped 80% in value since 2007 (Morgenson, 2007; Mildenberg, 2008). In April 2007, New Century Financial Corporation, another real estate investment trust and second biggest U.S. mortgage lender, filed for bankruptcy after effectively reducing its labor force by 54% (CNN, 2007). IndyMac—a Pasedena based savings banks specialized in Alt-A loans— also failed in July, 20008 and was seized by FDIC (Hudson, 2008). IndyMac’s collapse was estimated to ―cost the FDIC between $4 billion and $8 billion, potentially wiping out more than 10% of the FDCI's $53 billion deposit insurance fund‖ (Palette and Enrich, 2008). IndyMac’s failure was the third largest bank collapse in FDIC’s history after Continental Illinois in 1986 and First Republic Bank in 1988.

8

FINANCIAL REGULATION AND LOAN QUALITY PERFORMANCE IN THE S&L INDUSTRY S&Ls have remained key facilitators of home mortgages and savings deposits throughout American history. Yet there are few studies of the financial condition and performance of savings and loans associations (Brigham, 1964; Benston, 1972; Verbrugge, Shick and Thygerson, 1976; Geehan and Allen, 1978; Bradley, Gabriel and Wohar, 1995; Kaushik and Lopez, 1996; Jahere, Page and Hudson, 2006). While the bulk of the research has concentrated on the causes and consequences of the S&L crisis, savings and loan studies have drawn largely from US banking experts. Since the S&L crisis, research has shifted away from this area. However, the collapse of Indy Mac in July 2008, the largest S&L in California specializing in Alt-A loans, have once again raised concerns about the viability of S&Ls in their present form.

There are numerous investigations of mortgage lending institution performance, particularly of commercial banks (Benston, 1972; Berger, 1995; Bourke, 1986). Most of these studies, however, have examined the determinants of profitability rather than loan or asset performance. Therefore, they used a number of indicators (return on assets, return on equity) as proxies for profitability and regressed it with bank-specific, industryspecific and macroeconomic (external) indicators (Rasiah, 2010). Additionally, the impact of market and financial structures on bank performance has received considerable attention among American and European researchers (Short, 1979; Gilbert, 1984; Bourke, 1986; Goldberg and Rai, 1996; Casu, Girardone and Molynuex, 2004; Demirguc-Kunt and Huizinga, 2000). Further applications of profitability have been

9

critical in orienting methodology towards ―single country studies‖ and ―panel country studies‖ (Naceur, 2003). Because of the emphasis on the role of market structures, most of the studies fall under the Structure-Conduct-Performance (SCP) paradigm. Within panel country studies, for example, explanatory variables unique to each country and explanatory variables unique to each bank were highlighted in order to address the relationship between market structures and bank performance.

Research on S&L performance has drawn primarily from US banking studies. Benston (1972) investigated scale economies in a panel study of 83 commercial banks and 3159 S&Ls. Deposit and loan structure were used as proxies for bank size. He found that greater operating costs were accompanied by larger size, but marginal cost increased at a decreasing rate for branch banking. This illustrated the benefits of scale economies, especially in demand deposit and real estate loans. A similar analysis of that period by Verbrugge, Shick and Thygerson (1976) showed the effects of both bank-specific and regulation variables on profitability measured as return on net worth. They claimed that usury laws lessened fee income in S&Ls purchasing rather than servicing loans. This led to less profits yet marginally decreasing operating costs. The authors also found loan composition (―multi-family and other higher-risk non-single family‖) to be positively correlated with operating costs.

In research into the effect of S&L crisis on ―mortgage-credit intermediation‖, Bradley, Gabriel and Wohar (1995) sought to examine the root causes of housing sales and purchase cycles. Their results indicated a highly correlated relationship between deposit

10

flows and S&L intermediation of mortgage credit. Their analysis also shed light on the increasing disintermediation as a function of maladjustment to credit flows due to deposit rate ceilings and other structural obstacles. This has led to sharp reductions in the provision of mortgage credit, serving to increase the mortgage-Treasury interest rate spread and facilitating the crash. Policy lessons are drawn for the impact of the crisis on the housing sector. Results cast doubts on fixing the S&L crisis through increased bailouts of remaining S&Ls.

While not directly related to the S-C-P paradigm, financial regulation is a structural variable in most banking studies. This type of analysis extends beyond market structures by taking into account the larger regulatory and policy environment of banks. The premise of structural analysis is that the organizational make-up of the US financial system impacts the viability of savings and loan associations. The claim is that inefficiencies leading to inadequate oversight and poor performance stem from a fragmented regulatory environment (Blair and Kushmeider, 2006; Kushmeider, 2005; Matasar and Pavelka, 1998). Unlike other countries, the US regulatory structure is marked by deference to competition, a dual banking system, federalist politics and the choice of charter. In this system, a bank can designate which agency serves as its primary regulator through its selection of a chartering source (federal or state) and choosing to join the Federal Reserve System. It has long been claimed that the choice of regulatory authority (ie, the ability of banks to choose their primary regulator) has produced a clientele relationship between the regulators and banks. Clientalism implies giving bank regulators the exceptional power and ―authority‖ from the reporting banks. Federal

11

regulators compete with each other to woo banks into choosing their chartering authority so as to increase revenues from the number of reporting banks—a phenomenon of regulatory capture known as ―competition in laxity‖ (Matasar and Pavelka 1998:57).

