«Vladimir Mukharlyamov * January 11, 2016 JOB MARKET PAPER Please click here for the most recent version. Abstract Using a novel dataset that allows ...»
Beyond the Corner Office:
Employee Characteristics and Bank Performance
Vladimir Mukharlyamov *
January 11, 2016
JOB MARKET PAPER
Please click here for the most recent version.
Using a novel dataset that allows me to capture the education and career trajectories of
over 250,000 employees of 224 bank holding companies, I find that banks with shorter
employee tenures and higher fractions of MBAs, top school graduates, and job jumpers performed more poorly during the Great Recession. This relationship is driven by the predisposition of these banks to take on greater risk. These same workforce measures also explain banks’ performance in the 1998 crisis. Taken together, my results suggest that investigating workforce measures could be a step towards quantifying components of risk culture or strategy that contribute to financial institutions’ vulnerability to crisis.
* Department of Economics, Harvard University. Email: email@example.com. I am deeply indebted to my advisors Effi Benmelech, Paul Gompers, and Andrei Shleifer for extensive guidance and support. I also thank Christopher Anderson, Malcolm Baker, John Campbell, George Chacko, Aubrey Clark, Lauren Cohen, Shawn Cole, Sanjiv Das, Robin Greenwood, Sam Hanson, Steve Kaplan, Divya Kirti, Josh Lerner, Yueran Ma, Filippo Mezzanotti, Jonathan Roth, Atulya Sarin, Natasha Sarin, David Scharfstein, Hersh Shefrin, Stas Sokolinski, Erik Stafford, Jeremy Stein, Lawrence Summers, Adi Sunderam, Yuhai Xuan, and seminar participants at Harvard for thoughtful comments. For making their unique risk management dataset available, I am grateful to Andrew Ellul and Vijay Yerramilli. Financial support from Harvard GSAS is greatly appreciated.
1. Introduction In this paper, I investigate whether characteristics of bank employees shed light on bank conduct and performance before and during the financial crisis. While previous studies have considered top executives’ characteristics and their impact on firm performance,1 virtually no attention has been paid to the characteristics of non- executives.2 My paper is the first to do a close study of the individual characteristics of a firm’s workforce as a whole in efforts to understand performance.
To this end, I use a novel data set created by merging 1) individual-level resume data from a major professional networking website and 2) bank-level performance and balance sheet data from Call Reports and CRSP.3 The advantage of my dataset of individual resumes is that it allows me to reconstruct snapshots of workforces and their characteristics in the 224 bank holding companies (BHCs) in my sample.4 I am able to observe how the workforces of banks evolved in the pre-crisis years and document the relationship between a variety of workforce measures (particularly employee education, experience, and stability of employment) and crisis performance.
In my analysis, I use several measures of bank performance, including stock returns, bank failure, and percentage of loans charged-off (overall and by loan category).
I focus on four workforce measures: 1) percentage of employees with an MBA; 2) percentage of employees who received a degree from a highly-ranked university; 3) percentage of employees with a high propensity to switch jobs in the past; and 4) average turnover of the workforce.5,6 I hone in on these four characteristics because extant literature has documented a relationship between such traits at the executive
1 See, for example, Bertrand and Schoar (2003), Malmendier and Tate (2005), Güner, Malmendier, and Tate (2008), Kaplan, Klebanov, and Sorensen (2012), Graham, Harvey, and Puri (2013), Minton, Taillard, and Williamson (2014), and Benmelech and Frydman (2015).
2 Some exceptions are Agarwal and Wang (2009), Hertzberg, Liberti, and Paravisni (2010), Berg, Puri, and Rocholl (2013), Tzioumis and Gee (2013), Agarwal and Ben-David (2014), and Cole, Kanz, and Klapper (2015). These studies are either experimental over a subset of banks’ employees or use proprietary data from a single lender.
3 I supplement the main dataset used in the baseline analysis with the information on geographic distribution of banks’ deposits from FDIC’s Summary of Deposits, house prices from Zillow, and the Risk Management Index from Ellul and Yerramilli (2013).
4 FDICs Summary of Deposits contains matching between BHCs and their commercial bank subsidiaries. I use it to aggregate commercial banks to the BHC level.
5 Turnover is defined as the negative of the average job tenure of a bank’s employees.
6 These variables are labeled in the regression tables as MBA, Top school, Job jumper, and Turnover, respectively.
1 level and firm performance. For instance, Bertrand and Schoar (2003) note that employees with MBAs appear to follow more aggressive strategies and Minton, Taillard, and Williamson (2014) find that financial expertise amongst directors is strongly related to lower performance during crisis. Perez-Gonzalez (2006) finds that family-appointed CEOs with degrees from selective academic institutions tend to outperform those without them. Berger, Kick, and Schaeck (2014) find that younger executive teams tend to be riskier; and Chernenko, Hanson, and Sunderam (2015) find that experience was a significant predictor of crisis performance, as inexperienced managers had almost twice as much subprime exposure as their seasoned counterparts.
Job mobility measures have received less attention in the finance literature because of their reliance on employer-employee matched data. The broader economics literature has explored the effects of job mobility on individual outcomes such as wage growth (Keith and McWilliams 1999, Altonji and Williams 2005) and has found positive wage effects for seniority, suggesting that workers acquire firm-specific human capital for which they are compensated.7 Individuals with a tendency to move jobs despite this added compensation plausibly 1) have a preference for finding novel job opportunities or
2) tend to get laid off by employers. It is natural to postulate, then, that a workforce with a documented predisposition to move jobs for either reason will be less stable and require excessive future investment in worker training. As such, I anticipate that job jumping will be negatively related to performance.
