The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Federal Reserve Bank of Chicago Perverse Incentives at the Banks? Evidence from a Natural Experiment Sumit Agarwal and Faye H. Wang WP 2009-08 Perverse Incentives at the Banks? Evidence from a Natural Experiment∗ Draft: August 20, 2009 Sumit Agarwala and Faye H. Wangb Abstract Incentive provision is a central question in modern economic theory. During the run up to the financial crisis, many banks attempted to encourage loan underwriting by giving out incentive packages to loan officers. Using a unique data set on small business loan officer compensation from a major commercial bank, we test the model’s predictions that incentive compensation increases loan origination, but may induce the loan officers to book more risky loans. We find that the incentive package amounts to a 47% increase in loan approval rate, and a 24% increase in default rate. Overall, we find that the bank loses money by switching to incentive pay. We further test the effects of incentive pay on other loan characteristics using a multivariate difference-in-difference analysis. JEL Classification: D3, G2, J3 Key Words: Incentive Compensation; Small Business Lending; Loan Officers; Piece Rate and Salaries ∗ The views expressed in this research are those of the authors and do not necessarily represent the policies or positions of the Federal Reserve Board or the Federal Reserve Bank of Chicago. The authors wish to thank audience members at AEA 2009 annual meeting, FIRS 2009 Prague conference, Federal Research Bank of Chicago, Summer Research Conference at the Indian School of Business, and University of Illinois–Chicago for comments. The authors are grateful to Bob Chirinko, Evgeny Lyandres, Rich Rosen, and Greg Udell for helpful comments. a Federal Reserve Bank of Chicago, sagarwal@chi.frb.org. b University of Illinois–Chicago, hfwang@uic.edu. “Despite the vast outpouring of commentary and outrage over the financial crisis, one of its most fundamental causes has received surprisingly little attention. I refer to the perverse incentives built into the compensation plans of many financial firms, incentives that encourage excessive risk-taking with OPM -- Other People's Money” Professor Alan Blinder, Wall Street Journal, May 28, 2009 1. Introduction The current financial crisis has led to much debate about incentive provision at financial institutions.1 While the causes of the mortgage meltdown are complex, many would argue that perverse economic incentives are an important factor contributing to the current mess. Specifically, many banks in recent years gave out incentive packages to encourage loan underwriting. While such financial incentives were designed to promote greater employee efforts, anecdotal evidence suggests that they also encouraged loan officers to make more loans to unqualified borrowers.2 In this paper, we study the effects of an incentive pay based on loan origination by formally modeling two tasks loan officers perform: their efforts to assess loan quality and their loan origination decisions. The model allows us to derive testable implications regarding the approval rates, default rates and other characteristics of booked loans under an incentive pay. In addition, we provide answers to two central questions in the incentive provision literature: 1) do incentives matter, that is, do agents respond to contracts that reward performance? 2) are these responses in the firm’s interest, or do such contracts induce perverse incentives? We employ a unique data set from a large, national commercial bank on loan officer compensation, which allows us to empirically study loan officers’ incentives and loan performances. In January 2005, the management of the bank switched half of the loan officers from fixed-wage compensation contracts to a new incentive compensation 1 2 See Fahlenbrach and Stulz (2009) and Bebchuck and Spamann (2009). See Morgenson, Gretchen, “Was there a loan it didn’t like?,” New York Times, November 1, 2008. 2 package based on loan origination. We examine the status of more than 30,000 small business loan applications received at the bank from 12 months before the compensation change to 12 months after and the performance of more than 140 loan officers. The effects of such incentive pay on approved loan characteristics are dramatic, and are largely consistent with economic theory. For this purpose, we develop a theory of loan officers’ incentives to assess loan quality, with emphasis on the predictions that pertain to the compensation changes at this bank. Consistent with our model’s predictions, we highlight the following empirical results: 1) A switch to incentive pay increases the loan approval rate by 47%, the total number of booked loans by 44%, the average dollar amount of booked loans by 45%, and the default rate by 24%. 2) The average amount of time spent per loan applications drops by 21%. 3) The above effects of incentive pay are stronger for larger and longer maturity loans, and loans that contain more soft-information. 4) Although the bank shares the gains in more loan origination, the large amount of loan defaults results in a welfare loss for the bank. These results are in line with the model we developed on loan origination, which takes into account the loan officer’s career concerns and different loan information regimes. The loan origination process starts when a loan officer receives a loan application that contains observable information. The loan officer then studies the credit risks of the borrower and investigates the loan quality by exerting costly effort. We assume that the probability of revealing the loan quality depends on the loan officer’s 3 unobservable effort. She then makes a loan origination decision based on the information she reveals. This information may be hard or soft information. One distinction between hard information and soft information is that soft information is a signal that cannot be verified ex post (Petersen, 2004). We study cases where the loan officer makes the loan approval decision based on unverifiable information, or soft information, and where she makes the decision based on verifiable information, or hard information. With soft information lending, the loan officer may lie about the information she reveals in order to approve the loan. Much of the literature on incentive provision has focused on the role of outputbased contracts in solving the agency problem for the principal: if effort cannot be monitored, output-based contracts can address the moral hazard problem induced by agents’ hidden actions. We show that the effectiveness of such contracts in mitigating moral hazard depends crucially on whether hard or soft information is used in the loan origination decision. If hard information is used, incentive compensation works well to align the loan officer’s incentive to search for good loans. High-powered incentive compensation, in this case, motivates loan officers to work harder at assessing loan quality and writing more and better loans. If, however, the lending decision is based on soft information, incentive compensation may distort the loan officer’s incentive to “overbook” risky loans. This conclusion is consistent with Inderst (2008).3 The loan officer’s career concern, on the other hand, also plays an important role as a disciplinary device in soft-information lending. The loan officer trades off a monetary bonus with her career concern. In this case, incentive compensation may distort 3 Inderst (2008) studies loan officers’ incentive to generate new loan applications and their loan origination decisions under soft- and hard-information lending. The focus is more on bank competition and the optimal lending standard. 4 the loan officer’s effort, depending on the strength of the incentives given. Indeed, we find that younger loan officers respond to the incentive pay by booking more loans, without sacrificing loan quality, compared to older loan officers whose loans have a much higher default rates under the incentive pay. This is consistent with the view that younger agents have more career concern at stake. This paper offers new insights into the process of originating small business loans by focusing on the incentives faced by loan officers and how this affects the underwriting process. The conflict of interest between the loan officers and the bank has rarely been studied in the literature, with the exception of Udell (1989). By surveying 140 midwestern banks, Udell finds that banks increase monitoring of their loan officers when the bank explicitly compensates loan officers for generating more new business. Our model provides a theoretical foundation for this result. Our paper also relates to a broader literature on incentive provision to individuals in organizations, which is a central issue in modern microeconomics (see Prendergast (1999) for an extensive survey). In the context of compensation contracts, the provision of incentives usually takes the form of pay-for-performance, or piece-rate contracts (Lazear and Rosen (1981); Stiglitz (1981); Holmstrom (1982); Green and Stokey (1983)). Researchers have analyzed the choice of one compensation system over another (see Gibbons (1998), literature review). In particular, piece-rate payment has the effects of inducing appropriate effort levels and sorting workers across jobs (Lazear (1986)). Alternatively, economists also argue that such incentive contracts may give rise to dysfunctional behavioral responses, where agents emphasize only those aspects of performance that are rewarded (Baker (1992)). For example, agents may choose quantity 5 over quality. Following Holmstrom and Milgrom (1990) and Baker (1992), this incentive problem has become known as multi-tasking, where agents will allocate effort toward those activities that are directly rewarded and away from the uncompensated ones. Loan officers’ compensation provides a perfect opportunity for studying multi-tasking. Incentives are provided on the dimension of quantity and not on quality. We show that distortion arises for soft-information lending in particular. Due to the lack of data, there has been little work documenting the effect of compensation policies on performance. Lazear (1996) studies the performance of auto windshield workers and documents the incentive and worker selection effects of piecerate contracts. Paarsch and Shearer (1996) provide similar evidence using data on Canadian tree planters. It is important to bear in mind that these studies document cases in which the jobs carried out are relatively “simple,” in the sense that 1) performance is easily measured and 2) the quality is easily observed. The loan officer’s job in our paper is much more complicated than those in the previous papers. Most importantly, the quality of the loan officer’s work is not easily measured due to unobservable randomness of other factors. Our data set is richer, in the sense that it allows us to further analyze the effects of incentive contracts on multi-tasking behavior of the agents. In addition to providing empirical evidence for the existing theories, we add to the banking literature by studying how loan officers’ incentives affect the process of small business loan underwriting. The rest of the paper proceeds as follows. Section 2 develops a model of incentive compensation. Section 3 provides a detailed description of the data. Section 4 provides empirical results. Section 5 concludes. 