«Alejandro Drexler and, Antoinette Schoar ∗ This version December, 2012 Abstract Using exogenous shocks to the relationship between borrowers and ...»
Do Relationships Matter? Evidence from Loan Officer
Alejandro Drexler and, Antoinette Schoar ∗
This version December, 2012
Using exogenous shocks to the relationship between borrowers and loan-oﬃcers,
we document that borrowers are less likely to receive new loans from the bank and are
more likely to apply for credit from other banks when their original loan oﬃcers are
absent. They also are more likely to miss payments or go into default. These eﬀects are more pronounced when turnovers are unexpected as in the case of sickness leaves or when oﬃcers do not have strong incentives to transfer information, e.g. terminated loan oﬃcers. However, when given the right situation, e.g. voluntary resignations of loan oﬃcers, it seems possible to transfer soft information between employees within the same institution.
Alejandro Drexler is at the McCombs School of Business at The University of Texas at Austin, An- ∗ toinette Schoar is at the Sloan School of Management at MIT, NBER, and ideas42. We are grateful for the comments and suggestions of Daniel Paravisini, Amiyatosh Purnanandam, Borja Larrain, Giorgo Sertsios, the participant in Texas Finance Festival, and the seminars at The University of Texas at Austin, University of Houston, and Universidad Cat´lica de Chile. We thank Bank Estado, especially Roxana Aravena, Jose o Luis Arriagada, Pablo Coto, Enrique Errazuriz, Carlos Hernandez, Soledad Ovando, Hector Pacheco, Oliver Prostran, Marco Sambuceti, German Texido, Emilio Velez, Victor Vera, and Pamela Zalduando for provid- ing the data and making us familiar with the internal HR processes of the bank. We also thank Manasee Desai, and Katherine Gordon for their help collecting and organizing the data. Finally we thank Bridgette Hayes, Jonathan Moore, and Rouzhna Nayeri for excellent work in editing the paper. All remaining errors are our own.
Electronic copy available at: http://ssrn.com/abstract=2144337 Introduction Credit to small and opaque borrowers often relies on extensive interaction between loan oﬃcers and the businesses they lend to. Loan oﬃcers perform the task of gathering soft, and often hard information about their clients to reduce the information asymmetry and the cost of lending to the bank. For examples of this, see Rajan (1992), Petersen and Rajan (1995), or Berger and Udell (2002). The positive eﬀect of relationship lending might be reinforced if clients develop personal loyalties with their loan oﬃcer, since these often help clients with many business decisions. One could imagine that borrowers are less likely to engage in moral hazard behavior if they feel a personal connection with the loan oﬃcer.1 However, on the downside, relying extensively on loan oﬃcers’ personal contacts with the borrowers might make them indispensable in the lending process. Stein (2002) or Berger, Miller, Petersen, Rajan, and Stein (2005) argue that soft information cannot easily be transferred within the bank and thus limits the organizational structure and the bank’s size.
In this paper, we test the impact of shocks to the relationship between loan oﬃcers and borrowers, by exploiting instances where loan oﬃcers are absent from their job for long periods of time. If soft information matters, and is diﬃcult to transfer between employees of the bank, we would expect that access to credit deteriorates when the original loan oﬃcer is absent. A replacement loan oﬃcer might initially rely more on hard information in her credit 1 See Schoar (2012) for evidence along these lines.
1 Electronic copy available at: http://ssrn.com/abstract=2144337 assessment and, as a result, ﬁrms with worse observable characteristics, like smaller and less proﬁtable ﬁrms, should see a bigger eﬀect from the switch. We use detailed transaction-level data on small business borrowers from BancoEstado, the largest public bank in Chile. These loans are issued as personal loans, and therefore without limited liability. However, de facto, it is very diﬃcult to seize any assets from these clients. Therefore, loans in this segment rely heavily on soft information. We obtain comprehensive data about the loan oﬃcers, their backgrounds and leaves, as well as the transaction details and repayment behavior of the clients in their portfolios.
