Bank Behavior and Social Capital

Bank Behavior and Social Capital Marcia Millon Cornetta*, Kristina Minnick a, Patrick J. Schornob, Hassan Tehranianc* a

Department of Finance, Bentley University, Waltham, MA 02452 USA b Ally Financial, Charlotte, NC 28202 c Carroll School of Management, Boston College, Chestnut Hill, MA 02467 USA January 2018 Abstract This paper examines the performance and policies of banks located in high social capital areas. Results show that higher levels of social capital are associated with safer, more profitable, yet less capitalized banks. Additionally, banks in areas with higher social capital display lower likelihoods of default and failure, yet higher likelihoods of receiving capital under the TARP Capital Purchase Program. Examining the relationship between social capital and bank/consumer relations, we find that banks in higher social capital areas pay more interest and charge fewer fees on deposits, and charge lower rates and fees on loans. The results suggest that banks located in higher social capital areas take actions that promote trust and therefore enable working relationships that have productive benefits for both banks and consumers. Keywords: Financial institutions, social capital JEL Classification: A13, G01, G21, D14, D71, Z13

* Corresponding author. Tel.: +1 617-552-3944. E-mail addresses: [email protected] (M.M. Cornett), [email protected] (K. Minnick), [email protected] (P.J. Schorno), [email protected] (H. Tehranian). The views expressed in this paper are those of the authors and do not necessarily reflect those of Ally Financial.

Acknowledgements: The authors are grateful to Ian Appel, Otgo Erhemjamts, Ali Fatemi, Iraj Fooladi, Pouyan Foroughi, Rawley Heimer, Saeid Hoseinzadeh, Qian Jun, Oguzhan Karakas, Shahriar Khaksari, Samer Khalil, Len Kostovetsky, Vladimir Kotonin, Alan Marcus, Hamid Mehran, Mahdi Mohseni, Cal Muckley, Ali Ebrahim Nejad, Jordan Nickerson, Vinh Nguyen, Ronnie Sadka, Assem Safeeddine, Yao Shen, Phil Strahan, and Anand Venkateswaran for their helpful comments.

0

Bank Behavior and Social Capital 1. Introduction Social capital can broadly be defined by the links, shared values, and understandings in society that promote trust and therefore enable working relationships that have productive benefits. Within the academic literature, much attention has been paid to the impact of social capital on the performance of local and national governments (e.g., Putnam, 1993; Laporta et al., 1997; Knack and Keefer, 1997). However, more recent literature has focused on the relationship between social capital and financial transactions, with analysis indicating that there exists a bidirectional relationship between borrowers and lenders driven by trust. The information asymmetry inherent within the relationship between banks and their consumers can potentially lead to moral hazard if bank decision-making and policy setting is driven in any part by characteristics not directly associated with consumer creditworthiness. By studying the relation between social capital and bank decision-making, we document the relation between social capital and bank behavior, as measured by regulatory risk ratios, capitalization, bank default risk and failure, loan performance, deposit rates and fee structures, and loan income. We use two broad measures of social capital: the Putnam Index and Social Capital County. The Putnam Index is computed using principal components analysis on a set of fourteen different factors of associational activities (e.g., number of club memberships, amount of volunteering and participation in Presidential elections, attendance at political meetings, and participation in election campaigns (Putnam, 1993 and 1995)). Social Capital County is a county level, survey-based measure of social capital from Rupasingha and Goetz (2006, 2008) which includes variables representing membership organizations at the county level (e.g., civic organizations, bowling centers, golf clubs) and associational activities (percent of the voting eligible population in each county who voted in presidential elections, county-level response

1

rates to Census Bureau’s decennial census, and per capita non-profit organizations). We also include social controls for crime, education, and church attendance. Our bank data set includes all U.S. banks with data available from 2000 through 2015. We first examine bank behavior using risk measurement ratios used by federal regulators to assess bank risk. To control for endogeneity, we use a two-stage least squares approach in which we first endogenize the social capital measures (using distance of a bank’s headquarters from the Canadian border and voter turnout as instruments) and then regress the dependent variable of interest on the fitted index and controls for bank size, balance sheet composition, loan performance and reserves, and liquidity. Our results suggest that higher levels of social capital are associated with safer, more profitable, yet less capitalized banks. Additionally, banks in areas with higher social capital display lower likelihoods of financial distress and failure. Aligned with lower capitalization levels, we find a positive and significant relationship between social capital and the likelihood of the receipt of capital under the TARP Capital Purchase Program, as well as the likelihood of having a positive return on assets, particularly during the financial crisis. To complete the analysis, we examine the relationship between social capital and bank loan rates and deposit rates and fees. We find that banks in higher social capital areas pay more interest and charge fewer fees on deposits and charge lower rates and fees on loans. The results suggest that banks located in areas with higher social capital take actions that promote trust and therefore enable working relationships that have productive benefits for both the banks and consumers. The remainder of this paper is organized as follows. Section 2 reviews related literature and motivation. Section 3 discusses the data. Section 4 presents methodology and results. Finally, Section 5 concludes the paper. 2. Related Literature and Motivation The idea of quantifying social capital first started in the sociology literature. Jacobs 2

(1961), Coleman (1990), and Burt et al. (2009) define social capital as social and network ties that benefit an individual. Putnam (1993) widens the definitions to classify social capital as “networks, norms, and trust that enable participants to act together more effectively to pursue shared objectives.” Knack and Keefer (1997) and Guiso et al. (2004) use the definition of social capital from Putnam (1993) and conclude that social capital based off trust is essential to wellfunctioning societies and the economic progress of those societies. This literature suggests that social trust and other social capital aspects play an important role in economic performance and may potentially lead to socially efficient outcomes and reduced information asymmetry. One of the mechanisms through which social capital impacts economic efficiency is by enhancing the prevailing level of trust. For example, Guiso et al. (2004) find that the effect of social capital is more pronounced among less educated people, who need to rely more on trust because of their limited understanding of contracting mechanisms. In high social capital communities, people may trust each other more because the networks in their community provide better opportunities to punish deviants. At the same time, in these communities people may rely more on others keeping their promises because of the moral attitude imprinted with education. Research also examines mechanisms by which social capital generates trust needed for financial transactions. For example, Guiso et al. (2004) suggest that a high level of social capital promotes participation of individuals in financial transactions. Guiso et al. (2008) find that less trusting individuals are less likely to buy stock and, conditional on buying stock, they buy less. Allen et al. (2016) find that well-governed firms that suffer less from agency concerns engage more in activities that improve social good. Carlin et al. (2009) find that when the value of social capital is high, government regulation and trustfulness are substitutes. In this case, government intervention may actually cause lower aggregate investment and decreased economic growth. In contrast, when social capital is low, regulation and trustfulness may be complements. Finally, 3

Philipp (2015) shows that investors respond strongly negatively to negative events concerned with a firm's social activities and weakly negatively to positive events. El-Attar and Poschke (2011) find that households with less trust invest more in housing and less in risky financial assets. Georgarakos and Pasini (2011) show that specific trust in advice given by financial institutions represents a prominent factor for stock investing compared to other tangible features of the banking environment. Finally, Duarte et al. (2012) use photographs of potential borrowers from a peer-to-peer lending site and find that borrowers who appear more trustworthy have higher probabilities of having their loans funded, better credit scores, and default less often. They conclude that impressions of trustworthiness matter in financial transactions. More recently, Gupta et al. (2018) examine U.S. firms and find evidence that the implied cost of equity is lower for firms with headquarters in areas with high social capital. Ostergaard et al. (2016) find that stakeholder-oriented savings banks located in communities with high social capital have a higher probability of survival, but no similar effect exists for equity holder-owned commercial banks. The results are a function of the level of trust savings banks engender and the level of civic engagement to which they commit in the communities they serve. The authors also find that social capital is positively related to altruistic bank behaviors. Jin et al. (2017) find that banks in high social capital areas experience fewer failures and less financial trouble during the 2007-2010 financial crisis. Additionally, they find that banks in high social capital areas are more stable, as indicated by decreased risk-taking and increased accounting transparency and conservatism. Lins et al. (2017) find that firms that entered the financial crisis with higher social capital earned higher stock returns and experienced higher margins, sales growth, and sales-peremployee, relative to firms with lower social capital. During the financial crisis, a time characterized by an erosion of trust in firms, markets, and institutions, a firm’s social capital, and 4

the trust that it engenders, paid off. Finally, Hasan et al. (2017) find that banks headquartered in U.S. counties with higher levels of social capital incur lower bank loan spreads and conclude that social capital is perceived by debtholders as a means to constrain opportunistic behavior. In contrast to papers that document beneficial results of high social capital, some research finds that higher social capital and trust may provide a feeding ground for self-serving behavior. Knack and Keefer (1997) provide evidence for conflicting influences of social capital and economic performance. That is, higher social capital provides individuals a way to capture private benefits at the expense of society in general. Similarly, Olson (1982) finds that networks can hurt economic performance because groups can act as lobbyists for their own causes and that, in general, may impose costs on society. Applying this finding to banks, there is the possibility that banks may take advantage of high social capital to capture private benefits, which may enhance the bank’s profits. The main business of depository institutions is to accept deposits from the public and create credit for a community. Since trust among members in a community is an important part of social capital, social capital should affect both the behavior of borrowers and the behavior of lenders, particularly with respect to their capital, loan rates, and deposit rate and fees decisions. For example, credit unions are depository institutions which place strong emphasis on building social capital and empowering both their customers (who are also owners) and the local community in which they are based. Credit unions exist primarily to serve their members with higher interest rates on deposits. They also charge lower interest rates than banks on different consumer loans such as mortgages, auto loans, and home equity lines of credit. Indeed, in 2008 and 2009, industry net income was negative for commercial banks, whereas, industry return on assets remained positive for credit unions. Like credit unions, social capital may reduce the cost of financial contracts for 5

commercial banks, which may increase banks’ profits. Putnam (1993) suggests social capital helps build stakeholder trust and cooperation. According to Arrow (1972), activities that require agents to rely on the future actions of others are accomplished at lower cost in higher trust environments. Coleman (1990) and Spagnolo (1999) suggest that individuals in high social capital areas make additional efforts to honor contracts because there is a high cost of violating the contracts. This research suggests that borrowers in high social capital areas may be less likely to default on loans, which would increase bank profit and may also reduce bank risk. As a result, banks in areas with high social capital may be able to operate with lower capital ratios because their borrowers are less likely to default. Likewise, because customers make additional efforts to honor contracts, banks in areas with high social capital may be less prone to financial distress and remain profitable during times of financial distress. High levels of social capital may also induce bank managers to be less selfish and more publicly minded (i.e., pay higher interest rates on deposits and charge lower interest rates on loans). Given findings of previous research, we examine the degree to which banks in high social capital areas pursue or fail to pursue policies that enhance social capital. Specifically, we document the relation between social capital and bank behavior, measured through regulatory risk measurement ratios, loan performance, bank risk, bank failure, loan rates, and deposit rate and fee structures. 3. Data and Univariates Data used in the analysis come from a number of sources. All variables used in the analysis are defined in Appendix A. We build a quarterly panel data set for the period 2000 through 2015 that includes all commercial banks. Quarterly financial statement data for financial institutions are obtained from the Consolidated Financial Statements for Bank Holding Companies (FR Y-9C) database from Federal Financial Institutions Examination Council 6

