Instrumental variables

Slides:



Advertisements
Similar presentations
Graduate Methods Master Class
Advertisements

Economics 20 - Prof. Anderson1 Multiple Regression Analysis y =  0 +  1 x 1 +  2 x  k x k + u 7. Specification and Data Problems.
1 IV/2SLS models Vietnam era service Defined as Estimated 8.7 million served during era 3.4 million were in SE Asia 2.6 million served.
Instrumental Variables Estimation and Two Stage Least Square
Christopher Dougherty EC220 - Introduction to econometrics (chapter 4) Slideshow: interactive explanatory variables Original citation: Dougherty, C. (2012)
Heteroskedasticity The Problem:
HETEROSCEDASTICITY-CONSISTENT STANDARD ERRORS 1 Heteroscedasticity causes OLS standard errors to be biased is finite samples. However it can be demonstrated.
TigerStat ECOTS Understanding the population of rare and endangered Amur tigers in Siberia. [Gerow et al. (2006)] Estimating the Age distribution.
EC220 - Introduction to econometrics (chapter 7)
Lecture 12 (Ch16) Simultaneous Equations Models (SEMs)
Sociology 601 Class 19: November 3, 2008 Review of correlation and standardized coefficients Statistical inference for the slope (9.5) Violations of Model.
Adaptive expectations and partial adjustment Presented by: Monika Tarsalewska Piotrek Jeżak Justyna Koper Magdalena Prędota.
Valuation 4: Econometrics Why econometrics? What are the tasks? Specification and estimation Hypotheses testing Example study.
1 Angrist/Evans Angrist/Krueger
1Prof. Dr. Rainer Stachuletz Multiple Regression Analysis y =  0 +  1 x 1 +  2 x  k x k + u 7. Specification and Data Problems.
Chapter 12 Simple Regression
Prof. Dr. Rainer Stachuletz
Econ 140 Lecture 131 Multiple Regression Models Lecture 13.
Chapter 9 Simultaneous Equations Models. What is in this Chapter? In Chapter 4 we mentioned that one of the assumptions in the basic regression model.
Multiple Regression Models
Introduction to Regression Analysis Straight lines, fitted values, residual values, sums of squares, relation to the analysis of variance.
1 Review of Correlation A correlation coefficient measures the strength of a linear relation between two measurement variables. The measure is based on.
So far, we have considered regression models with dummy variables of independent variables. In this lecture, we will study regression models whose dependent.
1 Research Method Lecture 11-1 (Ch15) Instrumental Variables Estimation and Two Stage Least Square ©
Christopher Dougherty EC220 - Introduction to econometrics (chapter 6) Slideshow: variable misspecification iii: consequences for diagnostics Original.
TESTING A HYPOTHESIS RELATING TO A REGRESSION COEFFICIENT This sequence describes the testing of a hypotheses relating to regression coefficients. It is.
Christopher Dougherty EC220 - Introduction to econometrics (chapter 3) Slideshow: precision of the multiple regression coefficients Original citation:
EDUC 200C Section 4 – Review Melissa Kemmerle October 19, 2012.
TOBIT ANALYSIS Sometimes the dependent variable in a regression model is subject to a lower limit or an upper limit, or both. Suppose that in the absence.
1 INTERACTIVE EXPLANATORY VARIABLES The model shown above is linear in parameters and it may be fitted using straightforward OLS, provided that the regression.
LT6: IV2 Sam Marden Question 1 & 2 We estimate the following demand equation ln(packpc) = b 0 + b 1 ln(avgprs) +u What do we require.
Econometrics 1. Lecture 1 Syllabus Introduction of Econometrics: Why we study econometrics? 2.
1 PROXY VARIABLES Suppose that a variable Y is hypothesized to depend on a set of explanatory variables X 2,..., X k as shown above, and suppose that for.
Returning to Consumption