We ﬁnd that the relationship between loan oﬃcers and their clients has ﬁrst-order eﬀects on the borrowers’ access to credit. If the original loan oﬃcer is absent, we observe an 18% drop in the unconditional probability that a client gets a new loan during that time period. When decomposing this drop into the application rate of the client and the approval probability of the bank, we see that not only does the approval rate drop by more than 6%, but also the rate at which clients apply for new loans falls by about 1%, which represents a 12% reduction in the unconditional probability of applying for a new loan. At the same time, we do not observe any signiﬁcant changes in credit terms after a loan oﬃcer leaves;
for example, interest rates and loan maturity are on average unchanged. However, there is a signiﬁcant increase in the probability that a client becomes delinquent or even defaults when the original loan oﬃcer is out. For example, clients in good standing increase their probability of becoming delinquent by 22% compared to the average probability of missing
default shoots up by 17% compared to the unconditional probability of defaulting.2 Finally, only 11% of clients who have been rejected for a loan by the replacement loan oﬃcer are able to borrow from the outside loan market. These ﬁndings suggest that borrowers whose loan oﬃcers are absent have reduced access to credit since new loan oﬃcers seem to rely on rationing clients who they perceive as riskier. In addition, borrowers demonstrate less loyalty towards the bank.
Next, we test if the documented negative impact of loan oﬃcer turnover can be mitigated if there is a possibility to transfer soft information to a replacement loan oﬃcer. For this purpose, we look at variations in (1) how well the absence of a loan oﬃcer can be planned in advance, since it should be more diﬃcult to transfer soft information in the case of completely unplanned leave (2) whether the departing loan oﬃcer has any incentives to collaborate in conveying information to a replacement loan oﬃce. We observe four diﬀerent types of leave: due to sickness, resignation, pregnancy, and termination. The timing of a sickness leave is diﬃcult to plan in advance, since we look at major illnesses like heart attacks or cancer, which are unexpected. Even though the oﬃcer might have incentives to convey soft information to a replacement, the severity of the disease usually prevents it.
Here the replacement loan oﬃcer might not be able to access any of the soft information the previous loan oﬃcer had acquired. In comparison, a loan oﬃcer who is dismissed might 2 A client is considered in default if he or she has late payments of more than 60 days.
pregnancy has a nine-month lead-time, where the bank could ensure that the replacement loan oﬃcer be given information on the soft factors of the borrowers. Furthermore, the previous loan oﬃcer has an incentive to collaborate since she gets her clients back after the maternity leave. Similarly, in the case of resignations, loan oﬃcers usually have to give a one-month notice before they leave, which is usually enough time to brief the replacement loan oﬃcer. If we see deterioration in the credit terms, even in the last two circumstances, it would suggest that soft information is diﬃcult to transmit, even when given enough time.
We ﬁnd that clients whose loan oﬃcer takes a sick leave are 1.2% less likely to get a new loan from the bank during the time of the absence, compared to the average probability of getting a new loan, which is 6.5%. This is driven by a similarly strong decrease in the likelihood that clients apply for a loan, which can be a sign that they feel less loyal to the bank. These clients also show a 2.1% increase in the probability of getting a loan outside of the bank, which is almost 13% higher than the probability for an average client in the sample.3 The fact that they are able to get outside ﬁnancing also suggests that they are of reasonable credit risk. Furthermore, these borrowers experience a very signiﬁcant increase of 0.95% in the probability of delinquency. Overall, these results suggest that the sudden leave of a loan oﬃcer has a signiﬁcant impact for the credit access but also the loyalty of clients. The sickness leave can be interpreted as a quasi baseline, since loan oﬃcers do not 3 The average client gets a loan from another bank with a 16.2% probability.
In comparison, the clients of loan oﬃcers who are on pregnancy leave show a similar decline in their likelihood to get a loan. However, the decline seems predominantly driven by a drop in the application rate during the loan oﬃcer’s absence,not a reduction in approval.