(FFIEC). Table 1 presents the sample breakdown of quarterly bank observations by state. We find a relatively even distribution across states with Illinois having the largest percentage of observations (8.36%) and only Alaska and Hawaii containing no observations. As mentioned above, we use two measures of social capital: the Putnam Index and Social Capital County. The Putnam Index uses 14 associational activity measures (including number of club memberships, amount of volunteering and participation in Presidential elections, attendance at political meetings, and participation in election campaigns) to produce a state-level composite index of social capital in the United States. Putnam’s (1993) principal component analysis constructs an index as a weighted sum of each of the components; calculating weights on each of the components that maximize the total sum of the squared correlations between the composite variable and the components. Thus, higher weights are given to components that are more highly correlated with each other and outlier components get lower weights. We collect Putnam Index data from the Bowling Alone database, www.bowlingalone.com. We match sample banks to the Putnam Index based on the state in which the bank is headquartered. Rupasingha et al. (2006) and Rupasingha and Goetz (2008) develop a county-based model of social capital covering the entire United States. They use a data set from the County Business Patterns (CBP), compiled by the Census Bureau, which includes an extensive and comprehensive set of variables representing membership organizations at the county level (e.g., civic organizations, bowling centers, golf clubs). In addition to associational activities, they include the percent of the voting eligible population in each county who voted in presidential elections, county-level response rates to Census Bureau’s decennial census, and per capita nonprofit organizations from the National Center for Charitable Statistics. Based on principal component analysis, they create overall social capital indices from these data for the years 1990, 1997, and 2005 (they later add 2009 to the database). The first principal component is interpreted 7

as the index of social capital (hereafter called Social Capital County). We collect this data from Penn State University’s Northeast Regional Center for Rural Development, http://aese.psu.edu/ nercrd.1 We match sample banks to the Social Capital County based on the county in which the bank is headquartered. Table 2 presents descriptive statistics for the social capital indexes (Putnam Index and Social Capital County). The mean (median) value for the Putnam Index is -0.141 (-0.216) and ranges from -1.15 to 1.29. The mean (median) value for Social Capital County is -0.173 (-0.230) and ranges from -1.71 to 1.74. These values are similar to Gupta et al. (2018), although they are slightly lower than Hasan et al. (2017). Higher values for both measures indicate higher levels of social capital. Social capital control variables include percent of population that attend church, percent of population affected by reported crime in a given year, and percent of high school graduates— all variables are measured as percent in the state in which a bank is headquartered. Putnam (1993) states that communities and regions rich in social capital suffer less crime, educate their children better, have higher church attendance, and have more smoothly functioning economies. Previous research has documented religious engagement as one factor that contributes to overall levels of social capital in a community (e.g., King and Furrow, 2004; Smith, 2003). Hilary and Hui (2009) find that firms located in counties with higher levels of religiosity display lower degrees of risk exposure, exhibit a lower investment rate, and have less growth, but generate a more positive market reaction, when they announce new investments. More specific to banking, Adhikari and Agrawal (2016) find that banks headquartered in more religious areas exhibit lower stock return volatility, lower tail risk, and lower idiosyncratic risk. Akçomak and ter Weel (2012)

1

Social Capital County data are available for the years 1990, 1997, 2005, and 2009. We follow existing literature and fill in gap years using the most recent values available.

8

find that higher levels of social capital are associated with lower crime rates. In the context of education, social capital in the forms of parental expectations, obligations, and social networks that exist within the family, school, and community are important for student success (Helliwell and Putnam, 1999). Pevzner et al. (2015) find that investor reactions to earnings announcements are significantly higher in more trusting countries. They also find that the positive effect of societal trust on investor reactions to earnings news is more pronounced when a country's investor protection and disclosure requirements are weaker (suggesting that trust acts as a substitute for formal institutions), and when a country's average education level is lower (consistent with less educated people relying more on trust in making economic decisions). We collect county-level data for percent of population that attend church from the Association of Statisticians of American Religious Bodies, http://www.thearda.com, and percent of population affected by reported crime in a given year from the Uniform Crime Reporting Statistics, http://www.ucrdatatool.gov. State-level data on percent of high school graduates is collected from the U.S. Census Bureau, http://www.census.gov. Table 2 presents descriptive statistics on the variables. On average, 60.30% of the population attends church, 7.40% is affected by crime, and 19.53% graduate high school. Table 3 presents a correlation matrix among the social capital variables. Confirming previous research, there is a negative correlation between social capital and reported crime (-39.53% using the Putnam Index and -34.53% using Social Capital County, both significant at 1%) and a positive correlation between social capital and education (26.98% using the Putnam Index, significant at 1%, and 1.90% using Social Capital County, insignificant). Correlations between social capital and church attendance are mixed: -1.35% for the Putnam Index and 3.45% for Social Capital County. However, neither is significant. FR Y-9C reports include data for risk measurement metrics: loan loss provision/total 9

loans, total loan net charge-offs (NCOs)/total loans, pre-provision net revenue (PPNR), and Tier 1 risk-based capital (RBC) ratio. Loan loss provisions, pre-provision net revenue, and Tier 1 risk-based capital ratios are three of the five ‘risk measurement metrics’ now used by regulators to determine capital adequacy (as part of the Comprehensive Capital Analysis and Review (CCAR) program). 2 Loan loss provisions and total loan NCOs are components of loan losses recorded by a bank in a given quarter. Loan loss provisions are the expected losses on the loan portfolio that are recognized in the quarter, while NCOs are any additional losses or recoveries received when a bad loan is finally removed from a bank’s balance sheet. The two measures together are a major driver of bank losses each quarter and are used by regulators to project net income. However, they are excluded from PPNR. Thus, we include them as dependent variables for regression analysis. The risk measures are calculated from bank financial statements as: 1. Loan loss provisions/total loans 2. Total loan NCOs/total loans 3. PPNR/total assets = (net interest income + noninterest income – noninterest expense)/total assets 4. Tier 1 RBC ratio = Tier 1 capital/risk-weighted assets Data reported in Table 2 suggest fat-tailed distributions for the risk measurement metrics. For example, the mean (median) loan loss provisions/total loans ratio for the sample is 0.31% (0.10%), total loan NCOs/total loans ratio is 0.30% (0.10%), PPNR/total assets ratio is 8.30% (4.20%), and Tier 1 risk-based capital (RBC) ratio is 12.80% (12.00%). The values in our sample are aligned with regulatory expectations for capital adequacy: 12.00% Tier 1 RBC ratio. To examine the relation between social capital and productive benefits for both banks and consumers, we collect data on bank deposit rates and fees, and loan rates and fees from FR Y9C’s. Reported in Table 2, the mean (median) interest expense on core deposits/core deposits is

2

CCAR also includes the Tier 1 leverage ratio and total RBC ratio. The paper reports only results using the Tier 1 RBC ratio. Results using the other CCAR capital ratios are similar and lead to the identical conclusions.

10

5.60% (2.44%). Deposit interest rate and fee variables in our sample include automated teller machine (ATM) fees/core deposits and check income/core deposits. Table 2 shows means (medians) for the two measures are 0.09% (0.00%) and 0.01% (0.00%), respectively. On the lending side, we look at loan fee and interest income/total loans: the mean (median) value for the sample is 4.36% (4.26%). Combining interest income and interest expense, we calculate net interest income (total income on investment securities and loans minus total interest expense)/total loans: the mean (median) value for the sample is 3.19% (3.00%). An alternative source of income to consumer-oriented loan interest and fees is noninterest income, which results from on- and off-balance-sheet activities. Noninterest income has become increasingly important to banks as the ability to attract core deposits and high-quality loan applicants becomes more difficult. Included in this category is income from fiduciary activities (for example, earnings from operating a trust department), trading revenues (gains [losses] and fees from trading marketable instruments and off-balance sheet (OBS) derivative instruments), fees from other-than-banking activities (such as security brokerage, investment banking, and insurance), servicing fees (from mortgages, credit cards, and other assets), and gains and losses from the sale of investment securities. The mean (median) noninterest income/total income ratio for the sample is 2.45% (0.67%). Additionally, from FR Y-9C’s we collect data on bank-specific independent variables used to control for operating differences between banks. These measures include total assets, loan performance (nonperforming loans/total loans and loan loss reserve/total loans), portfolio composition controls (percent of total loans for commercial and industrial, agricultural, consumer, foreign government, real estate, and depository institution), and liquidity ratio (cash and investment securities/total assets). Table 2 shows the descriptive statistics for the financial statement variables. The average bank in the sample has approximately $9.10 billion in assets 11

