Internet Investment Advice: Investing with a Rock of Salt

Abstract

To investigate the origin and quality of investment advice on the Internet, I collected and categorized the stock recommendations (buy, sell, and neutral) distributed on major newsgroups (online discussion groups) for April 1999 and February 2001 and examined the return characteristics of the recommended stocks. I found, first, that Internet stock recommendations are overwhelmingly positive, with a ratio of buy advice to sell advice greater than 7:1. Moreover, most advisors on the Internet follow a momentum strategy: They recommend stocks after the stocks have experienced sharp price increases. In addition, the stock market does not appear to react to these recommendations. I found the two-day cumulative abnormal returns to be mostly insignificant, regardless of the recommendation characteristics. The longer-horizon returns are not significantly above the market's corresponding performance. In summary, I found no evidence that any new information is exchanged in these forums; the recommendations have no informational value. The fears raised by the media about the destabilizing power of such traders who participate in these discussions are thus groundless.

With the advent of online trading came a multitude of online resources for individual investors, including websites, chat rooms, newsgroups, and mailing lists, through which investors exchange stock recommendations. The popular media view is that these activities are a destabilizing element in the equity markets. I investigated the origin and the quality of investment advice in a specific segment of the Internet.

I collected and categorized the stock recommendations (buy, sell, and neutral) distributed on major newsgroups and examined the return characteristics of the recommended stocks. I chose a month during an up market (April 1999) and a month during a down market (February 2001). An analysis of 33,355 postings yielded 1,106 distinct recommendations during these months. The best-performing sector of the economy during the up market—technology stocks, especially Internet and telecommunications companies—represented more than half of the sample in both months.

The reasons for recommending a company are diverse. Not surprisingly, the most commonly cited reason was a recent price upswing for the company's shares. Demands for additional information ranked next highest, and discussions about the prospects of the company's particular industry followed. The variety of reasons indicates a well-informed set of individuals who take advantage of the wealth of resources freely available online.

In analyzing these recommendations, I addressed the following questions: (1) What drives advisors in their stock picking? (2) Does the market respond to these Internet recommendations? (3) And do the postings have informational value; that is, what is the subsequent performance of the recommended stocks?

I found that newsgroup stock recommendations are overwhelmingly positive, with a ratio of buy advice to sell advice of more than 7:1. This ratio is commensurate with the empirical results for studies of professional analysts' behavior. My finding is striking because the individuals posting on newsgroups (“posters”) are not biased by the agency pressures facing analysts whose firms may be providing investment banking services to the companies they cover. Moreover, most advisors on the Internet follow a momentum strategy. They recommend stocks after the stocks have experienced sharp price increases. Also, the companies they added to and dropped from “coverage” between February 1999 and April 2001 differed significantly in industry-adjusted change in return on equity. Companies added performed significantly better than companies dropped.

The stock market does not appear to react to these recommendations. The two-day cumulative market-adjusted returns (CARs) are mostly insignificant, regardless of the recommendation characteristics. Returns for a longer postrecommendation horizon are not significantly higher than the market's performance. In April 1999, a month during which markets were appreciating, the 20-day CAR accruing from following positive recommendations was 2.50 percent. In February 2001, as the market declined, a similar strategy over the same length of time yielded a 0.44 percent return. When I differentiated between one-time and frequent posters, I found that recommendations from frequent posters earned a significantly higher return than those from one-time posters. Although the recommendations had no informational value, frequent posters did exhibit more skill in selecting securities than did casual posters.

In summary, the posters favored small technology companies with high price-to-book ratios and extremely good performance in the short term. Once returns were adjusted for the market's return, a strategy strictly following the recommendations did not generate either short- or long-term significant returns. Thus, I found no evidence that any new information is exchanged in these forums. Overall, the fears raised by the media about the destabilizing power of traders who participate in these discussions are unwarranted.

I would like to acknowledge financial support from a Harold S. Cooksey Fellowship and the University of Oklahoma, where part of this work was conducted.