Returns To Education (Angrist&Kruger Duflo, DDK) Flashcards
Mincerian returns to education expression
Log incomei = a + b(years educationi) + c(experiencei) + εi
I.e an extra year of education raises income by b%
What do we find the return to education to be?
b=8-12% i.e 8-12% increase in income per year of education
Econometric concerns with this regression (2)
Omitted variable bias - omits other factors influencing income e.g motivation, ability etc
Measurement error - quality/learning Varys by country, school, teacher etc
Data may also be unreliable in low income countries
Rich country schooling: when do children enter school, and compulsory till?
Enter school in year they turn 6.
Compulsory till 16
Angrist and Krueger
Compared adult income of those born in first and last quarter.
Since those born in 1st quarter get 6 months less compulsory schooling than those in last quarter!
Then to account for people to drop out at 16, scale up the income difference
What did they find for return to schooling
Return to an additional year of schooling is 10% (complies with Mincerian estimate of 8-12%)
What is important to note about this result tho
Returns may be different compared to average person e.g smaller if they dont get much out of school anyways so a lot dropped out of school when they could (hit 16)
I.e only shows variation for people dropping out vs staying,
Why is measuring returns to education in developing countries difficult (3)
Most are not in formal sector - e.g self employed like farmers with no official/regular wage, informal sector
Wages may not reflect productivity e.g government salaries
Hard to measure social externalities too
So what is the best way to find returns to education in developed countries
Use surveys asking how much people earn
Duflo experiment
If school construction can increase average years of schooling.
(Reducing direct cost r of getting to nearest school, since more around and closer!)
Effect of 61807 new schools built 1973-79
Enrolment from 69% 1973 to 83% in 1978
(They built more schools in places with lower enrolment)
Research design in Duflo - 3 categories
Differences in differences
Looked at:
Children younger than 6 at program start - full exposure (get full primary school benefits)
Children older than 12 at program start - not exposed at all (since finished primary school)
Exposure between targeted vs non-targeted areas
Difference in difference equation
ΔDiD = (y target region old - y target region young) - (y other region old - y other region young)
To cancel out region and cohort invariant differences (because olders will obviously earn more, and target regions are poorer orginally)
Findings (DiD figure)
2.6% higher wage for DID
What did Duflo find returns to education to be
And criticism/evaluation
10.4% (similar to Angruist and Kruher, again supporting Mincerian returns)
Eval: from a sample of wage earners (61k/153k total), so doesnt include non-formal wage earners. (Problem mentioned earlier when looking at developing countries)
So a criticism of the results is that the sample only uses formal wage earners:
When considering self employed as well, what are returns to education?
3.2%.
Another evaluation of Duflos experiment to consider
Done in Indonesia, experiencing fast growth at the time. Same results may not occur in other developing countries
So sample size small (10.4% formal, 3.6 for self earners), and rapid growth in Indonesia! (Generalisability)
What happened to student/teacher ratio during this program, what did this affect
While number of schools increased by 100%, teachers by 43%.
So class sizes increase, so potentially reducing quality of education
General equilibrium concerns:
We saw 2.6% increase in wage from DID. why might wage not actually increase following education?
Newly educated cohort can just displace older cohorts if finite amount of jobs. especially countries with mostly fixed government salaries
Supply of labour increases so wages may fall as a result of the higher educated workforce.
So that was primary education. Now consider secondary school returns
Reasons against pushing universal secondary education (3)
Over-educated - given limited number of skilled jobs (mentioned last slide)
Overcrowd schooling and lower quality - recall student/staff ratios rising in Duflo
Subsidises rich - who no longer have to pay to send kids to school
Experiment for secondary education: Duflo, Dupas & Kremer
Explain process
Gave scholarships to 682 people who passed entrance but couldnt afford enrolment immediately.
Follow up at age 28 (entry: 17 so long term!!)
Results of the scholarship
Enrolment and completion, and average years of schooling
Those who couldnt enrol immediately and got no scholarship - 50% ended up enrolling anyways
For those who couldnt enrol immediately but got given scholarship, 75% went (so 25% increase)
Completion of school increased by 27% too.
Scholarships achieved 1.2 more years of schooling
Was it cost effective?
So 50% enrol given no scholarship, and 75% if given scholarship.
So for 4 students, 2 will enrol, and 1 will enroll given a scholarship. So paying for 3 gets one extra, so not so cost effective!! (since 50% go anyways!)
Extra: if we offered scholarships to everyone: it would increase to paying for 15 people would create 1 more enrolment
Results (in terms of education)
Better maths, literacy and less risky sexual behaviour
Labour market results following scholarship (3)
Girls now 65% more likely to get a government job
More likely to have formal contract
No significant increase in earnings!!! (Links back to limited jobs, or increase in supply lowering wage!)
Why no increase in earnings? (2) (1 is new)
Over-education - increased competition for these limited formal sector jobs
More education means less experience, so lower wages first, so wait even longer (past the 10year experiment!)
So from these findings we have seen b (return to education) is around 10% for each additional year of primary school (not secondary as seen)
So a low return b doesn’t seem to be the problem! What is?
If parents are aware of the b (PERCEIVED VS TRUE!!)