(ranging from $0.18 million to $2.57 trillion). The mean value of nonperforming loans (which includes loans past due 90 days or more and still accruing interest and loans in nonaccrual status)/total loans for the sample banks is 1.40% (0.70% median). Nonperforming loans are still listed on a bank’s balance sheet as an asset. The reserve for loan losses is a contra asset account that serves as an estimate by the bank’s management of the amount of gross loans that will not be repaid to the bank. The reserve for loan losses is an accumulated reserve that is adjusted each period as management recognizes the possibility of additional bad loans and makes appropriate provisions for such losses. Although tax laws influence the maximum amount of the reserve, the bank’s management actually sets the level based on loan growth and recent loan loss experience. The mean (median) loan loss reserve/total loans ratio for the sample is 1.50% (1.30%). The distribution of banks’ loan portfolio shows that real estate loans are the most predominant (mean is 26.40%, ranging from 0.26% to 89.49%), followed by commercial and industrial loans (mean 15.90%, ranging from 3.31% to 34.38%) and consumer loans (mean 7.50%, ranging from 0.29% to 22.53%). Finally, a bank’s ability to absorb losses is also affected by the amount of liquid assets held. To reduce liquidity risk, banks hold cash and other liquid assets as part of their overall management strategy. We include a liquidity ratio (cash and investment securities/total assets) in our analysis. Table 2 shows that banks’ average liquidity ratio is 23.20% over the sample period, ranging from 5.95% to 46.12%. 4. Methodology and Results The econometric approach of this paper is two-fold. First, to understand which elements of social capital may drive bank behavior, we estimate regressions which establish relationships between dependent variables and measures of social capital. Specifically, we use OLS regression analyses to examine regulatory risk measures as dependent variables and social capital variables 12

and controls as independent variables, with bank-specific independent variables to control for operating differences between banks. If social capital reduces the cost of financial contracts, and thus increases banks’ profits, we expect banks in areas with higher social capital to have lower loan losses, higher profit, and lower capital ratios. We examine versions of the following regression: 𝑅𝑖𝑠𝑘 𝑀𝑒𝑎𝑠𝑢𝑟𝑒𝑖,𝑡 = ∝ +𝛽1 ∗ (𝑆𝑜𝑐𝑖𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡 ) + 𝛽2 ∗ (𝑆𝑜𝑐𝑖𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡 ) + 𝛽3 ∗ (𝐵𝑎𝑛𝑘 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡 ) + 𝜀

(1)

In all regressions in the paper, bank and year fixed effects control for unobserved heterogeneity within the variables. Risk measures include loan loss provision/total loans, total loan NCOs/total loans, PPNR/total assets, and Tier 1 RBC ratio. As noted above, loan loss provision/total loans, PPNR/total assets, and Tier 1 RBC capital ratios are now used by regulators to determine capital adequacy. Further, loan loss provisions and total loan NCOs are components of loan losses recorded by a bank in a given quarter. However, they are excluded from PPNR. Thus, we include them as dependent variables for regression analysis. Social capital variables include the statelevel Putnam Index and county-level Social Capital County. Social capital controls are percent of population that attends church, percent of population affected by crime, and percent of the state population that is educated. Bank control variables include bank size, nonperforming loans/total loans, reserve for loan losses/total loans, composition of loan portfolio, and liquidity ratio. Second, we use instrumental-variable, two-stage least squares regressions (2SLS) to mitigate endogeneity concerns. First stage instruments used to assist in proper identification of the fitted value of social capital measures are distance of the bank’s headquarters from the Canadian border and voter turnout in the state in which the is bank headquartered. Ln(Canada) is the log of the distance from the bank’s headquarters to the Canadian border, from https://www. freemaptools.com/measure-distance.htm. Voter Turnout is percent of the voting eligible 13

population in the state in which the bank is headquartered that voted for the highest office in a given election year, obtained from www.electproject.org. The numerator is the number of people who voted for the "highest office" in a given election. The denominator is the number of people eligible to vote. The use of these instruments is supported by prior research. Hasan et al. (2017) and Gupta et al. (2018) reference Putnam (2001) and argue that distance to the Canadian border is the best single predictor of the level of social capital within the United States, where being closer to the Canadian border means more social capital. Additionally, both participation in social activities (Alesina and La Ferrara, 2000) and level of trustworthiness (Glaeser et al., 2000) have been found to be lower in areas with lower voter turnout. Hasan et al. (2017) conclude that, as there are no material incentives to vote, the public only engages in voting activity as a civic responsibility, which should improve social capital. Thus, these two instrumental variables are related to social capital. However, neither variable should influence bank behavior. 4.1. Risk Measurement Ratios Table 4 presents OLS regression results, which suggest that high social capital is associated with lower loan risk and improved operating performance. The coefficients on the social capital variables (Putnam Index and Social Capital County3) in regressions 1 and 2 are all -0.001, while the coefficients in regression 3 are 0.002 and 0.001, respectively (all significant at 1%). The negative relation between loan loss provision/total loans (and total loan NCOs/total loans) and social capital suggests that high social capital is associated with lower risk loan portfolios. The positive and significant relation between PPNR/total assets and social capital suggests that higher levels of social capital (and the lower risk loan portfolio) enhance operating

3

In this table, we include both measures of social capital in the regressions and find that they remain negative and significant despite the significant positive correlation between the two variables. In the next set of tables, instrumental-variable two-stage regressions require that we separate the two measures.

14

performance. Finally, from Table 4, regression 4, high social capital is associated with lower Tier 1 RBC ratios (coefficients on Putnam Index and Social Capital County are -0.004 and -0.001, respectively, significant at 1%). These findings suggest that borrowers in high social capital areas are less likely to default on loans, which increases bank profit and reduces bank risk. As a result, banks in areas with high social capital may be able to operate with lower capital ratios because their borrowers are less likely to default. A test of economic significance shows moving from the 25th to the 75th percentile value of social capital areas (measured by Putnam) decreases loan loss provision/total assets by 16.97%.4 We see similar economically significant total loan NCOs/total loans, PPNR/total assets, and Tier 1 RBC ratios (-12.47%, 5.04%, -6.82%, respectively for a move from the 25th percentile to the 75th percentile values of Putnam social capital, holding all other independent variables constant at their mean values). From Table 4, we see that social capital controls are significantly related to the dependent variables. Church Attendance is associated with lower loan loss provision/total loans, total NCOs/total loans, and capital levels (coefficients in regressions 1, 2, and 4 are -0.001, -0.001, and -0.002, respectively, all significant at 1%). Crime is associated with higher loan loss provision/total loans and total loan NCOs/total loans, and lower capital levels (coefficients in regressions 1, 2, and 4 are 0.006, 0.002, and -0.138 respectively, significant at 1%). Education is associated with lower loan loss provision/total loans and total NCOs/total loans, and higher capital levels (coefficients in regressions 1, 2, and 4 are -0.001, -0.001, and 0.001 respectively, significant at better than 5%). The results are consistent with Jin et al. (2017) who find that banks in high social capital areas are more stable, as indicated by decreased risk-taking and increased

4

If we hold all of the independent variables constant, and set the value of the Putnam Index to the 25 th percentile value, loan loss provision/total assets is 0.1921%. If we set Putnam to the 75th percentile value, loan loss provisions/total assets are 0.1595%. This shows a 16.97% decrease in the value of loan loss provisions/total assets ((0.1595%/0.1921%)-1).

15

accounting transparency and conservatism. Finally, in line with Cornett et al. (2017), results suggest that larger banks have riskier loan portfolios, smaller PPNR/Total Assets, and hold less capital (e.g., coefficient on ln(Total Assets) is 0.001 in regressions 1 and 2, -0.001 in regression 3, and -0.004 in regression 4, all are significant at 1%). Further, banks with more commercial and industrial, consumer, and real estate loans have larger loan loss provision to total loans (coefficients on C&I Loans/Total Assets, Consumer Loans/Total Loans, and Real Estate Loans/Total Loans are 0.002, 0.003, and 0.000, respectively, in regression 1, significant at 1%). A similar trend is seen with total loan NCOs/total loans in regression 2. Banks with higher liquidity on the balance sheet have higher Tier 1 RBC ratios and less risky portfolios, yet lower operating profit (the coefficient on Liquidity ratio in regression 4 is 0.129, in regressions 1 and 2 are -0.002 and -0.001, respectively, and in regression 3 is -0.019, all significant at 1%). Table 5 presents results from 2SLS estimations in which the social capital index is fitted within stage one and used as an independent variable within stage two. For both the Putnam Index and Social Capital County measures, the instruments (ln(Canada) and Voter Turnout) satisfy the exclusion criterion based on the Hansen J-statistic and p-values corresponding to the Sargan C-statistic (reported in the bottom two lines of the table) reject the null hypothesis that the measures of social capital are exogenous. Additionally, coefficients on both instruments align with intuition and prior literature (i.e., negative for ln(Canada) and positive for Voter Turnout). The regressions confirm results of Table 4. Coefficients on social capital (Putnam Index and Social Capital County) in regressions 3, 4, 7, and 8 are all -0.001, while coefficients in regressions 5 and 9 are 0.002 and 0.001, respectively (all significant at better than 5%). Finally, high social capital is associated with lower Tier 1 RBC ratios (coefficients on Putnam Index and Social Capital County are -0.942 and -0.231 in regressions 6 and 10, respectively, significant at 16

5%). A test of economic significance shows that moving from the 25th to the 75th percentile values of social capital measured by Putnam, loan loss provisions decrease by 14.15%.We see similar economically significant total loan NCOs/total loans, PPNR/total assets, and Tier 1 RBC ratios (changes of -6.96%, 6.16%, and -18.12% respectively). It is important to note that the coefficient on Putnam rises sharply for the Tier 1 RBC 2SLS estimation in comparison to the OLS estimation. This implies that examining the link between social capital and Tier 1 RBC using OLS understates rather than overstates the effect of social capital. Overall, these findings suggest that, controlling for endogeneity, borrowers in high social capital areas are less likely to default on loans, which increases bank profit and reduces bank risk. As a result, banks in areas with high social capital may be able to operate with lower capital ratios because their borrowers are less likely to default. In Table 5, we also see that, in almost every case, bank control variables have the same signs and significance levels. 4.2. Bank Financial Distress and Failure Results from Tables 4 and 5 suggest that banks operating in high social capital areas are less likely to experience loan defaults, which reduces bank risk and increases bank profit. As a result, banks in areas with high social capital may be able to operate with lower capital ratios. Given these differences, a natural follow-up is to examine whether there is a relation between social capital and bank financial distress and subsequent failure. We first analyze the relation between the risk of financial distress and social capital for banks. To measure financial distress, we use the Z-score, initially developed by Roy (1952) and subsequently advanced by Boyd and Graham (1986), Hannan and Hanweck (1988) and Boyd et al. (1993), which relates a firm’s capital level to the variability in its returns to determine losses that can be absorbed without the firm becoming insolvent. The variability in returns is typically measured as the standard deviation of return on assets (ROA). Accordingly, we calculate 17

