Long-Run Performance Evaluation of Journalists' Stock Recommendations

Kerl, Alexander G.; Walter, Andreas Article Long-Run Performance Evaluation of Journalists' Stock Recommendations Kredit und Kapital Provided in Cooperation with: Duncker & Humblot, Berlin Suggested Citation: Kerl, Alexander G.; Walter, Andreas (2009) : Long-Run Performance Evaluation of Journalists' Stock Recommendations, Kredit und Kapital, ISSN 1865-5734, Duncker & Humblot, Berlin, Vol. 42, Iss. 2, pp. 213-243, https://doi.org/10.3790/kuk.42.2.213 This Version is available at: https://hdl.handle.net/10419/293610 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Long-Run Performance Evaluation of Journalists’ Stock Recommendations By Alexander G. Kerl and Andreas Walter, Tübingen* I. Introduction Private investors are having a hard time when it comes to investing their funds, especially in times when private pension planning is becoming increasingly important. Not only do private investors usually lack knowledge about capital markets, but it is also difficult to make informed choices among thousands of different investment opportunities. Therefore, a whole industry providing professional investment advice has emerged. In general, private investors receive investment advice from financial experts, most prominently from security analysts of brokerage houses and from journalists. Both groups of financial experts usually provide, among other things, direct stock recommendations to investors. Although the immediate market reaction to financial experts’ stock recommendations has been extensively analyzed for security analysts as well as for journalists, the question whether they provide valuable advice in the long run is far less intensely researched. Particularly, the question whether the second group of financial experts (journalists) has the ability to predict stock prices and, thus, publishes valuable recommendations in the long run is basically unexplored. In order to examine the role of journalists as a source of investment advice for private investors, we evaluate stock recommendations of German Personal Finance Magazines (PFMs) such as, for instance, the EffecKredit und Kapital, 42. Jahrgang, Heft 2, Seiten 213-243 Abhandlungen * We received very helpful suggestions from Joachim Brixner, Jens Grunert, Werner Neus, Anna Rohlfing, and Martin Weiss as well as seminar participants at the European Financial Management Association Annual Meeting (Vienna), the Midwest Finance Association Annual Meeting (Minneapolis) and the Campus for Finance (Vallendar). We also want to thank an anonymous referee for many helpful comments and suggestions. Alexander G. Kerl is especially grateful for financial support from the Deutsche Forschungsgemeinschaft (DFG). All errors are our own. Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 ten-Spiegel and Börse Online. In contrast to other business media like television shows or daily newspapers, which often merely re-transmit stock recommendations of security analysts or prominent money managers, PFMs claim to employ selfcontained research procedures in order to derive original buy and sell recommendations for their readers. Although journalists of PFMs would not be willing to disclose their particular research procedures, we do have information concerning the educational and professional background of journalists working for PFMs. One editor-in-chief revealed that his journalists usually possess university degrees in economics or business. Often, journalists are former security analysts at brokerage houses. Thus, the educational and professional background of these journalists is similar to the one of security analysts. Although they might have limited access to various information sources, journalists working for PFMs should consequently be almost as competent to issue meaningful recommendations as security analysts employed by brokerage houses. Our contribution to the literature is threefold: Firstly, we aim to close the gap in research concerning the long-run performance evaluation of journalists’ stock recommendations. Besides the apparent lack of empirical evidence for this group of financial experts for international markets in general and for Germany specifically, analyzing the long-run performance of journalists’ recommendations might be particularly interesting since this group of financial experts is, unlike security analysts, free from the usual conflicts of interest. Journalists do not have to consider a company’s interests like investment banking activities. Secondly, prior research on long-run performance evaluation which employs a market index as a benchmark adjustment has been attacked on methodological grounds. By creating characteristic-adjusted reference portfolios we not only control for common characteristics of recommended stocks but we also account for the new listing bias and the rebalancing bias.Inaddition, we remedy the skewness bias by using bootstrapped skewness-adjusted t-statistics. Thirdly, we address for the first time the question whether self-contained research procedures of journalists work equally well concerning specific characteristics of stocks (market capitalization, price-to-book, prior performance, and listing at the Neuer Markt) or during several sub-periods of our investigation period. Analyzing a large sample of buy and sell recommendations issued by PFMs on German stocks in the period from 1995 to 2003, our results indicate that stock recommendations of journalists seem to have substantial 214 Alexander G. Kerl and Andreas Walter Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 investment value for private investors on the sell side. Private investors would have been guided correctly by