6 2. Model Description The primary motivation behind the incentive scheme is to increase worker efforts. A central role that loan officers play in the process of loan origination is to assess loan quality. In particular, loan origination depends on a significant amount of soft and subjective information from loan officers (Udell (1989)). In this section, we study the loan officer’s choice of effort to detect bad loans and her loan origination decisions under the incentive pay. We find conditions under which such incentive scheme gives rise to loan officers’ responses that are not in the bank’s interest. 4 A loan application is characterized by (Y, T, q), where Y denotes the requested loan amount, T is the time to maturity and q captures the ex-ante observable risk profile of the loan. Assume that q is uniformly distributed on [0,1]. There are two types of loans: a loan is “good” with probability q, and “bad” with probability 1-q. A good loan is repaid with probability p—that is, with probability 1-p even a good loan may fail. A bad loan defaults with certainty. Therefore, the higher q, the lower the probability of default. In addition, we assume that the bank makes a profit on good loans only and loses money on bad and questionable ones. To focus on the loan officer’s choice of effort to assess loan quality, we model a risk neutral loan officer’s decision to exert unobservable effort, e, and her loan origination decision. The probability θ that the loan officer reveals the loan type depends on her effort e. With probability 1- θ, the loan type remains uncertain. We assume that θ’(e) >0, θ(0)=0 and θ(∞)=1. 5 We also assume that e' (θ ) ∈ [0, ∞) and e" (θ ) > 0 .6 4 5 See Inderst (2008) for an analysis of optimal compensation contracts. Assume that θ is continuous. Then the inverse function e(.) is continuous and e’(θ) > 0. 7 Assume that a compensation contract is given by a + b(Y), where a is the base salary and b(·) is the bonus based on the amount of loans originated. Assume that b(Y) ≥ 0, and b’(Y) >0. For notional simplicity, we normalize a to zero.7 In the case a loan is denied, the loan officer receives no bonus component. If a loan is booked, she receives b(Y) – e(θ) – ρc(T), where ρ is the probability that a loan defaults, and c(T) is the negative career consequences of a defaulted loan. Specifically, we assume that c(T) takes the functional form c(T ) = Ke −δτ (T ) , where K>0 is the negative career shock to the loan officer when a default occurs, and τ(T) is the expected default time of a loan with maturity T, and is an increasing function of T.8 δ is the discount rate. We focus on a loan officer’s decision to assess loan quality under an incentive scheme that rewards her for loan originations. The basic game proceeds as follows: a loan officer reviews a loan application characterized by (Y, T, q). She exerts costly effort e, and reveals the loan type with probability θ(e). For soft-information lending, the loan officer has private access to information about the borrowing firm that is “hard to quantify, verify and communicate through the normal transmission channels of a banking organization.” (Berger and Udell (2002).) She can conceal, or lie about, the information she reveals. On the other hand, if lending is primarily based on hard information, the loan officer cannot conceal the information she reveals and this information is verifiable ex post by the bank. The loan officer then makes an approval decision and receives payoffs accordingly. 6 One such example of the functional form of e(.) would be e(θ)= tan(θ*π/2). In an optimal contract, a is set such that the loan officer’s individual rationality condition is satisfied. 8 This assumption can be justified by a model with a constant Poisson intensity of default. 7 8 With hard-information lending, the loan officer no longer plays an active role at the loan origination stage, since only loans revealed to be good may be booked. The problem reduces to a pure moral hazard problem, in which the only decision factor is the loan officer’s hidden effort. Consistent with economic theory, incentive pay promotes greater effort without sacrificing the loan quality. We provide a formal analysis in Appendix A. 2.1 Perverse incentives with soft-information lending In this section, we study the loan officer’s decisions in a soft-information lending regime. The loan officer makes a lending decision based on her privately observed soft information, which cannot be verified by the bank. She may lie about the soft information she reveals in order to book the loan. In this case, whether monetary incentives create perverse incentives depend on the loan officer’s career concern and the strength of the incentives. We analyze the soft-information lending problem in two steps. We first focus on the approval decision, then derive the loan officer’s optimal effort level. The loan officer’s approval decision now depends on the information revealed and privately observed by her: the loan type is good (G), bad (B), or no information is revealed, that is, the loan quality remains questionable (Q). In order for the loan to be approved, the loan officer may revise her subject input of the loan application’s risk rating downwards for a bad or questionable loan. If a loan is booked, the loan officer gets the following payoffs based on revealed types: 9 P(G ) = b(Y ) − (1 − p)c(T ) − e(θ ), P( B) = b(Y ) − c(T ) − e(θ ), P(Q) = b(Y ) − (q (1 − p ) + (1 − q ))c(T ) − e(θ ). Depending on the loan officer’s career concern and the size of the monetary bonus, which is tied to the amount of the loan requested, it may be in the loan officer’s interest to not only book the good loans but also the questionable ones. In other words, monetary bonus may induce her to take more risks for these loans. If the incentives are small enough or the loan officer’s career concern is large enough such that the she only writes good loans, we find that the effect of incentive contract is exactly the same as with the hard-information case: incentive compensation promotes effort without sacrificing loan quality. If the monetary bonus gets too large, such that the loan officer will accept any loan applications even the bad ones, we show that the loan officer has no incentive to exert effort and that incentive packages are purely costly to the bank. We discuss these two cases in more depth in Appendix B. We focus on the case where the loan officer approves both the good loans and the questionable ones, but her career concern is large enough that she wants to avoid writing bad loans. Recall that if no information is revealed, the prior assumption that the loan is good with probability q, in which case, the bank loses money on such loans. Such questionable loans cannot be booked with hard-information lending, nor does the loan officer have incentives to book such loans when the monetary incentive is small relative to her career concern. However, as the monetary incentives get large, the loan officer may find it profitable to write such loans at the bank’s expense. This is where an incentive pay may induce perverse incentives. 10 The question becomes when will the loan officer be incentized to write questionable loans? The loan officer will revise her subjective input for the loan quality upwards for a questionable loan when the following constraints are satisfied: b(Y ) − (1 − p )c(T ) ≥ 0, b(Y ) − (q (1 − p ) + (1 − q ))c(T ) ≥ 0, b(Y ) − c(T ) < 0. We next analyze the loan officer’s choice of effort under the above conditions. Recall that when the loan officer exerts costly effort e, with probability θ the loan type will be revealed, and with probability 1- θ its type remains questionable, in which case the loan is good with probability q, and bad with probability 1-q. The loan officer has incentive to book both the good loans and the questionable ones if the monetary incentive is large or when her career concern is relatively weak. The loan officer gets payoff qθ[b(Y) – (1-p)c(T)]+(1–θ)[ b(Y) – (q(1-p)+1-q)c(T)] – e(θ), yielding a FOC: q[b(Y ) − (1 − p)c(T )] − [b(Y ) − (q(1 − p) + 1 − q)c(T )] = e' (θ ) ⇒ − (1 − q)b(Y ) + (1 − q )c(T ) = e' (θ ). From the above FOC, we can derive the approval rate, the default rate of booked loans, and other comparative statics. We summarize these results in the following proposition. All proves are in the appendix. Proposition 1 If the monetary incentive is large or the loan officer’s career concern is small, the loan officer will book a loan unless it is revealed to be bad. In this case, 11 1. The loan officer’s effort level to investigate loan quality decreases with the loan amount, the time to maturity, the ex-ante score of the loan and the strength of monetary incentives, and increases with her career concern. 2. The probability of loan origination decreases with the loan officer’s effort, increases with the loan amount, the time to maturity, the ex-ante score of the loan, and the strength of monetary incentives, and decreases with her career concern. 3. The probability of defaults decreases with the loan officer’s effort, increases with the loan amount, the time to maturity, and the strength of monetary incentives, and decreases with her career concern. We see that monetary incentives induce preserve incentives in these that they reduce the loan officer’s effort to assess loan quality. The intuition is as follows. The loan officer receives a monetary bonus only when a loan is booked. Since both the good loans and the questionable ones are booked, her effort to investigate loan quality only affects her booking decision of bad loans, which she will not book weighing in her career concerns. In other words, the loan officer’s effort decreases the likelihood that she will book a loan and receive the monetary bonus. Career concern, on the other hand, serves as an effective disciplinary device and motivates the loan officer to exert effort to avoid booking bad loans. Interestingly, the ex-ante quality of a loan, q, does not predict the likelihood of loan origination or the default probability of a booked loan. Taking the derivatives with respect to q to the FOC, we get b(Y ) − c(T ) = e' ' (θ ) ∂∂θq* . Since c(T) > b(Y), it is easy to see that the loan officer’s effort decreases with q. In this case, the loan officer increases her effort to investigate lower quality loans to avoid booking a bad loan. Although these 12 lower quality loans are more likely to be bad and will default with greater probability, the loan officer also spends more effort to investigate them and avoids making loan originations. Thus, there is no direct relationship between the score and the expected default probability. 