How do Lawyers Set fees?. Learning Objectives 1.Model i.e. “Story” or question 2.Multiple regression review 3.Omitted variables (our first failure of.
MultiCollinearity. The Nature of the Problem OLS requires that the explanatory variables are independent of error term But they may not always be independent.
EDUC 200C Section 3 October 12, Goals Review correlation prediction formula Calculate z y ’ = r xy z x for a new data set Use formula to predict.
What is the MPC?. Learning Objectives 1.Use linear regression to establish the relationship between two variables 2.Show that the line is the line of.
Instrumental Variables: Problems Methods of Economic Investigation Lecture 16.
Random Regressors and Moment Based Estimation Prepared by Vera Tabakova, East Carolina University.
CENTRE FOR INNOVATION, RESEARCH AND COMPETENCE IN THE LEARNING ECONOMY Session 3: Basic techniques for innovation data analysis. Part II: Introducing regression.
Public Policy Analysis ECON 3386 Anant Nyshadham.
Application 3: Estimating the Effect of Education on Earnings Methods of Economic Investigation Lecture 9 1.
. reg LGEARN S WEIGHT85 Source | SS df MS Number of obs = F( 2, 537) = Model |
Two-stage least squares 1. D1 S1 2 P Q D1 D2D2 S1 S2 Increase in income Increase in costs 3.
POSSIBLE DIRECT MEASURES FOR ALLEVIATING MULTICOLLINEARITY 1 What can you do about multicollinearity if you encounter it? We will discuss some possible.
Nguyen Ngoc Anh Nguyen Ha Trang Applied Econometrics Instrumental Variable Approach DEPOCEN.
(1)Combine the correlated variables. 1 In this sequence, we look at four possible indirect methods for alleviating a problem of multicollinearity. POSSIBLE.
Christopher Dougherty EC220 - Introduction to econometrics (chapter 6) Slideshow: exercise 6.13 Original citation: Dougherty, C. (2012) EC220 - Introduction.
Review Section on Instrumental Variables Economics 1018 Abby Williamson and Hongyi Li October 11, 2006.
Endogenous Regressors and Instrumental Variables Estimation Adapted from Vera Tabakova, East Carolina University.
1 BINARY CHOICE MODELS: LINEAR PROBABILITY MODEL Economists are often interested in the factors behind the decision-making of individuals or enterprises,
1 In the Monte Carlo experiment in the previous sequence we used the rate of unemployment, U, as an instrument for w in the price inflation equation. SIMULTANEOUS.
F TESTS RELATING TO GROUPS OF EXPLANATORY VARIABLES 1 We now come to more general F tests of goodness of fit. This is a test of the joint explanatory power.
WHITE TEST FOR HETEROSCEDASTICITY 1 The White test for heteroscedasticity looks for evidence of an association between the variance of the disturbance.
1 COMPARING LINEAR AND LOGARITHMIC SPECIFICATIONS When alternative specifications of a regression model have the same dependent variable, R 2 can be used.
VARIABLE MISSPECIFICATION II: INCLUSION OF AN IRRELEVANT VARIABLE In this sequence we will investigate the consequences of including an irrelevant variable.
The Instrumental Variables Estimator The instrumental variables (IV) estimator is an alternative to Ordinary Least Squares (OLS) which generates consistent.
VARIABLE MISSPECIFICATION I: OMISSION OF A RELEVANT VARIABLE In this sequence and the next we will investigate the consequences of misspecifying the regression.
INSTRUMENTAL VARIABLES Eva Hromádková, Applied Econometrics JEM007, IES Lecture 5.
Instrumental Variable (IV) Regression
STOCHASTIC REGRESSORS AND THE METHOD OF INSTRUMENTAL VARIABLES
Instrumental Variables and Two Stage Least Squares
Instrumental Variables and Two Stage Least Squares
Identification: Instrumental Variables
Instrumental Variables and Two Stage Least Squares
Instrumental Variables Estimation and Two Stage Least Squares
Introduction to Econometrics, 5th edition
Presentation transcript:

