ECON7IE Topic 1 Class Exercises

ECON7IE Block 1 Topic 1: Dummy explanatory variables

Class exercises on Part 1 of lecture: Qualitative characteristics

1. Demographic determinants of leisure time

Consider the following regression results, generated using the South African Time Use Survey of 2010:
reg leisure age agesq male coloured indian white

Source | SS df MS Number of obs = 39,018

-------------+---------------------------------- F(6, 39011) = 700.67
Model | 215180528 6 35863421.3 Prob > F = 0.0000
Residual | 1.9968e+09 39,011 51184.3355 R-squared = 0.0973
-------------+---------------------------------- Adj R-squared = 0.0971
Total | 2.2119e+09 39,017 56691.5098 Root MSE = 226.24

------------------------------------------------------------------------------
leisure | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | -13.52333 .2602827 -51.96 0.000 -14.03349 -13.01317
agesq | .1784276 .0030344 58.80 0.000 .1724801 .1843751
male | 43.75669 2.309296 18.95 0.000 39.23041 48.28296
coloured | 15.87811 3.769596 4.21 0.000 8.489607 23.26661
indian | -18.4928 7.862506 -2.35 0.019 -33.9035 -3.082088
white | -38.41716 4.561935 -8.42 0.000 -47.35866 -29.47565
_cons | 1144.691 4.931571 232.11 0.000 1135.025 1154.357
------------------------------------------------------------------------------
Where the variables are defined as follows:
leisure Total minutes spent in leisure per day (incl. sleeping & eating)
age Age in years
agesq Age squared
male 1 if male, 0 otherwise
coloured 1 if Coloured, 0 otherwise
indian 1 if Indian, 0 otherwise
white 1 if White, 0 otherwise

1.1 Interpret the coefficients on male and indian. [4]

1.2 What is the average difference in leisure (in minutes per day) between an Indian individual and a
coloured individual, ceteris paribus? [2]
1.3 Describe in words the shape of the relationship between age and leisure.
(Hint: think about a parabola.) [3]
1.4 How old do South Africans need to be before their leisure time starts increasing as they get older?
[4]

Page 1 of 5
 ECON7IE Topic 1 Class Exercises

2. Gender differences in wages

The following wage equation was estimated using labour market data. Standard errors are in brackets.
̂
𝑙𝑛ℎ𝑤𝑎𝑔𝑒 = 2.3 + 0.4𝑚𝑎𝑙𝑒 + 0.08𝑦𝑟𝑠𝑐ℎ𝑜𝑜𝑙 − 0.02𝑚𝑎𝑙𝑒 ∗ 𝑦𝑟𝑠𝑐ℎ𝑜𝑜𝑙
(0.03) (0.09) (0.01) (0.01)
𝑛 = 14271; 𝑅 2 = 0.17
Where the variables are defined as:
lnhwage natural log of hourly wage from employment
male 1 if male, 0 otherwise
yrschool number of years of education completed

2.1 Write out the predicted wage equation for:

• Women [2]
• Men [3]
2.2 Determine how many years of schooling a woman would need in order for the predicted value of
her wage to be the same as a man. (Hint: set the two equations equal to each other and solve for
yrschool).
Is it plausible that women’s wages will ‘catch up’ with men’s? [5]
2.3 Do you agree with the following statement: “Education raises wages for women, but lowers wages
for men”? Explain. [3]

Page 2 of 5
 ECON7IE Topic 1 Class Exercises

Class exercises on Part 2: Structural breaks

3. Determinants of the cost of rent in towns with large student populations

The variables in the dataset are defined as follows:
lrent log of average rent
avginc per capita income, $000s
pctstu percent of population that are students
y90 =1 if year is after 1990, 0 otherwise

3.1 Conduct a Chow test to determine whether the regression equation for lrent is stable before and
after 1990, using the three sets of regression results and the F-table extract shown below:

Regression A:
. reg lrent avginc pctstu

Source | SS df MS Number of obs = 128

-------------+------------------------------ F( 2, 125) = 222.35
Model | 10.9735353 2 5.48676767 Prob > F = 0.0000
Residual | 3.0845993 125 .024676794 R-squared = 0.7806
-------------+------------------------------ Adj R-squared = 0.7771
Total | 14.0581346 127 .110693974 Root MSE = .15709

------------------------------------------------------------------------------
lrent | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
avginc | .0427615 .0020328 21.04 0.000 .0387384 .0467846
pctstu | .0053061 .0010295 5.15 0.000 .0032686 .0073436
_cons | 4.792021 .0533775 89.78 0.000 4.686381 4.897662
------------------------------------------------------------------------------
Regression B:
. reg lrent avginc pctstu if y90==0