2.2 Empirical Implications In the previous section, we analyze whether incentive pay creates perverse incentives for loan officers by studying their choices of effort to investigate loan quality and their loan origination decisions. While the choice of effort is hidden, the probability of default, which decreases with the level of effort, is observable. We thus have the following predictions: Prediction 1 Under incentive pay, the likelihood that a loan is booked increases with the loan amount Y, the time to maturity T, and the ex-ante score q. Prediction 2 Under incentive pay, the default probability of booked loans increases with the loan amount Y and the time to maturity T, when the strength of monetary incentive is large. Prediction 3 The likelihood of booking a loan decreases with the loan officer’s career concern. Our analysis on information verifiability and compensation schemes sheds lights on how incentives affect the subsequent loan performance based on the amount of soft information used in the loan origination process. We show above that with hardinformation lending, incentive compensation encourages loan officers to investigate loan 13 quality and to avoid booking bad loans. Only when information becomes unverifiable does a monetary bonus distort incentives. Prediction 4 Both the likelihood of loan origination and the default probability are higher for more informationally opaque loans with soft-information lending than with hard-information lending. 3. Description of the Market and Data 3.1 The Loan Officer’s Job Function Loan officers play a central role in the process of loan origination. The process begins when the loan officers initiate contacts with the firms to determine their needs for loans. After the initial contact has been made, loan officers assist the clients through the process of loan application. The loan officer gathers personal and business information about the borrower and explains the different types of loans and credit terms available to the client. Loan officers then verify the basic information of the borrower to assess the creditworthiness of the borrower and the probability of repayment. Specifically, loan officers assign credit scores to the potential borrower and determine collateral requirements. A loan that would otherwise be denied may be approved if the client can provide the lender with appropriate collateral property pledged as security for the repayment of a loan. For a more detailed description of the process, see Agarwal and Hauswald (2007, 2008). Loan officer compensation usually takes the form of fixed payment salary or incentive plans based on loan origination. Neither of these compensation schemes is tied to loan repayment or failure and the eventual profitability of these loans. Such 14 compensation contracts may distort loan officers’ incentive and encourage them to make any loan, regardless of its quality. While bonuses based on loan profitability would have the advantage of giving direct incentives to search out good credit risks, such performance measure would also give the loan officers greater risk, because many things can happen to borrowers that are essentially unknowable when a loan is written. The additionally imposed risks on loan officers are costly to the bank through higher wages. Baker (2002) argues that the trade-off between risk and distortion in this case is made in favor of lower risk and higher distortion. 3.2 Data from a natural experiment The data set used in this paper comes from a large, national commercial bank. Starting January 2005, the management of the bank implemented a new incentive compensation package for half of the small business credit services approval officers (henceforth referred to as the treated group). The other half of the loan officers remained on fixed wage (henceforth referred to as the control group). The selection of the loan officers was quasi-random. The management had multiple legacy portfolios as a result of earlier merger and acquisition activities over the years. They were broadly being managed under two legacy database management systems. The portfolio of loans under both these management systems had identical underwriting standards, geographical focus, portfolio management practices, and loss outcomes (see Table 3a and 3b). To evaluate the success of the incentive compensation package, the bank implemented the change on one of the management systems while leaving the other on fixed wage compensation. 15 The incentive package provides a “pay for performance” bonus opportunity based on individual results. Before that, all loan officers were paid a fixed salary. Specifically, loan officers will receive an annual bonus based on the percentage of new money applications booked compared to the previous year, the type of decisions made, and the timeliness of the decision. The details of the incentive package are summarized in Table 1 and 2. The goal of this program is to “recognize and reward those associates whose performance most aggressively contributes to the overall success of small business credit services,” and “to attract and retain outstanding talent.” The incentive plan comes with a quality assessment. In order for a loan officer to be eligible to participate in the incentive program, their total of unsatisfactory underwriting must not exceed 5% of total approvals, reviewed in a post approval review process. The data cover 12 months before the compensation change and 12 months after. To study the effects of incentive compensation on loan officers’ incentives and the implications for subsequent loan origination decisions and characteristics of the booked loans, we employ two control groups: data on loan officers and loans of the group before the implementation of the incentive plan, and data from the other half of the loan officers whose salary remained fixed during the same period. Data from the control group allow us to better control for macroeconomic fluctuations over this period. Our sample contains data on more than 140 loan officers and the status of 15,784 loan applications in the treated group and 14,484 loan applications in the control group. The data are summarized in Table 3 and Table 4. 16 4. Empirical Evidence 4.1 Do incentives matter? 4.1.1 Loan Origination Decisions Not surprisingly, loan officers are motivated to book more loans under the incentive pay structure. Table 3a summarizes the status of loan applications for both groups in year 2004 and 2005. While there is no apparent increase in the number of new applications from year 2004 to year 2005, the number of booked loans increases by 1,132 in the treated group, a 44.4% increase. Approval rate in the treated group goes up from 32% to 47%, consistent with our model’s prediction. Also consistent with Prediction 1, in addition to booking more loans, loan officers in the treated group are booking larger loans and longer maturity loans. Table 3a shows that the average dollar amount of booked loans increases by $96,470, a 44.7% increase. Table 3d shows that loan officers are more likely to approve bigger loans than smaller loans and longer maturity loans than shorter maturity loans. The effect is stronger after the implementation of the incentive plan: Big loans, those with requested amount above $700,000, have an approval rate of 55% under the incentive pay, compared to an approval rate of 33% for small loans. Long term loans have a 52% approval rate compared to a 37% approval rate for short term loans. 4.1.2 Are Loan Officers Booking Riskier Loans? One potential concern of paying piece rates is that quality may deteriorate. In our case, does a piece rate contract distort incentives in a way that results in loan officers booking riskier loans? The model suggests that loan officers have stronger incentives to 17 investigate and approve loan applications with inferior ex-ante quality. One observable key risk factor that lenders use to assess qualifying borrowers for loans is the loan-tovalue ratio (LTV). Our subsequent multivariate analysis also confirms that a higher LTV value predicts higher loan default probability. Table 3b shows that while the average LTV of loan applications decreased slightly from 2004 to 2005 in the treated group, the LTV of booked loans increases from 76.24 to 84.10, a 10.32% increase, suggesting that loan officers were booking riskier loans. Table 3c shows a noticeable increase in the number of booked loans secured by collaterals after year 2005 in the treated group. The average percentage of booked loans without collateral goes down by 13%, a 55% drop compared to the average percentage prior to the implementation of the incentive plan, whereas the pool of applicants without collateral does not change. Berger and Udell (1990) find that collateral is associated with ex-ante observably riskier borrowers and riskier loans. The increase in the percentage of secured loans adds to the evidence that loan officers are approving loans from riskier borrowers. On the other hand, Table 3b shows that the average business scores and personal scores of approved loans go up in year 2005, and the internal risk ratings go down. Since the internal risk ratings reflect a large amount of soft information possessed by loan officers, this implies that loan officers have been identifying safer borrowers since the implementation of incentive compensation (also see Agarwal and Hauswald (2007, 2008)). 