I remember going out to dinner with Jon Levin when he was on the job market (I think we went to Topolobambo?). My food memory is way better than my seminar memory but I think Jon presented “Relational Incentive Contracts” as his job market paper. Many employees are paid not only a base salary but a bonus based on hard to verify details of their performance. Their employer realizes if she reneges on the promised performance-based bonus, there will be serious consequences. As an example of the benefits of keeping employees happy, Jon mentions a dispute between United and its pilots. During contract negotiations, the pilots “worked to rule”. The resulting cancelled flights and delays convinced United to pay generously.
The relationship between the employer (principal) and the employee (agent) is a repeated game. The agent might have unobservable information each period or exert unobservable effort each period. The principal’s wage offer and bonus payment between the two parties are observed by both parties. Levin studies an environment with transferable utility. If the principal deviates at any point in time, the equilibrium demands that they move to a static equilibrium where the principal offers a constant wage and the agent exerts zero effort. Hence, with observable principal behavior it is easy to keep the principal in line. The difficulty is keeping the agent working at the optimal level. In principle, the (incentive constrained) surplus maximizing contract can be quite complex. But with transferable utility it is simple. The optimal contract can be taken to be stationary: the principal offers the same wage and bonus as a function of observed agent output. Any non-stationary contract where continuation values vary over time can be replicated by a stationary contract: Any variation in continuation payoffs to the agent can be replicated by a transfer in the first period. Hence, a stationary contract suffices to give incentives to the agent. It might unravel incentives for the principal. But since the principal’s and the agent’s payments to each other are observable, incentives w.r.t. transfers can be maintained in the stationary contract.
Levin then uses this benchmark to study many other things but my impression as an outsider to the relational contracting literature is that the result on stationary contracts is the fundamental contribution of the paper. In many organizational economics seminars, I have seen presenters say “by Levin’s Theorem 2 I focus on stationary contracts” and then proceed to find the optimal contract in their setting under stationarity. The paper has around 500 citations according to Google Scholar.
Now, I am going to wander into (even) shakier territory given my (lack of) expertise: I am going to describe one of his recent forays into empirical work. Adams, Einav and Levin study the behavior of borrowers in the subprime market for auto loans. They have some amazing data from an auto sales company that also originates such subprime loans. They also track the behavior of borrowers over time. There are lots of interesting results.
First, demand is higher during tax rebate season. Second, a $100 increase in required down payments reduces demand by 9% keeping prices fixed. Keeping the down payment fixed, the same decrease in demand is generated by a price increase of $3000! The authors calculate that this implies borrowers are indifferent between paying $100 today or $1515 in one year!
The authors have the same data as the lender: they have data on applicants and on the performance of loans that were initiated. Applicants have a median FICO score below 500 (the US median is 700-750), an average income of $1200/month and live with their parents or rent. Around 2/3rds are turned away. Those who are successful make a downpayment of $1000 and buy a car costing around $11,000 Interest rates are 20-30% and the default rate is 60%.
The lender uses risk-based pricing and faces regulated interest rate caps. In theory, the latter should affect the down-payment. Credit rationing can also occur in equilibrium with moral hazard and adverse selection. The authors tease out various implications of the theory credit-rationing and see if they are backed up by the data. For example, the more likely you are to default, the larger is the loan you demand – after all you are not paying it back anyway! Also, the probability of default should rise with loan size for a given individual. The authors can disentangle the first (adverse selection) from the second (moral hazard). They argue that credit-scoring can go a long way towards mitigating the impact of adverse selection but moral hazard is more difficult to eradicate.
I still want to absorb the identification of moral hazard vs adverse selection in the data. As a b school teacher, I will give this paper the highest compliment possible : I hope to incorporate this somehow into my competitive strategy course. I hope the authors tracked not only the borrowers but the lender’s performance so we can determine whether they should have been lending in this market in the first place.
Jon has done lots of other work: see the AEA write-up.