the journalists if they sold respective stocks. With respect to the buy side, however, we have to conclude that buy recommendations do contain positive but economically and statistically insignificant investment value in general. This result of insignificant investment value on the buy side differs, however, from prior findings which predominantly document a negative investment value for buy recommendations transmitted through the business media. In contrast, we find that journalists seem to have some predictive abilities for subgroups of stocks on the buy side. In particular, buy recommendations on value stocks and on positive momentum stocks seem to contain investment value. In addition, if journalists had refrained from recommending Neuer Markt stocks for purchase, our results would allow us to assign to them predictive ability with respect to the remaining market segments. The remainder of the paper is structured as follows. Section II gives a brief review of the related literature and presents our hypotheses. Section III describes the database and provides some descriptive statistics. The employed methodology to calculate reference portfolios and abnormal returns is also characterized in this section. Section IV presents our empirical findings. Finally, we conclude in Section V. II. Related Literature and Hypotheses 1. Related Literature The literature on performance evaluation of financial experts’ advice can basically be separated into stock recommendations issued by security analysts of brokerage houses and stock recommendations distributed via the business media. With respect to the second category, one has to further distinguish between those recommendations which are mere restatements of, e.g., recommendations by security analysts (second-hand information) and those recommendations which are based on self-contained original research by journalists. The vast majority of research on financial experts concentrates on recommendations issued by security analysts which work for brokerage houses. Since brokerage houses employ huge departments to perform this kind of research for their clients, only significant abnormal returns would justify the costs of preparing the reports and to work out stock recommendations. Starting with the work of Cowles (1933), researchers Long-Run Performance Evaluation of Journalists’ Stock Recommendations 215 Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 have been eager to analyze the shortand long-run performance of such recommendations (see, among many others, Bjerring et al. (1983); Elton et al. (1986); Stickel (1995); Womack (1996); Francis/Soffer (1997); Barber et al. (2001, 2003); Mikhail et al. (2004); Agrawal/Chen (2005); Asquith et al. (2005); Fang/Yasuda (2005); and Jegadeesh/Kim (2006)). The studies almost unequivocally find a significant market reaction associated with the release of a recommendation in the short run. In terms of the longrun investment value, Womack (1996) analyzes for the US market abnormal returns up to six months subsequent to the publication of the recommendation. In contrast to modest returns following buy recommendations, he finds a significant negative price drift subsequent to the publication of sell recommendations. Thus, only sell recommendations seem to have significant investment value for investors. Similar evidence is reported by Agrawal/Chen (2005) who find an unambiguously significant continuing price drift over the subsequent twelve months for negative recommendations. Accordingly, Fang/Yasuda (2005) find more investment value in sell rather than in buy recommendations. They document that only high-profile All-American analysts who also work for top-tier banks are able to consistently earn abnormal returns with their buy recommendations, whereas all different kinds of analysts earn significant abnormal returns on their sell recommendations. With respect to international markets, Jegadeesh/Kim (2006) again document a more pronounced investment value for downgrades. In particular, in five of the G7 countries they find evidence for significant price drifts for downgrades, whereas only in two countries the price drift is significantly positive over the subsequent 132 trading days. For Germany, Gerke/Oerke (1998) and Henze/Röder (2005), among others, examine analysts’ recommendations by various brokerage houses. The authors of the latter study find that both buy and sell recommendations lead to significant excess returns in the long run. In line with international evidence, sell and strong sell recommendations lead to more pronounced excess returns compared to buy and strong buy recommendations. Although the literature on security analysts’ recommendations is quite comprehensive and the review above only scratches the surface, one can extract two major findings from prior research: Firstly, stock recommendations issued by security analysts seem to have investment value in the long run. Secondly, the investment value for sell recommendations is higher than for buy recommendations. As mentioned before, the business media regularly publishes stock recommendations. However, one has to distinguish between two strands of 216 Alexander G. Kerl and Andreas Walter Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 the literature. Firstly, there is a number of studies which evaluate the performance of second-hand information re-transmitted through the business media. Those studies do not analyze financial advice generated by journalists themselves, but examine the investment value of re-statements of other financial experts’ recommendations like those of security analysts or financial gurus published by the business media. For example, Lloyd-Davis/Canes (1978), Syed et