4.2 Who respond to incentives and when? 18 As suggested in the literature, incentive compensation may have a sorting effect of attracting more able workers (Lazear (2000)). Table 4a shows that the treated group attracts younger loan officers and more male loan officers after year 2005, who are likely to be more aggressive in their career paths – the average age of the loan officers in the treated group goes down from 41 to 37, and the percentage of males goes up from 68% to 74%. There is also evidence of higher turnover in the treated group, as it is reflected by lower average tenure.9 Indeed, in Table 5a, we find that loan officers in the age group 25-34 are most aggressive at approving loans, while they have the lowest default rates among all age groups. Consistent with our model, achieving a higher approval rate without sacrificing loan quality is possible if the loan officers put in more effort to investigate the loan quality. Thus, the evidence above suggests that younger loan officers, who are likely to have stronger career concerns, worker harder than older loan officers. This effect gets amplified by the incentive pay. We find similar evidence for male loan officers compared to female loan officers. Table 5b reports that after the implementation of the incentive pay, the gap in the approval rate between male and female loan officers goes up. While the female loan officers also approve more loans under the incentive pay, the default rate among these loans is much higher than in year 2004, and higher than their male colleagues. This evidence is consistent with the literature that female workers have short careers, and thus less career concerns than males. In the context of our model, such loan officers are most 9 Our results are both qualitatively and quantitatively very similar if we have a constant pool of loan officers before and after the treatment period. 19 likely to have distorted incentives under the incentive plan, in the sense that loan quality deteriorates the most. We find further evidence that loan officers’ career concern becomes an important disciplinary device under the incentive pay in our multivariate analysis. A less studied question is the response time of the agents to incentive contracts. Do agents respond to incentives immediately, as assumed in the theoretical literature on incentive contracts, or is there a learning curve? In other words, if incentive contracts leave rooms for agents to game the compensation system, will agents respond to it as predicted by the rational theory immediately? The analysis of this question also provides a robustness check for our results. That is, was the information leaked to the loan officers prior to the actual implementation of the incentive pay in January 2005 so that loan officers may hold back from approving loans before they can receive the bonus linked to those booked loans? If there was such an information leak and loan officers did hold back booking loans in year 2004, our results would be weakened. Table 6 provides a month by month break-down of the status of loan applications. We plot the loan approval status in Figure 1 and observe a significant increase in approval rate and a decrease in rejection rate since January 2005. Figure 2 shows that both the average dollar amount of booked loans and the percentage of loans booked increase since January 2005, and the structural change takes place in January 2005. Figures 3, 4 and 5 show similar structural breaks in January 2005 for LTV, days-spentper-loan-application, and internal risk rating. 20 In summary, we find evidence that loan officers respond immediately as predicted by economic theory to the incentive pay, and we find no evidence that loan officers hold back approving loans prior to January 2005. 4.3 Does an incentive pay induce perverse incentives? Much of the study in the literature is on whether agents’ responses to incentive pay are in the firm’s interest. Our analysis and preliminary empirical results above show that incentive pay may create perverse incentives to loan officers by encouraging overbooking of inferior loans, especially for larger and longer maturity loans, whereas loan officer’s careern concerns serve as a good disciplinary device. We provide further evidence with a multivariate analysis. 4.3.1 Multivariate Analysis We examine whether during the treatment period, (i) the treated loan officers are more likely to approve or decline loan applications; and (ii) the booked loans are more likely to default. We employ the standard logit model specification to estimate these models. Our results reveal that loan officers’ inputs of internal risk ratings, LTV of the loans, loan amounts, and collateral are important for loan officers’ approval decisions. Table 7a shows that these variables are statistically significant and marginally important for loan approvals. Consistent with our intuition, riskier loans and larger loans are less likely to be approved, whereas collateral requirements increase loan approval rates. Moreover, we see that the variable, Treated Dummy*2005 Dummy is significantly 21 positive and marginally large, indicating that the implementation of the new incentive package in the treated group increases loan approval rates. While larger loans possess risk of a greater loss, the implementation of the incentive plan encourages loan officers to book larger loans, consistent with our model’s prediction; we see in Table 7a that the variable, Log(loan amount requested)*treated*2005 is significantly positive. Similarly, loan maturity*treated*2005 becomes significantly positive, indicating that loan officers in the treated group are more likely to book longer maturity loans as predicted by the model. We further analyze the subsequent loan performance of approved loans by examining the default probability of the loans based on loan characteristics. The results are reported in Table 7b. We confirm that internal risk ratings, LTV of the loan, loan amount requested, and loan maturity are important factors in predicting loan defaults. Collateral requirements, however, decrease the probability that a loan defaults, consistent with the moral hazard view of collateral requirements. In addition, we also find that Days Spent per Loan is negative, suggesting that the longer a loan officer spends on a loan application, the less likely it will default. We can interpret the number of days spent on the loan application as a measure of the loan officer’s effort to investigate loan quality. The harder the loan officer works, the less likely it is that an approved loan will default. This variable becomes especially important after the implementation of the incentive plan. Furthermore, we see evidence that loan officers in the treated group are booking larger and longer maturity loans that are riskier and more likely to default. The variables Loanto-Value of the Loan * treated * 2005 and Loan maturity * treated * 2005 are both significantly positive. 22 We further study loan officer’s fixed effect on loan approval and default rates. Table 8 summarizes the results from logit regressions of loan approval decisions and defaults on loan officer’s characteristics. Consistent with our prior findings, internal risk ratings, LTV of the loan, loan amount requested, loan terms, and collateral requirements are the key risk factors that drive approval decisions and predict subsequent loan defaults. Moreover, Treated Dummy*2005 Dummy is highly significant in both regressions, suggesting that loan officers in the treated group are more likely to book loans in year 2005, and that these approved loans are more likely to default. Our detailed micro-level information on loan officers allows us to study questions such as how the incentive plan interacts with loan officers’ career concerns. Our model indicates that loan officers’ career concerns serve as a powerful control mechanism that mitigates the distortion of incentives caused by monetary bonus. A loan officer with greater career concern will be more conservative in making loan approval decisions. We find evidence of this from the results in Table 8. Using loan officers’ ages and number of years on the job (tenure) as proxies for career concerns, we see that loan officers’ career concerns become significant after the implementation of the incentive plan in year 2005. The career concerns are insignificant on their own, but become significant after interacting with the treated dummy and the year 2005 dummy. Following previous literature, we argue that career concern is strongest when a person is just starting her career, thus tenure measures the reverse strength of career concern. We find that the marginal effects of Loan officer tenure * treated dummy * 2005 dummy and Loan officer tenure (sq) * treated dummy * 2005 dummy are 7.24% and 3.98%, respectively. That is, controlling for a loan officer’s age, the fewer years on the job, the less likely that she 23 books a loan. Interestingly, tenure does not predict default probability linearly. We see that the marginal effect of Loan Officer Tenure * Treated Dummy * 2005 Dummy on loan default probability is 6.77%. The positive number is consistent with our findings of loan officers’ approval decisions that the longer the tenure, the smaller the career concern, and thus, the more likely that the loan officer books riskier loans motivated by monetary incentives. Loan Officer Tenure squared, however, has the opposite effect in predicting loan defaults. Loan Officer Tenure (sq), Loan Officer Tenure (sq) * Treated Dummy, and Loan Officer Tenure (sq) * Treated Dummy * 2005 Dummy have negative marginal effects. In particular, the marginal effect of Loan Officer Tenure (sq) * Treated Dummy * 2005 Dummy is -1.91%, much larger than the other two, confirming that loan officers’ tenure is an important factor in loan approval decisions after the implementation of the incentive package. We interpret this as a “learning-on-the-job” effect. The longer the loan officer is on the job, the more experience she gains on detecting loan quality, thus, the lower the likelihood of booking a bad loan. This learning effect, however, becomes important only when the time on the job is sufficiently long. We also observe that Days Spent Per Loan*Treated Dummy * 2005 Dummy is marginally important for both the loan approval decision and loan default probability. We interpret Days Spent per Loan as a proxy for the loan officer’s effort to assess loan quality. We see that the longer the time spent reviewing the loan application, the less likely that it will be approved, and the less likely that the loan will default. In addition, the effect of this variable is large only after the implementation of the incentive plan among the loan officers in the treated group. 