out-of-state non-local mortgages - olin business school

Out‐of‐State Non‐Local Mortgages: Features and Explanation

Yilan Xu Jipeng Zhang

[email protected] [email protected]

Department of Economics

University of Pittsburgh

Abstract

Using a sample from the Home Mortgage Disclosure Act data, we find that the home purchase mortgage loans made by out‐of‐state non‐local banks have a much higher proportion of subprime/high‐priced loans and are sold much more to the secondary market than loans made by other banks. Our explanation is the following. When searching for a loan, borrowers first apply to the banks with local branches; if this fails, they next seek loans from non‐local sources through brokers or the internet. Thus, less creditworthy applicants tend to end up with non‐local loans. Hence, we expect that neighborhoods with high denial rate by local banks have more non‐local lending. The empirical evidence supports this hypothesis. Additionally, we argue that banks have extra incentives to make non‐local loans to such less creditworthy applicants in other states if they manage to sell most of their loans to the secondary market. Consistently, we find that the banks with better ability to sell loans to the secondary market originate higher proportion of non‐local out‐of‐state loans. The subprime loans from out‐of‐state non‐local banks and the high sales rate of those loans to the secondary market can be one contributing factor of the subprime crisis, suggesting that regulators should be cautious about the out‐of‐state non‐local lending which are typically initiated by brokers.

JEL classification codes: G01; G21

Keywords: Mortgage, Non‐Local Lending, Interstate Banking, Subprime Crisis

2

1. Introduction

With the development of the communicating, credit‐screening, underwriting and monitoring techniques, banks can now originate loans from a distance and enter local markets without physical presence. In the home mortgage market particularly, lending through brokers and on the internet has been an increasing trend in recent years. Out‐of‐state banks benefit more from the remote entries than their in‐state counterparts because they face higher branching costs than the latter due to the interstate banking restrictions. To investigate the features of loans originated through different lending channels by different banks, we classify the mortgages into four categories‐‐by whether the lending bank is in‐state or out‐of‐state and whether it has branch offices in the local Metropolitan Statistical Area market‐‐which are in‐state local, in‐state non‐local, out‐of‐state local, and out‐of‐state non‐local loans.

Using a sample of home purchase loans from the Home Mortgage Disclosure Act (HMDA) data, we find that out‐of‐state non‐local loans are more likely to be subprime loans, i.e., have a high subprime rate; and are more likely to be sold to the secondary market, i.e., have a high sales rate. The subprime loans are those high‐priced loans whose annual percentage rate are equal or greater than 3 percentage points for the first lien loans comparing to the applicable treasury yield. This study attempts to explain the salient features of out‐of‐state non‐local loans and to investigate the relationship between subprime lending and secondary market sale.

The challenge to interpret the high subprime rate lies in the absence of the credit information and the loan outcome information in HMDA data (Avery, Brevoort and Canner 2007). The high prices charged by the non‐local lenders could be legitimate risk premium for high‐risk borrowers, or compensation for the high transaction cost associated with the long distance, or discrimination against non‐local borrowers. We provide the evidence to show that non‐local borrowers have poorer unobservable credit quality, due to the credit searching‐screening process. The explanation is the following. From the demand side, when trying to obtain a loan, most borrowers first apply to the banks with local branches; if this fails, they next seek loans from non‐local sources through brokers or the internet. Thus, borrowers with lower credit quality tend to end up with non‐local loans. In supporting this, we find that the demand for non‐local loans is greater in the neighborhoods where the lending standards of the local banks are high, that is, the denial rates of the local banks are high. From the supply side, there are evidences showing that local banks have advantage to get access to borrowers with good credit quality comparing to non‐local banks Ergungor (2007).

Since the non‐local lending business is exposed to low‐quality borrowers, why would some banks do this business? How do they manage the high risks to make profits? We find that nearly 80 percent of the out‐of‐state non‐local loans are sold to the secondary market, which is around 30 percent higher than others. Moreover, the subprime loans originated by the out‐of‐state non‐local banks are more likely to be sold to the secondary market than the prime counterparts, although in general subprime loans are less likely to be sold than prime loans. We argue that as a result of the low‐quality applicant pool for the out‐of‐state non‐local banks due to the reasons

3

argued above, the high prices are not sufficient to compensate for the high risks of such loans. Therefore, the banks dump the bad loans into the secondary market by taking advantage of their information of the loan quality. We argue that banks make this type of low‐quality loans only if they can manage to sell them to the secondary market. Consistently, we find that the banks with better ability to sell the loans to the secondary market originate a higher proportion of out‐of‐state non‐local loans.

Given that the out‐of‐state non‐local loans have a much higher subprime rate and an even higher sale rate to the secondary market, their lending practice can be a contributing factor to the subprime crisis. Gerardi, Shapiro and Willen(2007) find that homeownerships that begin with a subprime purchase mortgage end up in foreclosure almost more than 6 times as often as those beginning with prime purchase mortgages. As the securitization of home loans becomes a trend, the secondary market may take up too many loans with low qualities because the secondary market purchase decision are not based on the attributes of the individual borrowers, as suggested by Gabriel and Rosenthal (2007).

Understanding the salient features of the out‐of‐state non‐local loans, related to non‐local lending and secondary market activity, will contribute important insights on the ongoing regulation reform. In Governor Randall Kroszner’s testimony before the Committee on Financial Services, U.S. House of Representatives on Oct. 24th, 2007, he proposed a nationwide registration and licensing system for all mortgage loan brokers across states in response to the concern of predatory mortgage lending through brokers. Meanwhile, he also called for legislative measures to ensure the risk management of the secondary market purchase. The testimony reflects the concerns from the non‐local lending and the loose risk control for secondary market.

In previous literature, interstate banking and lending channels have been well studied respectively for small business lending. We are the first to analyze the home mortgages by combining both perspectives and their implications on the subprime crisis.

The deregulation of geographic expansion for banks both intra‐ and inter‐state happened in three steps: intrastate branching, interstate banking by Bank Holding Companies, and interstate branching by independent banks. The interstate branching by independent banks permitted by the Riegle‐Neal Act of 1994 is a huge advancement in the interstate banking because acquiring or establishing a branch office is much cheaper than acquiring a whole bank. There are some findings suggesting that the banks mainly develop their non‐local lending business in historically underserved markets, where the borrowers are less creditworthy in the conventional sense. For instance, Peterson and Rajan (2002) argue that the distant borrowers are not necessarily the ones with high quality credits, due to the improvement in lenders’ productivity and the increasing competition in credit market. DeYoung, Frame, Glennon, McMillen, Nigro (2008) finds out that increase in the distance between small business borrowers and their lenders during the 1990s was disproportionately large for borrowers located in low‐income and minority neighborhoods. However, little is known about the impact of extending credit to those less creditworthy borrowers

4

without the presence of local branches to collect local information and provide local services. This study fills this gap in the literature.

