Empirical methods take real-world data estimate size of relationship between variables two types regression analysis natural experiments take real-world data estimate size of relationship between variables two types regression analysis natural experiments
Regression analysis two variables: X and Y fit a linear relationship Y = a + bX + u X is independent variable Y is the dependent variable how does a change in X cause Y to change? two variables: X and Y fit a linear relationship Y = a + bX + u X is independent variable Y is the dependent variable how does a change in X cause Y to change?
Y = a + bX + u get data on Y, X multiple observations use regression analysis to estimate a and b Y = a + bX + u get data on Y, X multiple observations use regression analysis to estimate a and b
multiple regression many independent variables X1, X2, X3, X4, … how many indep. variables to include? tough question if wrong, estimates may be biased multiple regression many independent variables X1, X2, X3, X4, … how many indep. variables to include? tough question if wrong, estimates may be biased
Natural experiment In medical research: random assignment experiment one group gets treatment one group gets nothing people randomly assigned to group then different outcomes due to treatment In medical research: random assignment experiment one group gets treatment one group gets nothing people randomly assigned to group then different outcomes due to treatment
in economics: natural experiment like random assignment researcher did not plan experiment the experiment results from an event or policy natural experiment like random assignment researcher did not plan experiment the experiment results from an event or policy
example: EITC income subsidy to low income families expanded 1986 women w/ children get EITC women w/ out children do not EITC increases labor force participation income subsidy to low income families expanded 1986 women w/ children get EITC women w/ out children do not EITC increases labor force participation