If experience barely changes for any worker, FE estimates will be imprecise. Check within variation via xtsum.
regress gdp fdi trade gcf
Output (illustrative):
| Variable | Coef. | Std. Err. | t | P>|t| | |----------|-------|-----------|----|----| | fdi | 0.142 | 0.021 | 6.76 | 0.000 | | trade | 0.009 | 0.003 | 3.12 | 0.002 | | gcf | 0.047 | 0.011 | 4.27 | 0.000 | | _cons | 6.123 | 0.215 | 28.48 | 0.000 |
R-squared = 0.32
This test determines whether FE or RE is appropriate. stata panel data
Stata Workflow:
* 1. Run Fixed Effects
quietly xtreg y x, fe
* 2. Store the estimates
estimates store fixed
* 3. Run Random Effects
quietly xtreg y x, re
* 4. Store the estimates
estimates store random
* 5. Run the Hausman test
hausman fixed random
Question: Does joining a labor union increase wages, controlling for individual ability?
Data: N=5,000 workers, T=6 years (2015-2020). Variables: wage, union, experience, educ (time-invariant), id, year.
Step 1 – Setup
use union_panel.dta
xtset id year
xtsum wage union experience
Step 2 – Visualize
xtline wage, overlay
Step 3 – Model
* Pooled OLS (with clustering)
regress wage union experience i.year, vce(cluster id)
Step 4 – Interpretation
FE coefficient on union = 0.145 (p<0.01) → Union membership raises wage by 14.5% (semi-elasticity), holding experience and year constant.
Step 5 – Diagnostics
xtserial wage union experience
No serial correlation (p>0.05).
Conclusion: Using Stata panel data methods, we isolate the causal effect of unions.
xtreg ln_wage hours age tenure, fe
estimates store fe
xtreg ln_wage hours age tenure, re
estimates store re
hausman fe re
Rule: If p < 0.05, FE is consistent; RE is inconsistent.
xtreg wage hours tenure age i.year, re