For Review Only. The tax exemption to Subchapter S banks: Who gets the benefit? The Financial Review. Journal: The Financial Review

Transcription

1 The Financial Review The tax exemption to Subchapter S banks: Who gets the benefit? Journal: The Financial Review Manuscript ID FIRE--0-0.R Manuscript Type: Paper Submitted for Review Keywords: Small Business Job Protection Act of, Subchapter S Banks, Tax Exemption, Benefits from Tax Exemption

2 Page of The Financial Review The Tax Exemption to Subchapter S Banks: Who Gets the Benefit? Abstract The Small Business Job Protection Act of allows US banks to adopt the Subchapter S status. We investigate if the Subchapter S banks use tax benefits for the intended purposes of protecting jobs, creating opportunities and increasing take-home-pay of workers. We find that the tax benefits: a) are not used in expenses related to protection of jobs, b) do not lead to greater employment opportunities within the Subchapter S banks, and c) do not benefit the employees in the form of increased salaries and benefits. Our results indicate that bank owners are sole beneficiaries of the tax exemption benefits. Keywords: Small Business Job Protection Act of, Subchapter S Banks, Tax Exemption, Benefits from Tax Exemption. Introduction The Public Law 0- enacted by the 0 th U.S. Congress on th August, which was termed as the Small Business Job Protection Act of, has the following intents: To provide tax relief for small businesses, to protect jobs, to create opportunities, to increase the take home pay of workers, to amend the Portal-to-Portal Act of relating to the payment of wages to employees who use employer owned vehicles, and to amend the Fair Labor Standards Act of to increase the minimum wage rate and to prevent job loss by providing flexibility to employers in complying with minimum wage and overtime requirements under that Act. We thank the participants at the Southern Finance Association Conference in, Midwest Finance Conference in and the st Annual Conference of the Multinational Finance Society in, for their helpful comments and suggestions. We also thank two anonymous referees and the editor, Dr. Srinivasan Krishnamurthy, for their detailed reviews and constructive comments. All the remaining errors are ours.

3 The Financial Review Page of The Small Business Job Protection Act of allows banks to convert to Subchapter S corporations and avoid double taxation. The Subchapter S bank is treated as a partnership for tax purposes. Its shareholders pay federal income taxes on pass-through earnings and thus, avoid taxes at the corporate level. Allowing banks to adopt Subchapter S status is a political decision. The law does not state the reason for its implementation, only the intents for the law are provided. The intents are to protect jobs, to create opportunities and to increase the take home pay of workers. According to the standard economic theory, imposing tax on a product will move the product s supply curve to the left, raising the market equilibrium price and lowering the market equilibrium quantity, thus creating a deadweight loss in social welfare. Along the same line of reasoning, a tax relief would result in a right-shift of the supply curve, causing a reduction in market equilibrium price and an increase in market equilibrium quantity, thus producing a rise of welfare to the society as a whole. Since the bank s conversion to Subchapter S status creates a tax relief to the bank, one would expect a non-trivial welfare increase to the economy. Economic theory does not predict the forms in which the welfare increase would take. The language in the Small Business Job Creation Act of seems to suggest that the welfare increase should be in protecting jobs, creating opportunities, and increasing the take home pay of the workers. The tax exemption conferred to Subchapter S banks at the corporate level clearly offers Subchapter S banks a competitive advantage over C-Corporation banks. We examine the benefits from tax exemption to see if they are utilized for the stated intents of the Small Business Job Protection Act of, i.e., to protect jobs, to create opportunities, and to increase the take home pay of workers. Subchapter S banks can

4 Page of The Financial Review afford to offer higher deposit rates and lower loan rates to their customers and a lower spread would make Subchapter S banks more competitive. As a result of tax exemptions, Subchapter S banks have more cash on hand, which can be used to increase lending, provide more loan products, and offer better services to customers. The more the bank increases its competitiveness and improves the services it provides to its customers, the more it increases its probability of survival and decreases the probability of employee layoffs. Bank owners can also reinvest the available cash from tax reduction back into the bank which may lead to a higher growth in the banks that adopt Subchapter S status. A growing bank would need more employees to meet the demand and would not lay off employees. Hence, passing some of the tax benefits to bank customers or reinvesting tax dollars back into the bank would lead to the protection of the jobs for the bank employees. Another intention of the Act is to create opportunities which can be realized when the tax relief is utilized to hire and expand the competitive workforce. The findings of our paper contribute to the ongoing debate of whether tax exemptions to businesses lead to job creation. Yet, another intention of the Act is to increase the take home pay of the workers." We investigate if the tax relief results in an increase in salaries and benefits for Subchapter S bank employees. The following are our research investigations and findings of tax exemptions granted to Subchapter S banks. First, we investigate whether the tax subsidy is passed on to the borrowers (depositors) in the form of lower (higher) loan (deposit) rates. In our study, this research question, is similar to what has been done by Depken, Hollans, and Swidler (0). Similar to their results, we find that bank customers, both depositors as well as borrowers, do not receive benefits as a result of the adoption of Subchapter S

