A Test of Portfolio Crowding-Out and Related Issues in Finance
NBER Working Paper No. 1205 (Also Reprint No. r0692)
This paper tests hypotheses regarding the parameters in investors'asset demand functions. Most important is the hypothesis that federal bonds are closer substitutes for equity than for money; it is associated with the hypothesis of "portfolio crowding out" by federal borrowing. Previous regression studies of asset demand functions have not been able to obtain precise and plausible estimates for the parameters, without the imposition of prior beliefs. The present paper uses a MLE technique that dominates regression in that it makes full use of the constraint that the parameters are not determined arbitrarily, but rather are determined by mean-variance optimization on the part of the investor. The technique also dominates, on the other hand, previous estimates of the optimal portfolio from ex post return data, in that expected returns are not assumed to be constant over time, or to change slowly, but rather are allowed to fluctuate freely. Thus the framework is consistent with questions such as the effects of a sudden increase in federal debt on the expected returns of the various assets.Some hypotheses are tested where the answer seems clear in advance, such as a negative effect of the supply of money on the expected rate of return on equities. There the results of the MLE technique are much more plausible than the regression results. In the case of greatest controversy, a point estimate shows portfolio crowding in, not portfolio crowding out.
Published: Frankel, Jeffrey A. "Portfolio Crowding-Out Empirically Estimated," Quarterly Journal of Economics, Vol. 100, Supplement, (1985), pp. 1041-1065.