This dissertation analyzes a proprietary, comprehensive database of investor characteristics and transactions from all funds within one no-load mutual fund family between 1994 and 2000. It summarizes institutional details of fund distribution that are expected to be correlated with shareholder trading patterns.
This work contains three major contributions to the financial economics literature. The first stems from an application of the Cox proportional hazards model to fund holding periods. Results show that there are observable shareholder characteristics that enable the fund to predict reliably which shareholders will be low duration and which will be high duration. The product-limit mean (median) fund holding period is 2.61 (1.88) years, and fewer than one-quarter of the accounts are held for more than five and a half years.
The second contribution is an analysis of the investor's within-account trading choice (to wit, buy additional shares, sell shares, close the account, or make no trade) using a multinomial logit model. It is again shown that the fund can segment its shareholders into economically significant a priori trading clientele. The (non-parametric) quarterly probabilities of purchase, redemption, closure, and no trade are 8.20%, 1.88%, 7.34%, and 83.34%, respectively.
The third contribution is a discussion of the mutual fund pricing mechanism. Because the current pricing model assumes shareholder homogeneity, the proportionate allocation of benefits and costs of fund ownership may be inefficient in a fund where investors differentially trade shares. The shareholders in the database are divided into two groups based on their expected duration in the fund. Simulations show that the high-duration shareholders receive forty-five basis points more each year in a fund restricted to their type while the low-duration shareholders receive thirty-four basis points less in a fund restricted to their type. These pooling costs are conservative: splitting the shareholders into more than two groups substantially increases these estimates.
This work contains a discussion of why mutual funds choose to pool predictably different shareholders. Policy implications are noted.