al. (1989), Liu et al. (1990, 1992), Beneish (1991) and Huth/Maris (1992) find short-run abnormal returns based on stock recommendations issued in the “Heard on the Street” (HOTS) column of the Wall Street Journal (WSJ). Kiymaz (2002) performs a similar analysis with recommendations of the HOTS column of the Turkish magazine Ekonomik Trend, supporting U.S. results. Barber/ Loeffler (1993), Metcalf/Malkiel (1994), Wright (1994) and Liang (1999) report significant abnormal returns associated with recommendations issued in the “Dartboard”columnoftheWSJ. Whereas Pari (1987), Beltz/ Jennings (1997) and Ferreira/Smith (2003) analyze recommendations issued by panelists in the Wall Street Week television show, Desai/Jain (1995) focus on recommendations issued by prominent money managers at Barron’s Annual Roundtable. All studies find excess returns around the event triggered by price pressure. A recent study by Brixner/Walter (2007) has also confirmed the existence of price pressure due to secondhand information for Germany. The study finds that the market reacts to re-statements of stale security analysts’ recommendations in the column Tendenzen & Tips of the daily newspaper FAZ. However, when it comes to long-run analyses, various studies suggest that second-hand information have negative investment value (see, e.g., Shepard (1977); Dimson/ Marsh (1986); Pari (1987); Desai/Jain (1995); Sant/Zaman (1996); and for an excellent review, Schuster (2003)). As a consequence, private investors lose money if they follow second-hand information distributed via the business media. Apart from studies on second-hand information and gossip re-transmitted via the business media, empirical evidence on stock recommendations issued by journalists using self-contained research procedures is rather limited. Some studies exist on the short-run market reaction associated with the initial publication of stock recommendations. Lidén (2007), for example, finds a market reaction on the publication day in accordance with the type of recommendation for the Swedish market. For the German market, Pieper et al. (1993), Röckemann (1994), and Kerl/ Walter (2007) analyze the short-run investment value of stock recommendations issued by Personal Finance Magazines (PFMs). All studies Long-Run Performance Evaluation of Journalists’ Stock Recommendations 217 Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 find positive abnormal returns around the event day for buy recommendations. In terms of the long-run performance of journalists’ stock recommendations, Lidén (2006) compares stock recommendations from security analysts and PFMs. He finds that for the Swedish market buy recommendations from PFMs mislead investors. In particular, the mean market-adjusted return over a two-year period is (insignificantly) negative at 6.01%. Thus, buy recommendations of Swedish journalists do not seem to contain investment value at all. Sell recommendations, however, have investment value as stock prices display a continuous negative drift in the months subsequent to the publication. Yazici/Muradog ˘lu (2002) focus on recommendations of the Investor Ali column of the weekly economics journal Moneymatik and thus on the Turkish market. In their long-run study of buy recommendations, they state that the recommendations do not add any long-term value to small investors. In contrast, the average two-year cumulative abnormal return is –13.9%. 1 2. Hypotheses If the efficient market hypothesis (EMH) proposed by Fama (1970) holds, we should not observe any price drifts in the months subsequent to the release of the recommendations; no matter whether journalists are capable of producing relevant information or not. In contrast, stock prices should adjust instantaneously or at least rapidly to new information. As a consequence, in the absence of a price drift private investors should not be able to profit from the recommendations in the long-run. Thus, building on the foundations of the EMH, we predict in our first hypothesis that stock recommendations of journalist do not contain investment value. In particular, we predict buy and sell recommendations to yield abnormal returns in the months subsequent to the release of the respective recommendation which are indistinguishable from zero. Although the traditional view on capital markets assumes market efficiency, the EMH has come under attack from both the theoretical as well as the empirical side recently. As far as theoretical papers are concerned, 218 Alexander G. Kerl and Andreas Walter 1For Germany, to the best of our knowledge, no academic study exists which analyzes the investment value of buy recommendations in the long run. However, Reinhart Schmidt earned merits for sensitizing private investors with respect to the performance of stock recommendations. Based on his work, the Manager-Magazin published a series of studies on the profitability of buy recommendations issued by German PFMs in the early 1990s. The analyses, which focused more on practical issues, found a poor long-run performance for the analyzed magazines. Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 the literature primarily argues that the premises for market efficiency are not fulfilled. In contrast, researchers argue that the existence of limits of arbitrage and investor sentiment prevents markets from being efficient. 2 The empirical evidence against the EMH can be separated into two categories. On the one hand, a large number of empirical studies have found an initial underreaction to news since long-run post-event returns are significantly positive, including dividend initiations (Michaely et al. (1995)), earnings announcements (Ball/Brown (1968); Bernard/Thomas (1990)), share repurchases (Lakonishok/Vermaelen (1990); Ikenberry et al. (1995); Mitchell/Stafford (2000)); and stock splits (Dharan/Ikenberry (1995); Ikenberry et al. (1996)). 