##
This **preview** shows page *1*
out of 2 **pages**.

*View Full Document*

End of preview. Want to read all 2 pages?

Upload your study docs or become a GradeBuddy member to access this document.

View Full Document**Unformatted text preview:**

Econ 3120 1st Edition Lecture 13Outline of Current Lecture I. Goodness of FitII. UnbiasednessCurrent LectureIII. Unbiasedness of Multivariate OLS EstimatorsUnbiasedness of Multivariate OLS Estimators In order to show that the OLS estimators described above are unbiased, we need to make a number of assumptions, along the same lines as SLR.1-SLR.4: • MLR.1: The model is given by y = β0 +β1x1 +β2x2 +...+βkxk +u • MLR.2: Random sampling: our sample {(x1, x2,..., xk , y) : i = 1,...,n} is a random sample following the population model in MLR.1 • MLR.3: No perfect collinearity: In the sample (and in the population): 1) Each independent variable has sample variation 6= 0 2) None of the independent variables can be constructed as a linear combination of the other independent variables. • MLR.4: Zero conditional mean: E(u|x1, x2,..., xk) = 0 Under these assumptions, multivariate OLS estimates areunbiased; that is: E( ˆβj |x1, x2,..., xk) = βj The proof is somewhat involved algebraically, so we will not cover it here. See Appendix 3A in Wooldridge for the details. 6 Omitted Variable Bias 6.1 Comparison of Bivariate and Multivariate Estimates In order to understand the effects of omitting relevant variables, let’s return to regression anatomy. We’ll use this to describe how theestimated β’s change when the regression is run without one of the x’s, i.e., we have an omitted variable. Take the example where the true model is of the form: y = β0 +β1x1 +β2x2 +u (3) 6 but we instead run the regression omitting x2 y = β0 +β1x1 +u (4) Let ˆβ1 and ˆβ2 represent the OLS estimators of β1 and β2 using (3), and ˜β1 be the OLS estimator of β1 using (4). How can we use regression anatomy to figure out what’s going on? It turns out that ˜β1 = ˆβ1 + ˆβ2 ˜δ1 where ˜δ1 is the estimate from equation (2) above. Proof: 6.2 Omitted Variable Bias Formula We’re now in a position to show what happens to the expectation of our OLS estimators when we exclude a relevant variable in the analysis. Suppose the true model is y = β0 +β1x1 +β2x2 +u (5) but we instead estimate y = β0 +β1x1 +u (6) We know that estimates from (5) are related to the estimates from (6) by the formula ˜β1 = ˆβ1 + ˆβ2 ˜δ1 7 Taking expectations (with implicit conditioning on the x 0 s), E( ˜β1) = E( ˆβ1 + ˆβ2 ˜δ1) = E( ˆβ1) +E( ˆβ2 ˜δ1) = β1 +E( ˆβ2) ˜δ1 = β1 +β2 ˜δ1 This implies that Bias( ˜β1) = E( ˜β1)−β1 = β2 ˜δ1 = β2 cov(dx1, x2) vard(x1) (7) If we have omitted a relevant variable, the bias of our estimator is going to depend on two relationships: 1) the the effect of the excluded independent variable x2 on the dependent variable, and 2) the correlation between the included independent variable (x1) and the excluded variable (x2). Example. Suppose you know that log wages depend linearly on both These notes represent a detailed interpretation of the professor’s lecture. GradeBuddy is best used as a supplement to your own notes, not as a substitute.education and ability: logwagei = β0 +β1educi +β2abilityi +ui (8) But you only include education as a regressor: logwagei = β0 +β1educi +ui (9) Formula (7) suggests that your estimation will be biased upwards if 1) ability has a positive effect on wages (positive β2), and 2) education and ability are positively correlated (positive ˜δ1). We can illustrate this using the 1980 (!) National Longitudinal Study which has information on earnings, education and IQ score (a proxy for ability). Estimating Equation (9) yields logdwagei = 5.97+0.06educi While estimating Equation (8) yields logdwagei = 5.66+0.039educi +0.0059IQi 8 where IQi is the person’s IQ score. We can see the positive correlation between IQ and education through the regression: IQci = 53.7+3.53educiWe can use these three equations to confirm the omitted variable bias formula: 0.06 = 0.039+0.0059 ∗

View Full Document