S&L industry performance is largely shaped by this contradictory relationship between the state and federal level of regulation. Similar to the commercial banking industry, the S&L industry organizes under a dual charter system. States provide a savings and loan association (S&L) charter while some states also provide a savings bank charter. This has allowed financial interests to switch charters and ―shop for‖ the regulator that sets the lowest standards. Policing banks runs the risk that regulatory agencies will lose the fees they rely on from the very institutions they are charged with regulating. At the federal level, the OTS offers both a federal S&L charter and federal savings bank (FSB) charter. All state-chartered thrifts (S&Ls and savings banks) are subject to regulation by their state chartering authority (such as state banking department) and also by a federal regulator—the OTS in the case of state-chartered S&Ls and FDIC in the case of statechartered savings banks. State-chartered banks enjoy the advantage of being more leniently regulated than national banks under OCC. This is mainly because state authorities alternate with federal regulators (FDIC and Federal Reserve) who do not charge fees for examining the books of state banks. The OCC, on the other hand, must cover the full cost of supervision. State-chartered saving banks, not charged by FDIC for examination, are more cost efficient than state-chartered S&Ls which pay a double supervisory assessment fee–both to their state-chartering authority and the OTS (Blair and Kushmeider, 2006). Cost effectiveness highlights why S&Ls became savings banks

12

with the hope of more leniently regulated. It also highlights the decreasing number of mortgage providers or increasing proliferation of lending categories in the S&L industry.

DATA AND ECONOMETRIC ANALYSIS Most of the structure-performance studies have looked at profitability as a measure of a bank’s financial performance or overall condition. While this relationship is recognized in research that associates profitability with a variety of internal and external factors, the discussion of loan performance received relatively little attention. However, loan performance is difficult to measure due to the diversity of bank assets. Since S&L associations are technically speaking providers of loan and deposit accounts, it is important to understand their function in the financial system. Primarily, S&Ls are ―specialized mortgage lenders with considerable expertise in evaluating potential borrowers, establishing long-term relationships with customers and designing loan agreements that minimize adverse selection‖ (Bradley, Gabriel and Wohar, 1995:478). Although governmental policies supported the diversification of S&L assets through deregulation in order to address the ongoing crisis, S&Ls historically remained as originators of home mortgages. As Chart 2 illustrates, the average amount of single and multifamily loans between 1978 and 2009 was $521.7 billion comprising nearly 76% of all loans in those categories (OTS, 2010).

Other measures of S&L credit provision (the share of non-home mortgage loans in construction, non-residential, consumer, commercial and land) are shown in Chart 2. These series diverged significantly for the second time during the cyclical downturn of

13

the early 2007, largely affecting the loan performance and resulting in a sharp decline in S&L provision of mortgage credit (Chart 1). This certainly reflected the overall decline in the supply of credit relative to credit demand and contraction in the housing market. The reduction in the supply of credit resulted in higher mortgage interest rates but also in higher ratio of non-performing to total loans or total assets. Therefore, the loan portfolio performance appears as one of the most import proxies for financial well being of this industry. To highlight the significance of this issue, the following section puts forward an equation using a set of financial ratios that are expected to affect the loan performance of the S&L industry during the period 1978-2009.

Chart 2: Average Assets by Loan Type, 1978-2009 ($ In Millions) 600000 500000 400000 300000 200000 100000 0

The data for this study is obtained from the Office of Thrift Supervision (OTS) database as well as the World Bank and Pen World Table at the University of Pennsylvania. The 14

data set consists of aggregate information for which a variety of financial ratios were calculated. Most savings associations, also known as thrifts, are small, private and customer oriented institutions committed to providing home mortgages and other retail lending to communities. The OTS, a bureau of the Department of the Treasury, is the federal regulator of S&Ls (state or federal charter) and the holding companies that own them (such as AIG). In 2009, the OTS supervised 765 thrifts with assets of $941.7 billion. Although there are sizable indicators of loan performance going back to 1964, we were only able to collect information between 1978 and 2009 for which data was found: Ratio of non-performing loans to total loans (DML_TLOANS) Return on assets (ROA), leverage ratio (LEV), Real GDP capita income (RGDPL), growth rate of Real GDP (GROWTH_RATE_GDP), ratio of net worth to total assets (NW_TA) and ratio of total liabilities to new worth (TLI_NW). The source for the financial ratios is Office of Thrift Supervision (OTS), 2009 Fact Book: A Statistical Profile of the Thrift Industry. Timeseries for RGDPL and GROWTH_RATE_GDP were obtained from the Pen World Table of the University of Pennsylvania (1978-207) and World Development Indicators & Global Development Finance (2008-2009) of the World Bank database.