5 The Financial Review Page of status, in terms of more favorable rates. Second, we investigate if the conversion to a Subchapter S bank leads the bank to offer better services to its customers. The improvement in services to customers is not explicitly observable, but it may be measured by using the increase in operating expenses (excluding the wage/salary expenses) as a proxy. An increase in operating expenses (excluding the wage/salary expenses) may signal an improvement in service to the customer. We find that the tax subsidy does not result in any increase in operating expenses. Third, we investigate if adoption of Subchapter S status leads to an increase in a bank s lending or an increase in the number of products offered by the bank. Our results show that adoption of Subchapter S status does not result in any increase in lending or in the number of products offered by the bank. Fourth, we study if the conversion to Subchapter S results in higher growth in assets. We find that the tax subsidy does not significantly impact the growth in the bank s assets. Together these four findings indicate that the tax benefits are not used by banks in expenses related to the protection of jobs. Fifth, we investigate if the conversion to Subchapter S status leads to an increase in the number of employees at the bank. Our results show that the change in the number of employees for the Subchapter S banks is Apart from reduced loan rates and higher deposit rates banks may provide non-pricing terms of loans, a number of additional services, and relationship banking. Depken, Hollans, and Swidler (0) argue that banks may practice some type of bundling strategy (include offering a free credit card to savings customers or granting credit card customers additional services such as fraud protection and online account management) to maximize profits, then the tax status of the bank may not have any effect on deposit and loan rates. Koderisch, Wubker, Baumgarten and Baillie (0) state that, for customers, bundling of products and services may lead to increased satisfaction, clarity of banking transactions, and ultimately a reduction in the full cost of purchasing financial services. By comparing the average loan rates and average deposit rates for the banks that adopt Subchapter S status with the non-subchapter S banks we cannot conclusively establish that customers do not receive any benefits from bank s adoption of Subchapter S status. We thank an anonymous referee for pointing out this limitation of our analysis. As banks don t publicly disclose information about bundling of services they provide, and the SNL database doesn t have any information on bundling of services, we are unable to make direct comparison of bundling services between banks.

6 Page of The Financial Review not statistically different from the change in the number of employees for the non- Subchapter S banks. Sixth, we examine if the tax subsidy is used to benefit the Subchapter S bank employees in terms of an increase in salaries and benefits. We find that a bank s conversion to Subchapter S status does not cause any changes to its employees salaries and benefits. Finally, we investigate the effect of the tax exemption on the owners of the bank. The owners benefit if the conversion to Subchapter S results in a higher return on equity. We find that the return on equity increases significantly after the bank adopts the Subchapter S status, indicating that the tax benefit is passed on to the owners of the Subchapter S banks in the form of higher return on equity. From our findings, we infer that banks do not use the tax savings from Subchapter S status conversion in expenses related to the protection of jobs, to increase employment opportunities, and to increase employee s wages. Overall, we find that the tax exemption to Subchapter S banks benefits only the owners of the bank.. Literature Review and Hypothesis Formation In the initial legislation of, the maximum limit on the number of shareholders was set at for banks wanting to incorporate as Subchapter S corporations. Later on, the American Jobs Creation Act of 0 raised this ceiling to one hundred. The 0 legislation allows multi-generational families to be counted as one shareholder, making it easier for family-owned banks to attain Subchapter S status. This has resulted in a steady increase in the number of Subchapter S banks from 0 banks in to

7 The Financial Review Page of banks in (see Figure ). Subchapter S banks are generally much larger and more profitable than an average S corporation. At the end of the average asset size of Subchapter S banks was $ million with average profits of $. million. The median asset size and median profits for these banks was respectively $ million and $. million. In, there were Subchapter S banks with profits greater than $0 million, Subchapter S banks with profit greater than $ million, and, Subchapter S banks with profit more than $ million. The highest profits earned by a Subchapter S bank in was $ million. Prior research in this area has mainly focused on why and which banks adopt Subchapter S status. In other words, the focus has been mainly on identifying the factors that influence a bank s decision to adopt Subchapter S status. Harvey and Padget (00) find that banks that adopt Subchapter S status tend to be well capitalized, but slowergrowing than other banks prior to conversion, and have a history of higher levels of dividends. After conversion, there is an increase in dividend payout rates leading to reduced levels of capital. Hodder, McAnally, and Weaver (0) find that banks elect Subchapter S status to avoid double taxation of dividends as well as to avoid alternative minimum taxes. They show that banks are less likely to convert when conversion restricts The favorable tax treatment should interest all banks in adopting the Subchapter S status, but there are certain conditions that have to be satisfied in order to qualify as a Subchapter S institution. The restrictions include limits to one class of stock as well as on the type and the number of shareholders. The Subchapter S bank can have no more than 00 shareholders and it is confined to individual shareholders, estates, and exempt organizations described in section (a), 0(c)() and certain trusts described in section (c)()(a). We would like to thank an anonymous referee for suggestion that the adoption of subchapter S status by bank may be due to a change in ownership structure of the bank. However, we are not able to exploit this issue further since the SNL database does not carry the bank ownership data. According to IRS statistics of income for tax year 0 there were approximately. million S corporations which generated $. trillion in annual revenues. Thus the average revenue of S corporations in the year 0 was about $. million.