At the same time, these clients show no propensity of going to a bank outside of the current relationship. We ﬁnd that one of the reasons for this outcome is that borrowers in this group are more likely to take out a loan in the month before the loan oﬃcer goes on pregnancy leave. This eﬀect is only observable for pregnant loan oﬃcers, but none of the other types of absentee spells. It appears that pregnant loan oﬃcers prepare for their absence by setting their clients up with a loan before they leave, possibly because they anticipate that the soft part of the information is diﬃcult to transfer. On the other hand, it is likely that pregnant oﬃcers have an incentive not to transfer soft information to their replacement, since they do not want to lose clients when they come back. Moreover, clients show an increased propensity to be late on their loans, which might underscore that these clients feel less loyalty to the interim loan oﬃcer.
In contrast, in the case of retiring loan oﬃcers (who usually retire because they have received an outside oﬀer...), conditions should be optimal to transfer information since there is enough lead time and the departing loan oﬃcer has no incentives to withhold information from the successor.4 Interestingly, in this case we see no drop in the access 4 Anecdotal evidence suggests that the incentives to transfer information are mostly explained by career concerns. Indeed, the job market of loan oﬃcers is specialized; they get a 6 month formal training plus
show no propensity of approaching an outside bank, which underscores that their access to ﬁnance does not change. While the likelihood of a client missing one month of payments also increases when their loan oﬃcer is hired away, the likelihood of outright defaulting on a loan does not increase. This could be a sign of transitory adjustment costs rather than a situation where the portfolio is permanently deteriorating when the previous loan oﬃcer leaves. Overall, these results suggest that given the right circumstances soft information seems to be transferable between loan oﬃcers.
Finally, for the portfolio of loan oﬃcers who are terminated, we see a much stronger drop in the probability of getting a new loan compared to all other spells of absence, which is equally driven by a reduction in approval rates as well as applications. There is also a signiﬁcant increase in the probability that clients fall late on their loans and default. In fact, in the two months before the loan oﬃcer is dismissed, we see a sharp increase in delinquency rates. It appears that the dismissed loan oﬃcers made bad loans and thus these clients do not get credit after the turnover. The new incoming loan oﬃcer has incentives to report poorly performing borrowers to start with a clean slate of clients. For a similar argument see Hertzberg, Liberti, and Paravisini (2010).
As a ﬁnal step, we investigate whether the magnitude of the reported eﬀects varies with important training in the ﬁeld. The market is also small and people from diﬀerent banks know each other.
Therefore, when loan oﬃcers switch banks, they want to keep their reputation in the industry, which maximizes their future outside opportunities. In particular they do not want to be perceived as un-loyal by stealing clients, or a poor performer if their old portfolio defaults just after they leave the bank.
with more reliable hard information, we should see a smaller eﬀect for these ﬁrms when the original loan oﬃcer leaves. We ﬁnd interesting heterogeneity depending on the type of leave.
For loan oﬃcers who are out due to sickness and thus did not have time to transmit any soft information to their replacement, we see a sharp decline in credit to small and low credit score clients, while there is almost no reduction in the access to credit for large borrowers and those that have high credit scores (which are usually seen as less opaque borrowers).
We do also see that larger and high credit score clients are signiﬁcantly less likely to fall behind in their repayments. For the portfolio of pregnant loan oﬃcers, we ﬁnd qualitatively similar, but quantitatively weaker heterogeneous treatment eﬀects.
In contrast, these heterogeneous treatment eﬀects are not present for the portfolios of loan oﬃcers who either resign or are terminated. For resigning loan oﬃcers, we see no diﬀerentiation based on observable information. Most importantly, we see no reduction in credit for borrowers with worse observable characteristics. This might indicate that resigning oﬃcers were able to successfully brief their replacements about the soft (and hard) information of the clients. Lastly, for clients of terminated loan oﬃcers, we see a drastic decline in access to ﬁnance for all types of borrowers independent of observable characteristics. We conﬁrm the same dynamics for female borrowers compared to male borrowers: women clients are often perceived as riskier and more opaque, since they have fewer assets and thus have to rely on soft information lending. Women experience a stronger
strong during temporary leaves such as sickness and pregnancy.