ECO 441K Introduction to Econometrics

University of Texas Prof. Stephen Donald

In-Class Midterm Exam (March 2, 2023)

NAME: ______________________________________________

There are a total of 55 points on this exam. The exam is closed book. You may only refer
to the exam and the provided “cheat sheet.” Do not spend too much time on any one
question; skip it if you get stuck. Be sure to show any work (for partial-credit purposes).

Good luck!

1
Part 1: True/False Questions (5 points)

(1 points each. Circle “T” (true) or “F” (false). No partial credit. You should only answer “T”
if you think that the statement is always true.)

1. T F A regression of y on x has exactly the same R squared as the regression of

2y on x.

2. T F In the simple linear regression model if add a constant c to y and a

constant d to x then the new slope will be different from what I get when I
regress y on x.

3. T F In the regression model with two regressors the R squared will be the
average of the squared correlation between each regressor and the
dependent variable.

4. T F In the regression of y on log(x) the slope coefficient divided by 100 tells

me the unit change in y due to a 1% change in x.

5. T F The slope coefficient I get when I regress y on x will be the numerically

identical to the slope when I regress x on y when x and y have numerically
identical variances.

6. T F A biased estimator will be inconsistent.

7. T F An estimator will be biased when MLRM.5 is violated.

8. T F When I add an irrelevant variable x2 to a model that includes an intercept

and the variable x1 the variance of the coefficient on x1 will not decrease.

9. T F The null hypothesis that two coefficients are equal has one restriction.

10. T F If two coefficients are individually insignificant (at say the 5% level) they
must also be jointly insignificant.

2
Part 2:

The remainder of the exam involves estimating a regression model in which the determinants of
standardized mathematics test scores using data at the school level. The dependent variable is the
percentage of students passing the test – math10. The explanatory variables are the natural log of
expenditure per student (lexpend) and percentage of students who are in the school lunch
program. (lnchprg) Summary statistics are as follows:
. summarize math10 lexpend lnchprg
Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
math10 | 408 24.10686 10.49361 1.9 66.7
lexpend | 408 8.370177 .1618824 8.111328 8.911799
lnchprg | 408 25.20147 13.61008 1.4 79.5

Here are the variances (diagonals) and covariances between the variables.
. correlate math10 lexpend lnchprg, covariance
(obs=408)

| math10 lexpend lnchprg

-------------+---------------------------
math10 | 110.116
lexpend | .292573 .026206
lnchprg | -59.0646 -.424572 185.234

Here are the results for a multiple regression.

. regress math10 lexpend lnchprg

Source | SS df MS Number of obs = 408

-------------+------------------------------ F( 2, 405) =
Model | 4031.91215 Prob > F =
Residual | 90.7490276 R-squared = 0.1799
-------------+------------------------------ Adj R-squared = 0.1759
Total | 44817.1805 407 110.115923 Root MSE = 9.5262

------------------------------------------------------------------------------
math10 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lexpend | 6.22969 2.972634 2.10 0.037 .3859705 12.07341
lnchprg | -.3045853 .0353574 -8.61 0.000 -.3740923 -.2350783
_cons | -20.36075 25.07288 -0.81 0.417 -69.64998 28.92848

3
1. (6 points) Using one sentence for each, interpret the three estimated coefficients in this
model (pay careful attention to how variables are measured).

2. (4 points) Note that the SSR and the SSE squared have been omitted. Using the output
determine the value of each of these (you may leave it as a fraction or other appropriate
form and you do not need to compute the final solution).

4
3. (4 points) Find the predicted effect of increasing both lexpend and lnchprg by one
standard deviation on the math10 variable. (You should be able to figure out the standard
deviations from the first table).

4. (4 points) Use the information provided to find the correlation between lexpend and
lnchprg. (You can leave this as an expression and you do not need to calculate it).

5. (4 points) Suppose we drop lnchprg from the model and just do a simple linear regression
of math10 on lexpend.

a. The coefficient on lexpend will (circle one): increase decrease

b. Find the R squared for this simple linear regression

5
6. (4 points) Suppose these data were collected via a survey and it was discovered that a
large number of schools declined to fill in the survey information. Which, if any, of the
assumptions MLRM.1-MLRM.5 would be questionable?

7. (4 points) Referring to the original multiple regression indicate which of the regression
coefficients are statistically significant at:
a) The 5% level

b) The 1% level

8. (3 points) Find the 99% confidence interval for the coefficient on lexpend (you can leave
it as a mathematical expression).

6
Part 3:

Using data from a previous class I was able to regress student’s height on the heights of their
parents (fheight and mheight for father and mother respectively). The results are below:
. regress height fheight mheight

Source | SS df MS Number of obs = 50

-------------+---------------------------------- F(2, 47) =
Model | Prob > F =
Residual | 446.825296 R-squared =
-------------+---------------------------------- Adj R-squared =
Total | 652.345 49 13.3131633 Root MSE = 3.0833

------------------------------------------------------------------------------
height | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
fheight | .4819863 .1429127 3.37 0.001 .1944831 .7694895
mheight | .3723857 .1967066 1.89 0.065 -.023337 .7681083
_cons | 11.70553 13.27912 0.88 0.383 -15.00861 38.41968
------------------------------------------------------------------------------

. vce

Covariance matrix of coefficients of regress model

e(V) | fheight mheight _cons

-------------+------------------------------------
fheight | .02042403
mheight | -.00937191 .03869349
_cons | -.82393459 -1.8388983 176.33503

1. (4 points) Using the information provided find an expression for the F test statistic for
testing the overall goodness of fit of the regression and write down an appropriate
decision rule for the test such that the size (significance level) of the test is guaranteed to
be no larger than 5%.

7
2. (4 points) Suppose I wanted to test the hypothesis that fheight and mheight have the same
population slope coefficients. Write down formally the null hypothesis and indicate how
you would conduct a t-test of this hypothesis. (Note: You have all the information needed
to do this so write down the appropriate calculations that you would do to find the t
statistic and write down an appropriate decision rule with a significance level no larger
than 5%).

3. (4 points) How would you do an F test of the hypothesis in 2. In doing this write down
how you would get the SSR (or R squared) for the restricted model and how you would
find the F statistic and write down an appropriate decision rule with significance level no
larger than 5%.