8 Page of The Financial Review access to equity capital and nullifies corporate tax-loss carry-forwards. Cyree, Hein, and Koch (0) study the other reasons beyond tax avoidance and examine the factors that affect the choice of Subchapter S status. They find that banks adopting Subchapter S status have higher dividend payout rates, higher profit growth, lower capital, and rely more on core deposits. Researchers have examined the effect of adoption of Subchapter S status on the bank s customers. Depken, Hollans, and Swidler (0) examine whether Subchapter S banks share the tax benefit with customers in the form of lower loan rates or higher deposit rates as compared to C-Corporation banks. They do not find any evidence of reduced loan rates or increased deposit rates by the banks that adopt the Subchapter S status. To the best of our knowledge, none of the papers so far, test if the intents of the law were met. We focus on a broad range of expenses by bank which may reduce the probability of employee layoffs. To test whether there is increase in expenses related to protection of jobs, we investigate if the adoption of Subchapter S status will lead to an increase in deposit rates, a decrease in lending rates, an increase in operating expenses, an Depken, Hollans, and Swidler (0) compare the rates for individual products for the Subchapter S banks and C-Corporation banks by using the second quarter data for the year 0. They find that Subchapter S deposit (loan) rates are equal to, or lower (higher) than, similar C-Corporation bank rates. In this paper we use average loan and deposit rates to investigate if banks share their tax benefits from the adoption of Subchapter S status with their customers. However, due to different bank strategies and customer targeted pricing, average values may not be that representative. Lending and customer portfolios of banks may differ significantly due to their self-designed strategies where some banks may focus on average-risk customers while others may focus both on high and low risk customers. Hence, Depken, Hollans, and Swidler s method of comparing the highest deposit rates and the lowest loan rates offered by banks may be more informative to assess if tax benefits are passed on to the customers of the bank. We thank an anonymous referee for pointing out that use of the highest and the lowest rate offerings (in terms of deposits and loans, respectively) may be more informative than the average deposit and loan rates.

9 The Financial Review Page of increase in lending, an increase in the number of products offered by the bank, or higher growth in assets of the bank. Formally we state our first hypothesis as: H: There is an increase in expenses related to the protection of jobs due to the adoption of Subchapter S status by the bank. The other intent of the law is to create opportunities. We investigate if the benefits from tax subsidy will lead to improvements in job opportunities i.e., we investigate if the adoption of Subchapter S status will lead to an increase in the number of employees at the bank. Formally, our second hypothesis is: H: There is an increase in expenses related to the creation of employment opportunities due to the adoption of Subchapter S status by the bank. The third intent of the law is to increase the take home pay of the employees for which we state the following hypothesis: H: There is an increase in the take home pay of workers due to the adoption of Subchapter S status by the bank. Though the law does not specifically mention benefits to the owners of the bank, it is likely that the tax savings from the Subchapter S conversion may be passed on to the owners. We state the fourth hypothesis: H: There is an increase in payouts to the owners of the bank due to the adoption of Subchapter S status by the bank. We investigate if the conversion to Subchapter S results in higher return on equity and return on assets of the bank. If the intents of the law are met along with benefits to the owners, then it is a Pareto welfare improvement to all the stakeholders of the bank. However, if only the owners capture the entire welfare increase, without benefiting other

10 Page of The Financial Review stakeholders (customers and employees), then the Small Business Job Protection Act of does not achieve its intended goals. Our paper is an attempt to help the policy makers determine if the law actually fulfills its intended purposes.. Data We collect the data on banks from the SNL Financial database ( for the period of to. SNL database provides detailed financial and other information on the U.S. financial institutions, including all publicly-traded banks and thrifts, and privately held institutions. We collect the yearly balance sheet and profit and loss items of banks along with the identification code on whether the bank was a Subchapter S bank for the year. To ensure homogeneity in characteristics of the treatment group and the control group, our entire sample consists of banks that adopted the Subchapter S status by the end. In Table we provide the number of Subchapter S banks at the end of each year, the number of C banks adopting Subchapter S status in the year, the number of banks starting operations as Subchapter S, and the number of Subchapter S banks closing their operations in the year. Strategy implementation in banks may take a prolonged period of time to see resulting changes in accounting figures. Hence, in difference-in-difference computation we use a three-year window to analyze the accounting performance results after the conversion. The banks in the treatment group are the banks that are non-subchapter S in the year (t-), convert to Subchapter in year (t) and maintain their Subchapter S status in year (t+). On the other hand, the banks in the SNL Financial database has been used by several studies including Acharya and Steffen (), Blankespoor, Linsmeier, Petroni and Shakespeare (), Wilson and Wu (), Nichols, Wahlen and Wieland (0), etc.

