“Why did U.S. banks invest in highly-rated securitization tranches?”

That is a new paper by Isil Erel, Taylor D. Nadauld, and Rene M. Stulz, and it is one of the more instructive papers on the financial crisis:

We estimate holdings of highly-rated tranches of mortgage securitizations of American deposit-taking banks ahead of the credit crisis and evaluate hypotheses that have been advanced to explain these holdings. We find that holdings of highly-rated tranches were economically trivial for the typical bank, but banks with greater holdings performed more poorly during the crisis. Though univariate comparisons show that banks with large trading books had greater holdings, the holdings of highly-rated tranches are not higher for banks with large trading books in regressions that control for bank size. The ratio of highly-rated tranches holdings to assets increases with bank assets, but not for banks with more than $50 billion of assets. This evidence is inconsistent with explanations for holdings of highly-rated tranches that emphasize the incentives of banks deemed “too-big-to-fail”. Further, the evidence does not provide support for “bad incentives” theories of holdings of highly-rated tranches. We find, however, that banks active in securitization held more highly-rated tranches. Such a result can be consistent with regulatory arbitrage as well as with securitizing banks holding highly-rated tranches to convince investors of the quality of these securities. Our evidence supports the latter hypothesis.

In other words, it was not “too big to fail moral hazard” and it also was not venal corporate incentives.  It was possibly some regulatory arbitrage but also just plain, flat stupidity and complacency.