The Principal – Agent Problem: Part I – RMBS Data Integrity
Back near the dawn of time when I was in business school, and the faculty was hard-pressed to find topics to fill up the curriculum, they introduced the Principal – Agent Problem. As future corporate managers and agents of the stockholders, I suppose they wanted to explain to us that our economic interests were not identical to those of the owners. This wasn’t exactly the most shocking news we had ever received, but that was all that was said about the issue, back then.
Of course, there is considerably more to this multi-faceted problem. According to Wikipedia, “The principal–agent problem arises when a principal compensates an agent for performing certain acts that are useful to the principal and costly to the agent, and where there are elements of the performance that are costly to observe,” primarily due to asymmetric information, uncertainty and risk.
Let’s look at the relationship between the RMBS bondholder (principal) and the mortgage loan servicer (agent) in this context. The bondholder relies completely on the servicer to collect principal and interest each month, remit cash collections to the trust, report monthly collateral performance data accurately, send out monthly bills to borrowers, encourage borrowers to pay on time, persuade them to catch up if they fall behind, and foreclose and sell the underlying property if all else fails. For these services the bondholder pays the servicer a fee which is a flat percentage of the aggregate unpaid balances of the loans owned by the trust.
Like all businesses, the goal of a servicer is to maximize its profits from the flat fee which is its revenue. The main route to this goal is to minimize its expenses, of which labor comprises approximately 80%.
The owner of the servicer has its own subset of Principal – Agent issues with respect to its employees. The incentives it provides its employees add several layers of complexity to the bondholder – servicer relationship. In this first part, I am going to discuss the effects of Principal – Agent and Owner – Employee relationships on the integrity of the data reported to the bondholder each month, using examples from my past experiences.
By 2001, MBIA, my former employer had about $3 billion of exposure to ten manufactured housing (“MH”) loan pools serviced by an affiliate of the bond issuer. We’ll call it MH Servicer I. In January 2002, MH Servicer I’s parent, had concluded from the rapidly increasing delinquency levels in the MH loan pools that it wanted to be out of the MH financing business. They present-valued MH Servicer I’s assets and liabilities over their remaining 25- to 30-year lives, and carried the net asset value as the residual value of the now discontinued business.
The parent had a problem to solve. How could it retain the management of MH Servicer I over an extended period of time, and how could it motivate the servicer to increase the value of a large run-off pool of badly performing MH loans? The solution was reasonably simple: the management team was awarded above-market base salaries, together with incentive compensation which paid them attractive bonuses if they could increase MH Servicer I’s residual value. If the residual value went up each year, they stood to make a lot of money.
Calculation of the residual value each quarter was performed by an outside vendor, and the methodology was based on delinquency trends. If delinquency went down, the value went up. Because MBIA was at risk if principal and interest collections of the ten trusts were inadequate to pay bondholders, we were delighted to see the elevated delinquency levels go down in 2002. We were happy, that is, until we found out why they were going down.
During a visit to MH Servicer I’s main collection site, I noticed a white message board which was posted with following two suggestions to the collections staff:
§ Ask for the payment in full.
§ If you can’t get it, offer an extension.
“What’s an extension?” I asked.
“Oh, if a borrower is 90 days past due, offer to extend the next due date by, say, 60 days,” responded a member of MH Servicer I’s staff.
“How is the resulting delinquency reported?”
“Extended borrowers are reported as current until they miss their next payment on the next due date.”
Extensions certainly brought reported delinquency down, while MH Servicer I’s residual value and management’s incentive compensation went up. Extensions also increased the amount of the trust’s non-earning assets without reducing the par value of the trust’s liabilities. This dynamic crushed the credit enhancement of each deal in accelerated fashion, and rendered delinquency reporting useless from an analytic perspective.
We had another $600 million of MH exposure coming from 11 trusts involved in the 2003 bankruptcy of another MH servicer we’ll call MH Servicer II. In this case, the delinquency situation was the reverse of the MH Servicer I story. These trusts issued bonds which were 3- to 5-years old at the time of the filing. Despite a vintage profile similar to the MH Servicer I trusts, the Conseco trusts consistently reported 30+ day delinquency in the 3% range. At least they did so until the month after the bankruptcy filing.
From that point on, delinquency increased each month for over a year, ultimately reaching peaks in the 18% range. I was not privy to the incentive compensation plans provided to MH Servicer II’s management and employees, but it is easy to infer that reported delinquency trends were somehow suppressed (i.e., held down) for several years. Once the bankruptcy filing occurred, delinquency increased to MH Servicer I levels and beyond, until a new management team came on board and began to exert control over the loan pools beginning in late 2003.
The point of these two examples is to illustrate how the Principal – Agent Problem, or its subset, the Owner – Employee Problem, can destroy the integrity of reported collateral performance data over extended periods of time.