on matched samples. For each publication, we match funds from families that advertised over the prior 12 months to non-advertiser funds with approximately the same propensity to be mentioned(based on predicted values from equation(1)when we set the coefficient on own-publication advertising to zero). In unreported tests, the coefficients on own-publication advertising are economically and statistically significant for positive entions in all three personal finance publications. Fourth, we attempt to control for any long-run affinity of a publications readers for a particular fund family by analyzing changes in advertiser status. We classify a family as starting to advertise in publication k in month t if the family has positive advertising expenditures in the publication in month t but not in months t-5 through t-1. Similarly, we classify a family as stopping advertising in publication k in month t if the family has positive advertising expenditures in month t-5 but not in months t-4 to t. In unreported tests, we find that families that start advertising in Money in the six months prior to the Money 100 list are more likely to appear on the list than families that do not start advertising. Similarly, we find evidence that families that start advertising in SmartMoney in month t are more likely to receive positive mentions in SmartMoney over the following six months relative to families that do not advertise. Finally, we find weak evidence(at the 20-percent level in a two-sided test)that families that stop advertising in Kiplinger's are less likely to receive positive mentions over the following six months Overall, our findings are consistent with own-publication advertising expenditures influencing fund kings at the three personal finance publications in our sample. Without purely exogenous variation advertising, we cannot entirely eliminate the possibility that the correlations we observe are being driven by the endogeneity of advertising. However, based on the robustness of the positive correlation between own-publication advertising and content for the personal finance publications-and the absence of a positive correlation for the newspapers, which receive derably less advertising revenues from mutual funds and otherwise serve as a falsification test--we believe that the most likely explanation is the causal one 15 As an alternative robustness test ether Money receives a dispropor- ionate share of mutual fund print advertising in the months preceeding the publication of the Money 100, as one might expect adverters were attempting to influence its on. We find that Money receives 8.9% of advertising in the three prior to the publication of the list, 11.7% in the publication month, and 7. 7% in the three months after. The difference between the pre and post-list window is statistically significant and robust to window lengths of 1, 2, and 4 months, or to the inclusion month fixed effects(that identify the month the list is published varies from year to year 11on matched samples. For each publication, we match funds from families that advertised over the prior 12 months to non-advertiser funds with approximately the same propensity to be mentioned (based on predicted values from equation (1) when we set the coefficient on own-publication advertising to zero). In unreported tests, the coefficients on own-publication advertising are economically and statistically significant for positive mentions in all three personal finance publications. Fourth, we attempt to control for any long-run affinity of a publication’s readers for a particular fund family by analyzing changes in advertiser status. We classify a family as starting to advertise in publication k in month t if the family has positive advertising expenditures in the publication in month t but not in months t − 5 through t − 1. Similarly, we classify a family as stopping advertising in publication k in month t if the family has positive advertising expenditures in month t − 5 but not in months t − 4 to t. In unreported tests, we find that families that start advertising in Money in the six months prior to the Money 100 list are more likely to appear on the list than families that do not start advertising.15 Similarly, we find evidence that families that start advertising in SmartMoney in month t are more likely to receive positive mentions in SmartMoney over the following six months relative to families that do not advertise. Finally, we find weak evidence (at the 20-percent level in a two-sided test) that families that stop advertising in Kiplinger’s are less likely to receive positive mentions over the following six months. Overall, our findings are consistent with own-publication advertising expenditures influencing fund rankings at the three personal finance publications in our sample. Without purely exogenous variation in advertising, we cannot entirely eliminate the possibility that the correlations we observe are being driven by the endogeneity of advertising. However, based on the robustness of the positive correlation between own-publication advertising and content for the personal finance publications—and the absence of a positive correlation for the newspapers, which receive considerably less advertising revenues from mutual funds and otherwise serve as a falsification test—we believe that the most likely explanation is the causal one. 15As an alternative robustness test suggested by an anonymous referee, we examined whether Money receives a dispropor￾tionate share of mutual fund print advertising in the months preceeding the publication of the Money 100, as one might expect if adverters were attempting to influence its composition. We find that Money receives 8.9% of advertising in the three months prior to the publication of the list, 11.7% in the publication month, and 7.7% in the three months after. The difference between the pre and post-list window is statistically significant and robust to window lengths of 1, 2, and 4 months, or to the inclusion of month fixed effects (that we can identify since the month the list is published varies from year to year). 11