Instrumental variables Anant Nyshadham

Instrumental Variables What is a natural experiment? “situations where the forces of nature or government policy have conspired to produce an environment somewhat akin to a randomized experiment” Angrist and Krueger (2001, p. 73) Natural experiments can provide a useful source of exogenous variation in problematic regressors But they require detailed institutional knowledge

Instrumental Variables and Natural Experiments Some natural experiments in economics Existing policy differences, or changes that affect some jurisdictions (or groups) but not others Minimum wage rate Excise taxes on consumer goods Unemployment insurance, workers’ compensation Unexpected “shocks” to the local economy Coal prices and the Middle East oil embargo (1973) Agricultural production and adverse weather events

Instrumental Variables and Natural Experiments Some potential pitfalls Not all policy differences/changes are exogenous Political factors and past realizations of the response variable can affect existing policies or policy changes Generalizability of causal effect estimates Results may not generalize beyond the units under study Heterogeneity in causal effects Results may be sensitive to the natural experiment chosen in a specific study (L.A.T.E.)

Instrumental Variables and Natural Experiments Some natural experiments used as IV which are of interest to development economists Acemoglu Johnson & Robinson (2001): settler mortality Paxson (1992): rainfall Schultz & Tansel (1997): healthcare prices

True Model Suppose true model is: Do not observe V Can only estimate: Y = a + bX + cV + e a, b, and c are parameters to be estimated; e is error term Do not observe V Can only estimate: Y = a + bX + e What do we do to get b instead of b?

Methods Y = a + bX + η; η = cV + e Differencing/FE Find groups with common V (assumption), but variation in X Subtract off V to remove it from error term Instrumental Variable Find instrument Z; X = j + kZ + i Predict portion of X which does not correlate with V Use this portion in original estimating equation

IV Criteria and Assumptions Step/Stage 1: X = j + kZ + I  X’ = k’Z Step/Stage 2: Y = a + bX’ + η; recover true b Criteria for Z Z must sufficiently predict X: k>>0 or k<<0 Testable using estimate of k from first stage Z must only impact Y through X Cov(Z,η)=0; Cov(Z,V)=0 & Cov(Z,e)=0 Z does not belong original estimation equation Assumption, untestable

An IV example: Angrist and Krueger (1991), J.L.E. Returns to education (Y = wages) Problem of omitted “ability bias” Years of schooling vary by quarter of birth Compulsory schooling laws, age-at-entry rules Someone born in Q1 is a little older and will be able to drop out sooner than someone born in Q4 Q.O.B. can be treated as a useful source of exogeneity in schooling

Angrist and Krueger (1991), J.L.E. People born in Q1 do obtain less schooling But pay close attention to the scale of the y-axis Mean difference between Q1 and Q4 is only 0.124, or 1.5 months So...need large N since R2X,Z will be very small A&K had over 300k for the 1930-39 cohort Source: Angrist and Krueger (1991), Figure I

Angrist and Krueger (1991), J.L.E. Final 2SLS model interacted QOB with year of birth (30), state of birth (150) OLS: b = .0628 (s.e. = .0003) 2SLS: b = .0811 (s.e. = .0109) Least squares estimate does not appear to be badly biased by omitted variables But...replication effort identified some pitfalls in this analysis that are instructive

Bound, Jaeger, and Baker (1995), J.A.S.A. Potential problems with QOB as an IV Correlation between QOB and schooling is weak Small Cov(X,Z) introduces finite-sample bias, which will be exacerbated with the inclusion of many IV’s QOB may not be exogenous (correlated with unobservable determinants of wages, e.g. family income) QOB may not satisfy exclusion restriction (e.g. age relative to peers changes social dynamics, competition, leadership skill etc.)

Bound, Jaeger, and Baker (1995), J.A.S.A. Even if the instrument is “good,” matters can be made far worse with IV as opposed to LS Weak correlation between IV and endogenous regressor can pose severe finite-sample bias And…really large samples won’t help, especially if there is even weak endogeneity between IV and error First-stage diagnostics provide a sense of how good an IV is in a given setting F-test and partial-R2 on IV’s

Useful Diagnostic Tools for IV Models Tests of instrument relevance Weak IV’s → Large variance of bIV as well as potentially severe finite-sample bias Tests of instrument exogeneity Endogenous IV’s → Inconsistency of bIV that makes it no better (and probably worse) than bLS Durbin-Wu-Hausman test Endogeneity of the problem regressor(s)

Tests of Instrument Relevance Diagnostics based on the F-test for the joint significance of the IV’s Nelson and Startz (1990); Staiger and Stock (1997) Bound, Jaeger, and Baker (1995) Partial R-square for the IV’s Shea (1997) There is a growing econometric literature on the “weak instrument” problem

