* Strictly speaking, the chart shows "cumulative average residuals," which is statistical jargon saying that many effects-in this case the market retum, the companys historical retum, risk, etc.-have been taken into account. For our pu oses, "cumulative above-market retum" will suffice.

The landmark study to show prices adjust to new information quickly was done by four leading researchers of the then-new faith-Eugene Fama, Lawrence Fisher, Michael Jensen, and Richard Roll. They examined all stock splits on the New York Stock Exchange from 1926 dirough 1960. The results the investigators arrived at, using extremely sophisticated statistical techniques for the time, appear in Figure 17-2. The chart is adjusted to illustrate the retums for stock splits relative to the market on a monthly basis for the 29 months prior to the split and the 30 months following it.* (The distribution date of the new stock from the split is month 0 on the horizontal axis, and the 29 mondis preceding it are marked from -29 to 0. After the split, prices are followed in an identical manner to month 30.)

The above-average retums are measured on the vertical axis, on a cumulative basis, starting at month -29. The more steeply die dotted line rises, the higher the retum relative to the market in the month. Figure 17-2 indicates that for the 29 months prior to the split, retums are high relative to the market compared to dieir historical relationship. After the splits, the retums are in line with their long-term relationships, and do not outperform die market. The authors conclude diat their work provides strong support for the hypothesis that the market is efficient.

This study has been cited hundreds of times in academic papers and has been taught to tens of diousands of graduate students as one of the major research works upholding market efficiency. All the same, die findings may not support the researchers conclusions but might present evidence that directly contradicts their claim that markets are efficient.

The researchers did not measure the right period in which news ofa split would affect stock prices. The information enters the market at the time of die announcement (t)etween months -4 and 0 on die chart), most often 2 to 4 months t)efore the split is distiibuted. As noted, according to the semi-strong form of EMH, the news of stock splits is factored into prices almost immediately. By the time of the distribution of the spht (point 0), the news was up to 4 months old, and the informational content was already fully reflected in the stock prices. The announcement date is die proper measuring point, as it is for eamings su rises, dividend increases or decreases, or other announcements that can have a major impact on stock prices.

<u >

Figure 17-2

Above-Market Retums for Stock Splits? 1926-1960

Months -4 to -2, the period in which most stock splits are annoimced, accoimt for 3 times the average above-maricet monthly letums of the other 27 months prior to the •

split distribution (point 0). *

°-30 -20 -10 -4 -2 0 10 20

Month Relative to Spht

Source: "The Adjustment of Stock Prices to New Information," International Economic Review, February 1969, p.l3. (Modified to show months -4 to -2.)

Instead, the academics used the month in which the stock spht was actually distributed (Figure 17-2, month 0) as the starting point. Measuring of stock retums up to 4 months after the announcement date is almost meaningless.

From examining the chart it is obvious that the steepest ran-up occurs during the time of the announcements of splits. In fact, the average extra monthly retum for the four months in which the splits are announced is almost double the above-market retums in the previous 25 months. Fully 27% of the above-average retum for the 29 months prior to the split distribution takes place in the 4 months in which the stock splits were announced. Moreover, 19% of this retum occurred in the third and fourth months prior to the stock distribution. The retums during this time-which, as noted, is normally the period in which most stock splits are announced-are 3.3 times the above-market monthly retum for the other 27 months before the splits were distributed.

* Except for a sample of only 52 of 904 splits, or under 6%, which the authors say supports their case.

This raises a difficult problem for the researchers. Given that by far the sha est rise for the 29 months before the spUt distribution comes in this period, (months -4 to -2 in Figure 17-2), they must claim that all the price run-up took place before the split announcements in this period, or that the adjustment to the new higher prices came almost immediately after the announcements, so that investors could not benefit from them.

What the chart appears to show, assuming the majority of split announcements occurred 2 to 4 months prior to the stock distribution, is that stocks may indeed have provided above-average retums after the announcement date. The spike in prices in this period-again, more than triple the average high residual retum of the previous 26 months- just seems too large to indicate otherwise. The adjustment to new information, then, appears not to have been immediate but to have taken place for some time, possibly weeks after the splits announcement. If this is the case dieir argument is invalid. The most logical conclusion is that the stocks continued to rise as a group for an extensive period after die split announcement. If tme, the researchers findings would support inefficiency, not efficiency in markets. However, the jury must be out until a clear-cut study supporting eidier this analysis or the authors conclusion is carried out.

Why did the researchers not start their measurements when the split was announced, since they explain several times in the paper that this is the date that the new information enters die market? The answer was that the announcement date simply was not in their database.* Perhaps this was fortunate. If it were possible to place the split at the correct point, as the above analysis indicates, the conclusion might have been different.

The researchers dont stop there. They attempt to show an investor cannot benefit from the rapidly rising prices during the period of the split announcements. Not knowing the date of the announcements for the overall sample, or being able to measure from this point, makes this conjecture. Finally, in examining the retums after the split is distributed, the investigators say that since investors get no extra retums from this time on (point 0 on the chart) its strong evidence that markets are efficient.

The statistics in the study seem to be inte reted differently for the varying time periods. For the original mn-up from month -29 to months