Empirical Tools

Question 1

Identification problem

Answer 1

Correlation (move together) vs causation (movement causes movement of another)

Question 2

How to solve?

B) issue

Answer 2

Randomised trial - treatment vs control group

B) bias - ensuring they are the same without the treatment is hard
(Also can be time consuming expensive unethical)

Question 3

How to treat bias:

Answer 3

Larger sample sizes can eliminate consistent differences via law of large numbers: (odds fall as sample size grows)

Question 4

2 technical issues with randomised trials (even with gold standard and large sample to eliminate bias)

Answer 4

External validity: only valid for sample tested, may be different from population at large

Attrition: individuals may leave before experiment complete, if not random can create bias estimates

Question 5

2 Problems with time series analysis (comparing 2 variables movement overtime)

Answer 5

Does not mean causation, they may just move together

Other factors get in the way to test causation since they may be also correlated with the variables of interest; hard to control for all variables

Question 6

So when is time series useful for causation

Answer 6

SHARP BREAK IN DATA more likely to suggest causation

E.g in 1993 simultaneous changes in price increase of cigarettes and the fall in smoking rates

Question 7

Cross-sectional regression analysis

Answer 7

Analysis between 2 or more variables, exhibited by many individuals at ONE point in time

Question 8

Problems with cross-sectional regression analysis

Answer 8

Reverse causation - does X cause Y or Y cause X

Question 9

Control variables

Answer 9

Account for differences between treatment and control groups that can lead to bias

Question 10

Quasi-experiments (natural) uses what

Answer 10

Difference in difference estimator

Question 11

DID estimator

Answer 11

(Y tafter - Y cafter) - (Y tbefore - Y cbefore)

Difference between treatment and control after, minus difference between treatment and control before

See if the difference between treatment and control has changed since in the policy change

Question 12

Quasi-experiment problems

Answer 12

Never can be sure control variable has got rid off all bias

External validity

Question 13

How do quasi-experiments try to mitigate this (ensuring bias all removed)

Answer 13

Robustness checks e.g find alternative control groups, do a placebo comparing treatment

Question 14

Best way to check validity of DD estimator

Answer 14

Plot time series to see if there is a clear break between the 2 groups at the time of reform