Mutual funds recommendations are a good laboratory in which to test for advertising bias. Product recommendations are a form of content that advertisers might expect to benefit most from biasing. Muti funds are numerous and because they are financial assets their ex-ante and ex-post quality is relatively asy to observe. The availability of detailed data on funds'objective characteristics helps us control for differences in schools of thought about product quality in a way that would be difficult in other settings For example, suppose we found that gm advertised more and received better reviews than Toyota in Muscle Cars. Here one might conclude that Muscle Cars simply likes muscle cars, and GM advertises there to reach its readers. With mutual funds, differences in schools of thought about fund selection are largely over the relative importance of variables we observe in our data, such as past returns and expenses. What we interpret as possible evidence of bias is when, controlling for these and other factors, a publication is more likely to recommend funds from the mutual fund families that have advertised the most within its pages For mutual fund families to benefit from biased recommendations, at least some set of investors must rely upon them. Therefore, after testing for bias, we examine the impact that the mentions in our sample have on fund flows. controlling for past media mentions and a variety of fund characteristics, a single additional positive media mention for a fund is associated with inflows ranging from 7 to 15 percent of its assets over the following 12 months. While investors appear to respond to the media mentions, we find that the media mentions have little ability to predict future returns. Interestingly, this is not due to the pro-advertiser bias When we predict which funds would have been mentioned in the absence of bias, we find little difference in future returns. However, we do find that the personal finance publications would perform significantly better by simply recommending funds with the lowest expense ratios in their investment objectives It is a puzzle why readers respond to recommendations that do not predict future returns. Conditional on their behavior, however, publications selecting funds to mention from the large number available have a valuable set of favors to bestow. In deciding whether to reward an advertiser, publications trade off the benefits(encouraging future advertising) with the costs(harming one's reputation by mentioning a lower quality fund). When there is a wide selection of advertisers funds to recommend, the cost of advertising bias AReuter [2002 finds that advertisers in Wine Spectator receive slightly higher ratings than non-advertisers, controlling for ratings in Wine Advocate, which does not accept advertising. While Wine Spectator claims to use blind tastings to rate wines-a claim intended to increase reader confidence in the ratings-approximately half of the ratings difference is associated with the selective retasting of advertisers'wines. The rest of the rating difference may, in fact, be due to different schools of thought about how to rate wineMutual funds recommendations are a good laboratory in which to test for advertising bias. Product recommendations are a form of content that advertisers might expect to benefit most from biasing. Mutual funds are numerous and because they are financial assets their ex-ante and ex-post quality is relatively easy to observe. The availability of detailed data on funds’ objective characteristics helps us control for differences in schools of thought about product quality in a way that would be difficult in other settings. For example, suppose we found that GM advertised more and received better reviews than Toyota in Muscle Cars. Here one might conclude that Muscle Cars simply likes muscle cars, and GM advertises there to reach its readers.4 With mutual funds, differences in schools of thought about fund selection are largely over the relative importance of variables we observe in our data, such as past returns and expenses. What we interpret as possible evidence of bias is when, controlling for these and other factors, a publication is more likely to recommend funds from the mutual fund families that have advertised the most within its pages. For mutual fund families to benefit from biased recommendations, at least some set of investors must rely upon them. Therefore, after testing for bias, we examine the impact that the mentions in our sample have on fund flows. controlling for past media mentions and a variety of fund characteristics, a single additional positive media mention for a fund is associated with inflows ranging from 7 to 15 percent of its assets over the following 12 months. While investors appear to respond to the media mentions, we find that the media mentions have little ability to predict future returns. Interestingly, this is not due to the pro-advertiser bias. When we predict which funds would have been mentioned in the absence of bias, we find little difference in future returns. However, we do find that the personal finance publications would perform significantly better by simply recommending funds with the lowest expense ratios in their investment objectives. It is a puzzle why readers respond to recommendations that do not predict future returns. Conditional on their behavior, however, publications selecting funds to mention from the large number available have a valuable set of favors to bestow. In deciding whether to reward an advertiser, publications trade off the benefits (encouraging future advertising) with the costs (harming one’s reputation by mentioning a lower￾quality fund). When there is a wide selection of advertisers’ funds to recommend, the cost of advertising bias 4Reuter [2002] finds that advertisers in Wine Spectator receive slightly higher ratings than non-advertisers, controlling for ratings in Wine Advocate, which does not accept advertising. While Wine Spectator claims to use blind tastings to rate wines—a claim intended to increase reader confidence in the ratings—approximately half of the ratings difference is associated with the selective retasting of advertisers’ wines. The rest of the rating difference may, in fact, be due to different schools of thought about how to rate wines. 2