Our results confirm that the securitization in the secondary market provides bad incentives for the banks to expand the risky out‐of‐state non‐local lending. Loutskina and Strahan(2009) show that securitization in the secondary market reduces the effect of lenders’ financial condition on credit supply. Using the HMDA data between 1992 and 2004, Gabriel and Rosenthal (2007) show the secondary market purchase helps to expand the credit supply in terms of higher origination rates, and the effect is even greater for by the subprime lenders. Keys, Mukherjee, Seru and Vig (2010) and Dell'Ariccia, Igan and Laeven. (2008) points out that securitization adversely affect the screening incentives of lenders, and therefore the banks lowers the lending standard when making home mortgage loans with securitization. Mian and Sufi (2008) argue that the sharp increase in the fraction of loans sold to secondary market expands the home mortgage credit supply to the neighborhoods of low creditworthiness in terms of low denial rates and high origination rates.

2. Industry Background and Data Resources

2.1 The Home Mortgage Market

The home mortgage market has grown rapidly in the past decade. The home mortgage debt as percentage of GDP has increased from 40‐50% in 1990s to more than 70% in 2003 and 2004 (Green and Wachter 2005). This growth arose from the complex interaction of many factors, such as the homeownership encouragement policy, financial innovation, and secondary market development, which are closely related to our study.

Several government programs were established to foster mortgage lending, construction and encourage home ownership. These programs include Ginnie Mae, Fannie Mae and Freddie Mac. As a result of readily available funding for home mortgages, denial rate for conventional home purchase loans in 2002 and 2003 decreased to 14%, half of the denial rate in 1997 (Source: Federal Financial Institutions Examination Council, Press Release). The lax lending also helped the U.S. homeownership rate to peak with an all time high of 69.2% in 2004 (Source: U.S. Census Bureau.). As the demand for housing increased, the housing prices also soared. In 2004 and 2005, Arizona, California, Florida, Hawaii, and Nevada record price increases in excess of 25% per year.

The innovation in the home mortgage market also helped the growth of the home mortgage lending. Many mortgage products were introduce to the market, such as Adjustment Rate Mortgage, balloon loans, interest‐only loans, piggy back loans. More importantly, the borrowers with poor credit in the conventional standard can now get loans in the subprime segment of the home mortgage market. The subprime market is meant to overcome the credit rationing (Stiglitz and Weiss, 1981; Woosley, 2004). However, there are evidences indicating that the subprime loans are at least six times more likely to default than prime loans, more than 5 times more likely to delinquent,

5

and more than 10 times likely to be foreclosed (Chomsisengphet and Pennington‐Cross, 2006) .

Thirty years ago, the deposits of banks are the only source of the home mortgage lending. The banks kept the loan on its balance sheet until the loan was repaid. Nowadays, both banks and mortgage companies can originate loans, and they can sell the loans to a third party in exchange for funding. The third party can be a government agency like Ginnie Mae, or Government‐sponsored enterprises (GSEs) like Fannie Mae and Freddie Mac; or a private sector financial institution. The third party then packages the mortgages and sells the payment rights to investors, whose gains are based on the payments of a collection of individual mortgages. The investors of the mortgage baked securities (MBS) can be individual investors or institutional investors like insurance companies, mutual fund companies, unit and investment trusts, commercial banks, investment banks, pension fund, and private banking organizations.

As broader investors get involved, in the event of huge volume of mortgage defaults, the whole chain breaks down: the originators fail to collect the payment from the borrowers, hence profits of MBSs drop, investors sell out their bonds, the stoke holders of the third parties sell their stokes, the third parties who issues the MBSs face illiquidity. By 2008, the loss of bank capital was around $150 billion and a large number of mortgage institutions had been sold or went bankruptcy (Kregel 2008).

In our study, the mortgage market is defined at the Metropolitan Statistical Area (MSA)/Metropolitan Division (MD) level. A bank is defined to be non‐local if the has no branches in the local MSA market. Compared to the local banks, the non‐local banks have no access to borrowers’ information through deposit and transaction accounts (Mester, Nakamura and Renault, 2007), and they cannot form up relationship lending with the local borrowers through in‐person interactions. Compared to the in‐state banks, out‐of‐state banks have little or no commitment to local prosperity, have higher cost of branching in the local market due to the regulation, yet they may be exempt from some of the state banking laws.

2.2 Data Resources

Our understanding of the home mortgage market is largely derived from the Home Mortgage Disclosure Act (HMDA) data. The home mortgage lending institutions include depository institutions such as commercial banks, saving and loan associations, and credit unions; and the non‐depository institutions such as impendent mortgage companies and mortgage companies that are subsidiaries of commercial banks or Bank Holding Companies. It is estimated that the more than 8,800 lenders currently covered by the law account for approximately 80% of all home lending nationwide. For more analysis of HMDA coverage, see Bercovec and Zorn (1996).

As a starting point, the bench mark sample from the Loan Application Register in HMDA 2005‐2008 includes the applications of conventional home purchase loans for 1‐ to 4‐family housing units secured by first lien. The lenders include only state chartered commercial banks, but not the national banks. Unlike the state banks whose

6

lending practices may be different inside and outside the home states due to the state regulations, the national banks run the business in a more uniform way across states, and therefore are less relevant in our analysis. For homogeneity, we also exclude the state banks’ mortgage banking subsidiaries from the baseline analysis, because those subsidiaries are specialized only in mortgage lending. Including the subsidiaries in the robustness test also get the same pattern.

Each observation in the data set is a loan application which may result in rejection or acceptance, and further origination. The interest rates spreads are reported if the difference between the annual percentage rate and the applicable treasury yield are equal or greater than 3 percentage point for the first lien loans. The loans with interest rates spreads reported are categorized as “subprime”. Only complete applications initiated by a nature person with the banks are included. The applications withdrew by the applicants and the loans purchased by banks are excluded, as well as the loans initiated by a corporation, partnership or other entity that is not a nature person.

