Wall $Treet Week : Information Or Entertainment

"Wall $treet Week": Information or entertainment?

Eurico J Ferreira; Stanley D Smith
4,638 words
1 January 2003
Financial Analysts Journal
45
Volume 59, Issue 1; ISSN: 0015-198X
English
Copyright (c) 2003 ProQuest Information and Learning. All rights reserved. Copyright Association for Investment Management and Research Jan/Feb 2003

The purpose of the study we report was to determine the information content of the recommendations made by panelists during 1997 on "Wall Street Week with Louis Rukeyser." Using event-study methodology for the short term, we found a statistically significant positive abnormal return of 0.65 percent for the recommendations on the first trading day after the show on Friday. To determine the abnormal long-term (one- and two-year) average holding-period returns, we used two matching processes. Using industry and size matching, we found not only that the portfolio of recommended stocks improved in value during the following eight quarters but that its increase in value was higher than for the matched sample in all eight quarters and statistically significantly higher in half of the eight quarters. Using industry, size, and book-to-market matching, we found similar results. Overall, this study's results suggest that the panelist recommendations have significant information content.

"Wall Street Week with Louis Rukeyser" (WSW) is a major television show that started in 1970 on the Public Broadcasting System. In discussing the growth in business news programs on television, Smillie (1998) said:

Only two business programs draw networksize viewership, and both air on public television stations: "Wall Street Week with Louis Rukeyser," with 4.7 million households each week; and "Nightly Business Report," which draws more than I million viewing households each evening.1

A rich literature has examined the information content of a variety of business media. For example, Lloyd-Davies and Canes (1978), Syed, Liu, and Smith (1989), Liu, Smith, and Syed (1990, 1992), Beneish (1991), Barber and Loeffler (1993), and Liang (1999) found abnormal price performance from recommendations reported in the "Heard on the Street" and "Dartboard" columns published in the Wall Street Journal.2 Desai and Jain (1995) examined the performance of stock recommendations made by prominent money managers at Barron's annual roundtable from 1968 to 1991. The buy recommendations earned significant abnormal returns from the recommendation day to the publication day, a period of about 14 days. The abnormal returns were essentially zero, however, for holding periods of one to three years after the publication day. Desai, Liang, and Singh (2000) documented that stocks recommended by Wall Street Journal all-star analysts outperform benchmarks controlled for size and industry.

In the past, three articles have looked at the performance of the analysts on WSW. Using eventstudy methodology, Pari (1987) found an excess return of 0.66 percent on the first trading day after the show for a 1983-84 sample; Griffin, Jones, and Zmijewski (1995) found a 1.10 percent excess return for a 1972-90 sample; and Beltz and Jennings (1997) found a 0.52 percent excess return for 734 positive recommendations and a -0.62 percent excess return for 67 negative recommendations for a 1990-92 sample.3 In the Beltz-Jennings study, all excess returns for the event day were statistically significant at the 0.05 level except those following negative recommendations. Generally, these studies found an information effect on the first trading day after the show.

The same three studies also examined the results for longer periods-with mixed findings. Pari found excess returns of -4.18 percent and -7.23 percent for, respectively, 6 months and 12 months after the show. Griffin et al. found excess returns of approximately 4 percent for one year after the show. Beltz and Jennings found a significant negative excess return after six months for the positive recommendations (-1.03 percent) and a significant negative excess return after six months for the negative recommendations (-3.25 percent).

We extended these studies by using a more recent sample, by examining a longer performance time frame (i.e., quarterly results up to two years), and by using a more appropriate methodology to evaluate the longer-term returns. The earlier three studies used cumulative abnormal returns based on event studies, with "abnormal" measured against a market model, such as an equally weighted index used as a benchmark. The methodology we used was suggested by Barber and Lyon (1997). They documented that cumulative abnormal returns are biased predictors of buy-and-hold abnormal returns, that there are significant biases in test statistics when long-run abnormal returns are calculated by using a reference portfolio (such as a market index), and that matching sample companies to control companies of similar size and with similar ratios of book value to market value (BV/MV) yields well-specified t-statistics in virtually all sampling situations that they considered.

The purpose of the study we report was to determine the information content of the recommendations made in 1997 by panelists on WSW. We considered the question: Could investors have profited by using the recommendations of the 1997 shows? If they could not, then we would consider the show one of primarily, or entirely, entertainment, not information. Keep in mind that we were measuring "information" only in the sense that positive excess returns were earned. Other information on the show may be useful in many ways, but not in our narrow use of the term here.

To examine the long-term value of the recommendations, we used two matching processes to determine the abnormal average holding-period returns each quarter, up to eight quarters. We also used these matching methods to analyze the value of individual panelists' recommendations.

