Did the U.S. Treasury’s capital purchase program (CPP) help bank lending and business activity?

Cecchetti (2009) provides a detailed chronology of the U.S. financial crisis and the Federal Reserve’s responses.

Under the CPP the U.S. Treasury provided U.S. $205 billion in capital to 707 institutions. The final investment under CPP was made in December 2009. As of August 31, 2011, the U.S. Treasury has received a total of U.S. $208.46 billion in repayments, dividends and interest while 243 banks have exited the program.

The normalized bank total asset variable by construction contains positive and negative values for the full sample and the four subsamples. Note that if the average bank total assets in log form for the full sample would have been used to compute normalized bank total asset values for our money center sample (small sample), then all normalized bank total asset values would have been positive (negative).

We report below the estimates under the SGMM estimator, following recent Monte Carlo results by Hauk and Wacziarg (2009), who show that not only fixed effects but also DGMM estimators overstate the speed of convergence in growth regressions. Results with fixed effects estimators and the DGMM procedure are available upon request. Our preferred SGMM estimators herein provide conservative results.

We report below estimates of the models with the SGMM procedure, following the original DGMM developed by Arellano and Bond (1991). The SGMM estimator ensures consistent parameter estimates by choosing instruments for the lagged dependent variable in such a way that the sample correlations between the instruments and the model’s error term are as close to zero as possible. To test the validity of the over-identifying restrictions, we employ the Sargan test. We also test the hypothesis that there is no second-order serial correlation in the disturbance of the first-difference equation using the Arellano and Bond test, AB (2). The tables below contain both sets of statistics—Sargan and AB (2)—along with AB (1), which just shows commonly found first-order correlation in the disturbances.

Allen and Santomero (2001) explain that the securitization process initially targeted the traditional mortgage loans and eventually expanded to other kinds of credit transactions including commercial loans portfolios through the emerging collateralized loan obligation (CLO) market. Altunbas et al. (2009) explain that securitization has altered bank characteristics (size, liquidity and capitalization) commonly used to identify shifts in loan supply in the bank lending channel literature.

Since the CPP variable contains dollar amounts only in the periods when funds are received and zero otherwise, the CPP variable is measured as ln (CPP dollars +1).