Source | SS df MS Number of obs = 64

-------------+------------------------------ F( 2, 61) = 74.41
Model | 1.06639842 2 .533199212 Prob > F = 0.0000
Residual | .437094208 61 .007165479 R-squared = 0.7093
-------------+------------------------------ Adj R-squared = 0.6997
Total | 1.50349263 63 .023864962 Root MSE = .08465

------------------------------------------------------------------------------
lrent | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
avginc | .0426749 .0035119 12.15 0.000 .0356524 .0496974
pctstu | .0039507 .0008049 4.91 0.000 .0023411 .0055603
_cons | 4.756772 .0611158 77.83 0.000 4.634563 4.87898
------------------------------------------------------------------------------
Regression C:
. reg lrent avginc pctstu if y90==1

Source | SS df MS Number of obs = 64

-------------+------------------------------ F( 2, 61) = 39.02
Model | 1.42056308 2 .710281538 Prob > F = 0.0000
Residual | 1.11043626 61 .018203873 R-squared = 0.5613
-------------+------------------------------ Adj R-squared = 0.5469
Total | 2.53099934 63 .040174593 Root MSE = .13492

------------------------------------------------------------------------------
lrent | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
avginc | .024278 .0027571 8.81 0.000 .0187648 .0297912
pctstu | .0040037 .0012976 3.09 0.003 .001409 .0065984
_cons | 5.339289 .0845494 63.15 0.000 5.170222 5.508356
------------------------------------------------------------------------------

Page 3 of 5
 ECON7IE Topic 1 Class Exercises

[6]

3.2 Use the results shown below, which apply the dummy variable method, to answer the questions
that follow:
. g y90pct = y90*pctstu

. g y90inc = y90*avginc

. reg lrent pctstu avginc y90pct y90inc y90

Source | SS df MS Number of obs = 128

-------------+------------------------------ F( 5, 122) = 197.26
Model | 12.5106042 5 2.50212083 Prob > F = 0.0000
Residual | 1.54753047 122 .012684676 R-squared = 0.8899
-------------+------------------------------ Adj R-squared = 0.8854
Total | 14.0581346 127 .110693974 Root MSE = .11263

------------------------------------------------------------------------------
lrent | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
pctstu | .0039507 .001071 3.69 0.000 .0018306 .0060708
avginc | .0426749 .0046726 9.13 0.000 .0334249 .0519248
y90pct | .0000531 .0015232 0.03 0.972 -.0029623 .0030685
y90inc | -.0183969 .0052087 -3.53 0.001 -.028708 -.0080858
y90 | .5825175 .1076725 5.41 0.000 .369369 .795666
_cons | 4.756772 .081315 58.50 0.000 4.595801 4.917743
-------------+----------------------------------------------------------------

(a) Does the intercept change significantly before and after 1990, and if so, how? [3]
(b) Which slope parameter is stable and which is unstable? Explain. [4]

Page 4 of 5
 ECON7IE Topic 1 Class Exercises

4. The leisure function with an age spline

Using the same data from Question 1, we create the following new variables:
d = 1 if age>=35, 0 otherwise
z = age-35
dz = d*z

We then re-estimate the leisure model:

. reg leisure age dz male coloured indian white

Source | SS df MS Number of obs = 39,018

-------------+---------------------------------- F(6, 39011) = 721.73
Model | 221002763 6 36833793.9 Prob > F = 0.0000
Residual | 1.9909e+09 39,011 51035.0895 R-squared = 0.0999
-------------+---------------------------------- Adj R-squared = 0.0998
Total | 2.2119e+09 39,017 56691.5098 Root MSE = 225.91

------------------------------------------------------------------------------
leisure | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
age | -7.56778 .1613542 -46.90 0.000 -7.884039 -7.251522
dz | 14.4508 .2414576 59.85 0.000 13.97753 14.92406
male | 43.21195 2.305773 18.74 0.000 38.69258 47.73133
coloured | 12.86198 3.76313 3.42 0.001 5.486155 20.23781
indian | -20.41697 7.850233 -2.60 0.009 -35.80362 -5.03032
white | -41.79348 4.555947 -9.17 0.000 -50.72325 -32.86371
_cons | 1105.701 4.352787 254.02 0.000 1097.17 1114.233
------------------------------------------------------------------------------

4.1 On average, how does getting older by one year change a person’s amount of leisure time, ceteris
paribus, for someone who is:
(a) Younger than 35? [2]
(b) Older than 35? [3]
4.2 Do a quick sketch of the estimated relationship between age and leisure (not to scale!). Assume
all else is zero. [3]
4.3 Which model explains leisure better: the age quadratic (in Q1) or the age spline? [1]

Page 5 of 5