Data and Methodology

Our primary source of stock recommendations was the transcripts of the "Wall Street Week with Louis Rukeyser" shows between 27 December 1996 and 26 December 1997. Each weekly presentation on Friday has several panelists, typically 3 or 4 from a regular set of approximately 22 panelists, plus a special guest. Because the special guest's stock recommendations are the primary focus of that week's show, those recommendations are the primary focus of this research; the other panelists' recommendations are grouped under "others" in our reported results. (Note that some of the later regular panelists may have been a special guest on occasion.)

Of the total of 351 stock recommendations we observed, only 10 were to sell; the rest were to buy. Each sample includes 7 recommendations to sell the stock.

We included only recommendations with stock data available in CRSP and/or in Standard & Poor's Compustat. For the sample with data in CRSP, we also required that data for matched stocks with the same two-digit SIC code and of the closest size (equity capitalization) be available. The final sample contained 204 stocks for the one-year postrecommendation holding period and 200 for the two-year postrecommendation holding period. As indicated by their delisting code on CRSP, four companies in the one-year holding period merged before the end of the second holding period; hence, their stock returns were not included for the twoyear sample computation.

For the sample of recommended stocks requiring data from both CRSP and Compustat, we required that data be available for matched companies that had the same primary two-digit SIC code, that were within the same industry and size decile, and that were nearest in BV/MV to the test company. This sample consisted of 194 stocks for the one-year and 190 for the two-year postrecommendation holding periods.

Our primary technique for estimating long-run expected returns was to use two matched samples. One sample contained each recommended stock and the stock of a chosen matched company with the same primary two-digit SIC code and one that was closest in equity capitalization to the recommended stock as of the recommendation day. We call this sample the "industry-and-size matched" (ISM) sample. The second sample contained each recommended stock and the stock of a matched company stock with, in addition to the requirements of the first sample, the closest BV/MV to that of the recommended stock on the recommendation day. We called this sample the "industry-, size-, and BV/MV-matched" (ISBM) sample. Because book value per share is a Compustat item with annual periodicity, we chose fiscal-year 1996 book values per share for recommendations that were made on 27 December 1996 as well as between 3 January 1997 and 26 December 1997.4

Each holding period begins on the day after the recommendation and ends on the last day of the quarter, or combined quarters, under consideration. Eight quarterly holding periods constitute the two-year postrecommendation span. If the matching company ceased to trade during the estimation period, it was replaced (under the same criteria) by a new matching candidate. Moreover, given the small number of sell recommendations, we simply multiplied these stocks' daily returns by -1 rather than grouping them independently. Therefore, we report the results for one portfolio containing all the recommended (buy or sell) stocks.

We measured the market performance of sample stock i over T days in the event period by determining the holding-period return, HPR^sub iT^, as follows:

As suggested by Barber and Lyon, we used the Wilcoxon signed rank test to test whether the average holding-period abnormal return over T days was significantly different from zero.

Finally, applying a standard event-study approach (Mikkelson and Partch 1988), we determined abnormal returns around the recommendation day for a 30-day window centered on the event day, Day 0. We used the market model and estimated its parameters from daily returns starting at Day -140 and ending at Day -20. The index was a CRSP-NYSE/Amex/Nasdaq equal-weighted market index.

We report the samples' descriptive statistics and, for comparison purposes, the statistics for all 1997 CRSP companies in Table 1. The mean size of the companies in the ISM sample (the first row), $15.5 billion, is substantially larger than the mean size for all 1997 CRSP companies. The median size of $3.6 billion can be compared with the much smaller median for all 1997 CRSP companies (second row). These statistics indicate that the companies that were the object of WSW recommendations in the study period are larger than the average or median in the CRSP database. The prior three studies were also generally weighted toward the largercap stocks. Pari made no comment on the sample size, but he included only NYSE and Amex stocks. In Griffin et al.'s sample, 81 percent or more of the recommendations were in the largest three deciles for the NYSE/Amex/Nasdaq stocks and 83 percent of the recommendations were listed on the NYSE or Amex. Of Beltz and Jennings' sample, 84 percent were exchange listed.

For the ISBM sample, for which the variable of interest was BV/MV, the last column of Panel B in Table 1 shows that the sample size is smaller than the ISM sample. The mean BV/MV is higher than the mean BV/MV for the 1996 fiscal year of all companies in Compustat. Although the means may suggest that the sample companies have, on average, a higher average BV/MV than those in Compustat for the 1997 fiscal year, the medians indicate that the distribution center of BV/MVs is lower for the sample (0.30) companies than for all Compustat companies (0.43).

Results

We report in this section the average daily abnormal returns (ARs) and cumulative abnormal returns (CARs) around the event (recommendation) day, Day 0, and the returns in the eight-quarter postrecommendation period, which of the guest panelists during 1997 provided recommendations that led to superior results, and the performance of a portfolio of recommended stocks based on size and BV/MV.