11 The Financial Review Page 0 of control group are the banks that are non-subchapter S in the year (t-), do not adopt Subchapter S status in year (t) and continue to be non-subchapter S banks in year (t+). However, the banks in the control group will convert to Subchapter S sometime after year (t+). To ensure that the banks in the two groups are located in the same state, and are of similar size, we match the banks in the control group with banks in the treatment group based on state code and asset size. We eliminate all the banks in the treatment group for which we do not get a match, based on size and state code. Our final sample for the treatment group and control group comprises of, banks in each. In our robustness tests we form matched groups based on size, state code, and county code. This more stringent matching results in the final sample size of banks in each group. Whereas the previous studies have used propensity score to construct the sample of control group, our control group is composed of the banks that adopt Subchapter S status later in their life, i.e., our control observations are out of the firms that have not yet elected Subchapter S Status, but will do so later on. In the propensity score matching procedure, the control group consists of the banks that have a high probability of converting to S bank in the future. While in our control group, the banks will convert to S banks in the future, making this almost a certainty, making the not yet Subchapter S sample a better control group than that of the always C-Corporation group. To confirm State banking laws play an important role in the choice of incorporation; hence, in our sample, we match the Subchapter S banks with non-subchapter S banks within the same state and control for the local economic conditions. Depken, Hollans, and Swidler (0) point out that ten states contain more than % of all Subchapter S banks, while five states and four territories have no Subchapter S banks. They further state that differences in state banking laws may explain this disproportionate incidence. For example, Minnesota, Texas, and Illinois had a long history of prohibiting branch banking. This has created statewide banking systems with smaller and more independent banks, thereby increasing the likelihood of meeting the Subchapter S criteria of a limited number of shareholders. 0

12 Page of The Financial Review the validity of our matching procedure we compare the characteristics for the Subchapter S and non-subchapter S banks. Table shows the descriptive statistics for several characteristics and the comparison of the mean values for the sample of Subchapter S and non-subchapter S banks in the year before conversion (t-). For the state and size matched sample in Panel A, and the state, county, and size matched sample in Panel B, we find no significant and consistent difference between the two groups in each of the characteristics except for ROA and ROE. The higher ROA and ROE of the Subchapter S banks indicate the higher profitability of these banks in the year prior to conversion, and a greater potential for tax savings. It is possible that the creation of Subchapter S captivates a lot of banks that will convert, though the need does not exist, i.e., banks may be exploiting the loophole right upon the creation of the law. This may result in a difference in the impact of adoption of subchapter S status on the banks which adopt the subchapter S status earlier than those that adopt the subchapter S status in later years. It is also possible that effects of adoption of subchapter S status are stronger in a later time period that is not captured in the full sample of to 0, since the sample is front-loaded. Hence, we exclude the banks that adopted subchapter S status in the first three years of the implementation of the law (i.e.,, and ) and study the banks that adopt subchapter S status in the later years, i.e., from 00 to 0. In Panel C of Table, we report the descriptive statistics We thank an anonymous referee for making the suggestion of studying the Subchapter S conversions in the later years separately from the previous years on the pretext that it is possible that the creation of Subchapter S captivates a lot of banks that will convert though the need does not exist, i.e., banks may be exploiting the loophole right upon the creation of the law. We have also split the time period from to 0 in two equal sub periods and have done tests for the 0 to 0 sample separately. The results for the sample over 0 to 0 are exactly the same as those

13 The Financial Review Page of and compare the mean values for a subsample of the state, county, and size matched sample in the years from 00 to 0. We find there is no significant difference between the two groups in each of the characteristics for this subsample of firms. As we show in Panel A and Panel B of Table, the test of comparison of characteristics for the full sample (from to 0) confirms that our control group and treatment group have similar characteristics; however, the differences in ROA and ROE may raise the concern that the matching is not perfect. For the subsample of 00 to 0, even ROA and ROE are statistically similar for the treatment and control group in the year prior to conversion (t-), thus alleviating the concerns of imperfect matching for the full sample from to 0. Though we are able to match the characteristics of the treatment and control group banks, due to the limitation of the data we are unable to check if the bank that is being matched satisfies the eligibility requirements of the number of shareholders. 0 Hodder, McAnally, and Weaver (0) find that banks delay adoption of Subchapter S status if they have net operating losses (NOL) that they haven t exhausted/recovered as tax credits in subsequent years. It is clear that the benefits of these carry forwards is less than being completely tax exempt in the last year when the tax shield is exhausted. To improve transparency of the sample, we compute the average time to Subchapter S adoption in the matches used. From to, it takes a bank in the non-subchapter S sample an average of. years (standard deviation of. years) to adopt the Subchapter S status. for the sample over 00 to 0, but the sample size is very small. In the sample over 0 to-0, we are able to get only banks in each group, whereas we have banks in each group for the period 00 to 0. To conserve space, we report the results only for the sample over 00 to 0. 0 We thank an anonymous referee for pointing out this limitation of our matched samples.