Tests of Instrument Exogeneity Model must be overidentified, i.e., more IV’s than endogenous X’s H0: All IV’s uncorrelated with structural error Overidentification test: 1. Estimate structural model 2. Regress IV residuals on all exogenous variables 3. Compute NR2 and compare to chi-square df = # IV’s – # endogenous X’s

Application: Adolescent Work and Delinquent Behavior Prior research shows a positive correlation between teenage work and delinquency Reasons to suspect serious endogeneity bias 2nd wave of the NLSY97 (N = 8,368) Y = 1 if committed delinquent act (31.9%) X = 1 if worked in a formal job (52.6%) Z1 = 1 if child labor law allows 40+ hours (14.2%) Z2 = 1 if no child labor restriction in place (39.6%)

Regression Model Ignoring Endogeneity . reg pcrime work if nomiss==1 & wave==2 Source | SS df MS Number of obs = 8368 -------------+------------------------------ F( 1, 8366) = 6.33 Model | 1.37395379 1 1.37395379 Prob > F = 0.0119 Residual | 1815.97786 8366 .217066443 R-squared = 0.0008 -------------+------------------------------ Adj R-squared = 0.0006 Total | 1817.35182 8367 .217204711 Root MSE = .4659 ------------------------------------------------------------------------------ pcrime | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- work | .0256633 .0102005 2.52 0.012 .0056677 .0456588 _cons | .3053242 .0074009 41.26 0.000 .2908167 .3198318 Teenage workers significantly more delinquent Modest effect but consistent with prior research

First-Stage Model State child labor laws affect probability of work . reg work law40 nolaw if nomiss==1 & wave==2 Source | SS df MS Number of obs = 8368 -------------+------------------------------ F( 2, 8365) = 626.64 Model | 271.829722 2 135.914861 Prob > F = 0.0000 Residual | 1814.33364 8365 .216895832 R-squared = 0.1303 -------------+------------------------------ Adj R-squared = 0.1301 Total | 2086.16336 8367 .249332301 Root MSE = .46572 ------------------------------------------------------------------------------ work | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- law40 | .0688902 .0154383 4.46 0.000 .0386274 .099153 nolaw | .3818684 .0110273 34.63 0.000 .3602521 .4034847 _cons | .3655636 .0074883 48.82 0.000 .3508847 .3802425 State child labor laws affect probability of work This is a really strong first stage (F, R2)

Two-Stage Least Squares Model . ivreg pcrime (work = law40 nolaw) if nomiss==1 & wave==2 Instrumental variables (2SLS) regression Source | SS df MS Number of obs = 8368 -------------+------------------------------ F( 1, 8366) = 6.86 Model | -19.5287923 1 -19.5287923 Prob > F = 0.0088 Residual | 1836.88061 8366 .219564978 R-squared = . -------------+------------------------------ Adj R-squared = . Total | 1817.35182 8367 .217204711 Root MSE = .46858 ------------------------------------------------------------------------------ pcrime | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- work | -.0744352 .0284206 -2.62 0.009 -.1301466 -.0187238 _cons | .3580171 .0158135 22.64 0.000 .3270187 .3890155 Instrumented: work Instruments: law40 nolaw

What Do the Models Suggest Thus Far? Completely different conclusions! OLS = Teenage work is criminogenic (b = +.026) Delinquency risk increases by 8.5 percent (base = .305) 2SLS = Teenage work is prophylactic (b = –.074) Delinquency risk decreases by 20.7 percent (base = .358) Which model should we believe? We still have some additional diagnostic work to do to evaluate the 2SLS model Overidentification test

Overidentification Test from the Software Tests of overidentifying restrictions: Sargan N*R-sq test 0.509 Chi-sq(1) P-value = 0.4757 Basmann test 0.508 Chi-sq(1) P-value = 0.4758 IV’s jointly pass the exogeneity requirement Notice that -overid- provides a global test, whereas the regression-based approach allows you to test the IV’s jointly as well as individually