24 4.3.2 Soft-information vs. hard-information lending Our theoretical analysis suggests that whether incentive pay induces perverse incentives for loan officers depends crucially on the type of lending regime, soft-or hardinformation lending. Hard information contained in a loan application is captured by its observable risk factors, such as Experian scores, LTV, loan amount, loan terms, and maturity, whereas the internal risk rating contains loan officers’ subjective input, much of which is soft information. Following Agarwal and Hauswald (2008), we capture the residual soft information collected by the bank by orthogonalizing the internal risk rating with the above set of publicly available information controlling for branch, year, and industry effects. Hence, we estimate the following regression: IntRiskRating i = β 0 + β 1 XCI i + β 2 XPARi + u i , where XCI contains the Experian business score and personal score, and XPAR includes other publicly available risk factors such as LTV, loan amount, loan terms, loan maturity, and personal and business collateral. We refer to the residual from the above regression as “Internal Risk Rating Residual.” We then estimate the logit regression as in the previous sections by replacing Internal Risk Rating with the residual from the above regression. The results are reported in Table 9. We see that while the residual information itself is not significant in predicting default, it is significant for the treated group in year 2005. This suggests that under the incentive plan, loan officers book riskier loans that contain more soft information. This observation is consistent with the model’s prediction on the interaction between incentive compensation and soft information lending. 25 Furthermore, we form quintiles based on the size of the Internal Risk Rating Residual for both the treated group and the control group in year 2004 and 2005, with the top quintile containing loans with the largest residual, or the greatest amount of soft information. Our theoretical analysis suggests that under the incentive pay, loan officers are more likely to make reckless approval decisions for loans that contain more soft information. For these loans, both the approval rate and the default rate are higher than those with mostly hard information. Indeed, Table 10 reports the ratio of the approval rates and the default rates of the loans in the highest quintile to those in the lowest quintile. During the treatment period, loans that contain the greatest amount of soft information are 2.36 more likely to be approved than those that contain the least amount of soft information. Moreover, these loans are 3.05 times more likely to default than those with the least amount of soft information. 4.3.3 Welfare Analysis Finally, did the bank profit from the incentive pay? Although there are more defaults, loan officers are indeed booking more loans, of which the bank can make a profit from the fees. To answer this question, we carry out a simple welfare analysis. Welfare = Volume × fees – Wages – Loss given default – other unobservables Here the unobservables include the externalities of not making a loan, the cost of funding for the bank, and utility loss by extending additional effort. The marginal welfare, therefore, is ∆ welfare = increased volume × fees – increased wages – ∆ Loss given default 26 Table 4a shows that the average income of loan officers in the treated group increases by $6,597 from $42,422 to $49,019 from year 2004 to 2005. Under the incentive plan, $6,500 is amount of bonus that a loan officer gets if she reaches 100% of the performance goal (see Table 2). This suggests that the 100% goal may create a focal point for loan officers to aim for. Following industry standards, fees are assumed to be 2% of the loan originated and the loss given default is assumed to be 50% (see, Agarwal et. al. 2007). Table 11 reports the marginal welfare of the incentive pay. We see that the bank experiences a loss of $6,880,446 in year 2005. The program was eventually discontinued in the first quarter of 2006 due to losses to the bank. 5. Conclusion A central question addressed by much research on incentive compensation has been whether incentive contracts provide the right incentives. In this paper, we propose a model that studies two tasks that loan officers perform in the loan origination process: their efforts to investigate loan quality and their loan origination decisions. Our model demonstrates that monetary incentives may distort loan officers’ incentives to identify bad loans. The distortion is greater under a soft-information lending regime. Loan officers’ career concerns serve as a good disciplinary device to mitigate the agency problem. Using a unique dataset from a major national commercial bank that implemented an incentive compensation package for half of its loan officers, we are able to test many of the model’s predictions. We find that observable risk factors such as Experian scores, 27 LTV, loan amount, term, maturity, and collateral are important for loan officers’ approval decisions and predict subsequent loan defaults. Moreover, the internal risk ratings, which contain a large amount of soft information, also predict defaults. Motivated by the incentive package, loan officers in the treated group book more loans and, in particular, book larger and longer maturity loans, consistent with the model’s predictions. These larger and longer maturity loans are more likely to default within two years of loan origination, indicating that these are riskier loans on average. Using loan officers’ age and tenure as proxies for their career concerns, we find that loan officers with greater career concerns are more conservative in their approval decisions, and their booked loans have lower default rates. Also consistent with the model, we find evidence that loans with a greater amount of soft information are more likely to be approved under the incentive scheme. These loans, however, are also more likely to default. Our research suggests that hardening the soft information used in lending decisions will reduce distortion of incentives of piece-rate contracts. Moreover, counteracting incentives with more stringent lending standards may also reduce some of the agency problems. 28 References Agarwal, S., B. Ambrose, S. Chomsisengphet, and C. Liu, 2007, “An Empirical Analysis of Home Equity Loan and Line Performance,” Journal of Financial Intermediation, 2006, 15(4), 444-469 Agarwal, S., S. Chomsisengphet, and C. Liu, 2007, “Determinants of Small Business Default,” in Christodoulakis, G., and S. Satchell (eds.), The Analytics of Risk Model Validation, Palgrave-Macmillan Publishing, 1-12 Agarwal, S., and R. Hauswald, 2007, “Distance and Information Asymmetries in Lending,” Working Paper, American University Agarwal, S., and R. Hauswald, 2008, “The Choice Between Arms-Length and Relationship Debt Evidence from E-Loans,” Working Paper, American University Bebchuck, L. A., and H. Spamann, 2009, “Regulating bankers' pay,” Working Paper, John M. Olin Center for Law, Economics, and Business, Harvard University Bakes, G., 1992, "Incentive Contracts and Performance Measurement," Journal of Political Economy, 1992, 100(3), 598-614 Baker, G., 2000, “The Use of Performance Measures in Incentive Contracting,” The American Economic Review, 90(2), 415-420 Baker, G., 2002, “Distortion and Risk in Optimal Incentive Contracts,” The Journal of Human Resources, 37(4), 728-751 Berger, A., and G. Udell, 2002, “Small Business Credit Availability and Relationship Lending: The Importance of Bank Organizational Structure,” Economic Journal, 112, 3253 Fahlenbrach, R., and R. Stulz, 2009, “Bank CEO Incentives and the Credit Crisis,” Working Paper, Ohio State University. Gibbons, R., 1998, "Incentives in Organizations," Journal of Economic Perspectives, 12(4), 115-32 Holmstrom, B., and P. Milgrom. 1991. “Multitask Principal Agent Analyses: Incentive Contracts, Asset Ownership and Job Design,” J. Law, Econ., Organ., 7: special issue, pp. 24-52. Inderst, R., 2008, “Loan Origination under Soft- and Hard-Information Lending,” working paper. Lazear, E. P., 1986, “Salaries and Piece Rates,” The Journal of Business, 59(3), 405-431 29 Lazear, E. P., and Rosen, S., 1981, “Rank-order tournaments as optimum labor contracts,” Journal of Political Economy, 80, 841-64 Paarsch, Harry and Bruce Shearer. 1996. “Fixed Wages, Piece Rates, and Incentive Effects,” Petersen, M. (2004), “Information: Hard and Soft,” mimeo, Northwestern University. Stiglitz, J. 1981. “Contests and cooperation: Towards a general theory of compensation and competition.” Unpublished manuscript, Princeton, N.J.: Princeton University, April Udell, G. 1989. “Loan Quality, Commercial Loan Review and Loan Officer Contracting,” Journal of Banking and Finance, 13(3), 367-82 30 Table 1: Performance Metric Metric Weighting Annual Goal Pull-through yield 50% 48% of new money applications booked based on applications received from January 1November 30, 2005. Decision Points 25% 1,080 points* Application to decision time (% 25% 68.5%** met) *Decision points are allocated as follows: (i) Score + (all products) = 1 point ; (ii) S/L – basic (term $500M - $1MM) = 2 points; (iii) S/L (term $1 - $3MM, lines of credit < $750M) = 3 points; (iv) S/L – complex (term > $3MM, lines of credit > $750M) = 5 points; (v) Letters of credit (S/L) = 2 points; (vi) Commercial card (S/L) = 2 points **Decision time guidelines are as follows: (i) Score + guideline is 3 days; (ii) S/L guideline is 5 days Table 2: Incentive Plan Total Score Incentive award Less than 80% of goal No award 80% of goal $4,000 + $125 per percentage point above 80% of goal 100% of goal 120% of goal $ 6,500 + $150 per percentage point above 100% of goal $ 9,500 + $175 per percentage point above 120% of goal Notes: A brief description of the incentive plan that outlines the score achievement and incentive award for each score band. 31 Table 3a: Descriptive Statistics of Loan Applications – Loan Status Variable Number of Loan Requests Number of Loans Booked Approval Rate Number of Defaults Default Rate Avg $ of Loans Requested Avg $ of Loans Booked Days Spent/Loan Requested 2004 (January - December) Control Group Treated Group Mean Std. Mean Std. 6920 7996 2192 2548 31.68% 31.87% 91 107 4.15% 4.20% $ 455,240 $ 336,805 $ 426,480 $ 378,698 $ 224,614 $ 279,361 $ 216,048 $ 229,403 1.38 0.85 1.35 0.70 2005 (January - December) Control Group Treated Group Mean Std. Mean Std. 7564 7788 2744 3680 36.28% 47.25% 119 192 4.34% 5.22% $ 454,141 $ 369,635 $ 444,137 $ 381,829 $ 253,219 $ 257,801 $ 312,518 $ 404,976 1.32 0.75 1.06 0.53 Table 3b: Descriptive Statistics of Loan Applications – Risk Profile Variable Internal Risk Ratings Business Score of Applicants Business Score of Booked Loans Personal Score of Applicants Personal Score of Booked Loans LTV of Applicants LTV of Booked Loans 2004 (January - December) Control Group Treated Group Mean Std. Mean 5.23 1.84 5.38 1.52 200.86 72.23 195.88 75.87 184.87 68.95 186.11 78.92 731.85 70.31 725.41 68.06 716.69 87.44 718.90 88.58 61.28 43.00 65.30 44.03 72.99 31.48 76.24 30.90 32 2005 (January - December) Control Group Treated Group Std. Mean Std. 5.44 1.3 4.93 1.53 195.99 75.27 200.36 68.47 185.50 93.09 196.09 87.01 725.91 74.39 728.06 76.72 719.54 98.25 725.77 66.51 65.16 46.87 63.05 43.48 74.90 33.10 84.10 50.10 Table 3c: Descriptive Statistics of Loan Applications – Collaterals Variable Avg % of Applicants with Personal Collateral Avg % of Applicants with Business Collateral Avg % of Applicants with No Collateral Avg % of Booked Loans with Personal Collateral Avg % of Booked Loans with Business Collateral Avg % of Booked Loans with No Collateral 2004 (January - December) Control Group Treated Group Mean Std. Mean 69% 46% 68% 47% 25% 44% 26% 44% 5% 23% 6% 24% 9% 35% 7% 26% 63% 48% 68% 47% 27% 42% 25% 44% 2005 (January - December) Control Group Treated Group Std. Mean Std. 64% 48% 70% 46% 28% 45% 24% 43% 8% 28% 6% 24% 4% 20% 19% 29% 67% 47% 69% 49% 28% 45% 11% 46% Notes: Panels a, b, and c outline the loan statistics, risk profile, and collateral requirements for the control and treated samples during and control and treatment time period. 