quarterly values of a bank’s Z-score as the sum of the equity capital ratio (common equity/total assets) and ROA divided by the standard deviation of ROA, where standard deviation is the quarterly deviation over the three prior years. This measure of financial distress links a bank’s capitalization with its ROA and risk (volatility of ROA), and indicates the allowable drop in ROA before the bank becomes insolvent. In other words, Z-score represents a bank’s distance from insolvency. A higher value of Z-score indicates a lower probability of financial distress. The average Z-score for the sample banks (reported in Table 2) is 12.22, which is in line with Zscore estimates in Bouvatier et al. (2017). We relate bank Z-scores to social capital and bank specific control variables5 using the following 2SLS estimation: 𝑍 − 𝑠𝑐𝑜𝑟𝑒𝑖,𝑡 = ∝ +𝛽1 ∗ (𝑆𝑜𝑐𝑖𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡 ) + 𝛽2 ∗ (𝐵𝑎𝑛𝑘 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡 ) + 𝜀 (2) Results presented in Table 6 use the same instruments as Table 5 (the Putnam Index in regression 1 and Social Capital County in regression 2), with both estimations satisfying the Hansen and Sargan tests for exclusion criterion and exogeneity of social capital. Results in Table 6 indicate that high social capital areas are associated with higher Z-scores (coefficient on Putnam Index is 2.556 (regression 2) and on Social Capital County is 3.275 (regression 4), both are significant at 5% or better) and therefore lower financial distress. This result aligns with Adhikari and Agrawal (2016), Ostergaard et al. (2016), and Jin et al. (2017) in that banks in high social capital are safer. Next, we analyze the relation between bank failure and social capital using the following 2SLS estimation: % 𝐹𝑎𝑖𝑙𝑒𝑑 𝐵𝑎𝑛𝑘𝑠𝑖,𝑡 =∝ +𝛽1 ∗ (𝑆𝑜𝑐𝑖𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 𝑖,𝑡 ) + 𝛽2 ∗ (𝑀𝑒𝑑𝑖𝑎𝑛 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝐵𝑎𝑛𝑘 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡 ) + 𝜀

5

(3)

To conserve space, we do not report coefficient results for bank control variables in the remaining tables. However, in all cases, the signs and significance levels are similar to those reported in Tables 4 and 5.

18

% Failed Banks is calculated as the number of failed banks in the county (from the Federal Reserve) divided by the total number of banks in that county. We use the median quarterly values of bank-specific financial ratios in each quarter to control for operating differences between banks in different counties. Table 7 presents the regression results. Similar to Tables 5 and 6, we use instrument variables to estimate Putnam Index in regression 1 and Social Capital County in regression 3. In both regressions the Hansen and Sargan test results confirm appropriate exclusions and exogeneity. Results in Table 7 indicate that not only are banks in high social capital areas safer (Table 6), but they also fail at a lower rate (coefficient on Putnam Index is -0.002 (regression 2) and on Social Capital County is -0.003 (regression 4), significant at 10% and 5%, respectively). This is consistent with Jin et al. (2017) who find that banks in high social capital areas experience fewer failures and less financial trouble during the 2007-2010 financial crisis. 4.3. Financial Crisis Performance Related to financial distress and failure, we next look at the relation between social capital and bank performance during the financial crisis. First, we estimate 2SLS regressions similar to those in Table 5, but examine separately bank risk measures during (2008-2009) versus outside (2000-2007 and 2010-2015) the financial crisis. Table 8 presents results from the estimations in which the social capital index is fitted within stage one and used as an independent variable within stage two. Panel A shows regressions in which social capital is measured using the Putnam Index and Panel B shows regressions using Social Capital County. For both the Putnam Index and Social Capital County measures, the instruments (ln(Canada) and Voter Turnout) again satisfy the exclusion criterion based on the Hansen J-statistic and p-values corresponding to the Sargan C-statistic (reported in the bottom two lines of the table) reject the null hypothesis that the measure of social capital is exogenous. 19

Table 8 regression results confirm results of Table 5 showing that high social capital is associated with lower loan risk and improved operating performance. Further, the results are consistent both within and outside the financial crisis. Coefficients on the Putnam Index in regressions 1 and 2 are both -0.001 (significant at 10% and 1%, respectively, and not significantly different from each other), in regressions 3 and 4 are also both -0.001 (significant at 10% and 1%, respectively, and not significantly different from each other), and in regressions 5 and 6 are 0.001 and 0.002, respectively (significant at 1% and not significantly different from each other). Further, high social capital is associated with lower Tier 1 RBC ratios in regressions 7 and 8 (coefficients on Putnam Index are -0.001 and -0.006 in regressions 7 and 8, respectively, significant at 1% and not significantly different from each other). Identical results and conclusions are found in Panel B using Social Capital County. These findings suggest that, both during and outside the financial crisis years, borrowers in high social capital areas are less likely to default on loans, which increases bank profit and reduces bank risk. As a result, banks in areas with high social capital may be able to operate with lower capital ratios because their borrowers are less likely to default. We next analyze the association between the receipt of TARP Capital Purchase Program (CPP) funds and social capital. For this test, we only look at the 2008-2009 period. We split this sample according to whether a bank received TARP funds during the crisis. Using an instrumental variables approach within a probit regression framework (IVPROBIT), we first endogenize the social capital measure and then use that in the second stage as an independent variable when regressed on a TARP indicator variable (one if the bank received TARP funds and zero otherwise) as shown below: 𝑇𝐴𝑅𝑃𝑖,𝑡 =∝ +𝛽1 ∗ (𝑆𝑜𝑐𝑖𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 𝑖,𝑡 ) + 𝛽2 ∗ (𝐵𝑎𝑛𝑘 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡 ) + 𝜀 20

(4)

Results reported in Table 9 suggest that banks in high social capital areas are associated with a greater likelihood of receiving TARP funds (coefficient on Putnam Index is 0.031 in regression 2 and on Social Capital County is 0.358 in regression 4, both significant at 1%). In evaluating the banks that would receive these funds, the U.S. Treasury intended to use TARP CPP funds to inject capital into healthy banks as a way to stimulate lending and restore credit flowing in the economy. Thus, the receipt of TARP funds is a sign that regulators expected these banks to survive the financial crisis. Consistent with the results in Tables 6 and 7, we find that these banks that are more likely to survive the crisis are in areas with higher social capital. Finally, we examine the association between bank ROAs and social capital both overall and during and outside the financial crisis. As mentioned above, credit unions are depository institutions which place strong emphasis on building social capital and empowering their customers and the local community in which they are based. As a result, during the financial crisis, while industry average net income was negative for commercial banks, credit unions return on assets remained positive. This test allows us to exam whether banks in high social capital areas, as was the case with credit unions, are those more likely to have positive ROAs both during and outside the financial crisis. We split the sample according to whether a bank has a positive ROA for the quarter. In addition to looking at the full sample period, we also examine separately bank periods during (2008-2009) versus outside (2000-2007 and 2010-2015) the financial crisis. Using an instrumental variables approach within a probit regression framework (IVPROBIT), we first endogenize the social capital measure and then use that in the second stage as an independent variable when regressed on by an ROA indicator variable (one if the bank’s ROA is greater than 0 for the quarter and zero otherwise) as shown below: 𝑅𝑂𝐴𝑖,𝑡 =∝ +𝛽1 ∗ (𝑆𝑜𝑐𝑖𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 𝑖,𝑡 ) 21

+ 𝛽2 ∗ (𝐵𝑎𝑛𝑘 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡 ) + 𝜀

(5)

Results presented in Table 10 suggest that banks in high social capital areas are associated with a greater likelihood of having a positive ROA (coefficient on Putnam Index is 0.183 in regression 3 and on Social Capital County is 0.210 in regression 6, significant at 5% and 1%, respectively). Further, the results are more pronounced during the financial crisis. The coefficient on Putnam Index is 0.038 in regression 4 (non-crisis years) and is 0.216 in regression 5 (crisis years), the difference is significant at 1%. Likewise, the coefficient on Social Capital County is 0.013 in regression 7 (non-crisis years) and is 0.246 in regression 8 (crisis years), the difference is significant at 1%. Thus, banks in high social capital areas see significantly larger profit (ROA) during the financial crisis. 4.4. Social Capital and Bank Deposit Rates and Fees and Loan Income Results thus far have highlighted the relation between social capital and overall bank risk and performance. We next study the other half of the bi-directional relationship: that between banks and consumers of banking products (borrowers and depositors). As mentioned above, previous research documents beneficial results of high social capital. Ostergaard et al. (2016) find that social capital is positively related to altruistic bank behaviors. Hasan et al. (2017) find that banks headquartered in counties with higher levels of social capital incur lower bank loan spreads and diminished opportunistic behavior. In contrast, Knack and Keefer (1997) find that higher social capital provides way to capture private benefits at the expense of society in general. This section examines the relation between social capital and consumer/bank relations (i.e., deposit rate and fee structures and loan income). We test whether social capital is positively related to altruistic bank behaviors, or whether banks take advantage of high social capital to capture private benefits, which may enhance profits.

22

We again first estimate ordinary least squares regressions with deposit interest and fees, loan income, noninterest income, and net income as dependent variables on social capital index variables, social capital controls, and bank-specific independent variables to control for operating differences between banks. We then confirm results using a more robust 2SLS estimation framework in which we endogenize the social capital indices and use the fitted values as independent variables in the second stage. We use the following model for both the OLS and 2SLS approaches: 𝐼𝑛𝑐𝑜𝑚𝑒(𝐸𝑥𝑝𝑒𝑛𝑠𝑒)𝑖,𝑡 = ∝ +𝛽1 ∗ (𝑆𝑜𝑐𝑖𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡 ) +𝛽2 ∗ (𝑆𝑜𝑐𝑖𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠𝑖,𝑡 ) + 𝛽3 ∗ (𝐵𝑎𝑛𝑘 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡 ) + 𝜀

(6)

Income items include ATM Fees/Core Deposits, Check Income/Core Deposits, Fee and Interest Income/Total Loans, and Total Noninterest Income/Total Income. Expense items include Interest Expense on Core Deposits/Core Deposits. Combining income and expense, a final measure is Net Interest Income/Total Loans. Table 11 presents OLS regression results. Looking first at deposit trends, we see that banks in high social capital areas charge significantly lower ATM and checking fees and pay higher rates on core deposits (coefficients on Putnam Index in regressions 1 through 3 are -0.001, -0.001, and 0.001, respectively, all significant at 1%). Signs and significance levels for Social Capital County are similar. Further, regression 4 shows that banks in high social capital areas charge significantly lower loan rates (coefficient on Putnam Index is -0.001 and on Social County Capital is -0.001, both significant at 1%). Focusing on the Putnam Index, we find that holding all other independent variables constant at their mean values, moving from the 25th to the 75th percentile value of Putnam Index leads to a 20.48% decline in ATM fees, a 63% decline in checking fees, and a 10% increase in interest expense on core deposits.