14 Page of The Financial Review Appendix A provides the list of the sources of all the other relevant data we use in this paper. Appendix B lists all the variables we use in this paper.. Methodology. Difference-in-Differences (DID) estimation Since a bank s owners choose their form of incorporation and tax status, any statistical analysis that does not account for this choice variable will suffer from selection bias. If the Subchapter S banks are systematically different from C-Corporation banks and this difference is correlated with the types of variables included in the study, then the results could be biased. To alleviate this problem, Depken, Hollans, and Swidler (0) use a propensity score matching procedure to adjust for sample-selection bias. They construct a probit model to predict the odds that a particular bank would be a Subchapter S bank given its characteristics. They then compare the Subchapter S banks with C- Corporation banks with similar predicted probabilities or propensities using the nearest-neighbor single-matching scheme. The propensity score matching takes care of the observed differences in the two groups, but there may be many differences that are unobserved and can lead to dilution of the regression results. We attempt to match for both observed and unobserved differences. Our entire sample is comprised of the banks that adopt the Subchapter S status by the end of. The banks in the treatment group are the banks that convert to Subchapter S status in year (t), and the banks in the control group are the banks that continue to be non-subchapter S banks in year (t+). As the banks in the control group will convert to Subchapter S later [sometime after year (t+)], Besley and Case (00) state the prevalent use of difference in difference estimation for such endogeneity issues.

15 The Financial Review Page of they have all the observed and unobserved characteristics similar to those the treatment group. When we use the Difference-in-Differences methodology and compare these two groups and control for the economic and demographic effects, we are able to capture the effect of the tax benefits on the Subchapter S banks. The timeline figure, illustrates the two groups in our Difference-in-Differences analysis. Group I (the treatment group) consists of banks that are non-subchapter S in year (-), convert to Subchapter S in year (0) and are Subchapter S in year (+). Group II (the control group) consists of banks that are non-subchapter S in all of the periods under investigation; i.e., they remain non-subchapter S banks in corresponding years -, 0, + and +. We select the banks in the control group from the pool of banks that adopted Subchapter S status by the end of. The banks in Group I and Group II are matched based on state code and size. We compute the change in the variables of interest from one period to another for both groups. We examine if the changes in the variables of interest are significantly different for the two groups. The DID method removes biases from comparisons over time in the treatment group that could be the result of trends. It also helps us control for macroeconomic factors like changes in interest rates, state of economy, etc.. Regression analysis Similar to Mayne () and Thomson (), in our regression analysis, we provide control for the local economy in which the banks compete; namely, the county unemployment rate, the per-capita income of the county, and whether or not the bank is Besley and Case (00) assert that in difference-in-difference estimation the control group should have the same effect of economic forces as for the group of interest.

16 Page of The Financial Review in a metropolitan area. We control for the size of the bank because economies of scale are experienced by banks as they grow larger, which may affect the spread and other variables included in the study. In addition, we control for the delinquency ratio as it is an indicator of the quality of a bank s loan portfolio which may affect the spread and average loan rates for the bank. We control for the amount of the real estate loan, consumer loan, and the commercial and industrial loan, as variation in product mix can affect average loan and deposit rates. As the banks are matched with another from the same time period, the sample is time unbalanced. Hence, we include several macro control variables, such as inflation rate and the market risk premium in the regressions to control for time fixed-effect. Gilbert and Zaretsky (0) provide an extensive review of market concentration s influence on bank deposit and loan rates, and point out that as banks compete for deposits and loans, Subchapter S banks will share some of their tax savings with customers. We include competition as a control variable, where competition is measured as the number of other banks and credit unions in the same county where a bank is located. We use the number of full time employees of the bank as another control measure of competition in regression equations. The earnings volatility of the bank may also impact the performance measures of bank; hence, we control it in our regression equations as well. We measure earnings volatility in year t as the variance of earnings as a fraction of total assets over the previous three years (t-, t-, and t-). We also provide Hughes, Mester, and Moon (0) find that large banks face significant scale economies that increase with bank size. We thank an anonymous referee for making this suggestion. We also use the 0-year constant maturity Treasury rates as macro control variables but drop it from the regressions because of multicollinearity issues with inflation. We find the correlation between inflation and the 0-year constant maturity Treasury bond yield to be 0.. We also run the regressions by excluding the market risk premium and by using the yield on 0-year constant maturity Treasury bonds instead of inflation. All our results stay the same.