3 On the other hand, there is also ample evidence that stock prices overreact since long-run post-event returns are significantly negative. This evidence has been documented for IPOs (Ibbotson (1975); Loughran/Ritter (1995)), mergers (Asquith (1983)), dividend omissions (Michaely et al. (1995)), and new exchange listings (Dharan/Ikenberry (1995)). The empirical evidence concerning financial experts’ stock recommendations can be attributed to both camps. Whereas the literature on financial analysts primarily finds an initial underreaction as price drifts usually continue in the direction of the recommendation, second-hand information distributed via the business media is basically associated with an initial overreaction. Thus, in the case in which we have to reject our first hypothesis as long-term returns are different from zero, two scenarios have to be distinguished. On the one hand, if we observe a price drift in the subsequent months according to the direction of the recommendation, this would indicate that stock recommendations of journalists are somehow similar to stock recommendations of security analysts. On the other hand, if we find a significant long-run return contrary to the recommendation, this would indicate that original stock recommendations by journalists do not systematically differ from second-hand information distributed via the business media. Our second hypothesis is motivated by the finding in the literature that sell recommendations are usually associated with a higher investment value than buy recommendations. As far as security analysts are concerned, the literature (see, e.g., Dugar/Nathan (1995); Womack (1996); Long-Run Performance Evaluation of Journalists’ Stock Recommendations 219 2See, e.g., Kent et al. (1997); Barberis et al. (1998); and Hong/Stein (1999) for theoretical models which explain overand underreaction of stocks prices. 3For an excellent discussion concerning the issue of overand underreaction see Fama (1998). Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 Lin/McNichols (1998); Michaely/Womack (1999); Agrawal/Chen (2005); and Fang/Yasuda (2005)) frequently explains higher abnormal returns for sell recommendations by “conflicts of interest”. E.g., Agrawal/Chen (2005) state that returns for sell recommendations are higher for analysts with investment banking relations since respective sell recommendations tend to be more credible if they are willing to voice an unfavorable opinion. However, the “conflicts of interest” argument does not apply to our sample, since journalists of PFMs do not have to take into account a company’s interests, such as investment banking. However, the result of higher investment value of sell recommendations is also documented in the study of Lidén (2006) who analyzes the buy and sell recommendations for Swedish journalists. Obviously, journalists are not subject to the usual “conflicts of interest”. So, how can one explain this finding for journalists? Firstly, an explanation could be found in the potentially infrequent occurrence of sell recommendations. If the number of sell recommendations is smaller compared to the number of buy recommendations, each rare sell recommendation potentially contains more information value. Secondly, an explanation for a more pronounced initial underreaction for sell recommendations might be found in the fact that private investors are exposed to short sale constraints. 4 Hence, implementing sell recommendations is only possible if a stock is part of an existing portfolio which might only be the case for a rather restricted number of investors. Thus, prices might adjust slowly to new information, since private risk arbitrageurs are restricted in their trading opportunities. This rationale is supported by a model of Diamond/Verrecchia (1987) who show the effects of short-sale constraints on the speed of adjustment to private information on stock prices. They find that these constraints reduce the adjustment speed of prices, especially with respect to bad news, thus sell recommendations. Hence, information efficiency is reduced. Hong et al. (2000) explain the obvious asymmetry between buy and sell recommendations through the analyst coverage of stocks. They claim that low-coverage stocks react more slowly to bad news than to good news since the former will only be revealed by analysts, whereas the latter will also be made public via increased disclosures, e.g., by the company itself. In consequence, in case we have to reject our first hypothesis as stock prices might initially underreact, we predict in our sec220 Alexander G. Kerl and Andreas Walter 4In Germany, online brokerage houses and commercial banks only rarely allow private investors to engage in short-selling activities. Hence, we conclude that within our investigation period, from 1995 through 2003, short-selling was not an option within reach of a common private investor. Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 justed t-statistics by drawing 10,000 resamples of size m=2 from the original sample of mrecommended companies. 