What are the determinants of loan performance as measured by the ratio of total nonperforming (delinquent) loans to total industry loans? Our analysis seeks to evaluate whether the specified financial ratios since 1979 have led to a loss of asset quality or poor loan performance specified by DML_TLOANS. The aim of this analysis is to establish which of these potential determinants of loan performance is dominant in S&L industry. The analysis is based on a two-steps least squares regression. In the first equation, we test

15

the existence of unit roots in levels for control variables that are used to predict the value of loan performance. In our second equation, we filter the effects of variables of unit roots on the dependent variable by conduct of an ADF test. The financial ratios are in accord with previously published studies in the field. We use the ratio of non-performing loans to total loans (DML_TLOANS) as dependent variable; explanatory or control variables are return on assets (ROA), leverage ratio (LEV), Real GDP capita income (RGDPL), growth rate of Real GDP (GROWTH_RATE_GDP), ratio of net worth to total assets (NW_TA) and ratio of total liabilities to new worth (TLI_NW). Meanwhile, macroeconomic indicators of Real GDP capita income (RGDPL), growth rate of Real GDP (GROWTH_RATE_GDP) are used as external variables. The calculation of the external ratios (macroeconomic variables) was obtained from Demirguc-Kunt and Huizinga (2000). We calculated industry endogenous ratios from the works of banking industry experts (Verbrugge, Shick, and Thygerson, 1976; Gallick, 1976; Berger, 1995; Chaudhry, Chatrath, and Kamath, 1995; Pervan, Pervan and Guadagnino, 2009; Papanikolaou and Wolff, 2010).

Inter-temporal factors might misrepresent variables on financial sector performance in a time-series analysis. Therefore, least squares trend fitting is necessary to capture the significance of variables under consideration. When data is observed over a defined time frame, autocorrelation may occur where the preceding and successive values of timeseries are highly correlated. This is due to the unit root in levels that cause non-stationary trend in the mean. Variables with unit roots exhibit non-stationary or trending behavior that also cause ―serial correlations‖ over time (Cromwell, Hannan, Labys, Terraza,

16

1994:23). Under panel data estimation, on the other hand, cross-sectional differences across units of observation may not be easily estimated due to variations in industry reporting standards. Since S&Ls service relatively homogeneous products (mainly real estate and mortgage loans), the shortcoming in both types of estimations can be filtered with a proper statistical technique. An ADF test (Augmented Dickey–Fuller Test) is a well-known co-integration procedure that tests the existence of a unit root in a time-series data. Since much of time-series theory is concerned with stationary time-series, an ADF test is conducted to filter non-stationary behavior by means of first or second differencing equations.

Our data temporal reference, t, in this case for a year, and i for parameter estimates, variables, with autoregressive model of order.

t

captures the random error in time

denoted by white noise (residual) and Y and x are the observed value of time-series at time t. Many macroeconomic variables such as Real GDP have unit roots in levels. As a result, they exhibit trending behavior that results in high R Square values and t-ratios with insignificant statistical meaning. To achieve our goal of transforming a nonstationary series into a stationary one, an ADF (Augmented Dickey Fuller) test is applied to the regression residuals of auto-correlated time-series. This is done by inclusion of lagged values of Y (DML_TLOANS) where

Y is the first difference of the variable Y,

indicating Y minus its one period prior value. The model parameters of the study are indicated below where regression is performed in terms of

YDML _ TLOANS Yt

0

x

0

i t i

x

1 t,1

x

2 t,2

t

i 6

17

x

3 t,3

4

Y rather than Y:

x t,4

x

5 t,5

x

6 t,6

t

Our regression equation estimates loan performance in terms of YDML _ TLOANS . As the proxy for asset quality, the loan performance ratio is computed by dividing the amount of delinquent loans by total loans. Delinquent mortgage loans are a form of non-performing loans for which the borrower has failed to make payments as specified in the loan agreement. If the borrower can’t pay the mortgage within a certain time period, the lender can start foreclosure proceedings. Foreclosure starts only after the borrower has completely defaulted on his or her payments. As a result of this lag factor, which may misrepresent the true extent of delinquency, mortgage foreclosures were omitted from the analysis. ROA is computed by dividing the net income over total assets. The leverage ratio (LEV) is calculated as the ratio of debt to equity capital. Due to lack of debt figures, debt was calculated as the ratio of (total liabilities minus equity) to equity. The ratio of net worth to total assets (NW_TA) signifies the amount of debt a company has, as net worth is calculated by the difference between total assets and total liabilities. The higher the ratio the lower the amount of debt a bank has. The ratio of total liabilities to new worth (TLI_NW) is an indicator of long-term debt since it implies the extent to which the net worth of the enterprise can offset its liabilities. The higher the ratio, the lower the ability of the enterprise to retain net worth or balance debt. On the other hand, macroeconomic variables such as real GDP per capital income (RGDPL) and the annual growth rate of Real GDP (GROWTH_RATE_GDP) are external factors that might affect the long-run performance or profitability.

The unit root test gives the researcher an opportunity to re-estimate the slope coefficients of variables in order to de-trendise a time-series or make it stationary. After identifying

