The effects of MiFID II on sell-side analysts, buy-side analysts, and firms

A directive is a legislative act that sets out a goal that all EU countries must achieve. Each individual member country devises its own laws on how to reach these goals (https://​europa.​eu/​european-union/​eu-law/​legal-acts_​en). We use Europe, European Union, and EEA interchangeably throughout the paper.

Of interest to researchers is also the fact that analyst forecast data in IBES Detail have changed due to MiFID II. Thomson Reuters, the owner of IBES, issued a product change notification on September 12, 2018, announcing that the contributor and analyst names of 88 pre-approval contributors will be anonymized for all clients, regardless of individual entitlements. In addition, estimates from UBS Equities will be removed from the IBES Detail History file and they explicitly tie this decision with MiFID II implementation (see https://​uk.​reuters.​com/​article/​uk-ubs-research-memo/​ubs-suspends-access-to-research-data-for-some-external-providers-idUKKBN1HG2O3). The IBES Summary History file does not change, and consensus estimates will continue to include all pre-approval brokers (as well as UBS Equities).

For example, the following studies examine the impact of these regulations: Barniv et al. (2009); Kadan et al. (2009); Clarke et al. (2011); Guan et al. (2012); Corwin et al. (2017); Hovakimian and Saenyasiri (2010, 2014). As the relationship with investment banking is the “primary aim” of these regulations (Groysberg and Healy 2013), this literature focuses on sell-side independence within investment banks. In comparison, MiFID II aims to address a different type of conflicts of interest (i.e., sell-side research used as an inducement to attract buy-side trade execution business), through a different mechanism (i.e., unbundling).

The relevant articles are 23, 24(4)(c) and 24(7)(b). The new regulations apply directly to financial-market players that are based in any of the EEA member states as well as to a European branch location of any company headquartered outside of the EEA.

While asset managers and stockbrokers usually rely on long-term contracts that specify the overall payment and negotiate commissions infrequently (Goldstein et al. 2009), the separate pricing of research required by MiFID II imposes such a negotiation. Anecdotal evidence from Cenkos Securities Plc shows a 7% decrease in research commissions in 2018, compared to 2017, that is “a direct result of MiFID II” (https://​www.​cenkos.​com/​~/​media/​Files/​C/​Cenkos-Securities/​reports-and-presentations/​2018/​Cenkos%20​Financial%20​Report%20​2018.​pdf, page 1). The lack of disaggregated disclosure of research and trading incomes precludes a systematic comparison.

https://​www.​sec.​gov/​divisions/​investment/​noaction/​2017/​sifma-102617-202a.​htm, last accessed November 27, 2019.

Large asset managers, such as Capital Group ($1.87 trillion), have declared they will absorb the cost of third-party investment research across their global business (Riding 2019a). This comes as U.S. clients, having noticed a drop in European transaction fees and commissions, are becoming concerned that they subsidize European clients (Riding 2019a). As a result, the Council of Institutional Investors has been lobbying the SEC to adopt regulations similar to MiFID II in the United States (Riding 2019b).

https://​www.​cnbc.​com/​video/​2017/​11/​03/​mfid-ii-treats-both-banks-and-clients-like-kids-erste-bank-ceo.​html, last accessed April 4, 2019.

The MiFID II requirements to unbundle research costs and justify the usefulness of research address a different type of conflict of interest, compared to the NASD Rule 2711 and the Global Settlement adopted in the United States in 2002–2003, which targeted the conflict of interest generated by funding equity research with investment banking revenue. Previously, sell-side analysts in investment banks were incentivized to cater to the investment banking business of their employer through overly optimistic stock recommendations, since their compensation was determined by their ability to attract investment banking business (e.g., Guan et al. 2012). After these regulations, the investment banking and research departments became physically and financially separated. To a large extent, these regulations applied internationally: the Global Settlement had spillover effects internationally, as the 12 banks involved adopted the same separation across all their subsidiaries (Hovakimian and Saenyasiri 2014), the International Organization of Securities Commissions (IOSCO) issued guidelines and best practices in the spirit of NASD Rule 2711, and the EU Market Abuse Directive contained similar recommendations. (See the chapter on Foreign Regulatory Initiatives in https://​www.​finra.​org/​sites/​default/​files/​Industry/​p015803.​pdf.)

