Global/Industrial Diversification and Analyst Herding

The study we report examined herding behavior among security analysts in the 1980–98 period. Using panel data, we analyzed the impact of industrial and geographical diversification on herding among analysts. We compared the propensity toward herding of analysts covering domestic companies that are industrially focused and analysts focusing on companies that are diversified geographically and/or industrially. We provide evidence that herding is more pronounced among analysts concentrating on diversified companies. This result is consistent with the notion that herding behavior increases with task difficulty. Our results also show that herding among analysts reduces the market value of companies they cover, which indicates that the market penalizes herding behavior among security analysts. This effect was more pronounced for companies that are industrially and/or geographically diversified. Given the recent market uneasiness about the quality of security analysts' services in the United States, an important issue is whether a relationship exists between the quality of analyst coverage and characteristics of the companies analysts choose to cover. The ability of security analysts to disseminate information and provide monitoring of corporate managements depends on a company's degree of organizational complexity and potential for agency conflicts. These characteristics, which are more common among diversified companies, could lead to large forecast errors coupled with unusually high levels of consensus in analyst forecasts. This phenomenon is referred to as “herding.” Because such behavior is synonymous with ineffective analyst coverage, herding could account for differences in valuations in a cross-section of companies with different degrees of diversification.We used security analysts' summary earnings forecast data for a large panel of companies in the 1980–98 period to detect whether security analysts' herding behavior is related to characteristics of the companies they follow. We operationalized herding among analysts as cases of clustering of analyst earnings forecasts coupled with large forecast errors. We maintain that, given the different nature of diversified and nondiversified companies, the extent and quality of analyst coverage of domestic companies that are industrially focused is different from the extent and quality of analyst coverage of diversified companies. Consequently, we expected analysts' propensity toward herding to be different when they were covering domestic/industrially focused companies from when they were covering geographically and/or industrially diversified companies.Diversification should exacerbate analysts' tendency toward herding because it increases the complexity and difficulty of the analyst's task. Diversified companies are generally larger, have more-complex organizational structures, and have less-transparent operations. In addition, they are more likely to exhibit agency conflicts and problems of informational asymmetry. Therefore, we argue that analysts following companies that are focused in a single line of business and that operate exclusively in the domestic U.S. market should display less herding than analysts covering companies that are industrially diversified, geographically diversified (multinationals), or both. If herding behavior serves as a proxy for analyst ineffectiveness (i.e., subpar monitoring and information dissemination), the differential valuations of diversified versus nondiversified companies is partly driven by differences in analyst propensity toward herding behavior.We used measures of forecast dispersion and forecast error to devise two measures of analyst propensity toward herding (herding indexes) for each company. We then used panel-data regressions to test the following hypotheses: (H1) Herding behavior is more pronounced among analysts following diversified companies than among those following domestic, industrially focused companies.(H2) Herding increases with the degree of both geographical and industrial diversification.(H3) Market value declines with analyst propensity toward herding.(H4) The impact of herding on market value is stronger for diversified companies.Our empirical results support all four hypotheses. Our findings indicate that analyst forecasts for geographically or industrially diversified companies display more herding, on average, than forecasts for domestic/focused companies. The results also provide strong support for the notion that herding increases with the degree of both industrial and geographical diversification. In addition, when we estimated two-stage least-squares regressions, we found that herding results in lower market valuations, which implies that the market penalizes security analysts' propensity toward herding. We found this effect to be stronger in the case of diversified companies.