23

Combining interest income and interest expense, Regression 5 shows that banks in high social capital areas have significantly lower net return on loans (coefficient on Net Interest Income/Total Loans is -0.163, significant at 1%). Moving from the 25th to 75th percentile value for Putnam Index decrease the ratio of net interest income by 7.39%. This supports previous findings that banks in higher social capital areas incur lower bank loan spreads and diminished opportunistic behavior. However, the result appears to be inconsistent with results from Tables 4 and 5 where we see a positive and significant relation between PPNR/Total Assets and social capital, suggesting improved operating performance. However, PPNR/Total Assets includes noninterest income while Net Interest Income/Total Loans does not. Regression 6 shows that banks in high social capital areas earn significantly more noninterest income (coefficient on Putnam Index is 0.001 and on Social County Capital is 0.018, significant at 5% and 1%, respectively). Moving from the 25th to 75th percentile value for Putnam Index increase the ratio of net interest income by 5.48%. As noted above, this category includes such things as income from fiduciary activities, trading revenues, fees from other-than-banking activities, servicing fees, and gains and losses from the sale of investment securities, and not consumer oriented lines of business. Thus, banks build social capital and empower their customers (borrowers and depositors) and the local community in which operate. Banks in high social capital areas commit to a high level of civic engagement in the communities they serve, which gives rise to increased profits. The results suggest that banks located in areas with higher social capital take actions that promote trust and therefore enable working relationships that have productive benefits for both the banks and consumers. Table 12 presents results from 2SLS regressions in which the social capital indices are fitted within stage one and used as independent variables within stage two. Across both social capital measures, the instruments (ln(Canada) and Voter Turnout) satisfy the exclusion criterion 24

based on the Hansen J-statistic and p-values corresponding to the Sargan C-statistic reject the null hypothesis that the measure of social capital is exogenous. Further, both sets of results confirm the results of Table 11: banks in areas with higher social capital pay more interest and charge fewer fees on deposits and charge lower rates and fees on loans. 5. Conclusion In this paper, we examine the degree to which banks in high social capital areas pursue or fail to pursue policies that enhance social capital by studying the relationship between social capital and bank decision-making, as measured by regulatory risk ratios, capitalization, bank default risk and failure, loan performance, deposit rates and fee structures, and loan income. Our results suggest that higher levels of social capital are associated with safer, more profitable, yet less capitalized banks. Additionally, banks in areas with higher social capital values display lower likelihoods of default and failure. Aligned with lower capitalization levels, we find a positive and significant relationship between social capital and the likelihood of the receipt of capital under the TARP Capital Purchase Program. To complete the analysis, we examine the relation between social capital and bank loan rates and deposit rates and fees. We find that banks in areas with higher social capital pay more interest and charge fewer fees on deposits, and charge lower rates and fees on loans. The results suggest that banks located in areas with higher social capital take actions that promote trust and therefore enable working relationships that have productive benefits for both the banks and consumers.

25

References Adhikari, B.K., and Agrawal, A., 2016. Does local religiosity matter for bank risk-taking? Journal of Corporate Finance 38, 272-293. Allen, F., Liang, H., and Renneboog, L., 2016. Socially responsible firms. Journal of Financial Economics 122(3), 252-606. Akçomak, S., and ter Weel, B., 2012. The impact of social capital on crime: Evidence from the Netherlands. Regional Science and Urban Economics 42(1-2), 323-340. Alesina, A., and La Ferrara, E., 2000. Participation in heterogeneous communities. The Quarterly Journal of Economics 115, 847-904. Bouvatier, V., Lepetit, L., Rehault, P.N., and Strobel, F., 2017. Bank insolvency risk and Z-score measures: Caveats and best practice. University of Birmingham Working Paper. Boyd, J. H., and Graham, S. L., 1986. Risk, regulation, and bnk holding company expansion into nonbanking. Quarterly Review (Spring), 2-17. Boyd, J. H., Graham, S. L., and Hewitt, R. S., 1993. Bank holding company mergers with nonbank financial firms: Effects on the risk of the failure. Journal of Banking and Finance 17(1), 43-63. Brinig, M.F., 2011. Belong and trust: Divorce and social capital. Brigham Young University Journal of Public Law 25, 271-286. Burt, R.S., Bartkus, V.O., and Davis, J.H., 2009. Network duality of social capital. Social capital: Reaching Out, Reaching In, Edward Elgar Publishing, Cheltenham, 39-65. Carlin, B.I., Dorobantu, F., and Viswanathan, S., 2009 Public trust, the law, and financial investment. Journal of Financial Economics 92, 321-341. Coleman, J.S., 1988. Social capital in the creation of human capital. American Journal of Sociology XCIV (1988), S95-S120. Coleman, J.S., 1990. Foundations of Social Theory. Cambridge, MA: Harvard University Press. Cornett, M.M., Minnick, K., Schorno, P.J., and Tehranian, H., 2017. An examination of bank behavior around Federal Reserve stress tests. Bentley University Working Paper. Duarte, J., Siegel, S., and Young, L., 2012. Trust and credit: The role of appearance in peer-topeer lending. Review of Financial Studies 25, 2455-2483. El-Attar, M., and Poschke, M., 2011. Trust and the choice between housing and financial assets: evidence from Spanish households. Review of Finance 15, 727–756.

26

Georgarakos, D. and Pasini, G., 2011. Trust, sociability, and stock market participation, Review of Finance 15, 693–725. Glaeser, E.L., Laibson, D.I., Scheinkman, J.A., and Soutter, C.L., 2000. Measuring trust. The Quarterly Journal of Economics 115, 811-846. Guiso, L., Sapienza, P., and Zingales, L., 2004. The role of social capital in financial development. American Economic Review 94, 526-556. Guiso, L., Sapienza, P., and Zingales, L., 2008. Trusting the stock market. Journal of Finance 63, 2557-2600. Gupta, A., Raman, K., and Shang, C., 2018. Social capital and the cost of equity. Journal of Banking and Finance 87, 102-117. Hannan, T. H., and Hanweck, G. A., 1988. Bank insolvency risk and the market for large certificates of deposit. Journal of Money, Credit and Banking 20(2), 203-211. Hasan, I., Hoi, C.K., Wu, Q., and Zhang, H., 2017. Social capital and debt contracting: Evidence from bank loans and public bonds. Journal of Financial and Quantitative Analysis 52(3), 1017-1047. Helliwell, J.F. and Putnam, R.D., 1999. Education and social capital. NBER Working Paper No. 7121. Hilary, G., and Hui, K.W., 2009. Does religion matter in corporate decision making in America. Journal of Financial Economics 93(3), 455-473. Jacobs, J., 1961. The death and life of great American cities. New York: Random House. Jin, J.Y., Kanagaretnam, K., Lobo, G.J., and Mathieu, R., 2017. Social capital and bank stability. Journal of Financial Stability 32, 99-114. Knack, S., and Keefer, P., 1997. Does social capital have an economic payoff? A cross-country investigation. Quarterly Journal of Economics 112, 1251-1288. King P.E., and Furrow, J.L., 2004. Religion as a resource for positive youth development: religion, social capital, and moral outcomes. Developmental Psychology 40(5), 703–713. Krüger, P., 2015. Corporate goodness and shareholder wealth. Journal of Financial Economics 115, 304- 329. Lins, K., Servaes, H., and Tamayowe, A., 2017. Social capital, trust, and firm performance: The value of corporate social responsibility in the financial crisis. Journal of Finance 72(4), 1785-1824.

27

Ostergaard, C., Schindele, I., and Vale, B., 2016. Social capital and the viability of stakeholderoriented firms: Evidence from savings banks. Review of Finance 20(5), 1673-1718. Pevzner, M., Xie, F., and Xin, X., 2015. When firms talk, do investors listen? The role of trust in stock market reactions to corporate earnings announcements. Journal of Financial Economics 117(1), 190-223. Putnam, R., 1995. Bowling alone: America's declining social capital. Journal of Democracy VI, 65-78. Putnam, R., 1993. Making democracy work: Civic traditions in modern Italy. Princeton University Press. Putnam, R., 2001. Social Capital: Measurement and consequences. Canadian Journal of Policy Research, 41 – 51. Ramseyer, M.J., 2014. Litigation and social capital: Divorces and traffic accidents in Japan. Journal of Empirical Legal Studies 11, 39-73. Roy, A.D., 1952. Safety first and the holding of assets. Econometrica 20(3), 431-449. Rupasingha A., Goetz S.J., and Freshwater D., 2006. The production of social capital in US counties. Journal of Socio-Economics 35, 83–101. Rupasingha, A., and Goetz, S.J., 2008. US County-Level Social Capital Data, 1990-2005. The Northeast Regional Center for Rural Development, Penn State University. Smith, C., 2003. Theorizing religious effects among American adolescents. Journal for the Scientific Study of Religion 42(1), 17–30.

28

Table 1 Sample Breakout by State This table presents the number of quarterly observations for the sample banks by state from 2000 through 2015 (79,448 total quarterly observations). Data for financial institutions are obtained from the Consolidated Report of Condition and Income database. State AL AR AZ CA CO CT DC DE FL GA IA ID IL IN KS KY LA MA MD ME MI MN MO MS MT

Number Percent of of Obs. Sample 1,440 1.81 1,948 2.45 143 0.18 3,640 4.58 1,091 1.37 525 0.66 75 0.09 354 0.45 2,337 2.94 3,153 3.97 2,302 2.90 234 0.29 6,643 8.36 2,103 2.65 1,867 2.35 2,088 2.63 1,357 1.71 2,317 2.92 875 1.10 654 0.82 2,092 2.63 2,353 2.96 3,011 3.79 1,328 1.67 510 0.64

State NC ND NE NH NJ NM NV NY OH OK OR PA RI SC SD TN TX UT VA VT WA WI WV WY Total

29

Number of Obs. 1,886 585 1,238 199 1,602 469 263 2,933 2,172 1,557 525 3,895 197 1,103 660 2,431 5,527 422 2,090 321 1,443 2,554 661 275 79,448

Percent of Sample 2.37 0.74 1.56 0.25 2.02 0.59 0.33 3.69 2.73 1.96 0.66 4.90 0.25 1.39 0.83 3.06 6.96 0.53 2.63 0.40 1.82 3.21 0.83 0.35 100