18

variables with unit roots in levels, one can apply the first or second difference operator to the auto-correlated time-series data. If the first operator shows the differenced time-series to be stationary, then one can apply ordinary least squares to these variables to estimate the slope coefficients. The ADF test specified that first level was required for all variables in order to induce stationary. In a series of unit root tests, the coefficients did not show the expected sign of significance in the level. Accordingly, our data was nonstationary and required transformation. A regression equation was then re-estimated (below) taking first difference of variables that had unit roots in levels. The testing procedure for the ADF test is the same as other autoregressive models, but it was mainly applied to the first difference operator. Re-estimated least squares (with ARMA) using DML_TLOANS as dependent variable: Estimation Equation: ===================== D(DML) = C(1) + C(2)*D(LEV) + C(3)*D(ROA) + C(4)*D(RGDPL) + C(5)*D(GROWTH_RATE_GDP) + C(6)*D(NW_TA) + C(7)*D(TLI_NW) Substituted Coefficients: ===================== D(DML) = 466.3003279 - 8116.863112*D(LEV) - 599187.3712*D(ROA) + 1.218213703*D(RGDPL) - 85.37358754*D(GROWTH_RATE_GDP) 473451.7964*D(NW_TA) + 7096.695556*D(TLI_NW) Re-estimated least squares (with ARMA) using DML as dependent variable: Estimation Equation: ===================== D(DML_TLOANS) = C(1) + C(2)*D(LEV) + C(3)*D(ROA) + C(4)*D(RGDPL) + C(5)*D(GROWTH_RATE_GDP) + C(6)*D(NW_TA) + C(7)*D(TLI_NW) Substituted Coefficients: ===================== D(DML_TLOANS) = 0.001572363527 - 0.01118915311*D(LEV) 0.6273853931*D(ROA) - 1.075641852e-006*D(RGDPL) - 6.21887169e006*D(GROWTH_RATE_GDP) - 0.3667931234*D(NW_TA) + 0.01006688646*D(TLI_NW)

19

DISCUSSION OF FINDINGS This section discusses loan performance in the US Savings and Loan Industry over the period 1978-2009. A review of variables shows some variations in our data. Trends in asset quality, earnings and profitability reflect the continuing US business cycle and housing market weakness. From 1978 to 2009, non-performing loans increased by 506.065% against an increase of 89.369% in total industry assets. On the other hand, total industry assets decreased by 21% over the period 2007-2008 to $1.20 trillion from $1.51 trillion, reflecting the loss of one big S&L that failed during the period (OTS, 2009). When loans are past due by 30 or 89 date, they indicate the borrower’s failure to pay monthly mortgage on due dates and therefore are classified as delinquent or nonperforming in this study. During the same period, the average number of delinquent loans stood at $17,189.12 (Table 1) while the amount of non-performing loans increased from $3.8 billion in 1978 to $23.1 billion in 2009. Max amount was $41.5 billion and minimum amount was $3.8 billion. Chart 3 discusses trends in non-performing loans as measured by the amount of delinquent loans in the same period and captures the impact of business cycle on asset quality and loan performance. The cyclical component of time-series indicates that non-performing loans reached the highest levels at the peak of the S&L and sub-prime mortgage crises, and then started to decrease. According to linear trend line, it seems that this trend will continue over time.

Reflecting the degree to which the credit supply was contracting, mortgage originations began to decrease at an increasing rate since the beginning of the subprime financial crisis (Chart 1). In 2008, total industry mortgage originations (multifamily and

20

nonresidential mortgages) were $404.9 billion, decreasing by 43% from $716.2 billion in 2007. In the fourth quarter of 2008, total mortgage originations decreased to $63.2 billion from $166.6 billion in the fourth quarter one year ago. Since they represent the largest category of loans in the S&L industry, single family (1-4) loans were impacted the worst. In the fourth quarter of 2008, 1-4 family mortgage originations by S&Ls were $52.4 billion, down 64% from $143.9 billion in the fourth quarter of one year ago (OTS, 2009:4).

Chart 3: Delinquent Mortgage Loans ($ In Millions) 45000 40000 35000 30000 25000 y = 217.1x + 13607

20000

DML Linear (DML)

15000 10000

Source: OTS, 2009

5000 0

Chart 4 shows average net charge-offs by loan type. In 2007-2009, while single and multi-family charge-offs constituted 55% of total charge-offs, net-charge offs in consumer and commercial loans were also significant (45%) highlighting the significance of non-collectible loans in non-mortgage category. Charge-offs arise when a bank is unable to collect some of its loans and therefore are subsequently written-off or reported as a ―bad debt expense‖ on a company’s financial statement. Net-charge off appears as a

21

form of debt or impairment of assets and negatively affect earnings, also resulting in a ―write-down‖ of some of the bank’s assets. Since total aggregated data goes back to 1998 only, we could not include net charge-offs into our regression analysis for estimating loan performance. In addition, net charge-offs are form of non-collectible loans so they are not essentially different from delinquent or non-performing loans. Therefore, including this variable might have caused multi-colinearity (very high R Square) essentially measuring the same variable as DML_TLOANS.

Chart 4: Avrg. Net Charge-Offs 2007-2009

Multi Fam Charge-Offs; 2%

Consumer Charge-Offs; 30%

Single Fam Charge-Offs; 53%

Commercial Charge-Offs; 15%

Source: OTS, 2009

Table 1: Basic Statistics of Variables Var. Mean Max.

DML_TLOA NS 0.0246223 0.0510772

Min.

Std. Dev.

