Week 4: Chapter 9&13 Flashcards
What’s a single cross-section?
Cross-sectional data collected by observing multiple subjects at one single point in time
Pooled-cross sectional data?
sampling randomly from the same population at different points in time
Panel data?
Many cross-sections that show data for the same respondents over different points in time.
Advantage of pooled CS?
Disadvantage and solution for this?
Adv: Samples are independent, also gives us a larger sample thus more precise estimates and more powerful test statistics
Disadv: Samples may not be identically distributed, solution is to control for the time period directly.
Differences in Differences Estimator (DiD) standard model
Useful for testing natural experiments
- see model in notes
Natural Experiments
Always have two groups:
Treatment group: T(dT = 1) and
Control group: C(dC = 0).
To control for systematic differences between T and C, we need two years o fdata
- Pre-policy change (d1=1) and post-policy change (d2=1)
- This gives us 4 groups: the control group before the change, the control group after the change, the treatment group before the change, and the treatment group after the change
How does panel data address OVB?
By splitting the unobserved factors into two groups (i) factors that are constant and (ii) factors that vary across time
Fixed-effects model, what does ai and uit mean?
Find model in notes
ai: the unobserved time constant factors that affect yit.
- Also known as heterogeneity
uit: idiosyncratic/time-varying error, unobserved factors that change overtime and affect yit.
When does heterogeneity bias exist within the FE model?
Corr(vit,xit) does not = 0, bias exists when omitting the ai
First difference model
notes
When is the FD worse than pooled OLS?
The FD estimator can be worse than pooled OLS if one or more of the explanatory variables is subject to measurement error
