Working Paper: NBER ID: w29148
Authors: Jason Sandvik; Richard Saouma; Nathan Seegert; Christopher T. Stanton
Abstract: In a field experiment, we find large differences in productivity treatment effects between voluntary and mandatory workplace mentorship programs. A significant portion of this difference is due to the best employees opting into the program when it is voluntary and these employees having the smallest treatment effects. Our findings suggest that pilot programs run on a voluntary group may obfuscate large potential gains. In our setting, the firm cannot rely on self-selection to help with program allocation because employees that benefit the most from the program are the least likely to participate. Our findings have implications for program evaluation, experimental design, productivity dispersion, and inequality.
Keywords: workplace programs; mentorship; field experiment; productivity; self-selection
JEL Codes: J24; L23; L84; M5; M53
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| if mentorship had been mandatory (M53) | substantial increases in aggregate productivity gains (O49) |
| voluntary mentorship programs (I29) | smaller treatment effects on productivity (D29) |
| mandatory mentorship programs (M53) | larger treatment effects on productivity (D29) |
| self-selection (C52) | smaller treatment effects on productivity (D29) |
| weaker workers opting out of voluntary programs (J26) | skewed results (C46) |
| opt-out workers (J79) | lower daily revenue and revenue per call (D49) |
| mandatory programs (H53) | greater productivity gains (O49) |