Table 4.
Determinants of Mobility: Demographic, Productivity & Peer Measures
| Movers + Stayers | Subsample w/Kid Info | ||||
|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | |
| Demographics | |||||
| Female | −0.0013 (0.0012) | −0.0005 (0.0013) | −0.0001 (0.0013) | −0.0001 (0.0013) | −0.0047+ (0.0028) |
| PhD (MD omitted) | −0.0014 (0.0009) | −0.0047** (0.0010) | −0.0066** (0.0010) | −0.0064** (0.0011) | −0.0081** (0.0017) |
| MD/PhD | −0.0007 (0.0015) | −0.0053** (0.0017) | −0.0056** (0.0017) | −0.0045** (0.0017) | −0.0048+ (0.0027) |
| Productivity Measures | |||||
| Ln(Pubs_t-1) | −0.0008 (0.0007) | 0.0005 (0.0007) | 0.0002 (0.0007) | −0.0009 (0.0011) | |
| Ln(Stock Pubs_t-2) | 0.0031** (0.0007) | 0.0045** (0.0008) | 0.0035** (0.0008) | 0.0048** (0.0014) | |
| Ln(NIH Funding_t-1) | −0.0010** (0.0001) | −0.0010** (0.0001) | −0.0010** (0.0001) | −0.0009** (0.0001) | |
| Ln(Stock NIH Funding_t-2) | 0.0009** (0.0001) | 0.0008** (0.0001) | 0.0006** (0.0001) | 0.0004* (0.0002) | |
| Collaborators | |||||
| Ln(Pubs), Colocated | −0.0043** (0.0003) | −0.0014** (0.0004) | −0.0011+ (0.0006) | ||
| Ln(Pubs), Close | −0.0020** (0.0005) | −0.0022** (0.0005) | −0.0012 (0.0009) | ||
| Ln(Pubs), Distant | 0.0017** (0.0004) | 0.0015** (0.0004) | 0.0016* (0.0007) | ||
| Non-collaborating Peers | |||||
| Ln(Pubs), Colocated | −0.0073** (0.0004) | −0.0064** (0.0006) | |||
| Ln(Pubs), Close | −0.0000 (0.0003) | 0.0000 (0.0005) | |||
| Ln(Pubs), Distant | 0.0046** (0.0008) | 0.0035** (0.0012) | |||
| Nb. of Observations | 189,884 | 174,144 | 174,107 | 174,107 | 62,181 |
| Nb. of Job Spells | 10,273 | 10,273 | 10,273 | 10,273 | 3,316 |
| Nb. of Scientists | 9,378 | 9,378 | 9,378 | 9,378 | 2,977 |
Notes: The dependent variable is a binary variable that takes on a value one in the year we observe the elite scientist moving to a new academic position located at least 50 miles away. Estimation is by logit and marginal effects are reported. All specifications include full age, vintage category, and year fixed effects. Robust standard errors are in parentheses, clustered at the individual level. See Section III for a full description of how the sample and variables were constructed.
+
p < 0.10
*
p < 0.05
**
p < 0.01