2 - Empirical tools of Public Econ Flashcards
empirical public finance
statistical methods to measure the impact of gov policies on individuals and markets
correlation
2 variables co move
difference between causation and correlation
- either A causes B
- B causes A
- 3rd aspect that causes A and B
causation
the movement of 1 variable causes the movement of another variable
what is an experiment used to test causality
randomised trial
what is a randomised trial
individuals randomly divided into treatment group and control group
- estimating the effects of something on something else
how is bias fixed
- large sample = eliminate differences between groups ‘law of large numbers’
problems with randomised trials
- external validity
- attrition
what is external validity
results are valid for the sample - but the sample is different from the population - doesnt apply to population
what is attrition
individuals leave the experiment before the end
- reduced size of sample
- biased estimates - not true to population
what are the 3 types of observational data
- observational data
- time series analysis = analysis over time
- cross sectional regression analysis = analysis of many individuals at one point in time
what are the problems with time series analysis
- cant tell if correlation is causation
- other factors may cause bias and effect causal interpretation = another variable that is also correlated with the DV
what is the regression line
measures best linear approximation to the relationship between any 2 variables
- if B is statistically significant = there is a relationship
problems with cross sectional regression analysis
- is the correlation causal?
- another variable that is effecting X = in the e = so bias - fix by adding more control variables
what are Quasi experiments
what does it test
natural experiments without randomisation
- treatment group
- control group
- outcome Y
tests the difference in average outcome between treatment and control after the change - but because T and C maybe different before the change you have to also compute difference before