F11 Panel data I

Question 1

What is panel data?

Answer 1

Multiple units (i = 1, 2, 3…) observed in multiple time periods (t = 0, 1, 2…).

Question 2

What is the upside to panel data?

Answer 2

Causal inference is possible under specific conditions: Units are the same for different time periods and both observed and unobserved confounders are time-invariant.

Question 3

What is time series analysis?

Answer 3

Analyzing one unit over time

Question 4

What is a balanced panel?

Answer 4

Complete data for all time periods (no attrition). If there is attrition then the panel is unbalanced

Question 5

Why is it important to differentiate between time variant and time invariant confounders?

Answer 5

Time invariant confounders are constant within a unit over time and can be discarded through fixed effects.

If a confounder is timevariant pose a risk for the causal estimate.

Question 6

How is timevariant confounders indicated in panel data models in DAG?

Answer 6

If the subscript is it and not only i

Question 7

What is fixed effects?

Answer 7

Each unit have an intercept that is fixed for all time periods. The intercept absorbs all timeinvariant confounders.

It allows for different baselines.

Question 8

How can you handled panel data? Two ways

Answer 8

Fixed effects (unit-specific intercept) or pooled regression (global intercept). The latter is problematic as you ignore the potential upside with panel data structure from a causal inference point of view.

Question 9

What does pooled regression not include?

Answer 9

Unit-specific fixed effects, period-specific fixed effects and no time trends

Question 10

How can you see whether an intercept is global or unit-specific?

Answer 10

Global: α (no subscript)
Unit-specific: u_i (subscript for unit)

Question 11

What challenges do panel data face? Or what does it not solve?

Answer 11

Reverse causality

Unobserved time variant confounders

Degrees of freedom (you need 10-15 observations per unit for FE)

Question 12

What is twoway fixed effects?

Answer 12

A baseline approach for panel data in political science. Fixed effects for both unit and time.

Controlling for both unit invariant confounders and exogenous time shocks for all units.

Could also include a time trend

Question 13

Why is degrees of freedom important with panel data?

Answer 13

For each unit- or time specific fixed effect estimated we lose degrees of freedom.

Therefore, leave out the global intercept. Year fixed effects could be left out if there are strong theoretical arguments.

Question 14

What is degrees of freedom?

Answer 14

The number of observations in relation to number of estimated parameters.

df = n - parameters

Question 15

What is assumed with two-way fixed effects?

Answer 15

That treatment effects is similar across time because the model estimates a single, average treatment effect across all time periods without distinguishing time-specific variation in treatment effects.

(may be solved with a time trend?)

Question 16

Explain fixed effects with demeaned variables

Answer 16

(Yit - Yi-bar) = delta (Dit - Di-bar) + (ui - ui) + (epsilon_it - epsilon_i-bar).

Yi demeand = delta Di demeand + epsilon i demeaned

Same as unit dummies

Question 17

What kind of standard error do you want to use with panel data?

Answer 17

Cluster robust by panel unit id.

This it to allow for correlation in the error term for the same person over time. This yields valid inference so long as the number of clusters is “large”.

Question 18

What happens if there is no variation in treatment status for a unit?

Answer 18

The estimator can not disentangle effect and unobserved confounding. BUT

Even if an individual’s treatment does not vary over time, as long as there is sufficient variation in treatment across individuals (and across time), the fixed-effects estimator can still use this to identify the causal effect.