Download presentation

Presentation is loading. Please wait.

Published byCamren Rivera Modified about 1 year ago

1
1 Instrumental Variables Regression (SW Chapter 12)

2
2 Two Conditions for Valid Instrument

3
3 Estimation 1 of via 2SLS

4
4 IV Regression, Graphically

5
5 IV Regression, Algebraically

6
6 Example #1: Supply and demand

7
7 So we need a variable which shifts supply but not demand!

8
8 2SLS in the supply-demand example

9
9 Example #2: Test scores and class size

10
10 Properties of

11
11

12
12 Example: Cigarette demand

13
13 Ignoring endogeneity of ln(Price). reg lpackpc lravgprs, r; Linear regression Number of obs = 48 F( 1, 46) = Prob > F = R-squared = Root MSE = | Robust lpackpc | Coef. Std. Err. t P>|t| [95% Conf. Interval] lravgprs | _cons |

14
14 First stage

15
15 Second stage

16
16 Combined 1st & 2nd stages Old “ivreg” command vs. “ivregress: Y X Z. ivregress 2sls lpackpc (lravgprs = rtaxso), vce(robust); Instrumental variables (2SLS) regression Number of obs = 48 Wald chi2(1) = Prob > chi2 = R-squared = Root MSE = | Robust lpackpc | Coef. Std. Err. z P>|z| [95% Conf. Interval] lravgprs | _cons | Instrumented: lravgprs This is the endogenous X Instruments: rtaxso This is the instrumental variable 2SLS is the estimator, as opposed to GMM or LIML Don’t abbreviate as “ivreg”!

17
17 The General IV Regression Model

18
18 Identification of

19
19 The General IV Regression Model

20
20 2SLS with a 1 endogenous X

21
21 Example: Demand for cigarettes

22
22 Example: 1 instrument Y W X Z. ivregress 2sls lpackpc lperinc (lravgprs = rtaxso), vce(robust); Instrumental variables (2SLS) regression Number of obs = 48 Wald chi2(2) = Prob > chi2 = R-squared = Root MSE = | Robust lpackpc | Coef. Std. Err. z P>|z| [95% Conf. Interval] lravgprs | lperinc | _cons | Instrumented: lravgprs Instruments: lperinc rtaxso

23
23 Example: 2 instruments Y W X Z1 Z2. ivregress 2sls lpackpc lperinc (lravgprs = rtaxso rtaxs), vce(robust); Instrumental variables (2SLS) regression Number of obs = 48 Wald chi2(2) = Prob > chi2 = R-squared = Root MSE = | Robust lpackpc | Coef. Std. Err. z P>|z| [95% Conf. Interval] lravgprs | lperinc | _cons | Instrumented: lravgprs Instruments: lperinc rtaxso rtaxs Differences when multiple instruments? Normal or inferior good? Luxury good or not? Elastic or inelastic?

Similar presentations

© 2016 SlidePlayer.com Inc.

All rights reserved.

Ads by Google