by Jon Frye of the Federal Reserve Bank of Chicago
October 17, 2000
Introduction: The current U.S. expansion is the longest and strongest in economic history. The expansion has been a boon for U.S. banking institutions, which have enjoyed a long period of relatively low default rates. And perhaps the economic good times will roll on forever, in a permanent economic high. But an eventual reversal--an economic downturn--seems more likely. If the downturn is severe, the misfortune for banks may be twofold: a higher rate of default among their borrowers, and a lower rate of recovery on defaulted loans.
Defending against economic downturns is the principal reason that banks hold capital. Increasingly, banks manage their capital with guidance from the new portfolio credit models such as CreditManager or CreditRisk+. But when these models look to the next economic downturn, they allow for only one misfortune, the increase in the default rate. The other misfortune, a possible simultaneous decrease in average recovery, evades analysis. If banks depend upon such models, they might enter a severe downturn holding too little capital.
The degree of potential shortfall is easy to grasp. Suppose one of the first-generation credit models accurately projects that in a severe downturn a bank will experience a 10% default rate. Overall credit loss is approximately equal to the default rate times the flip-side of loan recovery, loss-given-default (LGD). But in the first-generation models, LGD does not depend on default. If long-term average LGD is 25%, these models will project that LGD will equal 25% on average in any year, and that capital of 2.5% will withstand the downturn.
However, the same economic conditions that cause default to rise to 10% might also cause LGD to rise above its long-term average. If in the downturn LGD rises to 50%, the need for capital to withstand the downturn would equal 5.0% rather than 2.5%. If credit models overlook the possible doubling of LGD in a severe downturn, they understate capital by half.
This study examines data on U.S. corporate bonds and finds significant synchrony between default and recovery. It then fits the model presented in Collateral Damage (Risk, April 2000) to this data. Extrapolating the model to the conditions that produce a 10% default rate, recovery falls by 25% in absolute terms from its normal-year average. If that decline pertains to recoveries on bank loans as well on bonds, the LGD at banks could, in fact, double from the normal-year average.
Published in: RISK, Vol. 13, No. 11, (November 2000), pp. 108-111.
Download paper (61K PDF) 14 pages