The charter state of a bank is identified by the headquarter state according to the Institution Record. A loan is defined to be non‐local if the bank who takes the loan application does not have a branch office in the MSA where the property that secures the loan locates. The locations of their branches are identified by the MSA Office Information of HMDA.

3. Summary Statistics

3.1 Total Applications and Origination Rate

Table I shows the distribution of loan applications and originations among different types of banks and different lending channels. The total applications with in‐state banks are around twice of the applications with out‐of‐state banks. This indicates that the applications prefer in‐state banks to out‐of‐state banks when they apply for home mortgage loans. Among the applications with in‐state banks, more than two‐thirds are local applications; while for the application with out‐of‐state banks, only one‐third are local applications. The contrasting facts reflect less availability of out‐of‐state branches than in‐state branches because of the barrier facing out‐of‐state banks to enter the market of another state.

The majority of the local applications go to in‐state banks; however, more non‐local applications go to out‐of‐state banks than with in‐state banks. Except the branching restriction, such distribution might be attributed to the fact that the applicants who apply with local branches prefer in‐state banks to out‐of‐state banks, while the applicants who choose non‐local banks are indifferent between in‐state banks and out‐of‐state banks. Moreover, the out‐of‐state banks might be more specialized in conducting non‐local loans because of their disadvantage in local branching.

The share of out‐of‐state non‐local applications in total applications gets smaller over time, from around 30 percent in 2005 and 2006 to 16 percent in 2009; the share of out‐of‐state local applications is stable; the share of in‐state applications increases.

7

The out‐of‐state non‐local applications are more sensitive to time‐varying shocks, such as the subprime crisis.

The origination rate, that is, the share of applications that become loans, is the highest for in‐state local applications and the lowest for out‐of‐state non‐local applications. The origination rate of out‐of‐state non‐local applications is also sensitive to the shocks, decreasing from more than 70 percent in earlier years to 64 percent in 2008, while the origination rates of other types of applications remain almost the same over years.

3.2 Features of Loans and Borrowers

The most salient feature of loans from different banks is the following. Compared to other types of loans, the non‐local loans from out‐of‐state banks have a much larger proportion of subprime loans (or high‐priced loans) and are sold much more frequently to the secondary market. The summary statistics are shown in Table II.

The subprime rate of out‐of‐state non‐local loans drops substantially over the years, while the rate remains relatively stable for the other types, narrowing the gaps between the two. The sales rates drop for all types of loans other than the out‐of‐state local loans, indicating that the secondary market activities are sensitive to the shock of the subprime crisis. However, the sales rate of the out‐of‐state non‐local loans is still much higher than that of other types in 2008.

The borrowers of out‐of‐state non‐local loans have higher loan‐to‐income ratio and are less likely having co‐applicant (Table II); they are more likely to be minority groups, such as non‐white, female, and applicants from minority tract and low‐income tracts. (Table III.). These features indicate the low quality of the out‐of‐state non‐local loans. Given the charter of the banks (in‐state or out‐of‐state), the average income of the borrowers who file up non‐local application is lower than the average income of those who file up application through local branches (Table IV).

3.3 Market Structure and Banks’ Characteristics

Given the salient difference of out‐of‐state non‐local loans relative to other groups, we further investigate the features of out‐of‐state non‐local market and those lenders. First of all, we find a particular bank, Fremont Investment & Loan, originated a large amount of out‐of‐state non‐local loans, of which a great proportion is subprime.1 Around 90 percent of its loans are sold to the secondary market, and more than 90 percent of its loans are subprime. We check the stylized features documented in Section 3.2 without this major lender, and find that the sales rate and the subprime

1 FREMONT INVESTMENT & LOAN originated 67.77% (48.83%) of the out‐of‐state non‐local loans in 2005 (2006) and was closed in August, 2008. According to the Federal Deposit Insurance Corporation (FDIC) press release, "On March 8, 2007, the FDIC issued a cease and desist order against Fremont Investment & Loan, Brea, California (Bank), and its parent corporations, Fremont General Corporation and Fremont General Credit Corporation."

8

rate of out‐of‐state non‐local loans become smaller, but are still significantly higher than other types. The higher sales rate and higher subprime rate are not driven entirely by this particular bank.

The degree of market concentration is much higher for the out‐of‐state non‐local loans, that is, a small number of banks originate the majority of those loans. Including the biggest bank, the Herfindahl‐Hirschman Index (HHI) for this market is more than 0.3 if the biggest bank is included (HHI is more than 0.1 without this bank), while the HHI for in‐state local market is less than 0.01.

Table V lists the top 5 out‐of‐state non‐local lenders in each year, as well as the banks at 80 and 90 percentiles. In 2005 and 2006, less than 10 banks account for 80 percent of the total out‐of‐state non‐local loans. This again shows that the out‐of‐state non‐local lending is highly concentrated in a few banks. The market concentration decreases after the subprime crisis. In 2008, the top 70 banks originate 80 percent of the out‐of‐state non‐local loans. Before the subprime crisis, the biggest out‐of‐state non‐local lenders are not necessarily big banks. Those banks with small assets are more specialized in non‐local lending and sell extremely high proportions of their subprime loans to the secondary market. After subprime crisis, the out‐of‐state non‐local lenders are mostly big banks, and only a small proportion of the loans they sell are subprime loans.

4 Subprime Loans and Secondary Market Activity

The summary statistics only compare the unconditional means of the subprime loan incidence and the sales incidence for the four types of loans, however, the subprime status and the sales status may be correlated with the observed characteristics of the loans and the borrowers. To control for those factors, we run pooled logit regressions to see the determinants of the incidence of subprime loans and sales to secondary market. Due to the possible self‐selection into a particular type of loan based on unobervables, we do not interpret the results here as causal relations.

4.1 Features and Determinants of Subprime Loans

Table VI compares the features of the subprime loans with the prime loans. Subprime loans are more likely to be sold to the secondary market. In total, 70 and 62 percent of the subprime loans are sold to the secondary market in 2005 and 2006, which are 20 and 10 percentage more than the prime loans. This feature, however, is reversed in 2007 and 2008. The subprime loans are less likely to be sold to the secondary market than the prime loans, which indicates that the subprime crisis makes the subprime loans less favorable in the secondary market. Compare to prime loans, the subprime loans are less likely to have co‐applicants, more likely to be from such minority groups as female and minority borrowers and borrowers in low‐income and minority neighborhoods; have smaller average loan amounts and lower borrowers’ annual income.