33 Table 3d: Descriptive Statistics of Loan Applications – Loan Size and Maturity Treated Group: Loan Approval and Performance % approval Loan Size / Type Big loan (> $700K) 2004 % rejection % walk away % approval 2005 % rejection % walk away 35% 57% 8% 55% 36% 8% Medium loan ($250K-$700K) 31% 56% 13% 49% 40% 11% Small loan (< $250K) 27% 55% 18% 33% 51% 16% Long term loan (Larger than One Year) 32% 55% 13% 52% 36% 12% Short term loan (One Year) 29% 58% 13% 37% 51% 12% Treated Group: Risk Profile, Collateral Loan Size / Type Big loan (> $700K) Personal Collateral 2004 Risk Score LTV Personal collateral 2005 Risk Score LTV 4 717 77 20 713 79 Medium loan ($250K-$700K) 4 720 74 19 720 84 Small loan (< $250K) 4 721 72 19 725 89 Long term loan (Larger than One Year) 4 721 74 20 717 81 Short term loan (One Year) 4 717 72 19 720 90 34 Table 3d: Statistics of Loan Applications – Loan Size and Maturity (Con’t) Control Group: Approval and Performance % approval Loan Size / Type Big loan (> $700K) 2004 % rejection % walk away % approval 2005 % rejection % walk away 35% 60% 8% 34% 59% 8% Medium loan ($250K-$700K) 31% 56% 13% 30% 57% 13% Small loan (< $250K) 27% 56% 18% 26% 57% 17% Long term loan (Larger than One Year) 32% 56% 13% 31% 57% 13% Short term loan (One Year) 29% 59% 13% 29% 58% 13% LTV Control Group: Risk Profile, Collateral Loan Size / Type Big loan (> $700K) Personal Collateral 2004 Risk Score LTV Personal collateral 2005 Risk Score 4 737 74 4 724 73 Medium loan ($250K-$700K) 4 751 73 4 723 71 Small loan (< $250K) 4 722 72 4 724 70 Long term loan (Larger than One Year) 4 727 71 4 752 73 Short term loan (One Year) 4 739 70 4 730 69 Notes: Panels d outline the loan approval, performance, and collateral requirements for the control and treated samples during and control and treatment time period for a given loan size and type. 35 Table 4a: Demographics of Loan Officers 2004 Variable Total Number of Loan Officers Gender – Male Income Age Tenure 2005 Control Group 63 58% $42,363 43 3.49 Treated Group 70 68% $42,422 41 3.66 Control Group 65 59% $42,976 43 3.58 Treated Group 65 74% $49,019 37 2.91 Table 4b: Percentage of Loan Officers in Each Age Group 2004 2005 Loan Officer Age Control Group Treated Group Control Group Treated Group 25-34 35-44 45-55 55+ 22.33 24.98 31.67 21.02 26.92 29.07 24.66 19.35 26.90 24.30 27.00 21.80 33.04 37.58 21.04 8.34 Notes: Panels a and b outline the loan officer demographics – gender, income, age, tenure, and fraction of loan officers by age groups for the control and treated samples during and control and treatment time period for a given loan size and type. 36 Table 5a: Loan Status in the Treated Group for Each Age Group – 2004 Loan Officer Age % of Loan Officers Approval Rate Loan Size Credit Scores Default Rate 25-34 26.92 34.91 142,029 5.32 3.46 35-44 29.07 33.01 166,083 5.37 4.36 45-55 24.66 29.78 211,327 5.4 4.84 55+ 19.35 26.00 232,022 5.37 4.46 Table 5a: Loan Status in the Treated Group for Each Age Group – 2005 Loan Officer Age % of Loan Officers Approval Rate Loan Size Credit Scores Default Rate 25-34 33.04 56.08 229,116 5.04 4.73 35-44 37.58 52.70 244,892 4.97 4.74 45-55 21.04 43.46 328,117 4.89 5.99 55+ 8.34 40.01 387,727 4.81 6.58 Notes: Panel a outline the fraction of loan officers by age groups, the approval rates, loan sizes, credit scores, and default for the control and treated samples during and control and treatment time period for a given loan size and type. 37 Table 5b: Percentage bookings in the treated group by gender groups – 2004 Loan Officer Gender % of Loan Officers Approval Rate Loan Size Credit Scores Default Rate Male Female 68.40 31.60 32.28 28.41 193,092 185,029 5.27 5.39 4.51 3.71 Table 5b: Percentage bookings in the treated group by gender groups – 2005 Loan Officer Gender % of Loan Officers Approval Rate Loan Size Credit Scores Default Rate Male 74.30 51.19 299,101 5.03 5.21 Female 25.70 40.27 280,583 4.79 5.26 Notes: Panel b outline the fraction of loan officers by gender groups, the approval rates, loan sizes, credit scores, and default for the control and treated samples during and control and treatment time period for a given loan size and type. 38 Table 6: Monthly Loan Approval Status Number of Loans in the Treated Group Months Jan-04 Feb-04 Mar-04 Apr-04 May-04 Jun-04 Jul-04 Aug-04 Sep-04 Oct-04 Nov-04 Dec-04 Jan-05 Feb-05 Mar-05 Apr-05 May-05 Jun-05 Jul-05 Aug-05 Sep-05 Oct-05 Nov-05 Dec-05 Received 548 582 688 679 747 742 759 639 618 649 692 653 584 593 638 531 764 783 662 642 643 635 688 625 Rejected 380 346 354 344 342 344 370 313 401 389 411 416 262 243 204 276 316 268 249 289 255 258 297 289 69.34% 59.45% 51.45% 50.66% 45.78% 46.36% 48.75% 48.98% 64.89% 59.94% 59.39% 63.71% 44.86% 40.98% 31.97% 51.98% 41.36% 34.23% 37.61% 45.02% 39.66% 40.63% 43.17% 46.24% Withdraw 76 133 71 92 75 83 76 88 54 107 84 99 93 74 71 73 57 66 61 74 75 75 87 96 13.87% 22.85% 10.32% 13.55% 10.04% 11.19% 10.01% 13.77% 8.74% 16.49% 12.14% 15.16% 15.92% 12.48% 11.13% 13.75% 7.46% 8.43% 9.21% 11.53% 11.66% 11.81% 12.65% 15.36% Booked 92 103 263 243 330 315 313 238 163 153 197 138 229 276 363 182 391 449 352 279 313 302 304 240 Number of Loans in the Control Group 16.79% 17.70% 38.23% 35.79% 44.18% 42.45% 41.24% 37.25% 26.38% 23.57% 28.47% 21.13% 39.21% 46.54% 56.90% 34.27% 51.18% 57.34% 53.17% 43.46% 48.68% 47.56% 44.19% 38.40% 39 Received 532 531 538 520 655 644 632 570 553 568 604 573 574 599 637 645 630 636 604 591 683 639 692 634 Rejected 252 327 386 258 217 323 391 301 334 283 371 261 311 310 345 335 394 333 280 353 284 337 258 378 47.37% 61.58% 71.75% 49.62% 33.13% 50.16% 61.87% 52.81% 60.40% 49.82% 61.42% 45.55% 54.18% 51.75% 54.16% 51.94% 62.54% 52.36% 46.36% 59.73% 41.58% 52.74% 37.28% 59.62% Withdraw 103 78 86 102 86 78 79 79 89 88 67 89 56 83 98 73 52 91 93 66 87 68 60 75 19.36% 14.69% 15.99% 19.62% 13.13% 12.11% 12.50% 13.86% 16.09% 15.49% 11.09% 15.53% 9.76% 13.86% 15.38% 11.32% 8.25% 14.31% 15.40% 11.17% 12.74% 10.64% 8.67% 11.83% Booked 177 126 66 160 352 243 162 190 130 197 166 223 207 206 194 237 184 212 231 172 312 234 374 181 33.27% 23.73% 12.27% 30.77% 53.74% 37.73% 25.63% 33.33% 23.51% 34.68% 27.48% 38.92% 36.06% 34.39% 30.46% 36.74% 29.21% 33.33% 38.25% 29.10% 45.68% 36.62% 54.05% 28.55% Table 7a: Loan Acceptance Decisions based on loan characteristics Variable Coeff Val. t-stats Marg Eff Intercept -4.0768 -2.99 Internal Risk Ratings -0.3046 -2.92 -2.93% Experian Business Score 0.2719 16.57 Experian Borrowers Score 0.1238 Loan-to-Value of the Loan Marg Eff Coeff Val. t-stats -3.7241 -2.84 ** -0.2837 -2.97 -2.89% ** 0.27% ** 0.2641 16.82 0.25% ** 13.31 0.30% ** 0.1188 13.36 0.30% ** -0.0373 -2.32 -4.06% ** -0.0344 -2.25 -3.92% ** log(Loan Amount Requested) -0.0406 -2.07 -5.39% ** -0.0395 -2.17 -5.15% ** Loan Term -0.0046 -5.14 -6.49% ** -0.0042 -4.87 -5.99% ** Loan maturity 0.6106 0.92 0.12% 0.6082 0.98 0.12% Treated Dummy 0.6479 0.99 5.21% 0.6124 1.00 4.96% 2005 Dummy 0.7218 1.07 1.20% 0.6757 1.05 1.13% Treated Dummy*2005 Dummy 0.7825 4.36 12.66% 0.7109 4.33 12.29% Days Spent Per Loan 0.5733 0.87 0.25% 0.5309 0.85 0.24% ** Internal Risk Ratings * treated Dummy -0.2282 -0.50 0.13% -0.2210 -0.52 0.13% Experian Business Score * Treated Dummy 0.4988 1.55 -0.18% 0.4945 1.59 -0.18% Experian Borrowers Score* Treated Dummy 0.0882 0.26 0.01% 0.0828 0.25 0.01% Loan-to-Value of the Loan* Treated Dummy -0.7004 -1.70 0.01% -0.6704 -1.79 0.01% log(Loan Amount Requested)* Treated Dummy -0.5060 -1.34 -0.05% -0.4795 -1.35 -0.05% Loan Term* Treated Dummy -1.1192 -1.83 -0.03% * -1.0905 -1.91 -0.03% Loan maturity* Treated Dummy 0.4422 1.75 0.08% * 0.4080 1.66 ** 0.08% * Days Spent Per Loan* Treated Dummy 0.1436 1.40 0.28% 0.1390 1.44 -0.1570 -0.26 -0.11% -0.1492 -0.24 * 0.28% Internal Risk Ratings * Treated * 2005 * -0.11% Experian Business Score * Treated* 2005 0.4035 0.97 0.21% 0.3675 0.94 0.19% Experian Borrowers Score* Treated * 2005 0.2342 0.49 0.25% 0.2110 0.45 0.23% Loan-to-Value of the Loan* Treated * 2005 -0.4229 -1.47 -0.29% -0.3949 -1.41 -0.28% log(Loan Amount Requested)* Treated * 2005 0.7490 3.57 2.92% 0.7322 3.82 2.87% Loan Term* Treated * 2005 -0.1440 -1.09 -0.08% -0.1320 -1.09 -0.07% Loan maturity* Treated * 2005 0.9435 3.81 5.53% ** 0.8917 3.75 5.19% ** Days Spent Per Loan* Treated * 2005 1.7321 6.20 0.06% ** 1.5733 6.24 0.05% ** Personal Collateral 0.5634 3.13 6.41% ** 0.5372 3.11 5.99% ** Business Collateral 0.5669 3.59 3.76% ** 0.5575 3.58 3.52% ** Personal Collateral*Treated Dummy 0.1743 1.41 0.30% 0.1720 1.40 0.29% ** Business Collateral*Treated Dummy 0.2528 1.64 0.25% 0.2296 1.51 0.25% Personal Collateral * Treated Dummy*2005 Dummy 0.1785 1.56 0.37% 0.1697 1.51 0.36% Business Collateral*Bank A Dummy*2005 Dummy 0.1726 1.32 0.26% 0.1590 1.22 0.25% SIC Dummy Yes Yes Loan Officer Dummy No Yes Number of Observations 30268 R-Square 17.28% Notes: We report the coefficients, the Std Err, the T-stats and marginal effects for the decision to deny credit (Y = 1). We obtain the marginal effects by simply evaluating Pr xj at the regressors’ sample means and coefficient estimates . Since the probabilities of offering and denying credit sum to 1 the marginal effects for the decision to reject a loan application are simply the opposite of the reported ones. The pseudo-R2 is McFadden’s likelihood ratio index. 