Econometrics Final
tsset timevar, timeint regress y x predict uhat, resid reg uhat l.uhat gen newy = y - #coef on l.uhat# * l.y gen newx = x - #coef on l.uhat# * l.y regress newy newx corrects for serial correlation
CMD: Cochrane-Orcutt estimation
tsset timevar, timeint regress d.varname l.varname dfuller varname null hypothesis is that there is high persistence
CMD: Dickey-Fuller test
xtset id time reg d.y d.x
CMD: Difference in differences
tsset timevar, timeint reg d.y d.x l.d.x l2.d.x l3.d.x l4.d.x l5.d.x
CMD: Finite Distributed Lags
xtset id time xtreg y x, fe
CMD: Fixed effects
regress x y predict yhat, xb gen yhat2 = yhat^2 gen yhat3 = yhat^3 regress x y yhat2 yhat3 test yhat2 yhat3 null is no evidence of functional form misspecification
CMD: Ramsey RESET Test
tsset timevar, timeint reg y i.timeint
CMD: Seasonality
tsset timevar, timeint
CMD: Time series
regress y x predict u, resid gen u2 = u^2 predict yhat, xb gen yhat2 = yhat^2 regress u2 on yhat yhat2 F test, null is homoskedasticity
CMD: White Test (special case)
regress y x predict u, resid gen u2 = u^2 regress u2 on all x, x^2, and interactions F test, null is homoskedasticity
CMD: White test for heteroskedasticity
False
T/F The long run propensity (LRP) in a finite distributed lag model is the average of all the coefficients on the included lags of the variable of interest plus the value of the contemporaneous variable of interest.
True
T/F The p-value associated with the Breusch-Pagan test is 0.0002 which is strong evidence against the null hypothesis of homoskedasticity.
True
T/F The statement "morekids is an endogenous variable" means that morekids and u are correlated.
True
T/F The validity of difference-in-differences estimation depends on the assumption that the change in the treated and control groups would have been the same had it not been for the treatment.
False
T/F Under the assumption of classical measurement error in the explanatory variable in the simple linear regression model, the form of the inconsistency in the OLS estimator is
True
T/F Under the classical measurement error assumption, measurement error in an explanatory variable causes attenuation bias.
True
T/F We do not need the normality of the error term assumption to perform valid statistical inference if the other multiple linear regression model assumptions hold and we have a large sample.
True
T/F When the error term in a regression model is heteroskedastic, the OLS estimator is not the best linear unbiased estimator (BLUE).
True
T/F With a large sample size, heteroskedasticity-robust standard errors are valid even if the error term is homoskedastic.
False
T/F With panel data, estimation in first differences and fixed-effects estimation are computationally identical.
False
T/F Both first-differenced estimation and fixed-effects estimation can be used to estimate causal effects if the unobserved factors that are correlated with the dependent variable of interest change over time.
True
T/F Classical measurement error in the dependent variable does not cause bias in the OLS estimator, although it does increase the variance of the OLS estimator.
False
T/F Consider the regression model y=B0+B1x+u, where z is a potential instrument for x. In order for z to be a valid instrument, it must be the case that x and u are uncorrelated.
True
T/F Consider the regression model y=B0+B1x+u, where z is a potential instrument for x. In order for z to be a valid instrument, it must be the case that z and u are uncorrelated.
False
T/F Consider the regression model y=B0+B1x+u, where z is a potential instrument for x. In order for z to be a valid instrument, it must be the case that z and x are uncorrelated.
True
T/F Consider the regression model y=B0+B1x+u, where z is a potential instrument for x. It is possible for z to be a valid instrument even if it is correlated with y.
False
T/F Consider the regression model y=B0+B1x+u, where z is an instrument for x. If the correlation between z and x is stronger, then the variance of the instrumental variables estimator of B1 will be larger.
False
T/F Consider the regression model y=B0+B1x+u, where z is an instrument for x. If z is a valid instrument for x, we replace x with z in the regression model to obtain a consistent estimate of B1.