17 The Financial Review Page of control for the bank s leverage and equity ratio as they may affect the bank s performance measures. The bank s performance measures may be impacted by the amount of its net deferred taxes; hence, we also provide control for the changes in the net differed taxes of the bank in our regressions. We compute the correlation matrix between all dependent and independent variables. Among independent variables, we find a high correlation between the changes in yield on 0-year constant maturity Treasury bonds and the changes in inflation, and between the changes in equity ratio and the changes in leverage ratio. To avoid multicollinearity issues our regressions do not include the yield on0-year constant maturity Treasury bonds and the equity ratio as control variable. We find a high correlation between dependent variables and several of the independent variables indicating that the choice of control variables are appropriate. We find Scode to be highly correlated only with ROA and ROE suggesting that adoption of subchapter S status affects only the ROA and ROE of a bank. below. We present the representative multivariate regression model in Equation () We thank an anonymous referee for suggesting to include the control variables for consumer loans, commercial and industrial loans, number of banks, number of bank employees, lending ratio, earnings volatility, leverage, and net deferred tax. To conserve space we don t include the correlation matrix in the paper. Results are available from authors on request.

18 Page of The Financial Review () = + + L ԑ On the left-hand-side of Equation () we have the change of dependent variable of interest. On the right-hand-side of the equation we include our variable of interest, Scode, and various control variables as described above. Scode is a dummy variable, which takes a value for Subchapter S banks (Group I) and a value 0 for non-subchapter S banks (Group II). We test for the significance of the dummy variable Scode. A negative-and-significant coefficient on Scode will indicate that the change of the dependent variable reduces significantly more for the Subchapter S banks than for the banks in the control group and vice-versa. To study the impacts of tax subsidy from adopting Subchapter S status we apply Equation () to thirteen different variables of interests and name them Model () to Model (). Models () to () are for testing the first hypothesis that tax benefits are used in expenses related to protection of jobs. First, we investigate if banks would pass some of the tax benefits to their customers after converting to Subchapter S status through offering higher deposit rates, or lower loan rates, or both. In Model () we use the change in spread ( Spread) as the dependent variable in Equation (); in Model () we use the change in average loan rate ( Alr) as the dependent variable; and in Model () we

19 The Financial Review Page of use the change in average deposit rate ( Adr) as the dependent variable. If Subchapter S banks use the tax benefit to increase lending, to provide better facilities and services, or more loan products to their customers, then we expect an increase in the operating expenses of Subchapter S banks as compared to the non-subchapter S banks. In Model () we use changes in the operating expenses ( Optexsal) as the dependent variable. The Optexsal variable includes expenses related to the use of premises, equipment, furniture, etc., but it does not include the salaries and benefits of bank employees. In Model (), we use the change in lending as a fraction of assets ( Lendingr) as dependent variable to examine if there is a significant increase in lending for Subchapter S banks as compared to the non-subchapter S banks. In model () we use the change in the number of loan products ( Products) as the dependent variable in Equation (). The owners of Subchapter S banks may utilize the tax subsidy to make higher reinvestments, which may in turn lead to higher growth in assets. In Model () we use the change of growth in asset ( Growth) as the dependent variable. Since the growth in assets is related to the asset size of banks, for Growth regression we do not include the change in size ( Size) as an independent variable. We use Models () and () to test for our second hypothesis that banks use tax benefits on expenses related to the creation of employment opportunities. In Model () we use the change in the number of employees ( Emp) as dependent variable and in Model () we use the change in number of employees per unit asset ( Emp_TA) as dependent variables. In both these models we exclude Emp from the list of independent variables. We then use Model (0) to test our third hypothesis that tax benefits are used in increasing the take home pay of workers. In Model (0), we use Wstexp as the

20 Page of The Financial Review dependent variable to examine if there is a significant increase in the wages, salary and travel expenses for the Subchapter S banks as compared to the non-subchapter S banks. Finally we test our fourth hypothesis that that the tax subsidy to Subchapter S banks is used in increasing payouts to the owners of the bank in the form of an increase in the return on equity and the return on assets. To study this, in Models () and () we use the change in return on equity ( ROE) and the change in return on assets ( ROA) as the dependent variable. Finally, one may argue that any increase in ROE and ROA may be due to increased efficiencies which in turn lead to higher operating profits. Hence in Model () we regress change in pretax ROA ( EBT_TA) as the dependent variable over the dummy variable Scode and all control variables as in Equation ().. Results and Analysis Table presents the relevant descriptive statistics for the independent variables, such as means and standard deviations of the changes in the basic characteristics for Subchapter S and non-subchapter S banks. The change in size for the banks in the control group and the change in size for the banks in the treatment group are both positive, indicating an increase in the size for both types of banks over time. The change in the proportion of real estate, commercial and industrial loans is positive for both the control group and the treatment group. This indicates an increase in the proportion of real estate, commercial and industrial loans over time. The change in the delinquency ratio is positive for both of the groups, indicating deterioration in asset quality over time. There is an increase in the amount of lending for both groups. As expected, the net deferred tax decreases for the banks that adopt Subchapter S, whereas it increases for the C banks.

