Empirical Tools of Public Economics. Part-2

Transcription

1 Empirical Tools of Public Economics Part-2

2 Outline 3.1. Correlation vs. Causality 3.2. Ideal case: Randomized Trials 3.3. Reality: Observational Data

3 Observational data: Data generated by individual behavior observed in the realworld, not in the context of deliberately designed experiments. Time-series analysis Cross-sectional analysis Difference-in-differences (DD)

4 Time-Series Analysis Time-series analysis: Analysis of the comovement of two series over time. TANF Example: We can collect data over time on the benefit guarantee each year and compare this to the amount of labor supply delivered by single mothers in those same years.

5 Time-Series Analysis TANF Example:

6 Time-Series Analysis Given this data, what can we say about the relationship between TANF benefits and the labor supply of single mothers: There is definitely a negative relationship between the hours worked and the amount of TANF benefits. Benefits declined dramatically from $991 in 1968 to $515 in 1998 whereas the hours worked increased substantially from 1063 hours in 1968 to 1294 in 1998.

7 Time-Series Analysis Is this relationship causal? Is it necessarily the decrease in the TANF benefits that is causing the increase in the hours worked? Not necessarily.

8 Time-Series Analysis Problems with Time-Series Analysis There might be other things changing over this time period that is leading to the increase in the labor supply. Increasing acceptance of women in the workplace. Better and more options for child care. Possibly more pressure on women to work.

9 Time-Series Analysis Examine the period

10 Time-Series Analysis Examine the period

11 Time-Series Analysis Examine the period Seems to support the negative relationship: fall in benefits by 10% and sharp rise (13%) in labor supply.

12 Time-Series Analysis Examine the period Alternative explanations to the increase in labor supply: Dramatic growth during this time period: unemployment rate falls from 7.3% in January, 1993 to 4.4% in December, Enormous increase in the Earned Income Tax Credit (EITC), another assistance program for single mothers. Therefore, the increase in labor supply might not be due to the decline in TANF benefits

13 Time-Series Analysis When is time-series analysis useful? When there are sharp breaks in the trends of the variables over a narrow period of time Price war Price settlement

14 Cross-Sectional Regression Analysis Cross-sectional analysis: Statistical analysis of the relationship between two or more variables exhibited by many individuals at one point in time.

15 Cross-Sectional Regression Analysis NBA Example: Draft Number vs. Annual Salary Question: Does being drafted higher cause higher annual salaries for NBA players? At a given point in time, observe the draft numbers and annual salaries for NBA players

16 Cross-Sectional Regression Analysis NBA Example: Annual Salary- Thousand $ Draft Number

17 Cross-Sectional Regression Analysis NBA Example: Linear regression line: The line that measures the best approximation to the relationship between any two variables. yi = β 0 + β1x i + εi where y i is the outcome of interest (annual salary), β 0 is the constant term, β 1 is the parameter of interest, X i is the draft number and ε i is the error term.

18 Cross-Sectional Regression Analysis NBA Example: β β0 1 How do we find and? Using ordinary least squares (OLS): finds the regression line that minimizes the sum of squared distances from the regression line to the observations (the sum of squared errors). 2 min εi min β0, β1 β0, β1 ( y β β X ) 2 i 0 1 i

19 Cross-Sectional Regression Analysis NBA Example-Regression Line Draft Number Fitted v alues Annual Salary- T housands $

20 Cross-Sectional Regression Analysis Problems with Cross-Sectional Analysis It is quite likely that the intrinsic ability of the player, which is not controlled for in the regression model, both causes the player to be drafted higher and earn higher salary. What can we do about this?

21 Cross-Sectional Regression Analysis Control variables: Variables that are included in cross-sectional regression models to account for differenced between the treatment and control groups. Treatment in our case: being drafted higher In this case, we can use points per game, rebounds per game and assists per game as proxies for intrinsic ability, which is likely to be different between the treatment and control groups.

22 Quasi-Experiments Quasi-experiments: Changes in the economic environment that create nearly identical treatment and control groups for studying the effect of that environmental change, allowing public finance economists to take advantage of randomization created by external forces. In short, these are randomized experiments with observational data

23 Quasi-Experiments Quasi-experiments-Example The impact of zero-tolerance laws on young adult fatal traffic accident rate Florida passed the zero-tolerance law in 1995 whereas Georgia passed it in Observe the fatality rates in Florida and Georgia in 1993 and 1997.

24 Quasi-Experiments Quasi-experiments-Example Treatment group: Florida Control group: Georgia Compare the changes in the fatalities between the two years for the two states

25 Quasi-Experiments Quasi-experiments-Example Difference Florida 25 per per per Georgia 34 per 33 per 1 per Difference-in- Differences (DD) 3 per 10000

26 Quasi-Experiments Quasi-experiments-Issue Is the treatment and control groups comparable? Economic growth in that period Increase in alcohol consumption among youth If growth affects Florida and Georgia differently, the DD estimator will give you the sum of the treatment effect and the bias from the economic growth.

27 Structural Estimates So far, we only looked at reduced form estimates, which measures the total impact of an intervention. On the other hand, there might be cases where the primary interest is how the intervention changes individuals behavior. Structural estimates: Estimates of the features that drive individual decisions

28 Structural Estimates Structural estimates: Estimates of the features that drive individual decisions, such as income and substitution effects or utility parameters. Assume the form of the utility function, solve the optimization problem, using the solution, estimate the parameters. More involved and restrictive.