9

We run pooled logit regression to see the determinants of the incidence of subprime loans.

1

I indexes for the loan, s indexes for the state, and t indexes for the year. H is a discrete variable which takes value of one if the loan is subprime, and zero otherwise. ON is the dummy for out‐of‐state nonlocal loans, IN is the dummy for the in‐state nonlocal loans, OL is the dummy for the out‐of‐state local loans. The baseline case is the in‐state local loans. Fremont is the dummy for the loans made by the major out‐of‐state lender, Border is the dummy for the loans made to borrowers in MSAs that cross the state borders. is the state dummy, and is the year dummy. is a vector of controls, such as the occupancy of the property, the presence of co‐applicants, the loan amount, the gender and ethnicity of the borrower, the neighborhood income, minority share and population. Both results account for state and year effects, the impact of the major out‐of‐state non‐local bank, and the impact of the loans being in a border MSA market.

The results are shown in column 1 and column 2 in Table VIII. The results report the marginal effects—the expected probability of becoming subprime loan relative to the in‐state local loans. The marginal effects for the out‐of‐state non‐local and in‐state non‐local indicators are both positive and significant, indicating that non‐local loans from both in‐state and out‐of‐state banks are more likely to be subprime loans than the in‐state local loans. However, the marginal effect for out‐of‐state loans is bigger. The marginal effect for the out‐of‐state nonlocal loans decreases from 12 percent in column 1 to 11.8 percent in column 2 after adding the controls.

4.2 Features and Determinants of Sales to the Secondary Market

Table VII compare the characteristics of sold and unsold loans. Among the loans sold to the secondary market, 35 and 31 percent of them are subprime loans for 2005 and 2006; the subprime share decreases to 9 and 4 percent in 2008 and 2009. This also reflects the impact from the subprime crisis. The loans sold to the secondary market have higher interest spread; less likely have a co‐applicant; have higher loan‐to‐income ratio; have more borrowers who earn lower annual income and are relatively poorer in their neighborhood.

We run the following pooled logit regression to understand the determinants of the incidence of selling to the secondary market.

1

10

S is a discrete variable which takes value of one if the loan is sold to the secondary market in the same year as the loan origination, and zero otherwise. The baseline case is again the in‐state local loans. I indexes for the loan, s indexes for the state, and t indexes for the year. The explanatory variables are the same as explained in the previous section.

Column 3 to column 5 in Table VIII shows the sales rate regression. Column 3 is the simple regression, column 4 includes the controls, and column 5 incorporates the interaction between subprime status and loan types. The marginal effects for the out‐of‐state non‐local loans are positive and significant, indicating that those loans are more likely to be sold to the secondary market. Moreover, the subprime loans from out‐of‐state non‐local banks are more likely to be sold to the secondary market, compared to subprime loans originated by other groups. According to the result in column 5, the prime loans originated by out‐of‐state non‐local banks are 20.1 percent more likely to be sold to the secondary market compared to the ones originated by in‐state local banks. The subprime loans originated by out‐of‐state non‐local banks are 35.4 percent (20.1+15.3) more likely to be sold to the secondary market compared to the prime loans originated by in‐state local banks, and 75.1percent (20.1+15.3+39.7) more likely to be sold than the subprime loans originated by in‐state local banks. All results account for state and year effects, and the impact of the major out‐of‐state non‐local bank, and the impact of the loans being in a border MSA market.

5. A Behavioral Explanation and Empirical Evidence

As documented in section 4, the subprime rate and the sales rate to the secondary market are much higher for the loans originated by out‐of‐state non‐local banks, compared to other groups, even after controlling for their observed characteristics. It is still unclear whether the high prices are premium for the information asymmetry due to the distance between borrowers and lenders, or premium for high risks of lending to low‐quality borrowers, or the discrimination of the non‐local banks against credit‐constraint borrowers. Moreover, what is the relation between the subprime rate and the sales rate? In this section we provide the evidences to support an explanation from both demand side and supply side of the mortgage market.

5.1 The Local Lending Standard and the Non‐local Credit Demand

In the home mortgage market, there are two types of credit demand: spontaneous and induced. The former is endogenous to the local economic conditions and borrowers’ financial conditions. The creditworthy borrowers want to buy a house and can afford the home mortgage. The latter is the demand solicited by the banks. The targeted borrowers may not be financially ready to afford housing, but they decide to buy a house when the banks offer seemingly favorable terms such as zero down payment and low teaser interest rates. Those borrowers have poor credit and are financially less sophisticated.

11

When applying for a loan, most borrowers with spontaneous credit demand first consider those banks with local branches‐‐if this fails, they next seek loans from non‐local sources through brokers or in the internet. Since the potential borrowers first apply to local sources, the local banks have the privilege to screen the relatively better qualified borrowers. The non‐local banks get the residual demand after the screening of the local banks, and therefore the credit quality of non‐local demand is lower than the quality of local demand. Furthermore, the non‐local banks have to target on the less creditworthy borrowers since the local banks have the advantage in the local market. They encourage demands by mail‐in offers, building network of brokers, online advertising, etc. Those solicited demand also contributes to the lower credit quality of the non‐local demand relative to the local demand.

Moreover, the borrowers who borrow from local banks and those who borrow from non‐local banks have different bargaining power. The former are on the top of the credit distribution, if they are unsatisfied with the prices they get, they can shift to the non‐local borrowers and get lower prices. However, those who seek for non‐local sources are credit‐constrained in local market, they have no other options but to accept the high price charged by the non‐local banks. This also contributes to the high incidences of subprime loans with non‐local banks.

To provide evidence of this searching and screening process, we investigate the relationship of current non‐local credit demand and the local credit supply in the past. If it turns out the non‐local banks mainly serve the neighborhoods where the past decline rates of the local banks were high, then it implies that the non‐local banks gets the residual demand after the local banks screen the potential applicants in the first applications .

Since banks’ branching decisions are largely endogenous to the neighborhood characteristics, the share of non‐local lending is also endogenous. To deal with this endogeneity problem, we constructed a panel data set of the neighborhoods to provide evidence of the above hypothesis. The HMDA loan applications are collapsed to the tract level for each year during 2005‐2008. Each observation in the panel is a tract. We estimated the following model.

is the share of loan applications with non‐local banks or out‐of‐state non‐local banks in tract n in year t. The share as the dependent variable is less sensitive to the size of the tract. Given the dispersed distribution of the tract size, this measure is more desirable. is the lagged denial rate of the loan applications with local banks, which is perceived as a proxy for the lending standards implemented by the local banks. The lagged variable is used because the borrowers make their loan application decisions based on past information, and it takes some time for them to know the results of their initial applications and to response by reapplication if the initial ones are not approved.