LEV

ROA

15.533 95 34.294 12

0.003364 9

0.0088065

7.3286 2

0.0118789

7.9055 26

0.012577 0.013206 8 0.006934 1

RGDP L 33057. 61 42897. 42

GROWTH_G DP 2.875 7

NW_TA 0.06706 18 0.10737 33

TLI_N W 16.558 2 35.294 12

17189. 72

DML

41494

24160. 93

-3

0.02755 27

8.3133

3825

6013.2 77

2.12132

0.02427 43

7.9765 05

10565. 89

Table 2 shows the parameters for an autoregressive model where DML_TLOANS is a

22

measure of loan performance/asset quality. The same variables appear in Table 3 where DML is a dependent variable of loan performance. We used DML without ratios in order to detect the degree of autocorrelation. In the first specification, where DML_TLOANS is the dependent variable, there is a low level of positive autocorrelation with DurbinWatson statistic of 2.272113. In the second model, autocorrelation is almost non-existent at the significance of 1.998551 Durbin Watson value. Using DML instead of DML_TLOANS as dependent variable allowed lower autocorrelation and relatively significant ratio of net worth to total assets (NW_TA). Both specifications, however, are significant at 1% level based on Probability (F Statistic). Overall, these measure the mutual relationship between the predictor variables and dependent variable in each model. Based on R Square values, the right hand side variables explain the dependent variable by almost 44% and 50% and the F statistic supports the regression. Probability (F-Statistic) suggests that both regression models are significant at a 1% level, so we can be reasonably confident that the good fit of the equation is not due to chance.

Table 2: Parameters for an autoregressive model Dependent Variable: D (DML_TLOANS) Method: Least Squares; first difference operator Sample (adjusted): 1979 2009 Included observations: 31 after adjustments Variable

Coefficient

Std. Error

t-Statistic

Prob.

C D (LEV) D (ROA) D (RGDPL) D (GROWTH_RATE_GDP) D (NW_TA) D (TLI_NW)

0.001572 -0.011189* -0.627385** -1.08E-06 -6.22E-06 -0.366793 0.010067**

0.000977 0.003416 0.269874 6.68E-07 0.000385 0.251581 0.003177

1.609487 -3.275475 -2.324735 -1.611253 -0.016149 -1.457953 3.168480

0.1206 0.0032 0.0289 0.1202 0.9872 0.1578 0.0041

R-squared

0.440618

23

Mean dependent var

0.001000

Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat

0.300773 0.006353 0.000969 116.8040 2.272113

S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)

0.007597 -7.084131 -6.760327 3.150752 0.020116

(*Significant at 1% or 0.01 level; ** Significant at 5% or 0.05 level; ***Significant at 10% or 0.1 level)

The regression analysis in Table 2 indicates that variables except RGDPL, GROWTH_RATE_GDP and NW_TA are significant in explaining loan performance at 1% and 5% respectively. This seems to indicate that macroeconomic variables like GDP per capita income have no impact on the likely of borrowers to default on mortgage loans. While this seems to be inconsistent with Table 3, where DML is dependent variable instead of ratio value, the coefficient of RGDPL is barely significant at 10% level or P value of 0.0927. The unimportance of GDP per capita and Growth rate of GDP is unexpected given that they are regarded as major financial crisis indicators at the country level (Economics of Crisis, 2011). The impact of GDP on bank performance has received attention in Demirguc-Kunt and Huizinga (2000) who found a link between economic development and bank profitability. Banks in well-developed markets face tougher competition but lower profitability. Yet, greater financial market development is correlated with higher bank profits and net interest margins in less developed financial systems. Applying this interpretation to our analysis, it is plausible to say that higher profits lead to higher mortgage originations because it allows lenders to borrow more capital in order to originate more mortgages; greater economic growth generates profitable banking. Yet it is not conclusive from this analysis whether the determinants of loan performance are bank-specific (profits) or macroeconomic (financial market

24

development). It was difficult to establish this factor due unavailability of appropriate data. Table 3: Parameters for an autoregressive model Dependent Variable: D (DML) Method: Least Squares, first difference operator Sample (adjusted): 1979 2009 Included observations: 31 after adjustments Variable

Coefficient

Std. Error

t-Statistic

Prob.

C

466.3003

969.2582

0.481090

0.6348

D(LEV) D(ROA) D(RGDPL) D(GROWTH_RATE_GDP) D(NW_TA) D(TLI_NW)

-8116.863* -599187.4** 1.218214** -85.37359 -473451.8** 7096.696*

2355.956 284622.0 0.695624 297.2539 260465.2 2174.002

-3.445253 -2.105204 1.751253 -0.287208 -1.817716 3.264346

0.0021 0.0459 0.0927 0.7764 0.0816 0.0033

R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat

0.497549 0.371936 5470.263 7.18E+08 -306.8397 1.998551

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion F-statistic Prob(F-statistic)

624.4194 6902.496 20.24772 20.57152 3.960970 0.006804

(*Significant at 1% or 0.01 level; ** Significant at 5% or 0.05 level; ***Significant at 10% or 0.1 level)

There is evidence that non-performing loan as a share of total loans is an important indicator of loan performance and subsequently liquidity crisis in the banking sector. Since lenders generate income by writing loans, the price of these loans cover operating costs and generates a profit. If a borrower fails to make payments for a long period of time, the bank loses income and categorizes the debt as non-performing loan. High ratio of non-performing loans can further cause contraction in lending, even causing the share prices to fall. Yet the larger determinants of poor loan performance seem to be industry and bank related as well as macroeconomic. Loss of income, for example, is a

25

macroeconomic factor that can certainly lead to foreclosure because its direct effect on a borrower’s ability to make mortgage payments. It is plausible to say that job loss or unemployment is inversely correlated with bank earnings through increased mortgage defaults and foreclosures during a financial crisis. Due to unavailability of data, however, borrower’s income and socio-economic variables had to be omitted from the analysis.

