Lecture 7
Lecture 7
------------------------------------------------------------------------------
lhw_5 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
Single Sex School ssex_3 | .2480322 .0290271 8.54 0.000 .1910879 .3049765
Father’s Years of Education payrsed | .002115 .0054614 0.39 0.699 -.008599 .012829
Mother’s Years of Education mayrsed | .0061507 .0057607 1.07 0.286 -.0051505 .0174518
Constant _cons | 1.548435 .0314873 49.18 0.000 1.486664 1.610206
------------------------------------------------------------------------------
The impact of single sex Schools on Female wages at age 33.
Including an ability indicator
------------------------------------------------------------------------------
lhw_5 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
Single Sex School ssex_3 | .2126029 .0286988 7.41 0.000 .1563027 .2689032
Low Ability lowabil | -.2159654 .0269518 -8.01 0.000 -.2688385 -.1630922
Father’s Years of Education payrsed | .0013796 .005336 0.26 0.796 -.0090883 .0118475
Mother’s Years of Education mayrsed | .0053448 .0056284 0.95 0.342 -.0056969 .0163865
Constant _cons | 1.649788 .0332586 49.60 0.000 1.584543 1.715034
------------------------------------------------------------------------------
The impact of single sex Schools on Female wages at age 33.
Including an ability indicator and School type
------------------------------------------------------------------------------
lhw_5 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
Single Sex School ssex_3 | .1288113 .0324787 3.97 0.000 .0650957 .192527
Low Ability lowabil | -.1793011 .0275525 -6.51 0.000 -.2333525 -.1252496
Selective School selecsch | .1979693 .0372037 5.32 0.000 .1249843 .2709543
Father’s Years of Education payrsed | .0003462 .0052851 0.07 0.948 -.010022 .0107143
Mother’s Years of Education mayrsed | .0050315 .0055714 0.90 0.367 -.0058982 .0159612
Constant _cons | 1.629437 .0331409 49.17 0.000 1.564422 1.694451
------------------------------------------------------------------------------
Correlations Between the Various Variables
Yi = b0 + b1 X i1 + b2 X i 2 + u i
•The OLS estimator of one of the coefficients b1 can be written as a
function of sample variances and covariances as
cov( X 1 X 2 )
cov(Y , X 1 ) − cov(Y , X 2 )
Var ( X 2 )
b1 =
ˆ
cov( X 1 X 2 )
Var ( X 1 ) − cov( X 1 X 2 )
Var ( X 2 )
• Now notice that cov( X 1 X 2 )
Var ( X 2 )
is the regression coefficient one would obtain if one were to regress
X1 on X2 i.e. the OLS estimator of b12 in
cov( Y , X ) − ˆ cov( Y , X )
b
bˆ1 = 1 12 2
Var ( X ) − bˆ cov( X X )
1 12 1 2
N N
∑ (X i1 − X 1 )( Yi − Y ) − bˆ12 ∑ ( X i 2 − X 2 )(Y i − Y )
bˆ1 = i =1
N N
i =1
=
∑ (X
i =1
i1 − X 1 ) 2 − bˆ12 ∑ (X
i =1
i1 − X 1 )( X i 2 − X 2 )
∑ [( X ]
N
i1 − X 1 ) − bˆ12 ( X i 2 − X 2 ) (Y i − Y )
i =1
∑ [( X ]
N 2
i1 − X 1 ) − bˆ12 ( X i 2 − X 2 )
i =1
• To derive this result you need to note the following
∑(X i2 − X 2 )( X i1 − X 1 )
bˆ122 ∑ ( X i 2 − X 2 ) 2 = bˆ12 ∑ (X i2 − X 2 )2 =
∑ (X i2 − X 2 )2
bˆ12 ∑ ( X i 2 − X 2 )( X i1 − X 1 )
∑[( X ]
N N N
− X 1 ) − bˆ12 ( X i 2 − X 2 ) = ∑( X i1 − X 1 ) − bˆ12 ∑( X i1 − X 1 )( X i 2 − X 2 )
2
2
i1
i=1 i=1 i=1
• Now the point of all these derivations can be seen if we note that
vˆi = ( X i1 − X1 ) − bˆ12 ( X i 2 − X 2 )
• This implies that the OLS estimator for b1 can be obtained in the
following two steps:
– Regress X1 on X2 and obtain the residuals from this
regression
– Regress Y on on these residuals
~ cov( X 1Y )
b1 =
Var ( X 1 )
• Instead, by including X2 in the regression we estimate b1 as
∑ [( X ]
N
i1 − X 1 ) − bˆ12 ( X i 2 − X 2 ) (Yi − Y )
i =1
bˆ1 =
∑[ ]
N 2
( X i1 − X 1 ) − b12 ( X i 2 − X 2 )
ˆ
i =1
~
• The Gauss Markov theorem directly implies that b1 cannot be
less efficient that b̂1
~ σ2
Var (b1 ) =
NVar ( X 1 )