40 ** Table 7b: Probability of Loan Default on Loan Characteristics Variable Coeff Val. Std. Err. t-stats Marg Eff Coeff Val. Std. Err. t-stats Marg Eff Intercept -2.3794 0.9712 -2.45 -2.2942 0.9288 -2.47 Internal Risk Ratings 0.1784 0.0426 4.19 8.78% ** 0.1769 0.0395 4.47 8.72% ** Experian Business Score -0.0847 0.0110 -7.71 -0.44% ** -0.0780 0.0102 -7.60 -0.40% ** Experian Borrowers Score -0.0847 0.0066 -12.41 -0.59% ** -0.0809 0.0065 -12.53 -0.53% ** Loan-to-Value of the Loan 0.0517 0.0120 4.30 1.28% ** 0.0482 0.0118 4.09 1.25% ** log(Loan Amount Requested) 0.0289 0.0075 3.87 1.77% ** 0.0263 0.0068 3.88 1.73% ** Loan Term 0.0012 0.0007 1.75 0.03% * 0.0011 0.0006 1.81 0.03% * ** ** Loan maturity 0.4728 0.1168 4.05 6.79% Treated Dummy 0.0686 0.0716 0.96 0.33% 0.4510 0.1080 4.17 6.76% 0.0623 0.0691 0.90 0.32% 2005 Dummy 0.4781 0.1385 3.45 3.38% ** 0.4522 0.1286 3.51 3.07% ** Treated Dummy*2005Dummy 0.4274 0.1073 3.98 6.51% ** 0.4115 0.1059 3.88 5.87% ** Days Spent Per Loan -0.4869 0.1912 -2.55 -1.21% ** -0.4723 0.1796 -2.62 -1.13% ** ** ** Internal Risk Ratings * Treated 0.1269 0.0427 2.97 0.10% Experian Business Score * Treated -0.0235 0.0225 -1.04 -0.09% 0.1218 0.0393 3.10 0.09% -0.0235 0.0223 -1.00 -0.08% Experian Borrowers Score* Treated -0.5274 0.0531 -9.94 0.02% ** -0.4857 0.0479 -10.14 0.02% ** Loan-to-Value of the Loan* Treated -0.2573 0.0944 -2.27 -0.55% ** -0.2544 0.0913 -2.78 -0.50% ** log(Loan Amount Requested)* Treated -0.0205 0.2780 -0.07 0.00% -0.0186 0.2597 -0.07 0.00% Loan Term* Treated -0.0645 0.0780 -0.83 -0.01% -0.0583 0.0722 -0.80 -0.01% Loan maturity* Treated 0.6061 0.1711 3.54 3.34% 0.5781 0.1623 3.56 3.05% ** Days Spent Per Loan* Treated 0.0807 0.0587 1.38 0.01% 0.0760 0.0555 1.36 0.01% Internal Risk Ratings * Treated * 2005 0.2441 0.1755 1.39 0.08% 0.2261 0.1718 1.31 ** 0.08% Experian Business Score * Treated * 2005 -0.1005 0.1595 -0.63 -0.06% -0.0942 0.1485 -0.63 -0.05% Experian Borrowers Score* Treated * 2005 -0.3635 0.2916 -1.25 -0.94% -0.3404 0.2706 -1.25 -0.91% Loan-to-Value of the Loan* Treated * 2005 1.4476 0.1115 12.98 -1.82% ** 1.3847 0.1015 13.60 -1.68% ** log(Loan Amount Requested)* Treated * 2005 0.3568 0.1045 3.41 4.05% ** 0.3356 0.0966 3.47 3.64% ** Loan Term* Treated * 2005 0.3231 0.0986 3.28 4.40% ** 0.3066 0.0924 3.31 4.12% ** Loan maturity* Treated * 2005 0.8975 0.1839 4.88 9.30% ** 0.8788 0.1808 4.86 9.26% ** Days Spent Per Loan* Treated* 2005 -0.2368 0.0746 -3.17 -1.70% ** -0.2330 0.0739 -3.15 -1.56% ** Personal Collateral -1.5637 0.1571 -9.95 -4.91% ** -1.4954 0.1434 -10.42 -4.44% ** Business Collateral -1.8806 0.3037 6.19 -1.29% ** -1.7135 0.2884 -5.94 -1.17% ** Personal Collateral*Treated Dummy -0.2745 0.2659 -1.03 -0.16% -0.2549 0.2543 -1.00 -0.14% Business Collateral*Treated Dummy -0.1077 0.1029 -1.05 -0.33% -0.1007 0.0994 -1.01 -0.30% Personal Collateral*Treated Dummy*2005 Dummy -0.1132 0.1702 -0.67 -0.11% -0.1019 0.1571 -0.64 -0.10% Business Collateral*Treated Dummy*2005 Dummy -0.1453 0.3771 -0.39 -0.35% -0.1439 0.3409 -0.42 -0.32% SIC Dummy Yes Yes Yes Loan Officer Dummy No Number of Observations 11164 R-Square 7.99% Notes: We report the coefficients, the Std Err, the T-stats and marginal effects for the decision to default on the credit (Y = 1). We obtain the marginal effects by simply evaluating Pr xj at the regressors’ sample means and coefficient estimates . Since the probabilities of offering and denying credit sum to 1 the marginal effects for the decision to reject a loan application are simply the opposite of the reported ones. The pseudo-R2 is McFadden’s likelihood ratio index. 41 Table 8: Loan approval decisions and defaults Acceptance Variable Coeff Val. t-stats Intercept -4.1326 Default Coeff Val. Marg Eff t-stats -2.5701 -2.73 -2.37 Marg Eff Internal Risk Ratings -0.3311 -3.15 -3.25% ** 0.1811 3.93 9.30% ** Experian Business Score 0.2904 16.46 0.28% ** -0.0932 -7.70 -0.46% ** Experian Borrowers Score 0.1317 13.70 0.30% ** -0.0920 -13.08 -0.64% ** Loan-to-Value of the Loan -0.0407 -2.24 -4.45% ** 0.0529 4.16 1.32% ** log(Loan Amount Requested) -0.0415 -2.01 -5.63% ** 0.0305 3.99 1.85% ** Loan Term -0.0049 -5.28 -6.92% ** 0.0013 1.84 0.04% * Treated Dummy 0.6606 0.94 5.34% 0.0750 0.94 0.36% 0.5172 3.38 3.69% ** 0.4503 3.99 6.77% ** 2005 Dummy 0.7966 1.14 1.24% Treated Dummy*2005 Dummy 0.8250 4.21 13.02% Days Spent Per Loan 0.6328 0.92 0.25% -0.4917 -2.34 -1.23% ** Loan Officer Gender (Female) 1.0382 1.44 0.14% -0.6228 -12.00 -4.79% ** Loan Officer Age 0.4458 0.68 0.58% 0.1607 2.66 0.27% ** Loan Officer Age(sq) -0.5601 -0.82 -0.21% -0.0618 -1.36 -0.21% ** Loan Officer Tenure 0.4179 0.62 0.01% 0.7065 3.18 0.15% ** Loan Officer Tenure (sq) 0.9105 1.39 0.01% -0.5330 -3.67 -0.60% ** Days Spent Per Loan*Treated Dummy -0.7350 -1.06 -0.27% -0.0801 -0.18 -0.06% Loan Officer Gender (Female)*Treated Dummy -1.2078 -1.66 -0.07% -0.2054 -1.60 -0.01% Loan Officer Age*Treated Dummy 0.4993 0.69 0.15% 0.4904 2.78 0.76% Loan Officer Age(sq)*Treated Dummy -0.5918 -0.83 -0.60% -0.1099 -0.70 -0.10% ** Loan Officer Tenure*Treated Dummy 0.4399 0.56 0.79% 0.3096 2.09 0.16% ** Loan Officer Tenure (sq)*Treated Dummy 1.0422 1.35 0.44% -0.9549 -3.89 -0.41% ** Days Spent Per Loan*Treated Dummy*2005 Dummy Loan Officer Gender (Female)*Treated Dummy*2005 Dummy -1.3716 -1.87 -4.31% * -0.5589 -3.17 -2.25% ** -1.9498 -2.53 -8.09% ** -0.5634 -4.83 -2.68% ** Loan Officer Age*Treated Dummy*2005 Dummy 1.8456 2.62 6.13% ** 0.3650 2.36 2.41% ** Loan Officer Age(sq)*Treated Dummy*2005 Dummy -1.4044 -1.87 -5.26% * -0.2437 -1.15 -0.43% Loan Officer Tenure*Treated Dummy*2005 Dummy 2.4137 3.05 7.24% ** 0.9385 2.73 6.77% ** Loan Officer Tenure (sq)*Treated Dummy*2005 Dummy 1.9412 2.80 3.98% ** -0.7134 -5.68 -1.91% ** Personal Collateral 0.6172 11.71 7.00% ** -1.6722 -10.52 -5.41% ** Business Collateral 0.5948 14.73 3.97% ** -2.0781 -65.00 -1.35% ** Personal Collateral*Treated Dummy 0.1867 1.37 0.30% -0.2997 -1.00 -0.17% Business Collateral*Treated Dummy 0.2620 1.62 0.27% -0.1158 -1.06 -0.33% Personal Collateral*Treated Dummy*2005 Dummy 0.1961 1.60 0.37% -0.1217 -0.67 -0.11% Business Collateral*Treated Dummy*2005 Dummy 0.1857 1.35 0.28% -0.1531 -0.39 -0.38% SIC 2 Digit Dummies Yes Number of Observations 30268 R-Square Yes 18.93% Notes: We report the coefficients, the Std Err, the T-stats and marginal effects for the decision to deny credit (Y = 1). We obtain the marginal effects by simply evaluating Pr xj at the regressors’ sample means and coefficient estimates . Since the probabilities of offering and denying credit sum to 1 the marginal effects for the decision to reject a loan application are simply the opposite of the reported ones. The pseudo-R2 is McFadden’s likelihood ratio index. 42 Table 9: Probability of Loan Default with Soft/hard Information Variable Coeff Val. Std. Err. t-stats Marg Eff Coeff Val. Std. Err. t-stats -2.2298 0.9151 -2.43 0.0762 0.0384 1.92 0.0102 -7.49 Marg Eff Intercept -2.3506 0.9608 -2.44 Internal Risk Ratings Residual 0.0703 0.0423 1.66 0.55% Experian Business Score -0.0843 0.0107 -7.87 -0.43% ** -0.0763 Experian Borrowers Score -0.0820 0.0064 -12.88 -0.57% ** -0.0788 0.0065 Loan-to-Value of the Loan 0.0498 0.0116 4.31 1.24% ** 0.0467 0.0116 * -12.22 0.39% 0.40% 0.51% ** 4.01 1.21% ** ** log(Loan Amount Requested) 0.0278 0.0072 3.83 1.71% ** 0.0253 0.0067 3.80 1.72% ** Loan Term 0.0012 0.0007 1.78 0.03% * 0.0011 0.0006 1.80 0.03% * ** ** Loan maturity 0.4560 0.1148 3.97 6.75% Treated Dummy 0.0664 0.0692 0.96 0.32% 0.4366 0.1043 4.18 6.71% 0.0607 0.0686 0.88 0.31% 2005 Dummy 0.4622 0.1375 3.36 3.27% ** 0.4479 0.1265 3.54 3.06% ** Treated Dummy*2005Dummy 0.4220 0.1048 4.02 6.37% ** 0.4042 0.1042 3.87 ** -2.54 -1.18% ** -0.4550 0.1770 -2.57 5.76% 1.12% Days Spent Per Loan -0.4770 0.1875 Internal Risk Ratings Residual* Treated 0.1066 0.0425 2.51 3.10% ** 0.1001 0.0389 2.57 Experian Business Score * Treated -0.0228 0.0225 -1.01 -0.09% -0.0228 0.0222 -1.02 Experian Borrowers Score* Treated -0.5116 0.0530 -9.64 0.02% ** -0.4817 0.0463 -10.39 ** Loan-to-Value of the Loan* Treated -0.2480 0.0943 -2.62 -0.54% log(Loan Amount Requested)* Treated -0.0203 0.2680 -0.07 0.00% Loan Term* Treated -0.0640 0.0750 -0.85 0.5935 0.1668 3.55 3.31% 0.0895 -2.75 0.2576 -0.06 4.09% 0.08% ** 0.02% 0.50% ** ** -0.0580 0.0721 -0.80 0.00% 0.01% ** 0.5591 0.1601 3.49 2.93% 0.0760 0.0549 1.38 0.01% ** 0.1985 0.0692 2.87 ** -0.01% Loan maturity* Treated -0.2465 -0.0180 ** ** Days Spent Per Loan* Treated 0.0805 0.0566 1.42 0.01% Internal Risk Ratings Residual * Treated * 2005 0.1840 0.0532 3.46 0.08% Experian Business Score * Treated * 2005 -0.0968 0.1560 -0.62 -0.06% -0.0926 0.1430 -0.64 Experian Borrowers Score* Treated * 2005 -0.3560 0.2805 -1.26 -0.91% -0.3347 0.2641 -1.26 Loan-to-Value of the Loan* Treated * 2005 1.4024 0.1074 13.06 -1.78% ** 1.3457 0.0981 13.71 0.08% 0.05% 0.88% 1.65% log(Loan Amount Requested)* Treated * 2005 0.3429 0.1009 3.39 4.04% ** 0.3302 0.0958 3.44 3.62% ** Loan Term* Treated * 2005 0.3136 0.0956 3.28 4.37% ** 0.2970 0.0897 3.31 4.10% ** Loan maturity* Treated * 2005 0.8749 0.1828 4.78 9.18% ** 0.8773 0.1756 4.99 ** Days Spent Per Loan* Treated* 2005 -0.2274 0.0719 -3.16 -1.65% ** -0.2321 0.0724 -3.20 Personal Collateral -1.5421 0.1562 -9.87 -4.80% ** -1.4719 0.1416 -10.39 Business Collateral -1.8410 0.3001 -6.13 -1.28% ** -1.6782 0.2866 -5.85 Personal Collateral*Treated Dummy -0.2715 0.2566 -1.05 -0.15% -0.2501 0.2538 -0.98 Business Collateral*Treated Dummy -0.1038 0.1022 -1.01 -0.32% -0.0979 0.0994 -0.98 Personal Collateral*Treated Dummy*2005 Dummy -0.1099 0.1656 -0.66 -0.11% -0.1015 0.1537 -0.66 Business Collateral*Treated Dummy*2005 Dummy -0.1440 0.3753 -0.38 -0.34% -0.1427 0.3400 -0.41 9.12% 1.52% 4.32% 1.12% 0.14% 0.30% 0.10% 0.31% SIC Dummy Yes Yes Loan Officer Dummy No Yes Number of Observations 11164 R-Square 8.27% 43 ** ** ** ** Notes: We report the coefficients, the Std Err, the T-stats and marginal effects for the decision to default on credit (Y = 1). We obtain the marginal effects by simply evaluating Pr xj at the regressors’ sample means and coefficient estimates . Since the probabilities of offering and denying credit sum to 1 the marginal effects for the decision to reject a loan application are simply the opposite of the reported ones. The pseudo-R2 is McFadden’s likelihood ratio index. 