The tract fixed effect, , can control for any unobservable tract characteristics that can potentially correlate with the share of non‐local lending, such as low income, high

12

minority share, and borrower pools with poor credit. The aggregation to the tract level overcomes the main disadvantage of the HMDA data: the lack of credit information for individual applicants. At the tract level, however, unless there are immigrations or emigrations of large scale, the credit pool will remain almost the same over the short observation span. The year fixed effect, , can be used to control for any changes in the macroeconomics and housing policies that affect all tracts in a given year. No other control variables are added to the specifications because the FE specification can only estimate the time‐varying variables, which are limited for the tracts because the census data are available every 10 years. Moreover, the tract characteristics do not displace much variation across time in such short time span.

The pooled OLS and FE results are shown in Table IX. In the first two columns, the dependent variable is the share of loan applications with non‐local banks in a given tract. In the third and forth columns, the dependent variable is the share of loan applications with out‐of‐state non‐local banks in a given tract. The coefficients for the lagged local denial rate are positive and significant in all specifications. The FE estimators are smaller than the OLS estimators. The FE estimation results show that as the local denial rate in the previous year increases by one percent, the share of non‐local loans will increase by 0.015 percent, and the share of out‐of‐state non‐local loans will increase by 0.017 percent. The coefficient for the out‐of‐state non‐local share regression is greater than the non‐local share regression, implying the out‐of‐state non‐local lending is more sensitive to the lending standards implemented by the local banks.

5.2 The Sale Ability and the Out‐of‐state Non‐local Lending

Since the emerging of the subprime market for home mortgage, the high prices charged for subprime loans have been justified by the high risks of the loans. However, as the lending standard lowers, credits are extended to borrowers of extremely low quality and high risk, the high interest rates may not be sufficient to compensate the risks given the interest rate constraints imposed by federal and state anti‐predatory lending laws. Thanks to the invention of securitization, the banks can now sell the loans to the secondary market. The lending risks taken by different types of banks determine different roles of securitization. The banks that deal with qualified borrowers utilize the secondary market as extra source of funding rather than a channel to deliver the risks. On contrary, the banks that extend poor credit will sell the loans to the secondary market and get rid of the high risks if the interest rates are not sufficient for the risks. Dumping bad loans is possible because the buyers in the secondary market have limited information about the true quality of the loans.

Among the non‐local banks, the out‐of‐state banks are more active in encouraging less creditworthy borrowers to apply for loans. This is partly due to the Community Reinvestment Act (CRA), a federal law that requires banks to supply credit to low‐income, minority neighborhoods. Passing the CRA assessment is a condition for banks to do acquisition and merge. In order to meet the requirement, some banks need to extend lending to those who are otherwise credit‐constraint borrowers outside their home states. Note that the non‐local in‐state banks do not sell high proportion of their

13

loans to the secondary market. The reason is that in‐state banks have more knowledge of the local market, so they still can get better customers than out‐of‐state banks. Moreover, the in‐state banks have lower operation costs than out‐of‐state banks because of the interstate banking laws.

Overall, we conjecture that banks have incentives to make non‐local loans to less creditworthy applicants in other states only if they can transfer the risk by selling most of those low‐quality loans to the secondary market. If this is true, we expect that the banks with better abilities to sell to the secondary market will originate more and thus have higher proportions of out‐of‐state non‐local mortgages. That is, the banks originate high‐risk out‐of‐state non‐local loans only if they anticipate they can sell the loans to the secondary market based on the previous sales activities. The results in Table X support this hypothesis.

We estimated the following baseline model with different specifications.

is the share of out‐of‐state non‐local loans for a given bank j in year t. The explanatory variable in interest is the sales rate of the bank in the previous year, . This measure is a proxy for the bank’s credit rating, loan packing and selling skills, as well as the connection with the purchasers in the secondary market. The bank fixed effect, , controls for any unobservable bank characteristics that correlate with the out‐of‐state non‐local lending strategy. The year fixed effect, , controls for shocks in a given year that influent all banks. is a vector of controls, such as the banks’ total assets, total application counts (including applications for government backed loans, refinance loans, which are not included in the sample), as well as shares of loans sold to different types of purchasers in the secondary market and the share of subprime loans.

The HMDA loan origination data are aggregated to the bank level to construct a panel set. The influential bank, Fremont Investment & Loan, is no longer a concern in this analysis because it only accounts for one observation among more than 3000 thousand banks. Given that the out‐of‐state non‐local lending business is highly concentrated in a few banks, we restrict the sample to banks which originate more than 10 loans in the sample, have a proportion of out‐of‐state non‐local loans greater than 10 percent, and sell at least one loan to the secondary market. This criterion restricts the bank sample to 407 state banks.

Both pooled OLS and FE specifications are used. Column 1 and column 2 do not use any control variables. Both OLS and FE estimators have the expected signs and are significant. Since the bank characteristics can change dramatically across years, we control for assets, loan counts, passed distribution of loan sales among different purchaser, and the subprime ratio in column 3 and column 4. Both OLS and FE estimators are positive and significant after adding the controls. We argue that the specifications with controls are more convincing because they control for the diverse features of different banks, and the time variation of the features. As the FE estimation

14

in column 4 shows, the share of out‐of‐state non‐local lending for a bank increases by 0.19 percent if the sales rate in the previous year increases by one percent.

The coefficient for banks’ total assets is negative and the coefficient for the banks’ total application counts is positive in both OLS and FE specifications. This is consistent with the statistics showed in Table V. The banks who originate a greater proportion of out‐of‐state non‐local loans are not necessarily the banks with big assets, since they can sell the loans to the secondary market. And as the banks expand their home mortgage lending business, they involve more in the out‐of‐state non‐local lending because it expands the market to the areas without bank branches.

6. Conclusions

This study explains the high subprime ratio and the sales rate of the out‐of‐state non‐local loans by proposing a searching and screening process for the home mortgage loans. The story is the following. When applying for a loan, most borrowers first consider those banks with local branches‐‐if this fails, they next seek loans from non‐local sources through brokers or in the internet. The empirical evidences show that the non‐local credit demand is greater in the neighborhoods where the borrowers get rejected by the local banks more often. This finding indicates that less‐creditworthy and riskier applicants tend to end up with non‐local loans because the local banks have the privileges to pick the borrowers of better qualities. Additionally, we explore the incentives to make non‐local loans to low‐quality applicants located in other states by investigating the market structure and the specialized banks. The empirical evidences show that banks with better abilities to sell the home loans in the past will originate a higher proportion of out‐of‐state non‐local loans regardless the high risks.