21 The Financial Review Page of Table shows the results of the univariate analysis for all the variables of interest. We find the difference in changes to be statistically significant only for ROE, ROA and EBT_TA. These univariate results indicate that changes in return on equity and return of assets are significantly higher for the treatment group as compared to those of the control group. The difference in the change in pretax ROA ( EBT_TA) is negative and significant at less than % level, which suggests that the pretax ROA reduction for the Subchapter banks is more than that for the non-subchapter S banks. This result suggests that the increase in ROE and ROA for the Subchapter S banks is caused by the tax subsidy and not attributable to any improved efficiencies and revenues at the Subchapter S banks. Table displays the results of the multivariate regression models described in Section. We find the coefficient on Scode to be statistically insignificant for the change in the spread ( Spread) in Model (), indicating that there is no difference in the change in the spread between the treatment group and the control group. We also find the coefficient on Scode to be insignificant for the regression Models () and (). This confirms the univariate results that there is no difference in the changes in average deposit rates and changes in average loan rates between treatment and control groups. The coefficient on Scode is insignificant for change in the operating expense regression Model (), change in the amount of lending to bank customers (Model ()), and for the change in number of products offered by bank (Model ()). The result suggests that tax exemptions conferred to Subchapter S banks are not applied toward better services for customers or do not increase the amount of services provided to customers. The coefficient of Scode is not significant for the change-in-growth ( Growth) regression

22 Page of The Financial Review Model (), which indicates that the growth in assets for banks that convert to Subchapter S is not different from the growth in assets for banks in the control group. From the regression results of Model () to Model () we conclude that tax benefits from adoption of Subchapter S are not used in expenses related to the protection of jobs. We find the coefficient of Scode to be statistically insignificant for the change in the number of employees ( Emp) and change in the number of employees per unit asset ( Emp_TA) (Models () and ()), suggesting that the change in the number of employees for the treatment group is not significantly different from the control group. The regression results show that tax subsidy to Subchapter S banks does not result in any creation of employment opportunities. We find that the coefficient of Scode is insignificant for change in the wage, salary, and travel expenses regression (Model (0)). The result suggests that the tax exemptions to Subchapter S banks have not been passed on to the employees of the Subchapter S banks in the form of an increase in salaries and benefits. We regress the change in the return on equity ( ROE) and change in return of assets ( ROA) on the Scode dummy and the set of control variables (Models () and ()) and find the coefficient of Scode to be positive and significant at less than onepercent level. The regression results confirm the univariate results that the change in the return on equity for the treatment group is significantly higher than for the control group. Finally, the negative and significant coefficient of Scode for the regression on the changes in pretax ROA ( EBT_TA) (Model ()) indicates that the decrease in pretax operating profits for Subchapter S banks is more than that for the non-subchapter S

23 The Financial Review Page of banks. This confirms that the increase in ROE and ROA is due to the tax benefits and not because of any improved efficiencies and revenues at the Subchapter S banks. Overall, the univariate results in Table and the regression results in Table show that the tax benefits to Subchapter S banks are not used for the intended benefits of: Protection of jobs: there is no increase in Subchapter S bank s competitiveness as it does not pass the tax benefits to customers in the form of lower loan or higher deposit rates, or in providing better or greater services to customers, or in increasing lending, or in increasing the number of products offered. The tax benefits are not utilized on the growth in assets either. Creation of employment opportunities: there is no difference in the changes in number of employees at the bank that adopts Subchapter S status and comparable C bank. Increase in the take home pay of workers: there is no difference in the changes in wage salary and benefits for the employees of C bank and the bank that adopts Subchapter S status. The results suggest that the owners of the Subchapter S banks, with increased return on equity, are the sole beneficiaries of tax exemptions.. Robustness Tests We conduct a series of robustness tests to confirm the validity of our results. First, we repeat our analysis using a sample with more stringent matching. Apart from the state code and size matching done earlier, we now match the banks in the control group with banks in the treatment group based on county codes as well. This ensures that the banks in the two groups are located not only in the same state but also in the same

24 Page of The Financial Review county. There are now banks in each group in the new sample. We find the changes in competition, per-capita income, and unemployment rate to be exactly the same for the two groups, confirming that the banks in the two groups have been matched perfectly based on geographical location. For the new sample matched on the basis of size, state, and county, our results are similar to the ones we report in Table. We do not find any significant difference in the variables of interests except for the changes in lending ratio, growth, return on equity, and return on assets, which are significantly higher for the treatment group. We find the changes in the pretax return on asset to be significantly lower for the Subchapter S banks. In Table, we regress each of the variables of interests - Spread, Alr, Adr, Optexsal, Lendingr, Products, Growth, Emp, Emp_TA, Wstexp, ROE, ROA and EBT_TA - on the set of control variables and the Scode dummy. We do not find Scode to be significant for any regression (Models () to ()) except for the Lendingr, ROE and ROA regressions (Models (), () and ()) in which we find Scode to be positive and significant at less than % level, indicating a higher increase in lending, ROE and ROA for the treatment group. For the regression on the changes in pretax ROA ( EBT_TA) (Model ()) we find the coefficient of Scode to be negative and significant, indicating a more severe reduction in pretax operating profits for Subchapter S banks as compared to the non-subchapter S banks. Our robustness tests, using stringently matched samples, generally validate our previous results that tax exemption to Subchapter S banks does not result in any of the intended benefits of the Small Business Job Protection Act of To conserve space we do not report the tables for the descriptive statistics and the univariate results. These are available from authors on request.