15 We then use the percentile confidence intervals of the empirical bootstrapped distribution as critical value for the lower and upper bounds. IV. Empirical Results 1. Investment Value of Stock Recommendations Table 2 displays (adjusted) returns for several periods prior to and past the publication of a recommendation. The first vertical panel addresses actual returns, whereas the second and the third panel address marketadjusted and characteristic-adjusted returns, respectively. We first discuss some interesting findings regarding the prior performance of recommended stocks. To this end, we focus on the second vertical panel of Table 2 where market-adjusted returns for both buy and sell recommendations are displayed for the 6-month period and 3-month period prior to the month of publication. For buy recommendations, the table reveals a tendency for journalists to recommend those stocks for purchase which performed better compared to the market in the months prior to publication. For example, the 6-month market-adjusted return prior to the publication is significantly positive at 1.79%. 16 The analogous tendency of the editorial staff to put underperforming stocks on the sell list is even more apparent. Referring to the 6-month market-adjusted return prior to the publication, journalists recommend stocks for sale which underperform the market by a significant 11.03%. Thus, we find evidence of editors following momentum investment strategies while recommending stocks both for purchase and sale. However, this tendency is much more pronounced for sell recommendations. When looking at market-adjusted returns subsequent to publication, we observe, for buy recommendations, a modest but significant marketadjusted return of 4.83% in the long-run, i.e., in the 24-month period Long-Run Performance Evaluation of Journalists’ Stock Recommendations 227 15 Lyon et al. (1999) state that the sample size of m=4andm=2 yield well-specified results. However, in absolute terms, they use a sample size ranging from 200 to 4,000. Thus, for computing the bootstrapped t-statistics, we use the sample size of m=2 or at least 200. 16 For the remainder of the text, we will refer to a return as being statistically significant if the respective skewness-adjusted t-statistics is statistically significant at least at the 5%-level (two-tailed test) when comparing it to the bootstrapped, empirical distribution. Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 228 Alexander G. Kerl and Andreas Walter Table 2 Long-Run Investment Value for Buy and Sell Recommendations Actual Returns Market-Adjusted Returns (CDAX) Characteristic-Adjusted Returns (Size, Price-to-book) Mean t-skew bs Mean t-skew bs Mean t-skew bs Panel A: Buy Recommendations ÈNã2;637ê 6-Month Period Before Event 4.03% 7.02 *** 1.79% 3.41 *** n/a n/a 3-Month Period Before Event 2.41% 6.04 *** 1.17% 3.26 *** n/a n/a Eventhmonth 1.12% 6.48 *** 0.96% 6.03 *** n/a n/a 3-Month Period After Event 0.61% 1.45 –0.91% –2.38 ** –0.19% –0.49 6-Month Period After Event 1.98% 3.11 *** –1.04% –1.84 0.22% 0.42 12-Month Period After Event 7.39% 7.76 *** 0.80% 0.91 1.66% 1.96 ** 18-Month Period After Event 13.38% 11.23 *** 3.60% 3.12 *** 2.29% 2.03 ** 24-Month Period After Event 20.05% 14.53 *** 4.83% 3.24 *** 2.10% 1.42 Panel B: Sell Recommendations ÈNã1;168ê 6-Month Period Before Event –9.83% –2.42 ** –11.03% –2.50 ** n/a n/a 3-Month Period Before Event –5.84% –2.03 –6.91% –2.14 n/a n/a Eventhmonth –2.13% –4.07 *** –2.04% –4.13 *** n/a n/a 3-Month Period After Event –5.38% –4.42 *** –6.81% –5.85 *** –6.32% –5.69 *** 6-Month Period After Event –5.67% –3.32 *** –8.86% –5.26 *** –8.34% –5.19 *** 12-Month Period After Event –1.19% –0.48 –8.86% –3.52 *** –9.24% –3.78 *** 18-Month Period After Event 4.14% 1.42 –8.62% –2.51 ** –11.41% –3.22 *** 24-Month Period After Event 13.06% 4.03 *** –7.43% –2.01 –12.63% –3.35 *** ***, ** indicate statistical significance at the 1%- and 5%-level (two-tailed test) according to a bootstrapped empirical distribution (based on 10,000 resamples). Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 after the publication. For sell recommendations, we calculate strictly negative market-adjusted returns for all investigated periods. In particular, the market-adjusted return in the long run is –7.43% but insignificant. However, all market-adjusted returns between three and 18 months are similar in magnitude and statistically significant. Thus, our results support the finding in Lidén (2006) that sell recommendations do contain investment value when investment value is measured by market-adjusted returns. However, as mentioned before, abnormal return calculations using a broad market index are subject to several biases discussed above. Thus, in the remainder of the paper we focus exclusively on characteristic-adjusted returns to measure the investment value of journalists’ recommendations. For buy recommendations, we observe less pronounced characteristic-adjusted returns in the long run compared to market-adjusted returns. The 24-month characteristic-adjusted return, for instance, drops to 2.10% compared to the market-adjusted return of 4.83%. In addition, characteristic-adjusted returns are now, although still mostly positive, statistically insignificant for the majority of analyzed periods (with the exception of the 12-month and 18-month period after publication). Hence, we now find much weaker evidence for an investment value in buy recommendations compared to a naïve benchmark adjustment with a broad market index. Thus, with respect to buy recommendations we find support for our first hypothesis since abnormal returns are not consistently significantly positive in the months subsequent to the release of buy recommendations. However, unlike the finding in Lidén (2006), who documents for Swedish journalists negative market-adjusted returns while employing value-weighted industry indexes as benchmarks, journalists of German PFMs at least do not lead readers in the wrong but in a rather neutral direction. With regard to sell recommendations, employing characteristic-adjusted returns emphasizes that sell recommendations contain tremendous investment value, hence, that journalists have