Papers Flashcards
RCT example
Murallidharan and Sundararaman (2011) - Teacher Performance Pay
Teacher Performance Pay methodology
RCT across 500 schools in an Indian state
- random allocation across 5 different groups
- 4 different treatments with 1 control
Teacher Performance Pay balance test
shows randomisation worked
want a high p-value so there isn’t a statistically significant difference in means
Teacher Performance Pay effect on student test scores
0.149 showing 15% std increase if in an incentive school compared to a control school (statistically significant at the 99% level)
Teacher Performance Pay overall findings
2 years after RCT, students in incentive schools do 0.22 std better on test scores
gains are not driven by a selected group but broad-based
no negative consequences on other learning outcomes
individual incentives more effective than school incentives
teacher behaviour affected in terms of effort rather than attendance
performance-pay more cost-effective compared to school transfers
IV/TSLS example
Acemoglu, Johnson and Robinson (2001) - The Colonial Origins of Comparative Development
The Colonial Origins of Comparative Development methodology
Y variable - GDP per capita (PPP-adjusted) in 1995
X variable - average protection against risk of expropriation during 1985-1995
Z variable - colonial settler mortality rate
The Colonial Origins of Comparative Development first-stage regression
statistically strong first stage
The Colonial Origins of Comparative Development reduced form regression
clear strong negative relationship between settler mortality and current institutions
The Colonial Origins of Comparative Development main results table
with one unit increase in institution index score, there’s a 94% increase in economic performance in 1995
The Colonial Origins of Comparative Development overall findings
statistically significant causal effect of institutions on levels of economic development
also economically meaningful (institutions make a big difference)
first stage regressions show there is strong historical persistence in institutional quality
DiD example
Card and Krueger (1994) - Employment Effects of Minimum Wage Policies
Duflo (2001) - Schooling and Labour Market Consequences of School Construction
Employment Effects of Minimum Wage Policies methodology
effect of increasing minimum wage on the employment of low-wage workers (natural experiment)
assumption that pre-existing differences before minimum wage policy would have stayed the same had it not been the minimum wage policy
Employment Effects of Minimum Wage Policies results
minimum wage policy increased employment slightly by about 3 workers
Schooling and Labour Market Consequences of School Construction methodology
DiD with unusual policy experiment with increased oil windfalls and increase of primary schools constructed
using DD to see the causal effect of an extra school on wages
also using construction as an instrument for years of education in 1995
Schooling and Labour Market Consequences of School Construction overall findings
one additional school per 1000 children led to 0.15 additional years of education
one additional school per 1000 children increased average wage by 2%
additional year of education led to 10% higher wages
panel data/FE example
S&W Textbook - Beer Taxes and Traffic Fatalities
Beer Taxes and Traffic Fatalities methodology
can’t just regress fatality rate on beer tax since there’s OVB problems with cov(error, X) not being equal to 0
FE estimation
- taking out of the estimation equation that some states have higher fatality rates or higher beer taxes on average and using mean deviation over time
- using dummies
Beer Taxes and Traffic Fatalities overall findings
state FE has a substantial impact on the relationship between beer tax and fatality rate while time FE does not
- sign changes and R^2 increases a lot for state FE
stiffer punishments and higher drinking ages don’t reduce fatalities but higher taxes are effective
RD example
Lemieux and Milligan (2005) - Welfare Payments and Labour Supply in Canada
Abdulkadiroglu, Angrist and Pathak (2014) - Elite Schools in Boston
Welfare Payments and Labour Supply in Canada methodology
exploiting a discontinuity in social assistance in Quebec
- until 1989, unemployed individuals under 30 without children received lower welfare benefits than those over 30
Welfare Payments and Labour Supply in Canada overall finding
higher levels of social assistance reduce work incentives
employment rates dropped by 4.5% once you double benefits across the cutoff
Elite Schools in Boston methodology
three exam schools that are public but have higher competitive entrance exams
instrumenting peer quality for average test scores of peers
Elite Schools in Boston overall findings
substantially better peer quality does not improve math scores
for the instrument to be valid, attending boston latin school should only affect outcomes through changes in peer quality at the threshold but this could be violated since the school might have better and more motivated teachers
- overestimate the true effect of better peer quality
