Why Do I Predict That TARP Will Cost Less ?

Robert Waldmann

I am generally reluctant to make predictions. However, I am willing to predict that the cost of TARP will be less than forecast by the CBO. I should point out that forecasts of the cost have declined.

The reason is that the CBO values TARP assets at “fair market value” which means the value for which they could be sold. It assumes that risky assets are not systematically worth much more to the Treasury than to private investors. Thus the cost is the cost if the Treasury liquidiated its portfolio (without considering the huge market pressure and reputational effects). It is not equal to the expected value of additional debt due to TARP — it is assumed that variance in returns on assets (which by definition doesn’t affect the expected value of those returns) increases the cost to the Treasury.

I think this is backwards. It is good for the country for the Treasury to bear risk, so the risky returns act as automatic stabilizers.

After the jump, I quote Doug Elmendorf explaining what they do (and indirectly noting that Barney Frank objected). Barney Frank is an excellent congressman.

In its August 2010 baseline projections, CBO included an estimated $53 billion in costs for new mortgage guarantees that Fannie Mae and Freddie Mac will make over the 2011–2020 period. That estimate was made using a so-called “fair-value” basis of accounting, which differs from the way most federal credit programs are reflected in the budget. In a letter sent today to Congressman Barney Frank, CBO discusses that estimate and compares it with the budgetary impact that would be estimated using the procedures specified in the Federal Credit Reform Act of 1990 (FCRA), which governs the accounting for most federal direct loans and loan guarantees. CBO estimates that on a FCRA basis, Fannie Mae’s and Freddie Mac’s new mortgage guarantees made over the 2011–2020 period would generate total budgetary savings of $44 billion.


FCRA estimates are based on Treasury rates (which are generally viewed as risk-free), whereas fair-value estimates employ discount rates that are adjusted to match the risk of the specific credit obligation.

Why is the appropriate adjusment for risk a higher required return not a lower required return ? Elmendorf doesn’t say. He does make it clear that TARP costs are calculated in a way no other costs are calculated (just think what the expected cost of invading Iraq would have been if it were assumed that financial risk must be rewarded). Notice the little correction of $ 97 billion dollars. That is (part of) the difference between the reported expected “cost” of bailing out the GSEs and the expected effect of that bailout on national debt in 2020 (assuming they aren’t reprivatized in which case the national debt will be higher in expected value).

I think that answer is that using standard FCRA accounting for Treasury purchases of risky assets would imply that the Treasury should buy risky assets even when there isn’t a crisis and we just can’t have that. That would be socialism. The standard accounting would still be conservative given the desirability of automatic stabilizers (deficits in downturns and surpluses in expansions). If the numbers weren’t calculated with non standard rules, socialism would look sensible, and that just can’t possibly be allowed.