I find that banks with a higher proportion of workers having the aforementioned traits (MBAs, top school degree holders, job jumpers, and high turnover) tended to perform worse in the crisis than their counterparts. A one-standard deviation increase in the proportion of MBAs lowers the bank’s stock return during the recent crisis by 4.5 percentage points. Similar statistics for the fraction of top school degree-holders, job jumpers, and worker turnover are 9.4, 12.2, and 13.5 percentage points respectively.8 I also show that these same workforce measures are positively related to bank failure as well as the fraction of loans charged-off in the crisis years.
7 In fact, early-career wage growth is strongly associated with higher job mobility of young employees (Topel and Ward 1992), but frequent job switching over the entire course of one’s career is associated with lower earnings (Fuller 2008).
8 These workforce measures are aggregated to the bank level and adjusted by bank size.
2 I use a host of observable bank characteristics (for example, pre-crisis growth, securitization activity, compensation, and composition of the loan portfolio)9 known to be related to performance and risk-taking to show that my workforce measures are not proxies for other commonly used indicators of banks’ vulnerability to crisis. The baseline relationship between crisis performance and workforce measures is virtually unchanged in the presence of additional controls. I also show that the relationships between the workforce measures and different bank characteristics are, by and large, flat.
Accordingly, my workforce measures are unlikely to be picking up the effect of some known measures of banks’ vulnerability to crisis. Rather they provide meaningful information orthogonal to that which can be derived from other observable traits.
These baseline results lead me to develop three hypotheses about the nature of the relationship between my workforce measures and bank performance during crisis that I can then take to my data.
My first hypothesis is that firms with higher proportions of MBAs, top school employees, job jumpers, and workers with relatively short tenures10 had such workforce compositions because they looked to grow aggressively in the pre-crisis period, and these were the types of workers that were available to hire. If this is the case, then the relationship I document between these measures and performance can be explained by pre-crisis hiring.
I reject this hypothesis by showing that the characteristics of pre-existing employees–those recruited before the pre-crisis expansion kicked off–are driving the relationship I document between crisis performance and workforce measures. In other words, the hiring implemented by banks to accommodate the growth that preceded the financial crisis does not explain the baseline relationship between how well or poorly banks performed in crisis and the composition of their workforces.
9 The full list of variables includes: pre-crisis stock returns volatility, tier-1 capital ratio, ratio of core deposits to assets, housing bubble exposure, securitization activity, ratio of private MBS to assets, assets growth, assets per employee growth, number of employees growth, ratio of loans to assets, fraction of noninterest income in total income, assets per employee, residual compensation, fraction of real estate loans in total loans, fraction of C&I loans, fraction of consumer loans, fraction of agricultural loans, fraction of other loans, and Herfindahl index for loan categories.
10 I define my turnover variable as the negative of average job tenure. I choose to work with turnover for expositional convenience so that all of my workforce measures move in the same direction. That is, high values of MBA, top school, job jumper, and turnover are all associated with poor performance in crisis.
3 My second hypothesis is that the relationship between workforce characteristics and bank performance is attributable to underlying bank quality. If this is the case, I expect some banks to always perform poorly–in booms and busts alike–and some banks to consistently perform well. I reject this hypothesis by showing that the workforce measures related to poor crisis performance are, if anything, indicative of superior performance in the pre-crisis period.11,12 My third hypothesis is that my workforce measures are related to risk-taking by banks. If this is the case, then I would expect that risk-taking will be highest in banks with the highest proportion of MBAs, top school graduates, high turnover employees, and job jumpers. I find support for this hypothesis in the data.
I measure risk by calculating holdings of highly rated securitization tranches that are not government or agency affiliated,13 computing the ratio of private mortgagebacked securities to assets, and calculating the average interest rate on loans. I find that banks with the highest proportion of MBAs, top school graduates, high turnover employees, and job jumpers took on more risk before the crisis. They also had higher volatility of stock returns and higher realized tail risk14 which is consistent with their taking more ex-ante risk that resulted in greater ex-post fluctuations of stock prices.
Finally, these banks perform more poorly on Ellul and Yeramilli’s (2013) risk management index,15 which assesses the strength of risk management functions at bank holding companies. Taken together, these findings are consistent with the hypothesis that my workforce measures are related to banks’ crisis performance because of differences in banks’ risk-taking before the crisis.
11 This finding is consistent with Beltratti and Stulz (2012) who show that banks that perform worst in crisis had above-average returns before the crisis.
12 I also show that banks that had the largest proportion of MBAs, top school degrees, job jumpers, and workers with the shortest tenures had higher compensation per employee than their counterparts.
13 I follow Erel, Nadauld, and Stulz (2014) who study holdings of highly rated tranches, which are highly rated securities issued in securitizations and held on BHC balance sheets, like subprime residential mortgages and collateralized debt obligations. They identify the amount of securities assigned an AA or AAA risk weight that are not government or agency-affiliated, and call this the “highly rated residual.” 14 I compute tail risk following Acharya et al. (2010) based on the stock’s average return over its 5% worst trading days.
15 To construct a risk management index (RMI), Ellul and Yerramilli (2013) hand-collect information on the organizational structure of the risk management function for each bank holding company from its 10K statements, proxy statements, and annual reports. Their measure is only available for a subset of the banks in my sample.