Online Finance Research Seminar: “Treatment of Outliers: Firm Heterogeneity in Managerial Incentive and Corporate Innovation” by Kam-Ming WAN – Hanken School of Economics
Speaker: Kam-Ming WAN
Hanken School of Economics
Date and Location – Thursday November 5th 2020 from 12:30 to 14:00 on ZOOM
Abstract:
We propose a 3D approach to treat outliers because outliers are research-question dependent (RD), model dependent (MD), and estimation dependent (ED). Our approach is applied to estimation of the effects of CEO equity incentives on corporate innovation to handle two data problems: Corporate innovation data have a discrete spike at zero and equity incentives data have outliers.
We focus our research question on only firms in innovative industries because equity incentives are relevant for corporate innovation only for firms doing business in industries where innovation matters. In contrast, many firms in noninnovative industries spend nothing in innovative activities, leading to a discrete spike at zero phenomenon. We use a mixture distribution model to distinguish firms in innovative industries from firms in noninnovative industries. We perform logtransformation on key variables of interest to mitigate influence of extremely large values. As outliers are also dependent on estimation method, we use robust regression techniques including MM estimation which is robust to vertical outliers (outliers in y direction) and high leverage points (outliers in x direction); and median regression which is robust to vertical outliers but fragile to high leverage points. After applying the 3D approach, our results indicate that higher sensitivity of CEO wealth to stock volatility (vega) enhances more corporate innovation only for firms within innovative industries but this positive relation disappears if outliers are not treated.