44 Table 10: Comparison of Hard- vs. Soft-information Lending Approval rate Default rate Ratio of Highest/Lowest Quintile of Soft Info 2004 2005 2004 2005 Control group 1.27 1.29 1.16 1.12 Treated group 1.24 2.36 1.19 3.05 Table 11: Welfare Analysis 2005 2004 Remarks Average loan size $312,518 $216,048 # of loans booked 3680 2548 Increased volume $599,575,936 Average loan size × ∆ loans booked ∆ Fees generated $11,991,519 2% × increased volume Average income $49,019 Increased wages $428,805 # of defaults 192 107 Defaulted loans $60,003,456 $23,117,136 ∆ Loss given default $18,443,160 $42,422 ∆ Average income × # of loan officers Average loan size × # of defaults 50% × ∆ defaulted loans ∆ Welfare = ∆ Fees generated – Increased wages – ∆ Loss given default = -$6,880,446 45 46 Dec-05 Nov-05 O ct-05 Sep-05 Aug-05 Jul-05 Jun-05 M ay-05 Apr-05 M ar-05 Feb-05 Jan-05 Dec-04 Nov-04 O ct-04 Sep-04 Aug-04 Jul-04 Jun-04 M ay-04 Apr-04 M ar-04 Feb-04 Jan-04 M onths Dec-05 Nov-05 O ct-05 Sep-05 Aug-05 Jul-05 Jun-05 M ay-05 Apr-05 M ar-05 Feb-05 Jan-05 Dec-04 Nov-04 O ct-04 Sep-04 Aug-04 Jul-04 Jun-04 M ay-04 Apr-04 M ar-04 Feb-04 Jan-04 M onths Figure 1: Monthly Loan Approval Status Treated G roup 80.00% 70.00% 60.00% 50.00% 40.00% R ejected W ithdraw 30.00% Booked 20.00% 10.00% 0.00% Control Group 0.8 0.7 0.6 0.5 0.4 R ejected 0.3 W ithdraw Booked 0.2 0.1 0 Figure 2: Loan Applications Booked in the Treated Group $500,000 $450,000 $400,000 $350,000 $300,000 $250,000 $200,000 $150,000 $100,000 $50,000 $- Avg $ Loan Requested Avg $ Loans Booked Oct-04 Nov-04 Dec-04 Jan-05 Feb-05 Mar-05 Apr-05 Figure 3: Loan-to-Value Ratios 90.00 85.00 80.00 75.00 LTV of Applicants 70.00 LTV of Booked Loans 65.00 60.00 55.00 Oct-04 Nov-04 Dec-04 Jan-05 Feb-05 Mar-05 Apr-05 47 Figure 4: Days Spent on Loan Requested in the Treated Group Day Spent/Loan Requested s 1.60 1.40 1.20 1.00 D ays S pent/Loan R equested 0.80 0.60 0.40 0.20 0.00 O ct-04 Nov -04 Dec-04 Jan-05 Feb-05 M ar-05 A pr-05 Figure 5: Internal Risk Rating Internal Risk Ratings 5.4 5.2 5 4.8 Internal Risk Ratings 4.6 4.4 4.2 4 Oct-04 Nov-04 Dec-04 Jan-05 Feb-05 48 Mar-05 Apr-05 Working Paper Series A series of research studies on regional economic issues relating to the Seventh Federal Reserve District, and on financial and economic topics. U.S. Corporate and Bank Insolvency Regimes: An Economic Comparison and Evaluation Robert R. Bliss and George G. Kaufman WP-06-01 Redistribution, Taxes, and the Median Voter Marco Bassetto and Jess Benhabib WP-06-02 Identification of Search Models with Initial Condition Problems Gadi Barlevy and H. N. Nagaraja WP-06-03 Tax Riots Marco Bassetto and Christopher Phelan WP-06-04 The Tradeoff between Mortgage Prepayments and Tax-Deferred Retirement Savings Gene Amromin, Jennifer Huang,and Clemens Sialm WP-06-05 Why are safeguards needed in a trade agreement? Meredith A. Crowley WP-06-06 Taxation, Entrepreneurship, and Wealth Marco Cagetti and Mariacristina De Nardi WP-06-07 A New Social Compact: How University Engagement Can Fuel Innovation Laura Melle, Larry Isaak, and Richard Mattoon WP-06-08 Mergers and Risk Craig H. Furfine and Richard J. Rosen WP-06-09 Two Flaws in Business Cycle Accounting Lawrence J. Christiano and Joshua M. Davis WP-06-10 Do Consumers Choose the Right Credit Contracts? Sumit Agarwal, Souphala Chomsisengphet, Chunlin Liu, and Nicholas S. Souleles WP-06-11 Chronicles of a Deflation Unforetold François R. Velde WP-06-12 Female Offenders Use of Social Welfare Programs Before and After Jail and Prison: Does Prison Cause Welfare Dependency? Kristin F. Butcher and Robert J. LaLonde Eat or Be Eaten: A Theory of Mergers and Firm Size Gary Gorton, Matthias Kahl, and Richard Rosen WP-06-13 WP-06-14 1 Working Paper Series (continued) Do Bonds Span Volatility Risk in the U.S. Treasury Market? A Specification Test for Affine Term Structure Models Torben G. Andersen and Luca Benzoni WP-06-15 Transforming Payment Choices by Doubling Fees on the Illinois Tollway Gene Amromin, Carrie Jankowski, and Richard D. Porter WP-06-16 How Did the 2003 Dividend Tax Cut Affect Stock Prices? Gene Amromin, Paul Harrison, and Steven Sharpe WP-06-17 Will Writing and Bequest Motives: Early 20th Century Irish Evidence Leslie McGranahan WP-06-18 How Professional Forecasters View Shocks to GDP Spencer D. Krane WP-06-19 Evolving Agglomeration in the U.S. auto supplier industry Thomas Klier and Daniel P. McMillen WP-06-20 Mortality, Mass-Layoffs, and Career Outcomes: An Analysis using Administrative Data Daniel Sullivan and Till von Wachter WP-06-21 The Agreement on Subsidies and Countervailing Measures: Tying One’s Hand through the WTO. Meredith A. Crowley WP-06-22 How Did Schooling Laws Improve Long-Term Health and Lower Mortality? Bhashkar Mazumder WP-06-23 Manufacturing Plants’ Use of Temporary Workers: An Analysis Using Census Micro Data Yukako Ono and Daniel Sullivan WP-06-24 What Can We Learn about Financial Access from U.S. Immigrants? Una Okonkwo Osili and Anna Paulson WP-06-25 Bank Imputed Interest Rates: Unbiased Estimates of Offered Rates? Evren Ors and Tara Rice WP-06-26 Welfare Implications of the Transition to High Household Debt Jeffrey R. Campbell and Zvi Hercowitz WP-06-27 Last-In First-Out Oligopoly Dynamics Jaap H. Abbring and Jeffrey R. Campbell WP-06-28 Oligopoly Dynamics with Barriers to Entry Jaap H. Abbring and Jeffrey R. Campbell WP-06-29 Risk Taking and the Quality of Informal Insurance: Gambling and Remittances in Thailand Douglas L. Miller and Anna L. Paulson WP-07-01 2 Working Paper Series (continued) Fast Micro and Slow Macro: Can Aggregation Explain the Persistence of Inflation? Filippo Altissimo, Benoît Mojon, and Paolo Zaffaroni WP-07-02 Assessing a Decade of Interstate Bank Branching Christian Johnson and Tara Rice WP-07-03 Debit Card and Cash Usage: A Cross-Country Analysis Gene Amromin and Sujit Chakravorti WP-07-04 The Age of Reason: Financial Decisions Over the Lifecycle Sumit Agarwal, John C. Driscoll, Xavier Gabaix, and David Laibson WP-07-05 Information Acquisition in Financial Markets: a Correction Gadi Barlevy and Pietro Veronesi WP-07-06 Monetary Policy, Output Composition and the Great Moderation Benoît Mojon WP-07-07 Estate Taxation, Entrepreneurship, and Wealth Marco Cagetti and Mariacristina De Nardi WP-07-08 Conflict of Interest and Certification in the U.S. IPO Market Luca Benzoni and Carola Schenone WP-07-09 The Reaction of Consumer Spending and Debt to Tax Rebates – Evidence from Consumer Credit Data Sumit Agarwal, Chunlin Liu, and Nicholas S. Souleles WP-07-10 Portfolio Choice over the Life-Cycle when the Stock and Labor Markets are Cointegrated Luca Benzoni, Pierre Collin-Dufresne, and Robert S. Goldstein WP-07-11 Nonparametric Analysis of Intergenerational Income Mobility with Application to the United States Debopam Bhattacharya and Bhashkar Mazumder WP-07-12 How the Credit Channel Works: Differentiating the Bank Lending Channel and the Balance Sheet Channel Lamont K. Black and Richard J. Rosen WP-07-13 Labor Market Transitions and Self-Employment Ellen R. Rissman WP-07-14 First-Time Home Buyers and Residential Investment Volatility Jonas D.M. Fisher and Martin Gervais WP-07-15 Establishments Dynamics and Matching Frictions in Classical Competitive Equilibrium Marcelo Veracierto WP-07-16 Technology’s Edge: The Educational Benefits of Computer-Aided Instruction Lisa Barrow, Lisa Markman, and Cecilia Elena Rouse WP-07-17 3 Working Paper Series (continued) The Widow’s Offering: Inheritance, Family Structure, and the Charitable Gifts of Women Leslie McGranahan Demand Volatility and the Lag between the Growth of Temporary and Permanent Employment Sainan Jin, Yukako Ono, and Qinghua Zhang WP-07-18 WP-07-19 A Conversation with 590 Nascent Entrepreneurs Jeffrey R. Campbell and Mariacristina De Nardi WP-07-20 Cyclical Dumping and US Antidumping Protection: 1980-2001 Meredith A. Crowley WP-07-21 Health Capital and the Prenatal Environment: The Effect of Maternal Fasting During Pregnancy Douglas Almond and Bhashkar Mazumder WP-07-22 The Spending and Debt Response to Minimum Wage Hikes Daniel Aaronson, Sumit Agarwal, and Eric French WP-07-23 The Impact of Mexican Immigrants on U.S. Wage Structure Maude Toussaint-Comeau WP-07-24 A Leverage-based Model of Speculative Bubbles Gadi Barlevy WP-08-01 Displacement, Asymmetric Information and Heterogeneous Human Capital Luojia Hu and Christopher Taber WP-08-02 BankCaR (Bank Capital-at-Risk): A credit risk model for US commercial bank charge-offs Jon Frye and Eduard Pelz WP-08-03 Bank Lending, Financing Constraints and SME Investment Santiago Carbó-Valverde, Francisco Rodríguez-Fernández, and Gregory F. Udell WP-08-04 Global Inflation Matteo Ciccarelli and Benoît Mojon WP-08-05 Scale and the Origins of Structural Change Francisco J. Buera and Joseph P. Kaboski WP-08-06 Inventories, Lumpy Trade, and Large Devaluations George Alessandria, Joseph P. Kaboski, and Virgiliu Midrigan WP-08-07 School Vouchers and Student Achievement: Recent Evidence, Remaining Questions Cecilia Elena Rouse and Lisa Barrow WP-08-08 4 Working Paper Series (continued) Does It Pay to Read Your Junk Mail? Evidence of the Effect of Advertising on Home Equity Credit Choices Sumit Agarwal and Brent W. Ambrose WP-08-09 The Choice between Arm’s-Length and Relationship Debt: Evidence from eLoans Sumit Agarwal and Robert Hauswald WP-08-10 Consumer Choice and Merchant Acceptance of Payment Media Wilko Bolt and Sujit Chakravorti WP-08-11 Investment Shocks and Business Cycles Alejandro Justiniano, Giorgio E. Primiceri, and Andrea Tambalotti WP-08-12 New Vehicle Characteristics and the Cost of the Corporate Average Fuel Economy Standard Thomas Klier and Joshua Linn WP-08-13 Realized Volatility Torben G. Andersen and Luca Benzoni WP-08-14 Revenue Bubbles and Structural Deficits: What’s a state to do? Richard Mattoon and Leslie McGranahan WP-08-15 The role of lenders in the home price boom Richard J. Rosen WP-08-16 Bank Crises and Investor Confidence Una Okonkwo Osili and Anna Paulson WP-08-17 Life Expectancy and Old Age Savings Mariacristina De Nardi, Eric French, and John Bailey Jones WP-08-18 Remittance Behavior among New U.S. Immigrants Katherine Meckel WP-08-19 Birth Cohort and the Black-White Achievement Gap: The Roles of Access and Health Soon After Birth Kenneth Y. Chay, Jonathan Guryan, and Bhashkar Mazumder WP-08-20 Public Investment and Budget Rules for State vs. Local Governments Marco Bassetto WP-08-21 Why Has Home Ownership Fallen Among the Young? Jonas D.M. Fisher and Martin Gervais WP-09-01 Why do the Elderly Save? The Role of Medical Expenses Mariacristina De Nardi, Eric French, and John Bailey Jones WP-09-02 Using Stock Returns to Identify Government Spending Shocks Jonas D.M. Fisher and Ryan Peters WP-09-03 5 Working Paper Series (continued) Stochastic Volatility Torben G. Andersen and Luca Benzoni WP-09-04 The Effect of Disability Insurance Receipt on Labor Supply Eric French and Jae Song WP-09-05 CEO Overconfidence and Dividend Policy Sanjay Deshmukh, Anand M. Goel, and Keith M. Howe WP-09-06 Do Financial Counseling Mandates Improve Mortgage Choice and Performance? Evidence from a Legislative Experiment Sumit Agarwal,Gene Amromin, Itzhak Ben-David, Souphala Chomsisengphet, and Douglas D. Evanoff WP-09-07 Perverse Incentives at the Banks? Evidence from a Natural Experiment Sumit Agarwal and Faye H. Wang WP-09-08 6