Economics of Education

Education as a Consuption good: Utility

1. Enjoyment of Learning

2. Personal Growth

3. Social Life/Network

Education as a Consumption good: Costs

1. Tuition and Fees

2. Efforts

3. Opportunity Costs

Education as an Investment: Benefits

1. Higher earnings

2. Greater probability of employment

3. Higher Career progression

4. Better Working Conditions (Stability)

Mincer Equation

ln(wage_i) = a + bS_i + cExp_i + dExp_i² + e_i

1960 Census Data yields b in mincer =

0.11. 1 additional year of schooling is associated with an 11% higher wage

(Mincer) Earnings Gap can arise from 2 different reasons:

1. education may raise earnings

2. higher earning people may be more likely to be educated

Angrist and Krueger (1991): The Instrument

1. Date of birth measured by quarter of birth. Two students, one born on January 2nd, another on December 31st, if they both dropped out at 16 one would have a full extra year of education.

2. Exploiting compulsory schooling laws

Angrist and Krueger (1991): The Assumption

The only reason income would vary off quarter of birth is cumpolsory schooling laws

Angrist and Krueger (1991): Results

Men born in early quarters earn 6-10% less. (AKA, those with less schooling make less money

Angrist and Krueger (1991): Interpretation

1. Estimated Local Average Treatment Effect

   1. returns to education for those affected by law change

2. compulsory laws dont affect everyone (only those who want to drop out w/ minimum schooling)

Ashenfelter and Krueger (1994): Contributions

1. identical twin sample controlling for genetic factors across individuals

2. Additional year of schooling associated with 9% increase

3. Removes measurement error in amount of schooling

Measuring Returns to Education: Takeaways

1. Difficult due to Ability Bias

2. natural experiments to do quasi-random variation in schooling

3. overall positive effect on income

Correlation is Not Causation: Reasons for Correlation

1. Causation: X = Y

2. Reverse Causation: Y → X

3. Simultaneuity: X→ Y, Y→ X

4. Endogeneity: Z → X, Z = Y

5. Spurious Correlation: ?

4 Research Designs

1. Randomized Experiments

2. Difference in Differences

3. Regression Discontinuity Design

4. Instrumental Variables

Omitted Variables Bias: Formula

B^₁ = B₁ + B₂(cov x₂, x₁)/Var(x₁)

Omitted Variables Bias Formula: Two Parts

1. Effect of the Omitted (B₂)

2. Omitted on the Included ((cov x₂, x₁)/Var(x₁))

Conditional Independence

The assumption that the error term is uncorrelated with the explanatory variables in the model

When is it justified to assume conditional independence

if x is randomized

Average Treatment Effect (ATE)

E[Y_1,i - Y_0,i]

Average Treatment Effect on the Treated (ATT)

(E[Y_1,i] D_i = 1) - (E[Y_0,i] D = 1)

Selection Bias

(E[Y_0,i] D = 1) - (E[Y_{0, i}] D = 0)

Observed Difference Formula

ATT + Selection Bias or (E[Y_i] D = 1) - (E[Y_i] D= 0).

Intention to Treat (ITT)

(E[Y_i] T = 1) - (E[Y_i] T = 0)