Estimating Distributions of Counterfactuals with an Application ... - UCL

EFFECTS OF UNCERTAINTY ON COLLEGE CHOICE 389
TABLE 2c
COVARIATES INCLUDED IN OUTCOME, COST AND TEST EQUATIONS
Cost of Schooling
Tuition and Foregone Psychic
Earnings Earnings Cost Test Scores
Intercept Yes Yes Yes Yes
Urban Yes Yes Yes Yes
South Yes Yes Yes Yes
Cohort dummies Yes Yes Yes –
Mean local unemployment rate Yes Yes – –
Average local wage Yes Yes – –
Local tution – Yes – –
Number of siblings – – Yes Yes
Mother’s education – – Yes Yes
Father’s education – – Yes Yes
Broken family – – Yes
Enrolled in school at test date – – Yes
Age in 1980 – – Yes
utility) in the first period (ages 19–29 years) and in the second period (ages 30 to
65 years). Let V 1,s be the period 1 gross utility of achieving schooling level s, and
V 2,s be the period 2 gross utility of obtaining schooling level s. Using (23), we write
the gross utilities as
V 1,s = ¯δ 1,s + X ′ ¯β 1,s + ᾱ ′ 1,s θ + ¯ε 1,s
V 2,s = ¯δ 2,s + X ′ ¯β 2,s + ᾱ ′ 2,s θ + ¯ε 2,s
These are the outcome equations for the model that we estimate. To see this, notice
that
V 1,s =
=
=
∑A 1
a=19
∑A 1
a=19
∑A 1
a=19
+
ln Y a,s
(1 + ρ) a
δ a,s + X ′ β a,s + α ′ a,s θ + ε a,s + η 1,s × experience a + η 2,s × experience 2 a
(1 + ρ) a
δ a,s + X ′ β a,s + η 1,s × experience a + η 2,s × experience 2 a
(1 + ρ) a
[
A1
∑
a=19
]
α a,s
′
(1 + ρ) a θ +
∑A 1
a=19
= ¯δ 1,s + X ′ ¯β 1,s + ᾱ ′ 1,s θ + ¯ε 1,s
ε a,s
(1 + ρ) a