(Gervais, Kaniel, and Mingelgrin, 2001) and high-volume stocks with extreme returns
experience subsequent price reversals at short horizons (e.g., 1 week; Conrad, Hameed,
and Niden, 1994). We do not claim that attention-driven buying causes either of these
phenomena nor that pricing effects of attention-driven buying are more powerful than
these phenomena. Indeed, it must be the case that when the pricing implications of
attention-driven buying run contrary to those of persistence in high abnormal volume
stocks or short-term mean reversion, the influence of attention-driven buying is less
powerful, since the other phenomena are already well documented. Thus the influence of
attention-driven buying on returns is an incremental effect, or an overlay, on price
movements driven by other short-term factors. To document this influence, we must
control for these other factors (just as we must control for other factors such as
momentum).
As noted in Section 7.3.3, our proxies for attention are not perfect. For example, a
stock sorted into the lowest return decile may be there because of a well-publicized
negative event that attracted individual investors’ attention or it may be there because an
institutional investor sells a large position more quickly than the stock’s liquidity can
absorb. When we observe a great imbalance of individual investor buying in a stock in a
high-attention partition, that buying is likely to be driven by attention rather than stock
fundamentals and is thus likely to create temporary price pressure resulting in subse-
quent underperformance. To control for other factors influencing stocks, we compare
the performance of a portfolio of the stocks in each partition weighted by how actively
investors bought each stock with a portfolio of stocks in that partition weighted by how
actively investors sold each stock. This approach has three advantages. (1) We control
for known effects of abnormally high trading volume and short-term price moves
because both the stocks purchased and those sold have experienced the same recent
trading volume or return effects. (2) Portfolios are weighted in proportion to how
actively individual investors are buying each high-attention stock, the very trading
we anticipate will affect future returns. (3) Weighting our portfolios in proportion to
the value of purchases and sales by individual investors gives us an estimate of the
welfare costs of attention-driven buying. Of course, our estimate considers only the
month subsequent to a trade and may overestimate costs to investors who hold positions
for less than a month.
To test the model’s prediction, we first sort stocks into decile partitions on the basis of
the current day’s abnormal trading volume and on the basis of the previous day’s return,
and then, for each partition, we form two portfolios: a portfolio of stocks purchased by
individual investors and a portfolio of stocks sold by them. We then calculate the
difference in the returns of these two portfolios for each partition. On each day, we
construct a portfolio comprised of those stocks purchased within the past month
(21 trading days). The return on the portfolio is calculated based on the value of the
initial purchase as:
Rbt¼Xnbti¼ 1xit RitXnbti¼ 1xitð 7 : 4 ÞwhereRitis the gross daily return of stockion dayt,nbtis the number of different stocks
202 News and abnormal returns