These arguments still hold for independent research firms that have always functioned as profit centers because MiFID II leads to a decrease in overall demand for research.

Less publicly available research might primarily affect retail investors. We thank Richard Sloan for this point.

Iceland did not implement MiFID II, and so we exclude it.

For example, if a firm has no sell-side coverage for 2015, has coverage for 2016, and no coverage afterward, we include it in the sample starting in 2016, with zero coverage for 2017 and 2018.

We use a pre-trend test to provide insight into the parallel-trend assumption by including indicators for the pre-period years interacted with TREAT (Angrist and Pischke 2009). Untabulated results show that, before treatment, the treated and control groups are parallel for the sell-side and buy-side dependent variables.

Our tests include firm fixed effects that control for industry. Inferences are unaffected if we exclude finance.

Panel D of Table 2 shows that the matching process removes the differences in analyst coverage, raising external financing, and the last six months’ stock returns between the two samples prior to the treatment and reduces to a large extent the economic differences on other dimensions (i.e., most variables are very close in terms of actual magnitudes). Our inferences are robust to stricter caliper choices, such as 0.01, 0.005, or even 0.001. Using a stricter caliper further reduces the differences between the two groups but also decreases the sample size significantly. For example, using a strict caliper of 0.001 removes the statistical differences between the treated and the control group prior to the treatment for size, coverage, ROA, firm age, leverage, R&D, stock returns, and earnings volatility. But the strict caliper choice reduces sample size to 7641 observations. Again, our inferences remain the same using this matching choice.

We do not tabulate analyst-firm level tests on a propensity-score-matched sample, as this requires a whole new set of variable selections for the stage-one model and balance checks. Matching on firm characteristics, analyst experience, number of covered firms, and brokerage size results in a sample of 95,129 observations. Using Forecast Error as the dependent variable in the most restrictive research design (with firm-year-analyst fixed effects and standard errors clustered by analyst and firm), the coefficient on TREAT×POST is −0.018, marginally significant (t-stat −1.85). Using a similar matching procedure and adding analyst coverage as matching dimension on the recommendation-level sample yields 43,621 matched observations. When Abs CAR (0;+1) is the dependent variable, the coefficient on TREAT×POST is not significant at conventional levels. Using Industry Recommendation as dependent variable in the same restrictive research design, the coefficient on TREAT×POST is 0.026, marginally significant (t-stat 1.75). Overall, the propensity-score-matched results lend further, albeit weaker, support to Prediction 3.

Groysberg et al. (2008) and Groysberg et al. (2013) provide the first insights into buy-side research using proprietary data from a large investment firm between 1997 and 2004. Brown et al. (2016) provide cross-firm evidence on buy-side research using surveys.

We remove 263 investment firms that operate simultaneously in European countries and the United States or Canada.

We implement the following procedure to identify the sell-side and buy-side analysts. (1) We match by name and employer the list of noncorporate participants with the list of sell-side analysts obtained from FactSet Contact Screening. (2) We match the remaining participants by name and employer with a list of buy-side managers and analysts obtained from FactSet Contact Screening, Thomson Reuters, and S&P Capital IQ. (3) We match the remaining participants by employer with the list of buy-side and sell-side firms from the same sources. (4) We isolate the remaining nonmatched participants and manually code them as sell-side, buy-side, or other by searching for their employment information via LinkedIn, Bloomberg, and corporate websites. We manually coded over 9000 individuals.

About half of our sample does not hold earnings calls. Inferences remain unchanged if we drop missing firm-years.

The United Kingdom makes up a third of our European sample (Panel B of Table 1). Consequently, one might argue that our findings could be affected by Brexit. Results based on continental European firms as the treated sample provide similar inferences as those reported throughout the prior tables. Hence it is unlikely that our findings are solely driven by U.K. firms.

Whether the newly hired buy-side analysts are former sell-side analysts is an interesting question. However, the costly data collection necessary to answer this question places it beyond the scope of this paper.

A new equilibrium may emerge in the long term, once the various market participants have adjusted to the new regulation. We leave this to future research.