Explaining the salient features of subprime lending and secondary market activities that are related to out‐of‐state non‐local lending can deepen our understanding of the subprime crisis and can contribute important insights on the ongoing regulation reforms. The risky lending made by out‐of‐state non‐local banks might be one contributing factor of the subprime crisis. Related to out‐of‐state non‐local mortgages, government might need to reinvestigate their regulation on interstate banking and brokers. As shown in the paper the secondary market provides bad incentive for lenders to originate low‐quality loans, such as the out‐of‐state non‐local mortgages.

For future studies, it seems interesting to investigate the market structure in the home mortgage lending business and its evolution in the past decade. This could contribute more insights on the subprime crisis. Another question is the impact of governmental policies on the development of the subprime market, for instance, the Community Reinvestment Act (CRA), a federal law that requires banks to supply credit to low‐income, minority neighborhoods, and the Affordable Housing Goals (HG) complied by the government sponsored enterprises (GSEs). By linking the HMDA data with local foreclosure data or the outcomes of mortgages, we can also identify the true qualities of the non‐local loans.

15

References

1. Avery, R. B., K. P. Brevoort, and G. B. Canner, "Opportunities and Issues in using HMDA Data," Journal of Real Estate Research, 29 (2007), 351‐380. 2. Berkovec, J., and P. Zorn, "How Complete is HMDA? HMDA Coverage of Freddie Mac Purchases," Journal of Real Estate Research, 11 (1996), 39‐56. 3. Chomsisengphet, S., and A. Pennington‐Cross, "The evolution of the subprime mortgage market," Federal Reserve Bank of St. Louis Review, 88 (2006), 31‐56. 4. Dell'Ariccia, G., D. Igan, and L. Laeven, "Credit booms and lending standards: Evidence from the subprime mortgage market," CEPR Discussion Papers, (2008). 5. DeYoung, R., W. S. Frame, D. Glennon, D. P. McMillen, and P. Nigro, "Commercial lending distance and historically underserved areas," Journal of Economics and Business, 60 (2008), 149‐164. 6. Ergungor, O. E., "Bank branch presence and access to credit in low‐to‐moderate income neighborhoods," FRB of Cleveland Working Paper No. 06‐16, (2007). 7. Frame, W. S., and L. Woosley, "Credit scoring and the availability of small business credit in low‐and moderate‐income areas," Financial Review, 39 (2004), 35‐54. 8. Gabriel, S., and S. S. Rosenthal, "Secondary Markets, Risk, and Access to Credit: Evidence from the Mortgage Market," Unpublished Working Paper, (2007). 9. Gerardi, K., A. H. Shapiro, and P. S. Willen, "Subprime outcomes: Risky mortgages, homeownership experiences, and foreclosures," Federal Reserve Bank of Boston Working Paper, (2007), 07‐15. 10. Green, R. K., and S. M. Wachter, "The American mortgage in historical and international context," Journal of Economic Perspectives, 19 (2005), 93‐114. 11. Keys, B. J., T. Mukherjee, A. Seru, and V. Vig, "Did Securitization Lead to Lax Screening? Evidence from Subprime Loans*," Quarterly Journal of Economics, 125 (2010), 307‐362. 12. Kregel, J. A., "Changes in the US financial system and the subprime crisis," Levy Economics Institute Working Paper No. 530, (2008). 13. Kroszner, R. S., "Legislative Proposals on Reforming Mortgage Practices," Testimony before the Committee on Financial Services, US House of Representatives (October 24), (2007). 14. Loutskina, E., and P. E. Strahan, "Securitization and the declining impact of bank finance on loan supply: Evidence from mortgage originations," Journal of Finance, 64 (2009), 861‐889. 15. Mester, L. J., L. I. Nakamura, and M. Renault, "Transactions Accounts and Loan Monitoring," Review of Financial Studies, 20 (2007), 529. 16. Mian, A. R., and A. Sufi, "The consequences of mortgage credit expansion: Evidence from the 2007 mortgage default crisis," NBER working paper, (2008). 17. Petersen, M. A., and R. G. Rajan, "Does distance still matter? The information revolution in small business lending," The Journal of Finance, 57 (2002), 2533‐2570. 18. Stiglitz, J. E., and A. Weiss, "Credit rationing in markets with imperfect information," The American economic review, 71 (1981), 393‐410.

16

Table I Application and Origination by Bank Type

In‐state banks Out‐of‐state banks All banks Local Non‐local Sub‐total Local Non‐local Sub‐total Total

Panel A: 2005‐2008 Number of applications 890,282 360,444 1,250,726 223,642 486,035 709,677 1,960,403 Share of total application 0.45 0.18 0.64 0.11 0.25 0.36 1.00 Share of applications become loans 0.86 0.82 0.85 0.78 0.71 0.73 0.81

Panel B: 2005 Number of applications 266,145 105,018 371,163 63,609 169,261 232,870 604,033 Share of total application 0.44 0.17 0.61 0.11 0.28 0.39 1.00 Share of applications become loans 0.87 0.83 0.86 0.79 0.75 0.76 0.82

Panel C: 2006 Number of applications 240,414 96,689 337,103 58,399 168,705 227,104 564,207 Share of total application 0.43 0.17 0.60 0.10 0.30 0.40 1 Share of applications become loans 0.85 0.82 0.84 0.78 0.69 0.71 0.79

Panel D: 2007 Number of applications 210,167 87,183 297,350 60,795 93,614 154,409 451,759 Share of total application 0.47 0.19 0.66 0.13 0.21 0.34 1.00 Share of applications become loans 0.86 0.82 0.85 0.77 0.70 0.73 0.81

Panel E: 2008 Number of applications 173,556 71,554 245,110 40,839 54,455 95,294 340,404 Share of total application 0.51 0.21 0.72 0.12 0.16 0.28 1.00 Share of applications become loans 0.85 0.81 0.84 0.80 0.64 0.71 0.80