As clearly seen in Table 2, leverage ratio (LEV), return on assets (ROA) as a proxy for profitability and ratio of total liability to net worth (TLI_NW) are significant in explaining S&L loan performance. While LEV and ROA have a negative coefficient sign, TLI_NW is positively correlated with DML_TLOANS. The most statistically significant variables in Table 3 are LEV and TLI_NW at 1% with p values of 0.0021 and 0.0033 respectively. Leverage coefficients vary in significance depending on the loan performance used but in both regressions carry negative signs, indicating inverse relationship with respect to loan performance. This seems to indicate that higher the nonperforming loans as a share of total loans (DML_TLOANS), the lower the leverage and vice versa. Similarly, higher the ratio of DML_TLOANS, lower the return on assets (ROA), in other words less bank profits.

Given that profits are integral part of bank earnings, we expect non-performing loans to decrease as profits increase. Everything else remaining equal, there is evidence that the ratio of total liabilities to net worth (TLI_NW) affect loan performance positively. Measuring the net worth of the enterprise to offset its liabilities (debt), TLI_NW reflects on the relationship between assets and liabilities. Higher ratio is an indicator of higher

26

debt and thus less net worth. A bank with higher debt is also likely to display a higher ratio of non-performing loans to total loans. As DML_TLOANS increases, the ratio of net worth to total assets of the company decreases, consistent with negative coefficient sign of NW_TA in Table 3. Overall, our analysis is coherent with the view that diminishing net worth and under-performing loans have the potential to render S&Ls vulnerable to financial shocks, thus contributing to financial instability.

POLICY

IMPLICATIONS

FOR

LOAN

PERFORMANCE

AND

PROFITABILITY We can make some recommendations at the industry and country level that will help improve the safety and soundness of loan portfolios at remaining S&Ls. A numbers of forces affect the performance of financial intermediaries. While some of these forces are external to institutions such as trends of interest rates and the strength of the economy and regulations, some of them are internal. Internal forces reflect managerial capabilities and the usefulness of operating policies and procedures. Return on Assets (ROA) and ratio of total liabilities to net worth are important internal factors since they were found to be complementary with loan performance. Another factor is loss provisions that represent the reserves set aside for potential loan defaults. One possible way of protecting the safety and soundness of S&Ls is to increase provisions for losses in interest bearing assets. While S&L loss provisions have been consistently increasing since 1991, from $4.9 billion to $19.5 billion in 2009, they were the highest level on record in 2007 ($11.6 billion), 2008 ($39.3 billion) and 2009 (OTS, 2010) Higher than average levels of loss provisions reflected the credit cycle and persistent declines in home prices rather than a

27

standard policy. It would make sense to keep the average loss provisions substantial in order to cushion the effect of loan defaults on bank soundness and safety. They key question is whether sizable additions to loan loss reserves dampen earnings/profitability or filter the impact of poor loan performance is difficult to predict. The impact of loss provisions could have been estimated if data existed back to 1978.

The quality of loan portfolio reflects the degree of credit risk associated with an asset (such as MBS) and maps to the overall riskiness of an institution. The improvement of the loan quality of S&Ls needs to be based on reinforcement of the supervisory standards through national regulation programs. Adequately regulating the proportion of nonperforming loans to total loans and monitoring the size of leverage are essential in controlling risk. It is necessary to frequently monitor the adequacy of Loan Loss Provisions concomitantly with risk management processes and internal regulations at financial institutions.

This study only looked at one type of mortgage as a proxy for loan performance— delinquent—when ―the borrower has missed one or more scheduled monthly payments‖ There are other types of classifications that were not included due to lag factor that might exaggerate the size of poor performance. For example, default ―happens when is 90 or more days delinquent. At this point, foreclosure proceedings against the borrower become a strong possibility‖. Foreclosure is when ―the borrower has been delinquent for more than 90 days, and the lender has elected to foreclose in what is an often lengthy

28

process with several possible outcomes. For instance, the borrower may sell the property or the lender may repossess the home‖ (GAO, 2009:6-70).

In addition to adequately regulating loan performance and soundness at S&Ls, policy interventions geared towards borrowers might be appropriate. Since the start of the housing crisis, the government undertook a number of loan modification plans that included making changes to the terms of loan agreement, reducing the interest rate, extending the loan term or using ―forbearance plans‖. Under the Home Affordable Modification Program (HAMP), for example, Department of the Treasury, Fannie Mae, and Freddie Mac were called on to use up to $75 billion to promote loan modifications. These plans were aimed at delaying foreclosure and making mortgage payments more affordable, especially in the non-prime category (GAO, 2009:13).

While lenders and borrowers must work together to improve loan performance and help reduce future home foreclosures, it is a challenge to determine the eligibility for loan modification programs. The policy implications of our analysis (the need for loan portfolio improvement) can be subject to a number of obstacles. As GAO noted, US Treasury has estimated that up to 3 to 4 million borrowers with high risk profile (at risk of default or foreclosure) could be part of the loan modification plan under HAMP. However, as GAO noted again in July 2009, ―Treasury’s estimate reflects uncertainty created by data gaps and the need to make numerous assumptions, and therefore may be overstated‖ (GAO, 2009:13-14).