Table 2 Descriptive Statistics This table presents descriptive statistics for the sample banks over the period 2000-2015. Quarterly financial statement data for financial institutions are obtained from the Consolidated Report of Condition and Income database (79,448 bank quarters). Several variables are not available for all bank quarters including: ATM Fees, Check Income, and Fee and Interest Income. For these variables, we require observations for at least half of the bank quarters to ensure a robust sample. Putnam Index data are from the Bowling Alone database, Social Capital County from Penn State University’s Northeast Regional Center for Rural Development, Church Attendance from the Association of Statisticians of American Religious Bodies, Crime data from the Uniform Crime Reporting Statistics, and Education data from the U.S. Census Bureau. ln(Canada) is the log of the distance from the bank’s headquarters to the Canadian border, from https://www.freemaptools.com/measure-distance.htm. Voter Turnout data are collected from the United States Elections Project database. The Z-score is calculated as the sum of the equity capital ratio and return on assets divided by the standard deviation of return on assets, where the standard deviation is the quarterly deviation over the three prior years. Failed Banks in County is the number of failed banks in a county (from the Federal Reserve) divided by the total number of banks in that county. TARP Money is an indicator that is equal to one if the bank received TARP money and zero otherwise. All variables are defined in Appendix A. Variable Putnam Index Social Capital County Church Attendance Crime Education Loan Loss Provision/Total Loans Total Loan NCOs/Total Loans PPNR/Total Assets Tier 1 RBC Ratio Total Assets (millions of $s) ln(Total Assets) Nonperforming Loans/Total Loans Loan Loss Reserve/Total Loans C&I Loans/Total Loans Agricultural Loans/Total Loans Consumer Loans/Total Loans Foreign Gov. Loans/Total Loans Real Estate Loans/Total Loans Depository Inst. Loans/Total Loans Liquidity Ratio ln(Canada) Voter Turnout Z-score Failed Banks in County TARP Money ATM Fees/Core Deposits Check Income/Core Deposits Interest Expense on Core Deposits/Core Deposits Fee and Interest Income/Total Loans Net Interest Income/Total Loans Noninterest Income/Total Income

Mean -0.141 -0.173 60.30% 7.40% 19.53% 0.31% 0.30% 8.30% 12.80% 9,098 13.51 1.40% 1.50% 15.90% 2.90% 7.50% 0.01% 26.40% 0.10% 23.20% 5.77 17.92% 12.22 1.80% 3.70% 0.09% 0.01% 5.60% 4.36% 3.19% 2.45%

30

Median -0.216 -0.230 60.70% 7.50% 18.94% 0.10% 0.10% 4.20% 12.00% 573 13.26 0.70% 1.30% 14.10% 0.20% 4.70% 0.00% 6.53% 0.00% 21.50% 6.29 1.68% 4.13 0.00% 0.00% 0.00% 0.00% 2.44% 4.26% 3.00% 0.67%

Std. Dev. Minimum Maximum 0.66 -1.15 1.29 1.06 -1.71 1.74 16.13% 34.83% 85.80% 1.88% 4.58% 10.38% 3.48% 13.63% 26.97% 0.84% 0.01% 1.95% 0.81% 0.00% 1.21% 13.96% 1.16% 12.18% 29.00% 7.41% 22.53% 87,300 181 2,570,000 1.39 12.11 21.67 2.33% 0.01% 5.57% 0.97% 0.75% 2.89% 10.19% 3.31% 34.38% 6.14% 0.00% 15.95% 9.34% 0.29% 22.53% 0.10% 0.00% 0.05% 33.39% 0.26% 89.49% 1.55% 0.00% 0.21% 12.34% 5.95% 46.12% 1.52 4.19 6.98 24.73% 0.51% 56.26% 168.11 0.11 25.30 11.13% 0.00% 2.70% 18.95% 0.00 1.00 3.43% 0.00% 0.81% 0.02% 0.00% 0.08% 309.88% 0.57% 12.69% 2.17% 1.53% 9.04% 2.77% 0.01% 9.06% 64.45% 0.01% 11.29%

Table 3 Correlations of Social Capital Variables This table presents correlations between social capital variables for the sample banks from 20002015. Putnam index is created using principal component analysis on a set of fourteen different factors (Putnam, 1993) and is collected from the Bowling Alone database. Social Capital County data is a survey-based measure of social capital based on Rupasingha and Goetz (2008) and is collected from Penn State University’s Northeast Regional Center for Rural Development database. Church Attendance is the percent of the population in the state in which the bank is headquartered that attends church, collected from the Association of Statisticians of American Religious Bodies. Crime is the percent of the overall population in the state in which the bank is headquartered that is affected by any reported crime in a given year, collected from the Uniform Crime Reporting Statistics. Education is the percent of high school graduates in the state in which the bank is headquartered, collected from the U.S. Census Bureau.

Putnam Index Social Capital County Church Attendance Crime Education

Putnam Index 1 0.5379 -0.0135 -0.3953 0.2698

Social Capital Church County Attendance 1 0.0345 -0.3453 0.0190

31

1 0.0007 -0.0386

Crime

Education

1 -0.1642

1

Table 4 Social Capital and Bank Risk Measures This table presents OLS regression results in which we examine the relation between bank regulatory risk measures and social capital with bank-specific independent variables to control for operating differences between banks. Quarterly financial statement data for financial institutions are obtained from the Consolidated Financial Statements for Bank Holding Companies database (79,448 bank quarters). Putnam Index data are collected from the Bowling Alone database, Social Capital County data from Penn State University’s Northeast Regional Center for Rural Development, Church Attendance data from the Association of Statisticians of American Religious Bodies, Crime data from the Uniform Crime Reporting Statistics, and Education data from the U.S. Census Bureau. All variables are defined in Appendix A. Bank and year fixed effects are included within the estimations. p-values are shown in parenthesis below the coefficient estimates. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. (1) (2) (3) (4) Loan Loss Total Loan Tier 1 Provision/Total NCOs/Total PPNR/Total RBC Loans Loans Assets Ratio Putnam Index -0.001*** -0.001*** 0.002*** -0.004*** (0.00) (0.00) (0.00) (0.00) Social Capital County -0.001*** -0.001*** 0.001*** -0.001*** (0.00) (0.00) (0.00) (0.00) Church Attendance -0.001*** -0.001*** -0.002*** -0.002*** (0.00) (0.00) (0.00) (0.00) Crime 0.006*** 0.002*** 0.065*** -0.138*** (0.00) (0.00) (0.00) (0.00) Education -0.001*** -0.001*** 0.001*** 0.001** (0.00) (0.00) (0.01) (0.03) ln(Total Assets) 0.001*** 0.001*** -0.001*** -0.004*** (0.00) (0.00) (0.00) (0.00) Nonperforming Loans/Total Loans 0.023*** 0.033*** -0.030*** -0.312*** (0.00) (0.00) (0.00) (0.00) Loan Loss Reserve/Total Loans 0.020*** 0.006*** 0.194*** 0.415*** (0.00) (0.00) (0.00) (0.00) C&I Loans/Total Loans 0.002*** 0.001*** 0.007*** -0.038*** (0.00) (0.00) (0.00) (0.00) Agricultural Loans/Total Loans -0.000** -0.000** 0.005** -0.016*** (0.01) (0.04) (0.01) (0.00) Consumer Loans/Total Loans 0.003*** 0.003*** 0.026*** 0.028*** (0.00) (0.00) (0.00) (0.00) Foreign Gov. Loans/Total Loans -0.011* 0.001 0.343*** 0.255** (0.06) (0.81) (0.00) (0.02) Real Estate Loans/Total Loans 0.000*** 0.000*** 0.001*** 0.019*** (0.00) (0.00) (0.00) (0.00) Depository Inst. Loans/Total Loans -0.002*** -0.000* -0.036*** -0.071*** (0.00) (0.07) (0.00) (0.00) Liquidity Ratio -0.002*** -0.001*** -0.019*** 0.129*** (0.00) (0.00) (0.00) (0.00) Constant 0.000*** -0.001*** 0.040*** 0.159*** (0.00) (0.00) (0.00) (0.00) Observations 79,448 79,448 79,448 79,448 Adjusted R2 0.204 0.416 0.023 0.297

32

Table 5 Social Capital and Bank Risk Measures Controlling for Endogeneity This table presents 2SLS regression results in which we examine regulator risk measures as dependent variables with bank-specific independent variables to control for operating differences between banks. Columns (1) and (2) report the coefficients of the first stage regressions, which are used to obtain the fitted social capital variables. The dependent variables in the first stage regressions are Putnam Index (1) and Social Capital County (2). The instruments are ln(Canada) and Voter Turnout. These instruments satisfy the exclusion criterion based on the Hansen J-statistic. The p-values corresponding to the Sargan C statistic reject the null hypothesis (in all columns of Table 5) that the measure of social capital is exogenous. Bank and year fixed effects are included within the estimations. p-values are shown in parenthesis below the coefficient estimates. ***,**, and * denote significance at the 1%, 5%, and 10% levels, respectively. (1)

ln(Canada) Voter Turnout

Putnam Index -0.227*** (0.00) 0.001*** (0.00)

(2) First Stage Social Capital County -0.215*** (0.00) 0.009*** (0.00)

Putnam Index

(3)

(4)

Loan Loss Total Loan Provision/ NCOs/ Total Loans Total Loans

-0.001*** (0.00)

-0.001** (0.02)

(5) PPNR/ Total Assets

0.002*** (0.00)

(6) (7) Second Stage Loan Loss Tier 1 RBC Provision/ Ratio Total loans

Nonperforming Loans/Total Loans Loan Loss Reserve/Total Loans C&I Loans/Total Loans Agricultural Loans/Total Loans Consumer Loans/Total Loans Foreign Gov. Loans/Total Loans Real Estate Loans/Total Loans Depository Inst. Loans/Total Loans Liquidity Ratio Constant Observations R2 Hansen Sargan

0.004*** (0.00) -2.038*** (0.00) 5.549*** (0.00) 0.501*** (0.00) 3.792*** (0.00) -0.376*** (0.00) 0.318 (0.86) 0.041*** (0.00) 0.422*** (0.00) -0.074*** (0.00) 0.883*** (0.00) 79,448 0.425

0.039*** (0.00) -2.237*** (0.00) 2.141*** (0.00) 0.110*** (0.00) 5.030*** (0.00) -0.295*** (0.00) 49.188*** (0.00) 0.041*** (0.00) 1.473*** (0.00) 0.018 (0.49) 0.224*** (0.00) 79,448 0.238