25 The Financial Review Page of Instead, it leads to an increase in return on equity for the owners of the bank. The significance of Scode in the regression for Lendingr, indicates an increase in lending by the banks that adopt Subchapter S status. In our regressions on the changes in pretax ROA ( EBT_TA), we find the coefficient of Scode to be negative and significant, which may indicate the possibility that the firms adopting Subchapter S status are in distress. Distress might not reflect the true nature of the difference in pretax ROA. If the bank is not operationally profitable and benefits show up on an after tax basis, then the bank would not hire or pay more to its employees; and the absence of any changes in our result may only reflect that the adoption of Subchapter S indeed prevents layoffs. To test if banks adopting Subchapter S status are operationally non-profitable we compute the earnings before tax (EBT) of the banks in the year prior to conversion. The average EBT of $. million indicates that the banks that adopt Subchapter S status are operationally profitable. As we discuss in the data section, our comparison of different characteristics in the year before conversion (t = -) for the state and size matched sample and state, county, and size matched sample confirms that our control group and treatment group have similar characteristics, except for ROA and ROE. We further find that for the subsample of the state, county, and size matched sample in the later years (from 00 to 0), there is no significant difference even in the ROA and ROE, making this subsample more suitable for matching purposes. Table shows the regression of each of the dependent variables on the set of control variables and the Scode dummy for the state, We thank an anonymous referee for pointing out this possibility.

26 Page of The Financial Review county, and size matched sample from 00 to 0. We do not find Scode to be significant for any regression (Models () to (0) and ()). However, for the ROE and ROA regressions (Models () and ()), we find Scode to be positive and significant at less than % level, indicating a greater increase in ROE and ROA for the treatment group. While one might expect a before-and-after difference of incorporation on bank rates, it may be argued that the same is not true for employment and salaries. There is a possibility that the bank may begin to change its ownership structure and banking operations in anticipation of converting to Subchapter S status. In such a case, any conversion effect might have already taken place in year (t-); hence, we compare the banks in the treatment group and the control group in the year prior to conversion. There is also a concern that changes in total employee expenses may not properly represent the average compensation to the individual employee. For example, if the number of employees in the control group (C-Corporation banks) increases more than the treatment group (Subchapter S banks), but there is no difference in changes in operating expenses for the two groups, then it implies that Subchapter S bank employees benefit from taxexempt status on average (on a per-capita basis). Hence, as a robustness test, we also compare the salary and wages on a per-employee basis. In Table, we present the results for a univariate test in the year (t-) for the banks in the treatment group and the control group, matched on the basis of state code, county code, and size. We do not find any significant difference in the number of employees, wage salary expenses, operating expenses, and wage salary travel expenses per employee between the two groups. In Table, we present the results of multivariate regression analysis for the variables described in Table. We find the coefficient of

27 The Financial Review Page of Scode to be statistically insignificant, indicating that the number of employees, wage salary travel expenses, operating expenses, and wage salary travel expenses per employee are statistically the same for the Subchapter S and the non-subchapter S banks in the year prior to conversion. There is a possibility that some problematic banks with structural, regulatory, or bankruptcy issues may be influencing our results. Therefore, we create a sub sample of banks by removing all the bankrupt and merged or acquired banks and removing banks with negative earnings. We use this sub sample and run the multivariate regression Models () to () for the state and size, and state, county, and size matched sample. Our results are exactly the same as before. Scode is significant for the ROA and ROE regressions in the state and size matched sample and is significant for Lendingr, ROA and ROE regressions for the state, county, and size matched sample. To further ensure that outliers are not the main drivers of our results, we winsorize the data for all variables at top and bottom % levels. In the regression for state and size matched sample and the state, county, and size matched sample we find Scode to be significant only for ROA and ROE regressions. We notice that Scode is no longer significant in the regression for Lendingr suggesting that the previously observed increase in lending in the state, county, and size matched sample may have been due to few outliers in the sample. There is also a possibility that the financial crisis may have influenced the Subchapter S banks and lending practices. Employment increase/retention/decrease We thank an anonymous referee for suggesting the robustness tests by excluding the merged and bankrupt banks and banks with negative earnings, for suggesting to winsorize the variables and for suggesting the analysis of the pre, post and during crisis period. To conserve space we don t provide the tables for these robustness test regressions. These are available from authors on request.