17

Table II Characteristics of Originated Loans by Bank Type

In‐state banks Out‐of‐state banks Local Non‐local local Non‐local

Panel A: 2005 Share of sold loans 0.48 0.47 0.43 0.80 Share of subprime loans 0.12 0.19 0.12 0.67 Interest rate spread 4.71 4.53 4.48 5.69 Share with a co‐applicant 0.55 0.59 0.51 0.36 Loan‐to‐income ratio 2.07 1.88 2.14 1.87 Loan amount(thousands) 168.98 143.24 199.36 152.78

Panel B: 2006

Share of sold loans 0.48 0.47 0.40 0.76 Share of subprime loans 0.14 0.22 0.17 0.58 Interest rate spread 4.77 4.56 4.59 6.11 Share with a co‐applicant 0.54 0.57 0.47 0.38 Loan‐to‐income ratio 2.01 1.84 2.02 1.88 Loan amount(thousands) 168.02 143.6 191.68 165.7

Panel C: 2007


Panel D: 2008


18

Table III Borrower and Neighborhood Characteristics of Originated Loans by Bank Type

In‐state banks Out‐of‐state banks Local Non‐local Local Non‐local

Panel A: 2005 Share of white borrowers 0.91 0.94 0.91 0.71Share of male borrowers 0.76 0.79 0.76 0.67Share of minority tract 0.31 0.3 0.34 0.54Share of low‐income tract 0.15 0.1 0.15 0.23

Panel B: 2006

Share of white borrowers 0.91 0.94 0.90 0.75Share of male borrowers 0.76 0.79 0.74 0.68Share of minority tract 0.31 0.29 0.35 0.51Share of low‐income tract 0.16 0.10 0.16 0.22

Panel C: 2007


Panel D: 2008


19

Table IV Income Information of Originated Loans by Bank Type

In‐state banks Out‐of‐state banks Local Non‐local Local Non‐local

Panel A: 2005 Annual income(thousands) 106.75 98.49 122.96 100.2Income to MSA median income ratio 1.79 1.96 2.14 1.70Share of low‐income borrowers 0.22 0.22 0.22 0.23Tract to MSA median income ratio 1.10 1.07 1.14 1.02

Panel B: 2006

Annual income(thousands) 110.83 102.5 121.91 111.02Income to MSA median income ratio 1.83 2.03 2.07 1.87Share of low‐income borrowers 0.22 0.22 0.21 0.20Tract to MSA median income ratio 1.10 1.07 1.13 1.04

Panel C: 2007


Panel D: 2008


20

Table V Ranking of the Biggest Out‐of‐state Non‐local Banks

Rank Bank name OS NL

counts**

Cum. Mkt. share

OS NL

Share**

Total Asset $b

Subprime rate of sold loans

Sales rate of subprime loans

Panel A: 2005 1 FREMONT INVESTMENT & LOAN 85988 0.68 0.82 9.91 0.89 0.91 2 REGIONS BANK 5970 0.72 0.29 49.21 0.06 0.31 3 BRANCH BANKING AND TRUST CO 3352 0.75 0.16 74.48 0.01 0.09 4 FIRST BANK 2152 0.77 0.51 8.71 0.78 0.77 5 1ST MARINER BANK 1727 0.78 0.52 1.20 0.33 0.88 6 RESOURCE BANK 1654 0.79 0.30 1.16 0.34 0.97 35 CAPITAL CITY BANK 168 0.90 0.15 2.36 0.01 0.29

Panel B: 2006 1 FREMONT INVESTMENT & LOAN 57027 0.49 0.82 11.32 0.97 0.89 2 GMAC BANK 20410 0.66 0.94 2.43 0.07 0.84 3 REGIONS BANK 3565 0.69 0.25 81.07 0.08 0.21 4 BRANCH BANKING AND TRUST CO 3041 0.72 0.16 80.23 0.03 0.14 5 FIRST BANK 2219 0.74 0.55 9.15 0.77 0.67 9 RESOURCE BANK 1328 0.80 0.34 1.33 0.44 0.99 41 HOMESTREET BANK 141 0.90 0.10 1.95 0.02 0.44

Panel C: 2007 1 GMAC BANK 24898 0.38 0.95 19.94 0.21 0.83 2 M &T TRUST 5643 0.46 0.54 56.38 0.11 0.63 3 REGIONS BANK 3443 0.52 0.2 138.67 0.07 0.42 4 BRANCH BANKING AND TRUST CO 3420 0.57 0.15 117.13 0.03 0.11 5 FREMONT INVESTMENT & LOAN 2752 0.61 0.67 12.76 0.95 0.93 29 JACKSONVILLE SAVINGS BANK 225 0.80 0.85 0.27 0 0 105 TOYOTA FINANCIAL SAVINGS BANK 41 0.90 0.95 0.18 0 .

Panel D: 2008 1 GMAC BANK 11004 0.32 0.95 28.40 0.11 0.87 2 BRANCH BANKING AND TRUST CO 2367 0.38 0.15 127.70 0.02 0.09 3 M&T BANK 2073 0.44 0.34 64.07 0.03 0.39 4 REGIONS BANK 1587 0.49 0.17 137.05 0.07 0.47 5 BANK OF THE WEST 939 0.52 0.28 61.83 0.03 0.01 70 HUNTINGTON STATE BANK 49 0.80 0.91 0.18 . 0 223 IOWA‐NEBRASKA STATE BANK 11 0.90 0.23 0.22 0 0