29

CONCLUSION This paper investigated the determinants of asset quality and loan performance in the S&L industry over the period 1978-2009. In particular, it sought to establish which potential determinants of loan performance prevailed in the US S&L industry. Furthermore, the paper discussed the significance of recent market and financial innovations and assessed the role of regulation in light of the present crisis. The development of the secondary mortgage market, especially private label MBS’s, has helped expand the business of securitization but also adversely affected loan performance and quality at S&Ls with fixed asset structures. The final section briefly discussed the implications of loan modification programs for asset quality management and drew strategic lessons for future researchers and practitioners in the field of risk management.

We obtained the data from the Office of Thrift Supervision (OTS) database between 1978 and 2009, thus covering both the years before the start of the present crisis as well as those that followed. Despite the sizable indicators of loan performance going far back to 1964, this paper only included the period for which data on loan performance were found. Applying ADF as a unit root test for filtering non-stationary effects by estimation of least squares, we were able to establish meaningful trends in the loan performance of the S&L industry from 1978 to 2009. To apply the test, we accepted the existence of a unit root assuming that time-series variables were non-stationary. The model was then reestimated applying the first difference operator to the series and stationary de-trending.

30

Overall, the results of our analysis indicated that industry characteristics explain a considerable part of the variation in loan performance measured by the ratio of nonperforming (delinquent) loans to total loans. The most statistically significant variables are return on assets, leverage ratios and ratio of total liabilities to net worth. Poor loan performance tends to be associated with banks holding less profits and net worth. Put differently, this indicates higher return on assets (ROA) and ratio of net worth to total assets leads to a lower ratio of poor loan performance. Higher the ratio of consumers to default on their loans, the lower the return on assets. In addition, the leverage ratio has a significant negative coefficient on loan performance. By and large, our analysis is confirmed by earlier studies that decreasing net worth and under-performing loans have the potential to render S&Ls vulnerable to financial shocks, thus contributing to financial instability.

One of the limitations of the study is the use of time-series data for loan performance indicators rather than cross-sectional data by sector. The other limitation is the exclusion of other categories of data for measuring loan performance such as provisions for losses on interest bearing assets. While this type of data is largely available for US commercial banks, it only goes back to 1991 for S&Ls, making it difficult to generalize from a limited period of time. Although this has made it difficult to examine the variations in loan performance across institutions, the co-integration statistical procedure (ADF test) was able to filter some of the trending behavior in our data.

31

Future research can benefit from the inclusion of exogenous variables in regression analysis, such as policy interventions and regulatory capital requirements that can affect the long-run performance of the S&L industry at the country level. Structure-conductperformance (S-C-P) theory highlights the contribution of market structures and financial system variables to financial institution performance. This theory is used to analyze the relation among firm performance, market conduct, and market structures. With an industrial organization approach, it has led to useful empirical modeling of financial changes, technological innovations, merger analysis, profitability and identification of market power. If properly integrated into the S-C-P paradigm, the policy reforms discussed above can be the starting point for regulators to design long-term policies that can enhance financial institutional stability and sound lending practices, improved schemes for asset quality management, and strengthened oversight of leverage, liquidity and risk management.

BIBLIOGRAPHY Angell, C. and Rowley, C.D. 2007. Breaking New Ground in U.S. Mortgage Lending, FDIC Outlook [On line, March 21, 2007] Available at http://www.fdic.gov/bank/analytical/regional/ro20062q/na/2006_summer04.html [accessed: April 6, 2011]. Balderston, F. E. 1985. Thrifts in Crisis: Structural Transformation of the Savings and Loan Industry. Real Estate Economics, Volume 14, Issue 1, pp.165-168, March. Benston, G.J. 1972. Economies of Scale of Financial Institutions. Journal of Money, Credit and Banking, Vol.4, No.2, pp.312-341. Berger, A. N. 1995. The relationship between capital and earnings in banking. Journal of Money, Credit and Banking, Volume 27, Issue 2, pp.432-456. Blair, C. and Kushmeider, R.M. 2006. Challenges to the Dual Banking System: The Funding of Bank Supervision. FDIC Banking Review, Volume 18, No.1.

32

Bourke, P. Concentration and Other Determinants of Bank Profitability in Europe, North American and Australia. Journal of Banking and Finance 13, 65-79. Bradley, M.G., Gabriel, S.A. and Wohar, M.E.1995. The Thrift Crisis, Mortgage-Credit Intermediation, and Housing Activity. Journal of Money, Credit and Banking, Vol.27, No.2, pp.476-497. Brigham, E. 1964. Economies of Scale in the Savings and Loan Industry. Economic Inquiry, Volume 3, Issue 1, September, pp.7-20. Casu, B., Girardone, C. and Molyneux, P. 2004. Productivity Change in European banking: A comparison of parametric and non-parametric approaches. Journal of Banking and Finance, Volume 28, Issue 10, pp. 2521-2540. Chaudhry, M., Chatrath, A. and Kamath, R. 1995. Determinants of bank profitability. American Journal of Business. 1995, vol. 10, issue 1, pages 41-46

Cromwell, Jeff B., Labys, W.C. and Hannan, M.J. 1994. Multivariate Tests for TimeSeries Models. Sage Publications.

