lOMoARcPSD|9059393

ECONOMETRICS EXAM 2021 SECOND PART OPEN WITH

SOLUTIONS
Econometria (Universitat Pompeu Fabra)

StuDocu is not sponsored or endorsed by any college or university

Downloaded by Anh Nguy?n ([email protected])
 lOMoARcPSD|9059393

ECONOMETRICS I
Final exam
26 June 2021

You have 40 minutes to answer the questions of this part. Use only the space in the boxes after each
question to provide the answers. Nothing outside the corresponding box will be taken into account for
the grading. Use the last page for calculations and proofs.

The following results are based on information on 1580 football matches of the First Division in Spain
corresponding to the seasons 1992-93, 1993-1994, 1994-1995 and 1995-96. The definitions of the variables
are:

ladult: Number of sold tickets (log)

lpminr: Price of the cheapest ticket in real terms (log)
intcas: Number of players of the home team who have played with the national team
intvis: Number of players of the away team who have played with the national team
temp: Season (starting year) (92, 93, 94 95)
clim: Weather conditions (1=Heat without rain, 2=Cold without rain, 3=Rain)
poscas: Standings (position) of the home team in the classification (1 to 20)
posvis: Standings (position) of the visiting team in the classification (1 to 20)
lcapaci: Number of tickets on sale (log) (capacity of the stadium)
bar: Dummy variable (1 = visiting team Barcelona, 0 = otherwise)
mad: Dummy variable (1 = visiting team Real Madrid, 0 = otherwise)
rivloc: Rivalry game (1 = Yes, 0 = No)

The Stata ouput and the corresponding commands are:

. reg ladult c.lpminr##i.rivloc ib3.clim poscas posvis bar mad lcapaci i.temp, r
Linear regression Number of obs = 1,580
F(13, 1566) = 86.97
Prob > F = 0.0000
R-squared = 0.4096
Root MSE = .80541

---------------------------------------------------------------------------------
| Robust
ladult | Coef. Std. Err. t P>|t| [95% Conf. Interval]
----------------+----------------------------------------------------------------
lpminr | -.8790332 .0705448 -12.46 0.000 -1.017405 -.7406609
1.rivloc | 2.276154 .7905361 2.88 0.004 .7255333 3.826775
|
rivloc#c.lpminr |
1 | -.4936821 .2531639 -1.95 0.051 -.9902579 .0028938
|
clim |
1 | .4809646 .0709093 6.78 0.000 .3418773 .6200518
2 | .4468882 .0715177 6.25 0.000 .3066076 .5871687
|
poscas | -.0503606 .0036553 -13.78 0.000 -.0575303 -.0431908
posvis | -.0254847 .0038279 -6.66 0.000 -.032993 -.0179763
bar | 1.12603 .086791 12.97 0.000 .9557912 1.296269
mad | 1.059366 .0859997 12.32 0.000 .8906791 1.228052
lcapaci | .2524832 .0293928 8.59 0.000 .1948299 .3101366
|
temp |
93 | .159155 .0591599 2.69 0.007 .0431141 .2751959
94 | .2502336 .061802 4.05 0.000 .1290101 .3714571
95 | .0775709 .0582192 1.33 0.183 -.0366249 .1917667
|
_cons | 7.985274 .3808372 20.97 0.000 7.238269 8.732278
---------------------------------------------------------------------------------

Downloaded by Anh Nguy?n ([email protected])

 lOMoARcPSD|9059393

. mat cov=e(V)

. mat cov_0912=cov[9..12,9..12] /* Selecting elements of rows and columns 9-12 */

. mat list cov_0912

symmetric cov_0912[4,4]
poscas posvis bar mad
poscas .00001336
posvis -3.362e-07 .00001465
bar 6.199e-06 .00012536 .00753268
mad .00002071 .00009974 .00145842 .00739594

/* TEST 1 */
. test lpminr+1.rivloc#c.lpminr=-1
( 1) lpminr + 1.rivloc#c.lpminr = -1

F( 1, 1566) = 2.31
Prob > F = 0.1285

/* TEST 2 */

. test lpminr+1.rivloc#c.lpminr=0
( 1) lpminr + 1.rivloc#c.lpminr = 0

F( 1, 1566) = 31.37
Prob > F = 0.0000

/* TEST 3 */

. test (93.temp=94.temp-93.temp) (95.temp-94.temp=93.temp)

