Intro Panel data

Question 1

How do the ALLBUS, PISA & SOEP differ?

Answer 1

If units are persons and time is years (e.g., SOEP) -> A row in the dataset is a person-year

• ordering

Question 2

Micro vs macro panel

Answer 2

• In micro panel much more units (20.000 persons) than individual timepoints
• In macro panels still more units (200 countries) but not as much as in micro, also, could be same country but different individuals every other year (ESS)

Question 3

Unbalanced vs balanced panel data

Answer 3

Question 4

What are benefits vs challanges of panel data?

Overview

Answer 4

Question 5

What are benefits vs challanges of panel data?

Study of change

Answer 5

-

Question 6

What are benefits vs challanges of panel data?

Identification of causal effects

Answer 6

better grip with panel data
1) time: u can determine what event came first (i.e. unemployment or divorce)
2) if u do not have panel data –> left with between person comparison -> extremel hard to infer causality from that only if statistical twins
3) even better would be: almost randomized control trials (did)

Question 7

Why can’t we measure this question by comparing radio and non-radio owners?

+ Solution

Answer 7

• self-selection into radio ownership (i.e. education, income) –> classic confounding problem
• there needs to be absolute unit homogeneity except for radio ownership -> quite strong assumption

Solution: Ideally observe the same individual in two different states, which you can never do –> best shot: Within-comparison
you still make an assumption: temporal homogeneity so that the individual did not change in time (much weaker assumption than arguing for statistical twins)

Question 8

What are benefits vs challanges of panel data?

Cost of data collection

Answer 8

Question 9

What are benefits vs challanges of panel data?

Reliability and validity of constructs

Answer 9

Reliability: different time points, measurement error on average equals out (sometimes higher, sometimes lower)

much better view whether concept is stable or variant (i.e. big five were assumed to be stable but they do change)

Practical problem: human error, personal change may also lead to change due to frasing question differently

Question 10

What are benefits vs challanges of panel data?

Inference to a population

Answer 10