http://clevelandfed.org/research/workpaper
Best available copy
(bought) the stock to (from) the investor from (for) his inventory at his ask-
ing (bid) price, the transaction price at which the investor buys (sells) the
stock is not a market-clearing price. The size of the bid-asked bias is dir-
ectly related to the width of the bid-asked spread.
Blume and Stambaugh show that the use of the one-eighth pricing convention
in security markets increases the degree of bid-asked bias for low-priced
stocks relative to high-priced stocks. For example, the one-eighth pricing-
convention sets the minimum bid-asked spread at one-eighth of one dollar.2
The minimum percentage spread for a stock priced at $2 per share is 6.252,
while the minimum spread for a stock priced at $20 per share is 0.625%. It is
clear that in the absence of trading volume and other considerations, the
relative width of the bid-asked spread decreases as share price increases. In
fact, the negative relationship between price and the relative width of the
bid-asked spread is empirically documented by Branch and Freed C5J and Demsetz
(91. Therefore, the degree of bid-asked bias in recorded prices is inversely
related to price.
II. Effects of Recorded-Price Errors on Measures of Risk and Return
Measurement errors in recorded stock prices can lead to biases in
holding-period returns when the returns are calculated over short holding
periods characterized by flow-supply (flow-demand) pressures. Flow-supply
(flow-demand) pressures can lead to nonrandom recorded-price errors. If the
recorded-price errors are sufficiently nonrandom, then returns computed from
recorded stock prices will be biased. A reduction in the length of the hold-
ing period over which the returns are computed increases the probability that
the prices used to compute returns are subject to measurement error and
thereby increases the likelihood that the holding-period returns are biased.