( 1) 2*93.temp - 94.temp = 0
( 2) - 93.temp - 94.temp + 95.temp = 0

F( 2, 1566) = 8.67
Prob > F = 0.0002

/* TEST 4 */
. test 93.temp=94.temp=95.temp

( 1) 93.temp - 94.temp = 0
( 2) 93.temp - 95.temp = 0

F( 2, 1566) = 4.61
Prob > F = 0.0101

/* TEST 5 */
. testparm i(93/95).temp

( 1) 93.temp = 0
( 2) 94.temp = 0
( 3) 95.temp = 0

F( 3, 1566) = 6.32
Prob > F = 0.0003

Downloaded by Anh Nguy?n ([email protected])

 lOMoARcPSD|9059393

Based on the previous information answer the following questions, indicating, when necessary, the number
of the test you are using:

1) Interpret the coefficient of the variable lcapaci. [5 points]

Increasing the number of tickets on sale (capacity of the stadium) by 1% (increasing lcapaci by 0.01)
increases the number of sold tickets by 0.25% (ladult increases by 0.01*0.25), keeping all the other
variables constant

2) Which is the effect on the number of tickets sold if the home team is second instead of third in the
standings (classification) keeping all the other variables constant? [5 points]

If the home team is second instead of third in the standings the variable poscas “decreases” by one
unit (increases -1) and this implies that the number of tickets sold increases by 5.04% (ladult increases
by 0.0504), keeping all the other variables constant

3) A researcher concludes that the price elasticity for rivalry games is not significantly different from -1
and that of non-rivalry games is not significantly different from -1 as well. Is this conclusion correct?
Explain why. [5 points]

For non rivalry games (rivloc=0) the elasticity is the coefficient of lpminr (-0.879) and the value -1 is
within the 95% confidence interval, meaning that we do not reject the null that the elasticity is -1.
For rivalry games (rivloc=1) the elasticity is the coefficient of lpminr (-0.879) plus the coefficient of the
interaction term 1.rivloc#c.lpminr (-0.494). We can test the null hypothesis of the elasticity for rivalry
games being -1 by using TEST 1, and since the p-value is 0.1285 we do no reject the null hypothesis at
the 5% (1% and also 10%) significance level.
The conclusion is CORRECT

Downloaded by Anh Nguy?n ([email protected])

 lOMoARcPSD|9059393

4) Test the null hypothesis that if the visiting team is Barcelona, the number of tickets sold is the same
as if the visiting team is Real Madrid. [5 points]

The null hypothesis to test is  bar= mad ( bar- mad=0) . Given the available information the only way of
testing this hypothesis is by using the t-statistic (bbar-bmad)/S.E.(bbar=bmad), where bbar and bmad are the
OLS estimates. Using the information of the cov_0912 variance-covariance matrix, the t-statistic is
(1.12603-1.059366)/sqrt(0.00753268+0.00739594-2*0.00145842) = 0.609 Since 0.609 is less than 1.96 (in
absolute value) we do not reject the null of non significant difference in the number of tickets sold
depending on the visiting team being Barcelona or Madrid

5) Another researcher has included temp instead of i.temp in order to capture the season effect. Can
you reject this alternative specification? Explain why. [5 points]

Using temp instead of i.temp means that moving from one season to next one have the same effect on
the number of tickets sold it does not matter what season we consider.
This null hypothesis tested by TEST 3 in which the coefficient of 93.temp measures the change from
season 92 to season 93, the change from 93 to 94 is measured by 94.temp-93.temp and the change from
94 to 95 is measured by 95.temp-94.temp, and these changes have to be the same under the null. Since
the p-value is 0.0002 then we reject the null hypothesis

6) A researcher concludes that the season effect is reduced to consider two different groups of seasons
(92 and 95) and (93 and 94). How can we test whether this conclusion is statistically correct? Write the Stata
command to test the null hypothesis based on the previous model, indicating the distribution of the statistic,
including the degrees of freedom if necessary. [5 points]

Since the base level is season 92, the fact that season 95 is in the same group as season 92 means to
have both dummies have the same coefficient and this means being zero. The other constraint is
coefficients of seasons 93 and 94 to be the same.
test (95.temp = 0) (93.temp = 94.temp)
The statistic will be a F statistic with 2 and 1566 degrees of freedom

Downloaded by Anh Nguy?n ([email protected])