So Where Do We Stand with the Work-Delinquency Question? Are child labor laws correlated with work? YES = first-stage F is large Are child labor laws good IV’s? YES = overidentification test is not rejected Is teenage work endogenous? YES = Hausman test is rejected Prior research findings that teenage work is criminogenic are selection artifacts

Now…What Happens if I Throw in a Potentially Bogus Instrument? Now there are three instrumental variables Z1 = 1 if child labor law allows 40+ hours (14.2%) Z2 = 1 if no child labor restriction in place (39.6%) Z3 = 1 if high unemployment rate in county (20.1%) A little more difficult to tell a convincing story that the unemployment rate is only related to delinquency through work experience But let’s see what happens

First-Stage Model So far so good and consistent with expectation . reg work law40 nolaw highun if nomiss==1 & wave==2 Source | SS df MS Number of obs = 8368 -------------+------------------------------ F( 3, 8364) = 427.28 Model | 277.229696 3 92.4098987 Prob > F = 0.0000 Residual | 1808.93366 8364 .216276144 R-squared = 0.1329 -------------+------------------------------ Adj R-squared = 0.1326 Total | 2086.16336 8367 .249332301 Root MSE = .46505 ------------------------------------------------------------------------------ work | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- law40 | .0636421 .0154519 4.12 0.000 .0333525 .0939317 nolaw | .3775975 .0110447 34.19 0.000 .3559472 .3992479 highun | -.0636009 .0127283 -5.00 0.000 -.0885517 -.0386502 _cons | .3808061 .0080759 47.15 0.000 .3649754 .3966368 So far so good and consistent with expectation

Two-Stage Least Squares Model . ivreg pcrime (work = law40 nolaw highun) if nomiss==1 & wave==2 Instrumental variables (2SLS) regression Source | SS df MS Number of obs = 8368 -------------+------------------------------ F( 1, 8366) = 5.47 Model | -16.0635514 1 -16.0635514 Prob > F = 0.0194 Residual | 1833.41537 8366 .219150773 R-squared = . -------------+------------------------------ Adj R-squared = . Total | 1817.35182 8367 .217204711 Root MSE = .46814 ------------------------------------------------------------------------------ pcrime | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- work | -.0657624 .0281159 -2.34 0.019 -.1208765 -.0106483 _cons | .3534516 .0156602 22.57 0.000 .3227537 .3841496 Instrumented: work Instruments: law40 nolaw highun

Post-Hoc Diagnostics Overidentification gives cause for concern Tests of overidentifying restrictions: Sargan N*R-sq test 5.301 Chi-sq(2) P-value = 0.0706 Basmann test 5.301 Chi-sq(2) P-value = 0.0706 . ivendog Tests of endogeneity of: work H0: Regressor is exogenous Wu-Hausman F test: 12.32811 F(1,8365) P-value = 0.00045 Durbin-Wu-Hausman chi-sq test: 12.31438 Chi-sq(1) P-value = 0.00045 Overidentification gives cause for concern The p-value shouldn’t be anywhere near 0.05

Conclusion from Diagnostic Tests 2SLS “work effect” is similar Without unemployment, b = –.074 (s.e. = .028) With unemployment, b = –.066 (s.e. = .028) But…the second model is invalidated because the unemployment rate is not exogenous If affects criminality through other channels We need to control for all other indirect pathways, or… It should not be used as an IV at all

Closing Comments about Instrumental Variables Studies In general, a lagged value of the endogenous regressor is not a good instrument Traditional structural equation model uses lagged values of X and Y as instruments to break the simultaneity between the current values of X and Y X1 X2 Y1 Y2 These models impose the awfully strong assumption that lagged values of X and Y only affect the outcomes through current values

Rules for Good Practice with Instrumental Variables Models IV models can be very informative, but it’s your job to convince your audience Show the first-stage model diagnostics Even the most clever IV might not be sufficiently strongly related to X to be a useful source of identification Report test(s) of overidentifying restrictions An invalid IV is often worse than no IV at all Report LS endogeneity (DWH) test

Rules for Good Practice with Instrumental Variables Models Most importantly, TELL A STORY about why a particular IV is a “good instrument” Something to consider when thinking about whether a particular IV is “good” Does the IV, for all intents and purposes, randomize the endogenous regressor?