**OS NL: out‐of‐state non‐local loans

21

Table VI Loan and Borrower Characteristics by Subprime Status

2005 2006 2007 2008

prime subprime prime subprime prime subprime prime subprime

Share of sales 0.50 0.70 0.52 0.62 0.54 0.33 0.48 0.13

Interest rate spread . 5.29 . 5.48 . 4.42 . 4.46

Share with a co‐applicant 0.57 0.34 0.55 0.34 0.54 0.45 0.55 0.53

Loan‐to‐income ratio 2.03 1.90 1.98 1.86 2.07 1.98 2.08 1.80

Loan amount 168.07 151.08 166.36 163.19 174.82 142.36 189.75 145.72

Annual income 109.90 93.14 112.78 104.41 116.76 91.35 130.74 113.45

Income to MSA median income 1.94 1.56 1.96 1.75 2.05 1.69 2.19 2.06

Tract median income 57.84 61.43 59.04 61.48 58.32 55.19 60.97 55.70

Share of owner‐occupied 0.78 0.84 0.78 0.85 0.78 0.82 0.74 0.80

Share of white 0.92 0.71 0.92 0.74 0.92 0.87 0.92 0.92

Share of male 0.77 0.66 0.76 0.68 0.76 0.75 0.77 0.79

Share of Low‐income 0.19 0.27 0.19 0.22 0.21 0.26 0.22 0.27

Share of minority Tract 0.28 0.61 0.28 0.59 0.27 0.41 0.27 0.32

Share of low‐income Tract 0.13 0.27 0.13 0.27 0.13 0.20 0.13 0.15

Owner‐occupied house ratio 0.86 0.80 0.83 0.81 0.83 0.75 0.87 0.78

22

Table VII Loan and Borrower Characteristics by Sales Status

2005 2006 2007 2008

unsold sold unsold sold unsold sold unsold sold

Share of subprime 0.19 0.35 0.23 0.31 0.20 0.09 0.22 0.04Interest rate spread 4.63 5.57 4.57 6.05 4.32 4.62 4.60 3.52Share with a co‐applicant 0.56 0.46 0.54 0.46 0.55 0.51 0.56 0.53Loan‐to‐income ratio 1.88 2.08 1.80 2.07 1.83 2.28 1.81 2.35Loan amount 167.63 160.02 164.43 166.41 172.63 167.90 185.2 180.93Annual income 119.27 94.12 123.89 99.30 133.33 94.33 151.09 98.27Income to MSA median income 2.13 1.60 2.18 1.66 2.38 1.65 2.58 1.65Tract median income 57.21 60.11 58.20 60.95 57.31 58.41 59.60 61.02Share of owner‐occupied 0.70 0.88 0.70 0.89 0.69 0.88 0.67 0.86Share of white 0.91 0.82 0.91 0.84 0.92 0.91 0.92 0.93Share of male 0.79 0.71 0.78 0.71 0.79 0.73 0.80 0.74Share of Low‐income 0.22 0.21 0.21 0.22 0.20 0.22 0.21 0.25Share of minority Tract 0.30 0.43 0.31 0.40 0.30 0.28 0.30 0.25Share of low‐income Tract 0.16 0.17 0.16 0.17 0.16 0.12 0.16 0.10Owner‐occupied house ratio 0.87 0.83 0.83 0.82 0.84 0.81 0.89 0.81

23

Table VIII The Determinates of Subprime Status and the Sales Status

Subprime status Sales statusLogit Logit Logit Logit Logit(1) (2) (3) (4) (5)

Out‐of‐state non‐local 0.120** 0.118** 0.194** 0.193** 0.201**

(0.0430) (0.0413) (0.0687) (0.0697) (0.0689)In‐state non‐local 0.100*** 0.105*** ‐0.0178 ‐0.0119 0.0229*

(0.00745) (0.00814) (0.0116) (0.0113) (0.0109)Out‐of‐state local 0.00986 0.00878 0.00609 ‐0.0126 ‐0.0314

(0.0463) (0.0435) (0.0417) (0.0431) (0.0406)Fremont 0.799*** 0.788*** 0.360*** 0.351*** 0.389***

(0.0241) (0.0271) (0.0460) (0.0496) (0.0427)Border MSA 0.0194 0.00855 0.0477** 0.0540** 0.0585***

(0.0122) (0.0110) (0.0173) (0.0166) (0.0169)Subprime ‐0.397***

(0.0246)Fremont*subprime 0.163**

(0.0554)Out‐of‐state non‐local*subprime 0.153*

(0.0632)In‐state non‐local*subpirme ‐0.0699**

(0.0250)Out‐of‐state local*subprime 0.156

(0.0822)State dummies Yes Yes Yes Yes YesYear dummies Yes Yes Yes Yes YesControls Yes Yes YesN 1567367 1507404 1567367 1507404 1507404

Coefficients and Marginal effects reported for OLS and logit, respectively; Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001

24

Table IX The Local Lending Standard and the Non‐local Credit Demand

Non‐local share Out‐of‐state non‐local share OLS FE OLS FE (1) (2) (3) (4) Lagged local denial rate

0.0923*** 0.0146** 0.0891*** 0.0166** (0.0046) (0.0055) (0.0043) (0.0053)

2007 ‐0.0630*** ‐0.0353*** ‐0.0591*** ‐0.0379*** (0.0021) (0.0018) (0.0020) (0.0017) 2008 ‐0.1172*** ‐0.0755*** ‐0.1134*** ‐0.0797*** (0.0021) (0.0021) (0.0020) (0.0019) Constant 0.3114*** 0.3033*** 0.2593*** 0.2549*** (0.0019) (0.0015) (0.0018) (0.0014) N 1.1e+05 1.1e+05 1.1e+05 1.1e+05 r2 0.0281 0.0236 0.0297 0.0294 rho 0.6016 0.6021 Each observation is a tract. Clustered standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001

25

Table X The Sale Ability and the Out‐of‐state Non‐local Lending Share

Share of Out‐of‐state Non‐local OLS FE OLS FE (1) (2) (3) (4) Lagged sales rate

0.0962* 0.1861** 0.1036** 0.1881** (0.0391) (0.0627) (0.0378) (0.0653)

2007 ‐0.0050 ‐0.0058 ‐0.0085 0.0035 (0.0148) (0.0103) (0.0145) (0.0110) 2008 ‐0.0180 0.0083 ‐0.0094 0.0171 (0.0168) (0.0124) (0.0162) (0.0135) Assets($billion) ‐0.0013* ‐0.0014*** (0.0006) (0.0004) Application count(thousands)

0.0019*** 0.0007***

(0.0006) (0.0001) Lagged GSE purchase 0.0134 0.0579

(0.0251) (0.0533)

Lagged private securitization

0.0337 ‐0.0229 (0.0627) (0.0260)

Lagged affiliation purchase 0.1868* 0.0339

(0.0918) (0.0273)

Lagged subprime ratio 0.1908** ‐0.0268 (0.0587) (0.0777)

Constant 0.1877*** 0.1271*** 0.1443*** 0.1105* (0.0220) (0.0370) (0.0249) (0.0476) N 407.0000 407.0000 407.0000 407.0000 r2 0.0336 0.0719 0.1968 0.1175 rho 0.8650 0.8639 Each observation is a state‐chartered commercial bank, mortgage banking subsidiaries excluded. The sample includes the banks that have out‐of‐state shares greater than 10%, originate more than 10 loans and sell at least one loan in a given year. Clustered standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001

out-of-state non-local mortgages - olin business school

Documents