0.001*** (0.00) 0.058*** (0.00) 0.473*** (0.00) 0.001*** (0.00) -0.001** (0.03) 0.008*** (0.00) -0.001 (1.00) -0.001 (0.24) -0.001 (0.88) -0.007*** (0.00) -0.005*** (0.00) 79,448 0.398 0.127 0.002

0.001*** (0.00) 0.062*** (0.00) 0.389*** (0.00) 0.001 (0.93) -0.002*** (0.00) 0.010*** (0.00) 0.017 (0.49) 0.001*** (0.00) 0.003** (0.03) -0.004*** (0.00) -0.007*** (0.00) 79,448 0.347 0.316 0.003

33

-0.001*** (0.00) -0.138*** (0.00) 0.889*** (0.00) 0.008*** (0.00) -0.004 (0.14) 0.076*** (0.00) 0.034 (0.81) 0.001 (0.77) 0.109*** (0.00) -0.020*** (0.00) 0.038*** (0.00) 79,448 0.280 0.202 0.000

(9)

(10)

Total Loan NCOs/ Total Loans

PPNR/ Total Assets

Tier 1 RBC Ratio

-0.001** (0.00) 0.001*** (0.00) 0.062*** (0.00) 0.389*** (0.00) -0.001 (0.87) -0.001*** (0.00) 0.010*** (0.00) 0.024 (0.33) 0.001*** (0.00) 0.003** (0.03) -0.004*** (0.00) -0.007*** (0.00) 79,448 0.347 0.372 0.003

0.001*** (0.00) -0.001*** (0.00) -0.136*** (0.00) 0.879*** (0.00) 0.007*** (0.00) -0.005* (0.07) 0.077*** (0.00) 0.100 (0.49) 0.001 (0.86) 0.110*** (0.00) -0.019*** (0.00) 0.038*** (0.00) 79,448 0.278 0.738 0.001

-0.231*** (0.00) -0.914*** (0.00) -0.021 (0.84) 0.788 (0.58) -0.383** (0.02) -0.219 (0.99) 0.063*** (0.00) 0.285 (0.19) 0.122 (0.18) -0.840*** (0.00) 0.408*** (0.00) -0.583*** (0.00) 79,448 0.202 0.577 0.007

-0.942** (0.01)

Social Capital County ln(Total Assets)

(8)

-0.669*** (0.00) -0.213 (0.82) 0.667 (0.63) -0.792** (0.02) -0.057 (0.94) 0.675*** (0.00) 0.450 (0.20) 0.086 (0.17) -0.125*** (0.00) 0.513*** (0.00) -0.054*** (0.00) 79,448 0.202 0.698 0.005

-0.001*** (0.00) 0.001*** (0.00) 0.058*** (0.00) 0.472*** (0.00) 0.001*** (0.01) -0.001 (0.28) 0.008*** (0.00) 0.019 (0.45) -0.001 (0.17) 0.001 (0.92) -0.007*** (0.00) -0.006*** (0.00) 79,448 0.398 0.3 0.002

Table 6 Social Capital and Financial Distress This table presents 2SLS regression results in which we examine social capital and bank financial distress using bank specific Z-scores as the dependent variable with bank-specific independent variables to control for operating differences between banks. The Z-score is calculated as the sum of the equity capital ratio and return on assets divided by the standard deviation of return on assets, where the standard deviation is the quarterly deviation over the three prior years. Columns (1) and (3) report coefficients of the first stage regressions, which are used to obtain the fitted social capital variables. The dependent variable in the first stage regression in column (1) is the Putnam Index and in column (3) is Social Capital County. The instruments are ln(Canada) and Voter Turnout. All variables are defined in Appendix A. Bank and year fixed effects are included within the estimations. p-values are shown in parenthesis below the coefficient estimates. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

ln(Canada) Voter Turnout

(1) Putnam Index -0.227*** (0.00) 0.001*** (0.00)

Putnam Index

(2) Z-score

(3) Social Capital County -0.215*** (0.00) 0.009*** (0.00)

Bank Control Variables Bank Fixed Effects Year Fixed Effects Observations R2 Hansen Sargan

Z-Score

2.556*** (0.00)

Social Capital County Constant

(4)

0.883*** (0.00)

5.857*** (0.00)

0.224*** (0.00)

3.275** (0.02) 4.017*** (0.00)

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

79,448 0.425

79,448 0.202 0.411 0.001

79,448 0.238

79,448 0.206 0.293 0.001

34

Table 7 Social Capital and County Level Bank Failures This table presents 2SLS regression results in which we examine the percent of banks that fail in a county (from the Federal Reserve) with median quarterly values of bank-specific independent variables in each county to control for operating differences between banks in different counties. Failed Banks in County is the number of failed banks in a county divided by the total number of banks in that county. Columns (1) and (3) report coefficients of first stage regressions, which are used to obtain the fitted social capital variables. The dependent variables in the first stage regression are Putnam Index (1) and Social Capital County (3). The instruments are ln(Canada) and Voter Turnout. All variables are defined in Appendix A. Bank and year fixed effects are included within the estimations. p-values are shown in parenthesis below the coefficient estimates. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

ln(Canada)

Voter Turnout Putnam Index

(1) Putnam Index -0.240*** (0.00) 0.001*** (0.00)

(2) Failed Banks in County

(3) Social Capital County -0.214*** (0.00) 0.007*** (0.00)

-0.002* (0.07)

Social Capital County Constant Bank Control Variables Bank Fixed Effects Year Fixed Effects Observations R2 Hansen Sargan

(4) Failed Banks in County

0.807*** (0.00)

0.043*** (0.00)

0.339*** (0.01)

-0.003** (0.02) 0.041*** (0.00)

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

12,398 0.469

12,398 0.203 0.818 0.001

12,398 0.249

12,398 0.203 0.881 0.001

35

Table 8 Social Capital and Bank Risk Ratios During versus Outside the Financial Crisis This table presents 2SLS regression results in which we examine bank regulatory risk measures both during (2008-2009) and outside (2000-2007 and 2010-2015) the financial crisis as dependent variables with bank-specific independent variables to control for operating differences between banks. The dependent variables in the first stage regressions are Putnam Index (1) and Social Capital County (2). The instruments are ln(Canada) and Voter Turnout. We use predicted values in the second stage and report Sargan C and Hansen-J 2SLS test statistics. These instruments satisfy the exclusion criterion based on the Hansen J-statistic. pvalues corresponding to the Sargan C statistic reject the null hypothesis (in all columns) that the measure of social capital is exogenous. All variables are defined in Appendix A. Bank and year fixed effects are included within the estimations. p-values are shown in parenthesis below the coefficient estimates. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. (1)

Panel A: Putnam Index Putnam Index Constant

Bank Control Variables Bank Fixed Effects Year Fixed Effects Observations R2 Hansen Sargan Panel B: Social Capital County Social Capital County Constant

Bank Control Variables Bank Fixed Effects Year Fixed Effects Observations R2 Hansen Sargan

(2)

(3)

(4)

(5)

(6)

Non-Crisis Crisis Loan Loss Provision /Total Loans

Second Stage Non-Crisis Crisis Non-Crisis Crisis Total Loan NCOs/ Total Loans PPNR/Total Assets

-0.001* (0.06) -0.003*** (0.00)

-0.001*** (0.00) -0.011*** (0.00)

-0.001*** (0.00) -0.005*** (0.00)

-0.001* (0.06) -0.007** (0.01)

0.001***

0.002***

(0.00) 0.042*** (0.00)

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

71,598 0.377 0.177 0.001

7,850 0.549 0.131 0.006

71,598 0.316 0.316 0.004

-0.001** (0.02) -0.003*** (0.00)

-0.001*** (0.00) -0.011*** (0.00)

Yes Yes Yes 71,598 0.363 0.155 0.001

(7)

(8)

Non-Crisis

Crisis

Tier 1 RBC Ratio

(0.00) 0.166*** (0.00)

-0.001*** (0.00) 0.093* (0.10)

-0.006*** (0.00) 0.216*** (0.00)

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

7,850 0.459 0.352 0.001

71,598 0.084 0.363 0.001

7,850 0.134 0.357 0.001

71,598 0.101 0.215 0.001

7,850 0.157 0.776 0.001

-0.001*** (0.00) -0.005*** (0.00)

-0.002*** (0.01) -0.007** (0.01)

0.001* (0.09) 0.040*** (0.00)

0.002*** (0.00) 0.153*** (0.00)

-0.002** (0.01) 0.060* (0.06)

-0.005*** (0.00) 0.207*** (0.00)

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

7,850 0.543 0.161 0.001

71,598 0.316 0.616 0.001

7,850 0.453 0.656 0.001

71,598 0.469 0.666 0.001

7,850 0.221 0.655 0.001

71,598 0.613 0.615 0.001

7,850 0.211 0.556 0.001

36

Table 9 Social Capital and TARP Money during the Financial Crisis This table presents IVPROBIT regression results in which we examine whether a bank received TARP money as the dependent variable with bank-specific independent variables to control for operating differences between banks. The dependent variable, TARP Money, is an indicator equal to one if the bank received TARP money and zero otherwise. We look only at the 2008-2009 period. Columns (1) and (3) report coefficients of first stage regressions, which are used to obtain the fitted social capital variables. The dependent variable in the first stage regression in column (1) is the Putnam Index and in column (3) is Social Capital County. The instruments are ln(Canada) and Voter Turnout. All variables are defined in Appendix A. Bank and year fixed effects are included within the estimations. p-values are shown in parenthesis below the coefficient estimates. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. (1)

ln(Canada) Voter Turnout

Putnam -0.211*** (0.00) 0.001** (0.02)

Putnam Index

(2) TARP Money

(3) Social Capital County -0.183*** (0.00) 0.011*** (0.00)

0.031*** (0.00)

Social Capital County Constant

Bank Control Variables Bank Fixed Effects Year Fixed Effects Observations R2 Hansen Sargan

(4) TARP Money

1.359*** (0.00)

-0.843*** (0.00)

1.336*** (0.00)

0.358*** (0.00) -0.840*** (0.00)

