lecture12

# lecture12 - Marginal eects Suppose our model for the latent...

This preview shows pages 1–4. Sign up to view the full content.

Marginal e/ects Suppose our model for the latent variable y is: y = + ±x + u We have studied a number of cases of interest: A. The econometrician observes y = y . Then @E ( y j x ) @x = ± B. Probit model: y = 1 ( y > 0) . The marginal e/ect of interest is: @E ( y j x ) @x = @ Pr( y = 1 j x ) @x = @ Pr( y > 0 j x ) @x = @ Pr( + ±x ) @x = @ + ±x ) @x = ±² ( + ±x ) C. Tobit model: y = + ±x + u y = y & 0 if y > 0 otherwise We may be interested in two marginal e/ects @E ( y j x;y > 0) @x and @E ( y j x ) @x x among "participants"; the second is the e/ect in the population at large. Recall that E ( y j x ) = E ( y j x;y > 0)Pr( y > 0 j x ) + E ( y j x;y = 0)Pr( y = 0 j x ) = E ( y j x;y > 0)Pr( y > 0 j x ) 1

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
Now in our case Pr( y > 0 j x + ±x ) E ( y j x;y > 0) = E ( y j x;y > 0) = + ±x + E ( u j u > ( + ±x )) = + ±x + ² ³ + ±x ² ± + ±x ² ± @E ( y j x;y > 0) @x = ± + ² @ h ³ + ±x ² ± = + ±x ² ±i @x = ± 8 < : 1 ³ + ±x ² ± + ±x ² ± 2 4 ² + ±x ² ³ + ³ + ±x ² ± + ±x ² ± 3 5 9 = ; The second marginal e/ect is @E ( y j x ) @x = @E ( y j x;y > 0) @x Pr( y > 0 j x ) + E ( y j x;y > 0) @ Pr( y > 0 j x ) @x = ± 8 < : 1 ³ + ±x ² ± + ±x ² ± 2 4 ² + ±x ² ³ + ³ + ±x ² ± + ±x ² ± 3 5 9 = ; ² + ±x ² ³ + ² 0 @ + ±x ² + ³ + ±x ² ± + ±x ² ± 1 A ± ² ³ ² + ±x ² ³ = ± ±³ + ±x ² ± ³ 2 + ±³ + ±x ² + ± ³ 2 = ± ² + ±x ² ³ the marginal e/ect can be decomposed into two parts: a) @E ( y j x;y> 0) @x Pr( y > 0 j x ) : This is the change in y among participants weighted by the proportion of participants in the population. This is known as change in the intensive margin. b) E ( y j x;y > 0) @ Pr( y> 0 j x ) @x : A change in x changes also the probability of participating, for example shifting some people±s decision to participate. This is known as change in the extensive margin. 2
Program Evaluation 2 The best use of the Heckman sample selection correction is in the program evaluation literature (see one of the previous lectures). 1 Restating the problem Consider an individual who is faced with the problem of choosing whether to take or not a given "treatment". Assume that d i = 1 if the individual takes the treatment, and 0 otherwise. Associated to each possible alternative there is an outcome, usually an eco- nomic return such as earnings. Assume that earnings in each possible state (0,1) are given by: y 0 i = & 0 + u 0 i (1) y 1 i = 1 + u 1 i (2) respectively in state 0 (no treatment) and 1(treatment). People select into regime 1 if ± 0 + ± 1 z i + " i > 0 . Hence, z i is an "instrument". This is a very simple model in which the outcome is equal to an e/ect

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
This is the end of the preview. Sign up to access the rest of the document.

## This note was uploaded on 07/28/2011 for the course ECON 102C taught by Professor Pistaferri,l during the Spring '11 term at Stanford.

### Page1 / 9

lecture12 - Marginal eects Suppose our model for the latent...

This preview shows document pages 1 - 4. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online