Short-Term Average ARs and CARs. The average daily ARs and CARs for Day -15 through Day 15 around the recommendations are presented in Table 2. For the first day following the day of the show with the recommendation (Day 1), the AR is 0.651 percent and statistically significant (Z-statistic = 4.55), but this return reverses over the next four days. Table 1.

These short-term results for the day following the show are similar to those found by Pari (+0.66 percent) and by Beltz and Jennings (buy = +0.52 percent and sell = -0.62 percent) and lower than those of Griffin et al. Griffin et al. found stock price reaction to the panelists' recommendations of +0.90 percent (on the NYSE/Amex) and to special guests' recommendations of 1.40 percent (on the NYSE/ Amex) and 2.20 percent (on the Nasdaq). All of these studies provide evidence that these stock recommendations have a positive impact on the companies' share prices on the first trading day after the show; Pari also found, however, that the return reversed over the next three days.

Extended Postrecommendation Period. Return results for postrecommendation periods up to eight quarters for the ISM sample are in Panel A of Table 3. Note that the holding-period return of the portfolio of recommended stocks increases each quarter. The average holding-period abnormal return is positive in each of the eight quarters and is significantly different from 0 at the 5 percent level in Q2, Q3, Q5, and Q6.5 Table 2.

Therefore, Panel A suggests that the ISM portfolio of recommended stocks not only improved in value during the following eight quarters but that its increase in value was higher than the matched sample in all eight quarters and significantly higher in half of the eight quarters.

Panel B of Table 3 presents similar results for the ISBM sample: The recommended stocks' holding-period return increases each quarter, and the average holding-period abnormal return is positive in each quarter and significantly different from zero at the 5 percent level in Q4, Q5, and Q6. Therefore, Panel B suggests that the portfolio of recommended stocks improved in value during the following eight quarters and this increase in value was higher than that of the matched sample in every quarter, significantly higher in three quarters.

For a comparison, note that Pari found a sixmonth average excess return of -4.18 percent for buy recommendations, Beltz and Jennings found a -1.03 percent return for buy recommendations and a -3.25 percent return for sell recommendations, whereas we found a return of 5.18 percent for the ISM sample and 3.28 percent for the ISBM sample.6 For a one-year average excess return, Pari found -7.23 percent for buy recommendations, Griffin et al. found 4 percent, and we found 8.44 percent for the ISM sample and 9.69 percent for the ISBM sample.7 None of the other studies examined twoyear average excess returns, but we found twoyear returns of 15.29 percent for the ISM sample and 16.68 percent for the ISBM sample.

The results of all these studies are mixed. We would put the most weight, however, on the studies that used the longer time periods-that of Griffin et al. and our study. In addition, our study used a superior methodology to examine the longerterm returns. These two studies are consistent, moreover, in that both found positive excess returns following the WSW recommendations.

Specific Special Guest Recommendations. So far, we have built a case that, on average, the stocks recommended by WSW panelists do exhibit above-average performance beyond the first day following the program day. In this section, we examine which of the guest panelists during 1997 provided recommendations that led to superior results. Of course, the main fault of such an analysis is that the number of stocks recommended and the number of years for each special guest's performance are not large enough for us to identify among the panelists those who are superior and those who are inferior individual performers. Based on the results summarized in Table 4, however, we can sort the special guest panelists who generated positive long-run abnormal average returns from those who generated negative returns.

Table 4 presents the stock performance results for the ISM sample following recommendations by special guest panelists invited to the program in the study period. The left-hand side contains the information for those panelists whose recommendations earned positive average holding-period abnormal returns over a six-quarter postrecommendation holding period; the information on the right-hand side is for those panelists whose recommendations earned negative abnormal average returns. Similar information is given in Panel B of Table 4 for the eight-quarter postrecommendation holding period. Table 3. Table 4.

Panel B shows that five special guest panelists made recommendations that led to a more than 100 percent AHPAR for the chosen stock in the twoyear postrecommendation period.8

Size and BV/MV Effects. Recent studies (among others, see Fama and French 1992, 1995) have indicated that both size and BV/MV have a significant impact on stock returns, even when other variables are included in the analysis. Consistent with the literature, we also found that long-run average abnormal returns for portfolios composed of our samples varied when the portfolios were formed on the joint basis of company size and BV/ MV. Table 5 presents the results of our testing for six-quarter holding-period returns for portfolios formed on the basis of size plus BV/MV. Size quintiles were formed of stocks that were the subject of WSW recommendations from small stocks (Quintile 1) to large stocks (Quintile 5), and similar portfolios were formed for low BV/MV (Quintile 1) to high BV/MV (Quintile 5).