Measuring the Cost of the TARP

Douglas J. Elliott
Douglas J. Elliott Former Brookings Expert, Partner - Oliver Wyman

January 23, 2009


Late last week, the Congressional Budget Office released an estimate for the expected cost of the TARP, the Treasury’s Troubled Asset Relief Program that is intended to help stabilize the financial system. Sadly, the CBO report received much less attention than it deserves. CBO estimates that just one quarter of the funds committed to the program will be actual losses — only $64 billion of the $247 billion committed at the time of their analysis would be true costs to the taxpayers. The rest would essentially be a loan from the public that would be repaid at a market rate of interest.

Unfortunately, the political discourse and headlines have focused on the maximum possible loss – that famous $700 billion figure – rather than the expected cost or even a conservative “probable worst case” number. Despite today’s grim economic circumstances, it would be virtually impossible to lose that whole amount.

Focusing on the wrong number is dangerous. It is a well-known axiom of politics and business that the answer to a question can be determined by how you frame it. In this case, the key framing issue is how you measure the cost. Telling the public they could lose $700 billion has naturally produced a firestorm of resentment. We might instead have gained at least grudging public support by framing the cost differently, given the dire circumstances. Consider the potential reaction to an estimate of $175 billion of expected losses with a chance it could rise to $300 billion in a very bad case, offset by a reasonable chance the ultimate cost would be lower than expected.

More insidiously, we are less sensitive to differences in likely costs and risk levels when we treat all the parts of the program as if they were going to lose the full amount of their funding commitments. This matters because some uses of TARP funds are expected to lose much more than others. CBO projects that funds for the automakers and AIG are likely to suffer losses of about 55 cents on the dollar while the main bank rescue package loses only 18 cents, less than one third as much. Perhaps the greater losses are outweighed by larger economic benefits to the nation, but we cannot honestly make that choice without the right cost estimates.

The government is being besieged by requests to support municipal bonds, small business loans, consumer loans, mortgages, banks, insurers, etc. Using the right ruler to measure the cost will strongly influence how much Congress and the voters are willing to commit to save our financial system, which programs are funded, and how we execute the rescues.

It will be critical to economic recovery and the long-term health of our financial system that we allocate money to the different rescue programs in a way that maximizes the “bang for the buck”. As the rest of this paper will show, this allocation is best done by comparing the expected costs of various programs, rather than focusing on their maximum possible losses. The most useful estimates of expected losses are calculated as the CBO did, using net present value analysis with market interest rates. Given the murkiness of the present environment, it may be useful as well to provide an estimate of the probable worst case loss as a supplemental risk measurement.