-l

rI

Final Examination
Econometrics I

f;nswer all the questions'

1. The following regression equation is estimaled as a production tunctron: . ,,

logQ - 1.37 + 0.632logK+0.4521ogL R2 :0.98

cov(b6, b) = -0.044. The sample size is 40. Test the following hypotheses at the
5Vo level of significance:
(a) b6:6r.
(b) There are constant returns to scale.

2. Indicate whether each of the following statements is tnre (T), false @, or uncertain
(U), and give a brief explanation or prooft
(a) Suppose that the coefficient of a variable in a rcgression equation is sipiftcantly
different from zero at the 207o level. If we drop this variable from the regression,
both n2 and R-2 will necessarily decrease.
(b) Compared with the unconstrained regression, estimation of a least squares
regression under a consrraint (say, h will result in a higher R2 if the
-Fl)
coustraint is true and a lower R2 if it is false,
(c) In a least squares regression of y on .r, observations for which:r is far from its
mean will have more effect on the estimated slope than observations for which
.r is close to its mean value.

g. The following estimated equation was obtained by ordinary l"ar, squares regression
using quarterly data for 1960 to 1979 inclusive (I :80):

,, :'O?3 * 0 J 04:1, * t'?rY * oi3.1fi'

standard errors are in parentheses, the explained sum of squares was 112.5, and the
residual sum of squares was 19.5.
(a) Which of the slope coefficients are significantly different from zero at the 5%
significance level?
(b) Calculate the value of R2 for this regression.
(c) Calculate the value of F2 1"ad.lusted R2").

4. A researcher tried two specifications of a regression equation:

y:q+Fx+u
!:o'*p'x*y'2+u'
Explain under what circumstances the following will be true. (A ,,hat" over a
parameter denotes its estimate.)
6) b=b'.
O) If ii and ij are rhe estimated residuals from the rwo equations
Dt? :Dfr,,?,
(c) F is statistically significant (at the SVo level) but p, i, not.
(d) F' is statistically significant (at the 57o tevet) but p is not.
5. suppose that the least sQuares regrcssion of Y on x:,xz,...,x1 yields coeffrcient
estimates b1 Q : L,2, . . ., ,t) none of which exceed their respective standard enors.
However, the F-ratio for the equation rejects, at rhe 0.05 level the hypothesis that
bt=bz=...=Dr:0.
(a) Is this possible? ",,
(b) What do you think is the reason for this?
(c) What further analysis would you perfonn?

6. what would be your answer to the following queries regarding multiple regression
analysis?
(a) I am trying to find out why people go bankrupt. I have gathered data from a
sample of people filing bankruptcy petitions. will these data enable me to find
j]lyryI !o qy_q!$q.oq?
(b) I want to study the costs of auto accidents. Ihave collected data from a sample
of police auto accident reports. Are these data adequate for my purpose?
(c) I am trying to estimate a consumption function and I suspect ttrat the marglnal
propensity to consume varies inversely with the rate of interest. Do I nrn a
multiple regression using income and interest rate as explanatory variables?
(d) I am fining a demand for food function for a sarple of 1000 faurilies. I obtain
an R2 of only 0.05 but the regression program indicates that the F-statistic for
the equation is very significant and so are the r-shtistics. How can this be? Is
there a mistake in the program?
(e) In the regression of Y on r and e, should I leave one of them out?
(f) I know that y depends linearly on .r but I am not sure whether or not it also
depends on another variable z. A friend of mine suggests that I should regrcss y
on.r first, calculate the residuals, and then see whether they are correlated with
z. Is the correct?

7. In a srudy of 27 industrial estabusnnents of varying size, y - fts qrrm[rg1 sf

supervisors and x : the number of supervised workers. y varies from 30 to 210 and t
ftom247 to 1650. The results obtained were as follows:

Variable Coefficient SE

x 0.1 15 0.011 9.30

Constant L4.48 9.562 1.51
n =27 g : 21.73 R2 = 0376
After the estimation of the equation and ploning the residuals against x, it was found
that the variance of the residuals increased with x. Plouing the residuals against l/x
showed that.there was no such relationship. Hence ttre assumption made was

var(a;) - o'i!
The esti:nated equation was

I = 0,l?-l + 3.803(l/.r) R2 = 0.03

.r (0.009) ll.sroj
In terms of the original variables, we have

y : 3.903 + 0, LZLx

T
7.
t.-
I

The esiinrauon gave the touowr-ug results:

Variable Coefficient SE t
x 0.121 0.009 L3.M
Constant 3.803 4.57 0 0.832
n :27 r - 22577 62 - 0.7597

(a) An iuvestigator looks at the drop in R2 and concludes that the first equation is
better. Is this conclusiou valid?
(b) What would the equation to be estimated be if var(u;) * &r, instead of oLfit
How would you determine which of these altemative hypotheses is the better one?
(c) Comment on the computation of R2 from the transformed equation and the R2
from the equation in terms of the original variables.

8. To find out if ado no),

Morgan and Teachman studied a sample of 342 adolescents from the Na tional
.survey af Children, 134 white rnales, L49 white females, 23 black males and 36
'
black females and obtained the fotlowing results from a logistic regression.22
The underlying model is:
p.
ln* = Br+ BzWhitei+ BzFemale,*I,ti,
r- Pt

where Pi = probabilify of sexual intercourse

Variable Slope coefficient se of slope coefficient p value

White - 1 .314 0.226 0.000
Female -0.648 0.225 0.004
Constant 0.192 0,226 0.365
LR statistic 37 ,459, df. = 2

Nrte: All the regressor are dummy variables. The base or comparison categories are
blacks and males, which take values of 0.
(a) How would you interpret the various coefficients?
(b) Are the estimated slope coefficients individually statistically significant? t
How can you tell?
(c) Can you compute the odds ratios from the estimated slopes? Show the
necessary calculations.
(d) How would you interpret the odds ratios obtained in (c)?
(e) Suppose you change the assignments of the dummies, lening blacks and
male take the value I instead of 0. Do you have to repeat the analysis or
can you get this information from the results presented above? (Hintt
Change the sign).

.-3