The difference in bias and inaccuracy between commissioned and traditional sell-side equity research : evidence from Nordic investment banks after MiFID II implementation

Abstract

This thesis investigates whether there is a difference in the bias and inaccuracy of the EPS estimates in sponsored and traditional sell-side equity research. As a result of the EU regulation MiFID II’s requirement that asset managers and broker-dealers unbundle the cost of investment research from the cost of trade execution, several Nordic investment banks have begun offering research that is paid for by the covered company. Naturally, some concern has been raised as to the independence and bias of this type of research and the conflicts of interest that may arise. This paper examines the validity of this concern by comparing bias and inaccuracy of EPS estimates in the two types of research made by five Nordic investment banks: Danske Bank, DNB, SEB, ABG, and Nordea. To perform this comparison, we constructed a proprietary dataset including EPS estimates and actuals, firm-characteristic variables, and a dummy variable indicating whether a specific estimate belongs to sponsored or traditional sell-side research. We then estimated five multiple fixed effects regression models on three different datasets to determine whether there was a significant difference in bias and inaccuracy between the sponsored and traditional sell-side samples. Since the sponsored sample is substantially smaller than the traditional sell-side sample, we entropy balanced the samples in our regressions. From our descriptive statistics, we find that companies paying for sponsored research are on average smaller, younger, have more volatile earnings and returns, are less levered, have less institutional ownership, and have less analyst coverage than companies covered by traditional sell-side research. When contrasting the frequency of industry observations in our sponsored and traditional samples, companies in the Software and Healthcare-related industries are among the most frequently observed in the sponsored sample. In contrast, Oil & Gas, Industrial Machinery, and Packaged Foods and Meat companies are among the most frequently observed in the traditional sample. Furthermore, the share of Swedish companies in the sponsored research sample indicates that sponsored research is more common in Sweden than in the other Nordic countries. Lastly, we find indications that the number of EPS estimates per company is lower for the sponsored sample than the traditional sell-side sample. Overall, in our regressions, we fail to find sufficient evidence indicating that the inaccuracy and bias in the sponsored research sample is significantly different from the inaccuracy and bias in the traditional sell-side research sample. There are some weak indications that, if anything, sponsored research is more accurate and less positively biased than traditional sell-side research. Followingly, we conclude that sponsored and traditional sell-side EPS estimates appear to be of similar quality.