predictive abilities when issuing sell recommendations. In particular, characteristic-adjusted returns in all analyzed periods display large negative and statistically significant returns with a peak in the long run corresponding to –12.63%. Consequently, our first hypothesis has to be rejected for sell recommendations. In contrast, stock prices seem to initially underreact to sell recommendations. As one looks at the magnitude of long-run returns, we find strong support for our second hypothesis which predicts the investLong-Run Performance Evaluation of Journalists’ Stock Recommendations 229 Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 ment value for sell recommendations to be higher than for buy recommendations. In particular, the absolute value of the investment value is about six times higher for sell recommendations compared to buy recommendations. With respect to the second hypothesis, the findings for German PFMs are in line with the international evidence for security analysts. One might wonder why the usage of characteristic-adjusted returns lowers the investment value for buy recommendations and increases the investment value for sell recommendations. This is due to the fact that the value-weighted broad market index CDAX is heavily dependent on large capitalized stocks. As small stocks perform better than large stocks during our investigation period, returns of characteristic-adjusted reference portfolios are usually higher than respective returns of the CDAX. Thus, employing characteristic-adjusted returns affects abnormal returns for buy and sell recommendations asymmetrically. 2. Determinants of Characteristic-adjusted Returns In this section, we analyze the determinants of characteristic-adjusted BHARs. This might not only be a decisive question from an academic point of view. Moreover, identifying characteristics of stocks for which journalists show the most predictive ability might help investors to make more educated investment decisions. In addition, although journalists are unable to generate investment value with their buy recommendations generally, it might be interesting to explore whether journalists show predictive abilities with respect to specific types of buy recommendations. This section is organized as follows. Firstly, in a univariate analysis in Table 3, we present BHARs for the 6-, 12and 24-month period for specific sub-groups (with regard to company size, price-to-book, prior performance, sub-periods and stock listings at the Neuer Markt) in order to determine the magnitude and significance of characteristic-adjusted returns. Secondly, results derived from the univariate analysis are complemented with evidence from a multivariate regression which can be found in Table 4. a) Company Size As has been shown in numerous previous studies (see, e.g., Banz (1981); Fama/French (1993)), company size plays a decisive role in ex230 Alexander G. Kerl and Andreas Walter Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 plaining (abnormal) returns. Thus, we partition our sample into SMALL stocks and BIG stocks, where SMALL stocks are defined as stocks belonging to the smallest quintile in terms of the market capitalization of the respective group of recommendations (e.g. buy recommendations) in a given year. Analogously, BIG stocks belong to the quintile with the largest market capitalization. As displayed in Panel A of Table 3, we find mixed evidence for buy recommendations concerning company size as a determinant for BHARs. In the first year after the publication, abnormal returns are slightly lower for SMALL stocks than for BIG stocks. However, in the long-run we report a positive but insignificant BHAR 24 for SMALL stocks with 3.69%, whereas buy recommendations on BIG stocks are associated with asignificantnegativeBHAR 24 of –4.52%. Interestingly, the three remaining quintiles (Others) display a similarly positive BHAR 24 of 3.76% compared to SMALL stocks, which is statistically significant. Multivariate results emphasize the finding that BIG stocks are associated with mediocre returns. As can be seen from Table 4, the respective coefficient is significantly negative for the 12-month and 24-month horizon. We find even clearer evidence in favor of small stocks for sell recommendations. As can be seen from Panel B of Table 3, the investment value for BIG stocks is negligible compared to SMALL stocks for all analyzed periods. Surprisingly, the long-run BHAR 24 is positive at 2.57% for BIG stocks. SMALL stocks, however, experience large negative but insignificant BHAR 24 at –16.63%. Similar evidence can also be documented for the three remaining quintiles (Others). 17 The key finding that BIG stocks do not have investment value is also supported by multivariate regression results where the coefficient for BIG stocks is positive for all analyzed periods. For the long run (thus BHAR 24 ), the effect even turns out to be statistically significant. b) Price-to-Book Previous research has documented a decisive role of the price-to-book ratio in explaining (abnormal) returns (see, e.g., Fama/French (1993); Fama/French (1995)). Thus, we separate recommendations according to Long-Run Performance Evaluation of Journalists’ Stock Recommendations 231 17 Due to the higher number of constituents, the results of this group, although similar in the level of return compared to SMALL stocks, are found to be significant. Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 232 Alexander G. Kerl and Andreas Walter Table 3 Characteristic-adjusted BHAR for Specific Sub-Groups BHAR 6 BHAR 12 BHAR 24 NMeant-skew bs Mean t-skew bs Mean t-skew bs Panel A: Buy Recommendations SMALL 523 –1.79% –1.32 –1.79% –0.75 3.69% 0.77 BIG 523 0.62% 0.68 –0.26% –0.19 –4.52% –2.34 ** Others 1591 0.75% 1.07 3.42% 3.14 *** 3.76% 2.17 ** VALUE 521 3.26% 2.77 *** 4.70% 2.54 ** 8.65% 2.88 *** GLAMOUR 521 –1.50% –1.08 –0.46% –0.19 –3.02% –0.85 Others 1595 –0.21% –0.31 1.35% 1.33 1.63% 0.84 POSPERF 1296 3.23% 4.36 *** 6.42% 5.16 *** 8.07% 3.61 *** NEGPERF 1341 –2.68% –3.48 *** –2.94% –2.66 *** –3.67% –1.94 1995–1997 739 2.50% 3.17 *** 7.08% 4.85 *** 9.29% 3.04 *** 1998–2000 837 –0.79% –0.80 1.42% 0.96 –0.03% –0.01 2001–2003 1061 –0.56% –0.60 –1.94% –1.40 –1.23% –0.48 NEUER MARKT 312 –9.68% –4.34 *** –6.59% –1.85 –21.94% –5.24 *** Others 2325 1.55% 2.96 *** 2.76% 3.28 *** 5.33% 3.44 *** Continue page 233 Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 Long-Run Performance Evaluation of Journalists’ Stock Recommendations 233 Table 3: Continued BHAR 6 BHAR 12 BHAR 24 NMeant-skew bs Mean t-skew bs Mean t-skew bs Panel B: Sell Recommendations SMALL 230 –6.51% –1.40 –9.12% –1.54 –16.63% –1.47 BIG 230 –3.34% –1.75 –2.11% –0.85 2.57% 0.60 Others 708 –10.55% –5.14 *** –11.59% –3.24 *** –16.27% –3.29 *** VALUE 227 –11.51% –2.33 ** –14.73% –2.37 ** –21.35% –1.75 GLAMOUR 227 –5.71% –1.79 –5.92% –1.42 –4.73% –0.67 Others 714 –8.16% –4.38 *** –8.54% –2.62 ** –12.37% –2.80 *** POSPERF 297 –2.23% –0.80 1.30% 0.34 3.95% 0.56 NEGPERF 871 –10.42% –5.35 *** –12.83% –4.08 *** –18.29% –4.41 *** 1995–1997 324 –4.36% –2.01 ** –6.50% –2.08 ** –7.75% –1.45 1998–2000 298 –12.51% –2.90 ** –18.62% –3.44 *** –25.29% –3.26 *** 2001–2003 546 –8.42% –3.41 *** –5.74% –1.46 –8.62% –1.37 NEUER MARKT 318 –15.10% –3.80 *** –12.16% –2.41 ** –16.74% –2.35 ** Others 850 –5.80% –3.49 *** –8.14% –2.91 *** –11.10% –2.49 ** ***, ** indicate statistical significance at the 1%- and 5%-level (two-tailed test) according to a bootstrapped empirical distribution (based on 10,000 resamples). Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 234 Alexander G. Kerl and Andreas Walter Table 4 Multivariate OLS Regression Results for Buy and Sell Recommendations BHAR 6 BHAR 12 BHAR 24 Coefficient t-stat Coefficient t-stat Coefficient t-stat Panel A: Buy Recommendations SMALL –0.0267 –1.76 –0.0566 –2.19 ** 0.0023 0.04 BIG –0.0141 –1.20 –0.0475 –2.72 *** –0.1178 –4.31 *** VALUE 0.0359 2.60 *** 0.0399 1.77 0.0503 1.23 GLAMOUR –0.0058 –0.38 –0.0219 –0.89 –0.0239 –0.60 POSPERF 0.0607 5.62 *** 0.1129 6.14 *** 0.1294 4.04 *** 1995–1997 0.0277 2.18 ** 0.1090 4.67 *** 0.0900 1.98 ** 1998–2000 0.0039 0.29 0.0540 2.54 ** 0.0216 0.68 NEUER MARKT –0.0935 –4.12 *** –0.0457 –1.26 –0.2477 –5.78 *** C–0.0234 –1.81 –0.0641 –3.00 *** –0.0276 –0.77 N2637 2637 2637 Adj. R 2 2.98% 2.56% 2.12% Prob (F-statistic) 0.0000 0.0000 0.0000 F-statistic 9.58 10.49 10.41 Panel B: Sell Recommendations SMALL 0.0647 1.31 0.0582 0.90 0.0365 0.38 BIG 0.0355 1.32 0.0581 1.51 0.1360 1.99 ** VALUE –0.0452 –0.93 –0.0580 –0.91 –0.0548 –0.55 GLAMOUR 0.0106 0.31 0.0004 0.01 0.0251 0.33 NEGPERF –0.0659 –1.91 –0.1249 –2.34 ** –0.1880 –1.96 ** 1995–1997 0.0173 0.54 0.0008 –0.01 0.0280 0.32 1998–2000 –0.0425 –1.18 –0.1246 –2.45 ** –0.1592 –2.05 ** NEUER MARKT –0.0719 –1.80 0.0084 –0.14 0.0100 0.12 C–0.0216 –0.62 0.0233 0.39 0.0158 0.14 N1168 1168 1168 Adj. R 2 1.02% 0.88% 0.74% Prob (F-statistic) 0.0013 0.0006 0.0002 F-statistic 3.21 3.47 3.80 ***, ** indicate statistical significance at the 1%- and 5%-level (two-tailed test) according to the parametric t-test employing robust standard errors (see White, 1980). Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 membership of the group of VALUE stocks or GLAMOUR stocks. VALUE stocks belong to the smallest quintile in terms of the price-to-book ratio of the respective group of recommendations (e.g. buy recommendations) in a given year. Analogously, GLAMOUR stocks belong to the quintile with the largest price-to-book ratio. For buy recommendations, Panel A of Table 3 documents strong evidence in favor of a superior investment value for recommended VALUE stocks. In particular, BHARs for VALUE stocks are consistently positive and statistically significant for all analyzed periods. In the long run an average BHAR 24 of 8.65% is found for buy recommendations. In contrast, recommended GLAMOUR stocks do not offer comparable returns, since respective buy recommendations earn an insignificant –3.02% in the long-run. Analogous results can be found for the three remaining quintiles (Others). The finding that recommendations on VALUE stocks exclusively earn positive characteristic-adjusted returns is supported by multivariate results. In particular, the coefficient on VALUE is positive for all analyzed periods and significantly positive for the 6-month period. With regard to sell recommendations, we find complementing evidence for a superiority of VALUE stocks over GLAMOUR stocks. For example, going short in sell recommendations on VALUE stocks will result in an average BHAR 24 of 21.35% in the long-run. In contrast, executing sell recommendations on GLAMOUR stocks will result in a respective characteristic-adjusted return of 4.73%. However, apart from short-selling recommended VALUE stocks, the remaining three quintiles (Others) are associated with high negative and statistically significant BHAR 24 of –12.37% in the long run. Consistently, according to multivariate regression results, sell recommendations on VALUE stocks are associated with negative but insignificant coefficients for all analyzed periods. One might find a reason for the superiority of value stocks over glamour stocks in the information environment of a firm. In particular, value stocks were pretty much out of favor during our investigation period, whereas glamour stocks attracted most of the attention from the financial community. Therefore, our results contradict the anecdotal evidence that