what makes underwriting and non-underwriting clients of brokerage firms receive different recommendations? an application of uplift random forest model

I explore company characteristics which explain the difference in analysts’ recommendations for companies that were underwritten (affiliated) versus non-underwritten (unaffiliated) by analysts’ brokerage firms. Prior literature documents that analysts issue more optimistic recommendations to underwriting clients of analysts’ brokerage employers. Extant research uses regression models to find general associations between recommendations and financial qualities of companies, with or without underwriting relationship. However, regression models cannot identify the qualities that cause the most difference in recommendations between affiliated versus unaffiliated companies. I adopt uplift random forest model, a popular technique in recent marketing and healthcare research, to identify the type of companies that earn analysts’ favor. I find that companies of stable earnings in the past, higher book-to-market ratio, smaller sizes, worsened earnings, and lower forward PE ratio are likely to receive higher recommendations if  they are affiliated with analysts than if they are unaffiliated with analysts. With uplift random forest model, I show that analysts pay more attention on price-related than earnings-related matrices when they value affiliated versus unaffiliated companies. This paper contributes to the literature by introducing an effective predictive model to capital market research and shedding additional light on the usefulness of analysts’ reports.