True
T/F Consider the regression model y=B0+B1x+u, where z is an instrument for x. The instrumental variables estimator of B1 is
True
T/F First differencing can be used to render a highly persistent time series weakly dependent.
True
T/F Fixed-effects estimation can be used to estimate causal effects if the unobserved factors that are correlated with the dependent variable of interest are time invariant.
False
T/F Functional form misspecification is when the model does not properly account for the relationship between the dependent and explanatory variables, often because the appropriate explanatory variables are not observed.
False
T/F Heteroskedasticity causes the OLS estimator to be biased.
False
T/F Heteroskedasticity causes the OLS estimator to be inconsistent.
True
T/F Heteroskedasticity causes the usual estimator of the variance of the OLS estimator to be inconsistent.
False
T/F Heteroskedasticity-robust standard errors are always larger than the usual standard errors.
True
T/F Heteroskedasticity-robust standard errors are valid only if the sample size is large.
True
T/F If a Ramsey RESET test gives a p value <0.05, we reject the null that the regression is correctly specified.
True
T/F If a Ramsey RESET test gives a p value >0.05, we fail to reject the null that the regression is correctly specified.
False
T/F If the average value of the outcome variable is different for the treated and control groups before the treatment, difference-in-differences estimation will not be able to provide an unbiased estimate of the effect.
False
T/F If x is correlated with x* and if x is uncorrelated with the error term, u, then we say that x is a good proxy for x* .
False
T/F RESET is useful in detecting functional form misspecification as well as general omitted variable bias.
True
T/F Regressing a highly persistent time series on another highly persistent time series produces spurious results.
False
T/F Serially correlated errors cause the OLS estimator to be biased and inconsistent.
True
T/F The Breusch-Pagan test and White's special test both give p-values <0.05, so we reject the null hypothesis that the standard errors are homoskedastic and should use heteroskedasticity-robust standard errors.
False
T/F The Breusch-Pagan test and White's special test both give p-values>0.5, so we fail to reject the null that there is heteroskedasticity and we can continue to use homoskedastic standard errors.
False
T/F The Cochrane-Orcutt estimation procedure should be used when regressing a highly persistent time series on another highly persistent time series in order to obtain unbiased parameter estimates.
True
T/F The Dickey-Fuller test can be used to determine if there is evidence that the specified time series is highly persistent.
False
T/F The F test is not useful in detecting functional form misspecification. Instead, one should use RESET or the Davidson-MacKinnon test.
A
Which of the following tests can be used to test if a series is highly persistent?
A
After running this regression , y=B0+B1x1+B2x2+u which of the following regressions is the correct form of RESET (the regression specification error test)?
D
An explanatory variable is said to be endogenous when
regress y x predict u, resid gen u2 = u^2 regress u2 on all x F test, null is homoskedasticity
CMD: Breusch-Pagan test for heteroskedasticity
D
Consider a model . We know x1 is an endogenous variable. Which of the following is true?
A and D
Consider a regression of y on x in which z is included as a control variable. With z included, you believe the zero conditional mean assumption is satisfied. Without z included, you believe the error term is likely correlated with x. Now, suppose that z is not observed, but you have a potential proxy variable w. Describe the characteristics needed for w to be a good proxy for z.
B
Consider the following model where the dependent variable is binary: y=B0+B1x1+B2x2+u Which of the following statements is true?
A
How does Stata calculate the estimate of the heteroskedasticity-robust variance of Beta1hat in the simple linear regression model? Write out the formula as a function of xi and ui.
Take partial derivative with respect to x2
Method: What is the predicted effect from a one unit increase in x2 holding x1 constant?
C
Under which of the following conditions is the difference-in-differences estimator not able to provide an unbiased estimate of the effect?
C
What is the null hypothesis for the Breusch-Pagan Test?
D
What is the null hypothesis of Dickey-Fuller test?
D
What would be an appropriate procedure to correct the standard errors when serial correlation is present in a time series regression model?
C
Which of the following is a consequence of serially correlated errors?