选择性偏差，比较优势和存在异质性的教育收益:

Size: px

Start display at page:

Download "选择性偏差，比较优势和存在异质性的教育收益:"

鼎利强
5 years ago
Views:

1 * 24 4 **

2 lny = β + γx + =, 2, L, n lny X β γ A Cov( A, ) E( ) I X 2

3 lny = β + γx + β X = = (lny, lny ) lny lny = γ X = γ X + + ( ( = ) = ) (3a) (3b) E( X ) = E( X ) = lny lny F(lnY X, = ) F(lnY X, = ) F(lnY X ) F(lnY X ) lny = lny = [( γ γ ) X = [( γ γ ) X = β + ( + γ X ] + ( + ) lny + γ X + [ )] + ( + γ X + ) ] (4a) (4b) (4c) (4d) β = ( γ γ ) X + ( ) 3

4 γ γ ( γ γ )X ( ) β β β = E β X ) = E[( γ γ ) X ] ( * = P ( Z ) = * * < P = P * Z Z X ( Z ) P Z ) = ln Y ln Y = = ( = β p lm( ˆ β OL ) = E(lnY = E( γ X = β X, + + [ E( = ) E(lnY X, = ) E( (ATE) ( ) X, = ) E( γ X = ) + = )] X, = ) ATE ATE = E X ) = E( X ) ( β = β 4

5 , ) ( OL ATE p lm( ˆ β OL ) = E(lnY = E( β X X,, = ) E(lnY = ) + [ E( X, = ) E( = )] (TT) ( ) TT TT = E( X, = ) = E( β X = β + E(, = ATE + E( = ) = )] = ) = ) ( ) E ( = ) E ( = ) E ( =) IV I β plm ˆ β IV Cov( I, ln Y) Cov( I, ) Cov[ I, ( ) ] = = β + + Cov( I, ) Cov( I, ) Cov( I, ) Cov[ I ( [ ( ) ] Cov I, ) ] P, = = β + = β + Cov( I, ) Cov( I, ) P Pr( =) = plm( ˆ β ) = IV β IV β OL IV 5

6 LIV MTE MTE X ln Y ln Y WTP MTE( X = x, = u) = E( X = x, = u) = E( β X = x, = u ) = ( γ γ ) x+ E( = u ). MTE LIV E(lnY X = x, P = p) MTE( X = x, = P = p) = LIV ( X = x, P = p) = p MTE ATE = MTE u ) du ( TT = MTE u ) h ( u ) du ( TT TT = MTE u ) h ( u ) du ( TT h ATE ( u ) = h h TT TT ( u ( u F ( ) P ( u ) u f p dp ) = = E( P ) E( P ) FP ( u ) f ( p) dp ) = = E( P ) E( P ) u TT TT = E( X, = ) = E( β X, = ) = β + E ( = ). 6

7 7

8 OL IV 2 exp( β X ) Pr( = X) = = Λ( β X) + exp( β X ) Pr( = X ) = = Λ( β X )[ Λ( β X )] β X 8

9 Pr( =) 5% 2% f(p) 9% 6% 3% % P u u u u IV > ATE > OL = E( β β X, = ) = E( β X, = ) β = TT ATE 9

10 3. γ γ.9.7 MTE

11 = OL ATE = OL TT = TT ATE 2..5 MTE MTE u

12 Bjorklund, A. and R. Mofftt (987), The Etmaton of Wage Gan and Welfare Gan n elf-electon Model, Revew of Economc and tattc, 69: Carnero, P. (22), Heterogenety and electon n the Return to choolng: Implcaton for Educaton Polcy Evaluaton, Ph.D. Dertaton, nverty of Chcago. Carnero, P. and J. Heckman (22), The Evdence on Credt Contrant n Pot-econdary choolng, Economc Journal, Volume 2, ue 482, pp: (23), Human Captal Polcy, n J. Heckman, A. Krueger, Inequalty n Amerca: What Role for Human Captal Polce, Cambrdge, MA: MIT Pre. Carnero, P., J. Heckman and E. Vytlacl (2), Etmatng the Return to Educaton When It Vare Among Indvdual, Workng paper, nverty of Chcago. Chna' rban Houehold Income and Expendture urvey (CHIE) (2), Natonal Bureau of tattc, rban oco-economc urvey Organzaton. Chow, G. (2), Chna Economc Tranformaton. Malden, MA: Blackwell Publher. Grlche, Z. (977), Etmatng the Return to choolng: ome Econometrc Problem, Econometrca, Volume 45, number, pp Heckman, J. (2), Mcrodata, Heterogenety and Econometrc Polcy Evaluaton, Nobel Memoral Lecture n Economc cence, Journal of Poltcal Economy, 9(4), Heckman, J., L. Lochner and P. Todd. (23) Ffty Year of Mncer Earnng Functon, npublhed 2

13 manucrpt. nverty of Chcago. Heckman, J. and. Navarro-Lozano. (23), ng Matchng, Intrumental varable and Control Functon to Etmate Economc Choce Model forthcomng n Revew of Economc and tattc. October. Heckman, J. and E. Vytlacl (999), Local Intrumental varable and Latent varable Model for Identfyng and Boundng Treatment Effect, Proceedng of the Natonal Academy of cence, 96: (2), Local Intrumental varable, n C. Hao, K. Mormune, and J. Powell, (ed.), Nonlnear tattcal Modelng: Proceedng of the Thrteenth Internatonal ympoum n Economc Theory and Econometrc: Eay n Honor of Takeh Amemya, Cambrdge: Cambrdge nverty Pre, (2), tructural Equaton, Treatment Effect and Econometrc Polcy Evaluaton, Fher chultz Lecture World Congre of the Econometrc ocety. Natonal Bureau of tattc (NB). (2), Chna Labor tattcal Yearbook. Bejng: Chna tattc Pre. 3

作为市场化的人口流动

作为市场化的人口流动 Workng Paper Seres No.30 2003 6 1 Lews 1954 Harrs et al. 1970; Felds 1974 50 1949 1957 70% 80% 1958 50 1949 1986 50 20.7% 60 17.0% 70 32.9% 1980 1986 29.4% 45.23% 1988 6 1978 2 Meng 2000 pp. 26-28 Cook