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

7,850 0.622

7,850 0.240 0.436 0.003

7,850 0.343

7,850 0.260 0.996 0.007

37

Table 10 Social Capital and Return on Assets – Crisis and Non-Crisis This table presents the IVPROBIT results in which we examine an indicator variable for positive or negative ROA (indicator equal to one for positive ROA and zero otherwise) with bank-specific independent variables to control for operating differences between banks. Columns (1) and (2) report coefficients of first stage regressions, which are used to obtain the fitted social capital variables. The dependent variable in the first stage regression is the Putnam Index (1) and Social County Capital (2). The instruments are ln(Canada) and Voter Turnout. These instruments satisfy the exclusion criterion based on the Hansen J-statistic. p-values corresponding to the Sargan C statistic reject the null hypothesis (in all columns) that the measure of social capital is exogenous. All variables are defined in Appendix A. Bank and year fixed effects are included within the estimations. p-values are shown in parenthesis below the coefficient estimates. ***,**, and * denote significance at the 1%, 5%, and 10% levels, respectively. (1)

(2) First Stage Social Capital Putnam County

(3)

(4)

All Years Non-Crisis ln(Canada) Voter Turnout

-0.227*** (0.00) 0.001*** (0.00)

(5) (6) Second Stage

ROA Crisis

0.183** (0.04)

0.038*** (0.01)

Observations R2 Hansen Sargan

Crisis

0.216** (0.04)

Social Capital County

Bank Control Variables Bank Fixed Effects Year Fixed Effects

All Years Non-Crisis

(8)

-0.215*** (0.00) 0.009*** (0.00)

Putnam

Constant

(7)

0.883*** (0.00)

0.224*** (0.00)

1.901*** (0.00)

0.546 (0.34)

2.065*** (0.00)

0.210*** (0.01) 1.792*** (0.00)

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

79,448 0.425

79,448 0.238

79,448 0.201 0.664 0.048

71,598 0.201 0.651 0.036

7,850 0.222 0.446 0.005

79,448 0.201 0.713 0.000

71,598 0.201 0.736 0.013

7,850 0.222 0.447 0.013

38

0.013*** (0.01) 0.552 (0.33)

0.246*** (0.01) 1.926*** (0.00)

Table 11 Social Capital and Bank Deposit Rate and Fees and Loan Income This table presents OLS regression results in which we examine bank deposit rate and fee structures, and loan income as dependent variables with bank-specific independent variables to control for operating differences between banks. Quarterly financial statement data for financial institutions are obtained from the Consolidated Report of Condition and Income database (79,448 bank quarters). Putnam Index data are collected from the Bowling Alone database, Social Capital County data from Penn State University’s Northeast Regional Center for Rural Development, Church Attendance data from the Association of Statisticians of American Religious Bodies, Crime data from the Uniform Crime Reporting Statistics, and Education data from the U.S. Census Bureau. All variables are defined in Appendix A. Bank and year fixed effects are included within the estimations. p-values are shown in parenthesis below the coefficient estimates. ***,**, and * denote significance at the 1%, 5%, and 10% levels, respectively. (1) ATM Fees/Core Deposits Putnam Index Social Capital County Church Attendance Crime

(2)

(3) (4) Interest Expense Fee and Check on Core Interest Income/Core Deposits/Core Income/Total Deposits Deposits Loans

(5)

(6)

Net Interest Income/Total Loans

Noninterest Income/Total Income

-0.001***

-0.001***

0.001***

-0.001***

-0.163***

0.001**

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.01)

-0.001**

-0.001***

0.001***

-0.001***

-0.001***

0.018***

(0.03)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

-0.001

-0.001***

-0.002***

0.001

-0.001

-0.001***

(0.37)

(0.00)

(0.00)

(0.15)

(0.40)

(0.00)

-0.006

0.001***

0.030***

0.024***

0.045***

0.020***

(0.51)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

-0.001**

-0.001***

0.001***

-0.001

-0.001***

0.001***

(0.05)

(0.00)

(0.00)

(0.32)

(0.00)

(0.00)

-0.003*

-0.001*

0.029***

0.052***

0.028***

-0.006***

(0.07)

(0.07)

(0.00)

(0.00)

(0.00)

(0.00)

Bank Control Variables

Yes

Yes

Yes

Yes

Yes

Yes

Bank Fixed Effects

Yes

Yes

Yes

Yes

Yes

Yes

Year Fixed Effects

Yes

Yes

Yes

Yes

Yes

Yes

63,537

63,398

79,448

48,747

79,448

79,448

0.202

0.228

0.130

0.253

0.258

0.273

Education Constant

Observations Adjusted R

2

39

Table 12 Social Capital and Bank Fees Controlling for Endogeneity This table presents 2SLS regression results in which we examine bank fee structures as dependent variables with bank-specific independent variables to control for operating differences between banks. Columns (1) and (2) report the coefficients of first stage regressions, which are used to obtain the fitted social capital variables. The dependent variables in the first stage regression are the Putnam Index and Social Capital County. The instruments are ln(Canada) and Voter Turnout. These instruments satisfy the exclusion criterion based on the Hansen J-statistic. The p-values corresponding to the Sargan C statistic reject the null hypothesis (in all columns of Table 10) that the measure of social capital is exogenous. Bank and year fixed effects are included within the estimations. p-values are shown in parenthesis below the coefficient estimates. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. (1)

(2)

(3)

(4)

(5)

(6)

Social Capital County -0.215*** (0.00) 0.009*** (0.00)

ATM Fees/Core Deposits

Interest Expense Fee and Check on Core Interest Income/Core Deposits/Core Income/ Deposits Deposits Total Loans

-0.001*** (0.00)

-0.001*** (0.00)

(7)

First Stage

ln(Canada) Voter Turnout

Putnam -0.227*** (0.00) 0.001*** (0.00)

Putnam

0.018** (0.05)

-0.001** (0.04)

(8) (9) Second Stage

Net Interest Noninterest ATM Income/Total Income/Total Fees/Core Loans Income Deposits

-0.001*** (0.00)

Bank Control Variables

Bank Fixed Effects Year Fixed Effects Observations R2 Hansen Sargan

0.883*** (0.00) Yes

0.224*** (0.00) Yes

(11)

(12)

Interest Expense Fee and Check on Core Interest Income/Core Deposits/Core Income/Total Deposits Deposits Loans

(13)

(14)

Net Interest Income/ Noninterest Total Income/Total Loans Income

0.012* (0.06)

Social Capital County Constant

(10)

-0.002 (0.13)

-0.001*** (0.00)

-0.679*** (0.00)

0.054*** (0.00)

0.027*** (0.00)

-0.141*** (0.00)

-0.001** (0.03) -0.002* (0.09)

-0.001*** (0.00) -0.001*** (0.00)

0.011*** (0.01) -0.662*** (0.00)

-0.001** (0.01) 0.054*** (0.00)

-0.001** (0.02) 0.027*** (0.00)

0.018*** (0.00) -0.152*** (0.00)

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

79,448 0.425

79,448 0.238

63537 0.202 0.757 0.001

63,398 0.225 0.282 0.001

79,448 0.201 0.551 0.001

48,747 0.252 0.595 0.001

79,448 0.243 0.342 0.001

79,448 0.220 0.703 0.003

63,537 0.201 0.894 0.001

63,398 0.215 0.754 0.001

79,448 0.201 0.551 0.001

48,747 0.252 0.338 0.001

79,448 0.244 0.980 0.001

79,448 0.219 0.176 0.001

40

Appendix A This appendix provides definitions and sources of variables used in the analysis. Variable

Putnam Index

Social Capital County

Church Attendance

Crime

Education Total Assets ln(Total Assets) Loan Loss Provision/Total Loans PPNR/Total Assets Tier 1 RBC Ratio Total Loan NCOs/Total Loans

Nonperforming Loans/Total Loans Loan Loss Reserve/Total Loans C&I Loans/Total Loans Agricultural Loans/Total Loans Consumer Loan/Total Loans Foreign Gov. Loans/Total Loans Real Estate Loans/Total Loans Depository Inst. Loans/Total Loans Liquidity Ratio ln(Canada)

Voter Turnout

Definition A county level social capital index based on 14 different social capital indicators and available from Bowling Alone database. A survey-based measure of social capital based on Rupasingha and Goetz (2008). It is constructed using principal component analysis based on social capital indicators at the county level and available from Penn State University’s Northeast Regional Center for Rural Development. Percent of the population in the state in which the bank is headquartered that attends church, collected from the Association of Statisticians of American Religious Bodies. Percent of the overall population the state in which the bank is headquartered that is affected by any reported crime in a given year, collected from the Uniform Crime Reporting Statistics. Percent of high school graduates in the state in which the bank is headquartered, collected from the U.S. Census Bureau. Quarterly total assets of the bank Natural log of total assets Ratio of loan loss provision to total loans Ratio of (net interest income + noninterest income – noninterest expense) to total assets Ratio of Tier 1 capital to risk-weighted assets Ratio of total loan net charge-offs/Total loans Ratio of loans past due 90 days or more and still accruing interest and loans in nonaccrual status/Total loans Ratio of reserve for loan losses/Total loans Ratio of commercial loans to total loans Ratio of agricultural loans to total loans Ratio of consumer loans to total loans Ratio of foreign government loans to total loans Ratio of real estate loans to total loans Ratio of loans to depository institutions to total loans Ratio of cash and investment securities to total assets The log of the distance from the bank's headquarters to the Canadian border Percent of voting eligible population in the state in which the bank is headquartered that voted for the highest office in a given election year. The numerator is the number of people who voted for

41

Data Source

www.bowlingalone.com

http://aese.psu.edu/nercrd

http://www.thearda.com

Uniform Crime Reporting Statistics http://www.ucrdatatool.gov/

U.S. Census Bureau Call report Call report Call report Call report Call report Call report

Call report Call report Call report Call report Call report Call report Call report Call report Call report https://www.freemaptools.com/measuredistance.htm

www.electproject.org/home

Z-score Failed Banks in County TARP Money ROA ATM Fees/Core Deposits Check Income/Core Deposits Interest Expense on Core Deposits/Core Deposits Net Interest Income/Total Loans Noninterest Income/Total Income Fee and Interest Income/Total Loans

the "highest office" in a given election. The denominator is the voting eligible population, defined as the number of people eligible to vote. Sum of the equity capital ratio and return on assets divided by the standard deviation of the return on assets, where the standard deviation is the quarterly deviation over the three prior years. Number of failed banks in a county divided by the total number of banks in that county. Indicator equal to one if the bank took TARP money and zero otherwise. Net income to total assets Ratio of ATM fees to core deposits Ratio of total income from checks to core deposits Ratio of interest income expense on deposits to core deposits Ratio of net interest income to total loans Ratio of non-interest income to total income Ratio of fee and interest income to total loans

42

Call report Federal reserve Federal Reserve Call report Call report Call report Call report Call report Call report Call report