Economics 308: Econometrics Professor Moody

Text Pindyck and Rubinfeld, Econometric Models and Economic Forecasts (PR) Moody, Basic Econometrics with Stata (BES) References on reserve: Wooldridge, Jeffrey M., Introductory Econometrics (W) Maddala, G.S. Introduction to Econometrics, Second Edition (M) HB139.M353.1992 Kennedy, Peter, A Guide to Econometrics (K) HB139 K45 2003 Belsley, Kuh, and Welsch, Regression diagnostics (BKW) QA278.2 .B44 Stock and Watson, Introduction to Econometrics (SW) Some journal articles are available electronically in the Course Documents section of Blackboard. Grading: Midterm Homework assignments Final Exam Project 25% 25% 25% 25%

Review: Correlation and Regression.

Describing the relationship between two variables

Scatter diagrams Correlation Simple regression Why is it called regression? Reference: BES Ch. 7 Assignment 1: reproduce the regression on page 11 of BES.

Theory of Least Squares

Properties of estimators
Small sample properties bias efficiency mean square error relative efficiency robustness Large sample (asymptotic) properties consistency mean square error consistency asymptotic efficiency asymptotic unbiased Note: consistency "carries over" transformations while unbiasedness does not

References: W 699-713, M Ch 2.6, PR 27-30, SW 56-60, BES Ch. 8, 58-60, W Ch. 19. Assignment 2

Gauss-Markov Theorem
Gauss-Markov assumptions Yi = + X i + U i
U i iid (0, 2 ) Which implies that is a linear function of Y is random variable with a distribution (the sampling distribution of )
is an unbiased estimator of Deriving the variance of beta: Var ( ) = 2 / x 2
u

Gauss-Markov theorem: OLS is BLUE OLS is also a maximum likelihood estimator

References: W Ch 1,2, SW 103-107, PR Ch.3, BES Ch. 8, 60-64.

Inference and Hypothesis Testing

Assume the error term is distributed normally, then the sampling distribution of betahat is also normal with

 E( )= (the truth) Var ( ) = 2 / x 2

u

however, we must estimate the variance of u: u2 = e 2 /( n 2)

A note on the Normal, Chi-square, t, and F distributions References: SW 32-39, BES 65-71. Testing hypotheses concerning References: PR Ch. 2; W 724-736, SW 108-117, BES Ch. 8, 65-75.

Maximum likelihood and the likelihood ratio test

References: M 118-129, K Ch. 4.4, BES Ch.8, 76-78.

Multiple Regression
Why? Because life is complicated: omitted variable bias Three variable regression model Interpretation of formulas Goodness of fit: R2 References: W Ch 3-6, M Ch 4, PR Ch 4-5, SW Ch 5, BES Ch 8, 79-83 Assignment 3

Omitted variable bias and modelling

Review: multiple regression formula determining the direction of bias. There is only one way to be right and there are many ways to be wrong. It is wrong to include an irrelevant variable (inefficiency) and it is wrong to leave out a relevant variable (bias). However, omitting a relevant variable whose value is less than its standard error will decrease mean square errors. Proxy variables References: M Ch 11.6, W Ch 9.2, PR Ch 7.3, 7.5.1; W Ch 3, SW Ch 5, BES Ch 8, 83-90. Digression: torturing the data until it tells you what you want to hear. Leamer, Lets take the con out of econometrics, American Economic Review, March 1983, 31-43 (Blackboard).

Dummy Variables References: W Ch 7, PR 104-108, 121-123, M Ch 8.1-8.3, K Ch 13, SW 119-122, BES Ch 8 90-94. Dey, Matthew S. "Racial Differences in National Basketball Association Players' Salaries: A New Look,"The American Economist, 41, Fall 1997, 84-90 (Blackboard)

Useful Tests
F-test Chow test Granger causality test J-test for non-nested hypotheses LM test References: W Ch 4.5, 237-240; PR 110-112, 115-117, 216-219; M 393-394, 443-446; SW 165-70, 448-9, 468-9; BES Ch 8, 94-102.

Regression Diagnostics
Influential Observations Multicollinearity References: BKW, M Ch 7; BES Ch 9.

Digression: torturing the data until it tells you what you want to hear: Leamer,
"Let's Take the Con out of Econometrics" American Economic Review, March, 1983, 3143. (Blackboard)

Econometrics: What if the Gauss-Markov Assumptions are Violated?

