This paper presents a comprehensive analysis of the determinants of buy
and sell transactions. With a variety of tests, it shows that past returns,
reference price effects, the size of the holding period capital gain or loss,
tax-loss selling, and, to a small extent, the smoothing of consumption over
the life cycle all are determinants of trading.
The regressions for the sell versus hold decisions suggest that the propensity
to sell stocks one holds is positively related to recent returns. The effect
of the past return on trading activity is much more important for positive
past market-adjusted returns than for negative past market-adjusted returns.
Investors also tend to be reluctant to realize their losses except in
December, when the urge to realize large losses for tax purposes tends to
eliminate this effect. We also present evidence that tax-loss selling primarily
arises in the last two weeks of the year and that reference prices matter.
Conditional on a trade, sophisticated investor classes place less weight on
past returns in deciding whether the trade is to be a buy or sell. By contrast,
the less sophisticated investors—households, general government, and nonprofit
institutions—are more predisposed to sell than to buy stocks with
large past returns. The buy versus sell results are largely consistent with
the results of Grinblatt and Keloharju ~2000a!, in that domestic investors—
particularly the less sophisticated investor categories—tend to be contrarians
and foreign investors tend to be momentum investors. They are also
consistent with the sell versus hold evidence that high past market-adjusted
returns generate sells.
Life-cycle considerations also may account for some of the trading. Investors
tend to sell ~primarily inherited stock! early in life, purchase stock in
the prime earning years of middle age, and then sell stocks in old age. However,
the results are economically unimpressive.
By looking at all participants in the stock market, we are able to generate
a more complete picture of the stylized facts of trading than can be achieved
by exploring only a small fraction of the participants in the market. Also,
our methodological design incorporates numerous controls to avoid spurious
conclusions based on omitted variables. We believe that the main conclusions
of this work are fairly robust. For example, when we break the sample
614 The Journal of Financeinto both odd and even months, or into separate calendar years, we find that
the results are largely unchanged. However, it remains for future research
on other stock markets and other time periods to fully verify this conjecture.
The main conclusion after compiling the stylized facts about trading is
that theoreticians are going to be challenged. Although many of the documented
facts have separate theoretical models to explain them, researchers
will have to come up with better models than those that currently exist to
explain these stylized facts in combination with one another.