Brokers' Recommendations: The Value of a Telephone Tip
基于Dimson和Marsh（1984）对英国券商预测的研究，发现未公开的预测有一定经济价值，但市场效率使公开推荐后收益中性，并讨论了研究设计可能低估经纪人能力的风险。
In a recently published study, Dimson and Marsh (1984) evaluated some 4,000 stock return forecasts produced by 35 leading UK stockbrokers. They found that after the date of the forecast there was a significant degree of out(under-) performance by those stocks with positive (negative) predictions of abnormal return. These observations conflict with the usual outcome from following the published recommendations of brokers and other advisors, whose performance from the date of receiving the recommendation is approximately neutral (see the review of some 30 published studies in Dimson and Marsh, 1984). Evidently the unpublished forecasts, collected by the institution which provided the data, had some real, albeit limited, economic value. The stock market appears not to be wholly efficient in this respect. A large institution which pays substantial commissions would seem to be able to recoup some of (or, sometimes, more than all of) its transaction costs through the superior performance of its purchases. Nevertheless, this violation of strict market efficiency should not be seen as a counter-intuitive result. Since investors compete to acquire, interpret and use relevant information, the stock market can be expected to be relatively efficient. But if investors cease to analyse securities once the marginal benefits of research are outweighed by its marginal cost, the market cannot be perfectly efficient. Consequently, it should be possible to identify research whose marginal benefits (before deducting costs) are positive. The economic value of information decays fast, however, and neutral postpublication performance of stock recommendations is to be expected. Most or all of a broker's or advisor's insights can be expected to manifest themselves in stock market price movements on or before the date of receipt of a published recommendation. Consequently, evaluation of the profitability of published recommendations imposes substantial risks of a Type II error in evaluating brokers' skills. One could easily conclude that brokers have no forecasting ability when, in fact, they have some limited ability. The risk of a Type II error is reduced when unpublished forecasts are investigated, as in the Dimson-Marsh study. However, there is still the possibility that brokers' skills may be understated when advisors are asked for their forecasts at a time of the researchers' choosing. This is because their current recommendations are likely to be pooled with other out-of-date, or as yet inadequately formulated recommendations, in order to respond to the researchers' request for immediate information. It could therefore be argued that the very modest (though statistically significant) level of forecasting skill disclosed in Dimson and Marsh (1984) does less than justice to the brokers. If a broker genuinely has some forecasting skill, he is likely to appear to