CNN. 2007. New Century files for Chapter 11 Bankruptcy, April 3. Curry, T. and Shibut, L.2000. The Cost of the Savings and Loan Crisis: Truth and Consequences. FDIC Banking Review, (12), 26–35. Demirguc-Kunt, A. and Huizinga, H. 2000. Financial Structure and Bank Profitability. The World Bank in its series Policy Research Working Paper Series with number 2430 [On line August 31, 2000] Available at http://ideas.repec.org/p/wbk/wbrwps/2430.html [accessed: March 5, 2011]. Economics of Crisis. 2011. Financial Crisis Indicators in the Literature. Available on line http://www.economicsofcrisis.com/economics_of_crisis/lit.html [accessed: April 17, 2011]. Geehan, R. and Allen, L. 1978. Measuring the real output and productivity of savings and credit institutions. The Canadian Journal of Economics, Vol.11, No.4, November, pp.669-679. Gilbert, R.A. 1984. Bank Market Structure and Competition: A Survey. Journal of Money, Credit and Banking, Volume 16, Issue 4, pp.650-656.

33

Golderberg, L.G. and Rai, A. 1996. The structure-performance relationship for European banking. Journal of Banking & Finance, Volume 20, October, Issue 4, pp. 745-771. Gramlich, E. M. 2004. Sub-prime Mortgage Lending: Benefits, Costs, and Challenges, Governor of the Board of thee Federal Reserve System, at the Financial Services Roundtable Annual Housing Policy Meeting, Chicago, Illinois, May 21. GAO. 2009. Loan Performance and Negative Home Equity in the Non-Prime Mortgage Market. US Government Accountability Office [On line, December 16] Available at www.gao.gov/new.items/d10146r.pdf [accessed: April 20, 2011). Jahera, J.S., Page, D.E. and Hudson, C.D. 2006. Financial Condition and Performance of Savings and Loans: A Retrospective Look at Mutual to Stock Conversions. Investment Management and Financial Innovations, Volume 3, Issue 2, 2006, pp.77-87. Kaushik, S.K. and Lopez, R.H. 1996. Profitability of Credit Unions, Commercial Banks and Savings Banks: A Comparative Analysis. The American Economist, Vol.40, No.1, Spring, pp.66-78. Kushmeider, R. M. 2005. The US Federal Financial Regulatory System: Restructuring Federal Bank Regulation. FDIC Banking Review, Volume 17, No.4, pp. 1-29. Matasar, A.B. and Pavelka, D.D. 1998. Federal Banking Regulators’ Competition in Laxity: Evidence from CRA Audits. International Advances in Economic Research, February, Vol.4, No.1, 56-69. Mike Hudson IndyMac: What Went Wrong? How An Alt-A Leader Fueled its Growth With Unsound and Abusive Mortgage Lending, Center for Responsible Lending, CRL Report, June 30, 2008. Mildenberg, D. 2008. Bank of America to Acquire Countrywide for $4 billion, Bloomberg, January 11. Morgenson, G. 2007. Inside the Countrywide Lending Sphere, New York Times, August 26. Naceur, S.B. 2003. The Determinants of the Tunisian Banking Industry Profitability: Panel Evidence. [On line] Available at www.mafhoum.com/press6/174E11.pdf [accessed: February 27, 2011]. OTS. 2009. THRIFT INDUSTRY HIGHLIGHTS FOURTH QUARTER 2008 [On line. February 26, 2009] Available at http://www.ots.treas.gov/?p=IndustryPerformance [accessed: October 16, 2010]. OTS. 2010. 2009 Fact Book: A Statistical Profile of the Thrift Industry. [On line. March, 34

2010] Available at www.ots.treas.gov/_files/481165.pdf [accessed: October 16 2010]. Paletta, D. and Enrich, D. 2008. Crisis Deepens as Big Bank Fails, Wall Street Journal, July 12. Papanikolaou, N.I and Wolff, C.C.P. 2010. Leverage and risk in US Commercial Banking in the light of the current financial crisis. Paper provided by Luxembourg School of Finance, University of Luxembourg in its series LSF Research Working Paper Series with number 10-12 [Online]. Available at http://ideas.repec.org/p/crf/wpaper/10-12.html [accessed: March 15, 2011]. Pervan, M., Pervan, I. and Guadagnino, A. 2009. Market Structure and Performance of Croatian Commercial Banks. Paper presented at July 2009 Business and Economics Society International Conference, July 15-18, Kona-Hawaii, USA. Rasiah, D.2010. Review of Literature and Theories on Determinants of Commercial Bank Profitability. Journal of Performance Management 23.1, 23-49. Short, B.K. 1979. The Relation Between Commercial Bank Profit Rates and Banking Concentration in Canada, Western Europe and Japan. Journal of Banking and Finance 3, 209-219. US Treasury Department. 2010. Overview of OTS-Regulated Mutual Thrifts. Available at files.ots.treas.gov/upload/files/49005.pdf [accessed: April 6, 2011]. Verbrugge, J.A., Shick, R.A. and Thygerson, K.J. 1976. An Analysis of Savings and Loan Profit Performance. The Journal of Finance, Vol.31, No.5, December, pp.14271442. World Data Bank. 2010. World Development Indicators & Global Development Finance [On line] Available at http://databank.worldbank.org/ddp/home.do?Step=12&id=4&CNO=2 [accessed: March 13, 2011].

35