Abstraction bias or bias bias?

Last month, Steven L. Schwarcz put out a paper, Regulating Financial Guarantors: Abstraction Bias As a Cause of Excessive Risk-taking, arguing that financial guarantors suffer from abstraction bias:

Financial guarantors commit to pay out capital only if certain future contingencies occur, in contrast to banks and other financial firms that pay out capital—for example, by making a loan—at the outset of a project. As a result, financial guarantors are subject to a previously unrecognized cognitive bias, which the author calls “abstraction bias,” that causes them to underestimate the risk on their guarantees.

Reading this paper reminded me of Gigerenzer’s paper (The Bias Bias in Behavioral Economics, Review of Behavioral Economics, 2018, 5: 303–336) arguing that:

[behavioral economics] is tainted by a “bias bias,” the tendency to spot biases even when there are none.

Let us look at the examples that Schwarcz presents of “abstraction bias”:

• The bond-CDS basis: Selling CDS protection (insuring a risky bond against default) is equivalent to buying the risky bond if we ignore the issue of how the purchase of the risky bond is funded. Therefore, in an idealized world, the CDS spread should be the same as the credit spread of the bond. Schwarcz argues that abstraction bias causes the CDS spread to be lower than the bond spread. Interestingly, Schwarcz concedes in footnote 89 that the discrepancy between the CDS and bond spreads has arisen only after the Global Financial Crisis. Does Schwarcz think that the Global Financial Crisis led to a rewiring of the human brain creating an abstraction bias where none exited before? The Jennie Bai & Pierre Collin-Dufresne paper that Schwarcz cites in footnote 89 tells a different story: it is all about funding risk and liquidity risk which did change after the Crisis.

• Bond Insurance: Financial Guarantors like MBIA got into big trouble during the Global Financial Crisis by insuring mortgage backed securities against default. Schwarcz’s claim that they charged too little premium for taking on this risk is based on the following analysis. Consider a mortgage backed security that would have been rated BBB without bond insurance and compare (a) the premium for insuring the bond, and (b) the spread of the BBB bond over US government bonds. The fact that (a) is lower than (b) is cited as evidence of abstraction bias. But this is a wrong comparison because even after bond insurance, the resulting AAA mortgage security trades at a significant spread over US government bonds. The correct way of measuring (b) would be to take the spread of the BBB mortgage security over a AAA mortgage security. My sense is that if Schwarcz’s Appendix 1 is corrected in this manner, much of the discrepancy would disappear.

• Standby Letters of Credit: I have great difficulty understanding the claim here because the author says:

  … standby letters evolved on a path-dependent progression from commercial letters of credit, which traditionally are prudent banking instruments.

  …

  … there is evidence that standby letters of credit are much riskier than commercial letters of credit

  The problem is that both standby and commercial letters of credit involve the same alleged abstraction bias. Even if it is true that banks have lost more money on standby than on commercial letters of credit, I fail to see what that tells us about abstraction bias.

I got the feeling that Schwarcz picks up examples where there is significant tail risk that takes the form of a contingent liability. It is the tail risk that makes assessment and valuation difficult, but the author seems to think that it is all about abstraction instead.

Posted at 3:39 pm IST on Sat, 21 Sep 2019         permanent link

Categories: market efficiency

Comments