Heteroskedasticity
Definition: nonconstant error variance, a common problem in cross sections Effects: (1) ols estimates remain unbiased, but (2) inefficient, (3) standard errors and t-scores are incorrect Tests: plot residuals, BreusCh-Pagan, White Cure: weighted least squares 1.known variances: weighted least squares 2.unknown variances: assume that the error variance is a function of an observable variable (the usual case) Heteroskedastic consistent (robust) standard errors References: M Ch 5, PR Ch 6.1, K Ch 7, SW 124-129, 139-140, 591-596; BES Ch 10 Assignment 3

Specification Bias

Rule: if one or more of the explanatory variables in a regression are correlated with the error term, the resulting OLS estimates are biased and inconsistent Causes of correlation between X and u incorrect functional form omitted variables errors of measurement in the independent variables simultaneous equations

Errors in variables
Definition Effects: ols is biased and inconsistent Cure: instrumental variables (two stage least squares) Problems: (1) Choice between a biased but efficient estimator (ols) and an unbiased but inefficient estimator (IV) (2) Where are the instruments? References: M Ch 11.1-11.3, 11.5-11.7, PR Ch 7; SW 248-250; BES Ch 11.

Simultaneous equations
When an equation is part of a simultaneous equation system, suCh that causation runs from Y to X as well as X to Y, then X is correlated with the error term and OLS is biased and inconsistent. Example: the consumption function Example: supply and demand Endogenous and exogenous variables, structural versus reduced form Consistent parameter estimation: instrumental variables (2sls) Indirect Least Squares The identification problem The order condition for identification Types of equation systems: general, recursive, block recursive Strategies: ols, ols with lags, reduced form, 2sls, VAR Standard tests Hausman test for mis-specification Basmann test for over-identification restrictions Bound-Jaeger-Baker test for weak instruments System estimation methods: ZELS, 3SLS References: M Ch 9, M Ch 12.10, PR Ch 11; KO Ch 7, K Ch 9; SW Ch 10; BES Ch 12. Bound, Jaeger, and Baker, "Problems with Instrumental Variables Estimation When the Correlation Between the Instruments and the Endogenous Explanatory Variable is Weak." Journal of the American Statistical Assocation 90 (430) June 1995, pp. 443-450.

Graded Assignment 5

Time Series Analysis

Time series data have advantage and disadvantages. The primary advantage is that we know time does not go backwards, so we can use lags to identify causal relationships (not possible in cross sections). The disadvantages are that we have to worry about certain problems that are unique to time series data, namely autocorrelation, unit roots, and cointegration.

Linear Dynamic Models

Autoregressive Distributed Lag (ADL) model. The L (lag) operator. The following models are special cases of the ADL. Static model AR model Leading Indicator model First Difference model Distributed Lag model Partial Adjustment model VAR model Common Factor model Error Correction model Note: the Error Correction model is not really a special case, since we did not restrict any coefficients. It is just a re-statement of the ADL after some algebra. References: SW 443-336, 485-486; BES Ch 13.

Autocorrelation
Definition: u(t) correlated with u(t-1) (and/or u(t-2), etc.) Effects: OLS remains unbiased variance of betahat will not be minimum (loss of efficiency) standard errors will be underestimated and t-scores overestimated (second order bias) If regressors include a lagged dependent variable, then ols estimators will be biased and inconsistent as well as inefficient. Tests: Durbin-Watson, Breush-Godfrey (LM).

There are two reasons for autocorrelation (1) serial correlation in the error term and (2) omitted variables with time components. If the autocorrelation is due to omitted lagged variables, then we can't fix it with Cochrane-Orcutt. We need to test to see if we have serial correlation or mis-specified dynamics. Likelihood ratio test for mis-specified dynamics Heteroskedasticity and autocorrelation consistent (HAC) standard errors (Newey-West) References: M Ch 6,PR Ch 6.2, K Ch 7.4; SW 504-517, 530-531; BES Ch 14.

Analysis of non-stationary data

Random walks and unit roots Spurious regressions Unit root tests: ADF, DF-GLS References: SW 457-467, 545-552; BES Ch 15. Assignment 6

Cointegration and long run equilibrium

Short-run relationships: first differences Cointegration and the long-run relationship Testing for cointegration Estimating the cointegrating regression Error correction model Dynamic Ordinary Least Squares References: P&R Ch 15.3, 15.4; BES Ch 16 Granger "Introduction." (Blackboard) Granger and Newbold, "Spurious Regressions in Econometrics" Journal of Econometrics 2, (1974) 111-120. (Blackboard)

Pooling Time Series and Cross Section Data

Motivation: cure for one kind of omitted variable bias, efficient use of data, increases degrees of freedom Fixed effects model Time series issues First difference model References: PR Ch 9.4; W Ch 13, 14; SW Ch 8; BES Ch 17.

Optional (extra credit): Panel Data Exercise in the Course Documents section of the Blackboard site.