profit opportunities arose for biotech and internet stocks. In fact, our results indicate quite the opposite. A reader of the analyzed magazines was well advised not to invest in glamour stocks but rather in value stocks, because the advice from journalists was particularly predictive for this sub-group. Long-Run Performance Evaluation of Journalists’ Stock Recommendations 235 Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 c) Prior Performance The literature on the momentum effect (see, e.g., Jegadeesh/Titman (1993); Rouwenhorst (1998)) shows that stock prices seem to be exposed to short-term and medium-term price drifts. As discussed in Section IV.1 of the paper, journalists seem to follow momentum investment strategies when deciding on stock recommendations, i.e., they have a tendency to recommend past winners for purchase and past losers for sale. Thus, we partition our sample into two sub-groups according to whether a stock has a positive (POSPERF) or negative (NEGPERF) market-adjusted return in the 6-month period prior to the month of publication. Notably, past performance is a highly selective criterion for buy recommendations. Whereas buy recommendations on past winners are associated with significantly positive characteristic-adjusted returns for all analyzed periods, buy recommendations on past losers are associated with negative returns, statistically significant for most periods. In particular, buy recommendations of stocks with a positive prior market-adjusted return earn a BHAR 24 of 8.07%, whereas we document a respective value of –3.67% for stocks with a negative prior performance. This result is supported by multivariate regression results, which reveal consistently positive and statistically significant coefficients for the dummy variable POSPERF. Analogously, past performance also serves as selection criterion with respect to the predictive ability of journalists for sell recommendations. For recommendations on past losers, we document both economically and statistically significant characteristic-adjusted returns for all analyzed periods with a peak for the 24-month period following the event. The respective BHAR 24 is –18.29%. For sell recommendations on past winners, however, characteristic-adjusted returns are close to zero and turn even positive in the long run with an insignificant BHAR 24 of 3.95%. Results are again backed by multivariate regressions as the dummy variable NEGPERF takes on negative and statistically significant coefficients for most analyzed periods. Our finding that only buy recommendations on past winners earn abnormal returns, whereas sell recommendations are only profitable if a stock performed below average prior to the publication might indicate a very pronounced momentum effect for the German stock market. A number of papers has documented a momentum effect in terms of price drifts for the German market (see, e.g., Schiereck et al. (1999); Glaser/Weber 236 Alexander G. Kerl and Andreas Walter Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18 mance of Buy and Sell Recommendations, Financial Analysts Journal 51, 25–39. – Syed,A./Liu,P./Smith, S. (1989): The Exploitation of Inside Information at the Wall Street Journal. A Test of Strong Form of Efficiency, Financial Review 24, 567–579. – White, H. (1980): A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity, Econometrica 48, 817–838. – Womack, K. (1996): Do Brokerage Analysts’ Recommendations Have Investment Value? Journal of Finance 51, 137–167. – Wright, D. (1994): Can Prices be Trusted? A Test of the Ability of Experts to Outperform or Influence the Market, Journal of Accounting, Auditing and Finance 9, 307–323. – Yazici,B./Muradog ˘lu,G.(2002): Dissemination of Stock Recommendations and Small Investors: Who Benefits?, Multinational Finance Journal 6, 29–42. Summary Long-Run Performance Evaluation of Journalists’ Stock Recommendations This paper evaluates the long-run performance of buy and sell recommendations issued by journalists at German Personal Finance Magazines for the first time. We find evidence for journalists providing significant investment value with their recommendations on the sell side since sell recommendations contain high investment value for readers. In contrast, buy recommendations generally contain only little investment value. However, we find that journalists’ predictive abilities differ with respect to specific types of buy recommendations. On the one hand, buy recommendations on value stocks and stocks with a positive performance prior to the publication date are associated with significant investment value for readers. On the other hand, executing buy recommendations on stocks listed at the Neuer Markt would have resulted in serious losses for private investors. (JEL G11, G14) Zusammenfassung Langfristige Performance von Aktienempfehlungen deutscher Börsenmagazine Der vorliegende Aufsatz untersucht erstmalig die langfristige Renditeentwicklung von Kaufund Verkaufsempfehlungen deutscher Anlegermagazine. Die Untersuchungsergebnisse legen nahe, dass die Empfehlungen der Journalisten werthaltig sind. Insbesondere ist das Befolgen von Verkaufsempfehlungen anzuraten. Kaufempfehlungen verfügen hingegen im Allgemeinen lediglich über geringen Wert für Privatanleger. Allerdings zeigen die Journalisten besondere prognostische Fähigkeiten bei einzelnen Typen von Aktien. So lassen sich mit Kaufempfehlungen für Substanzaktien und für Aktien mit einem positiven Renditemomentum signifikante Überrenditen erwirtschaften. Long-Run Performance Evaluation of Journalists’ Stock Recommendations 243 Kredit und Kapital 2/2009 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/kuk.42.2.213 | Generated on 2023-01-16 13:34:18