Introductory Econometrics

Exercises for tutorials (Fall 2014)

Jan Zouhar
Dept. of Econometrics, Uni. of Economics, Prague, [email protected]

September 23, 2014

Tutorial 1: Review of basic statistical concepts

Exercise 1.1 (Three M’s.) Assign each term in the 1–4 list a meaning from the a–d list.
Note: two of the 1–4 terms are actually synonymous.
1) Mean.
2) Median.
3) Mode.
4) Expected value.

a) The most likely/frequent value in a population.

b) The long-run average of the results of many independent draws from a population.
c) The value separating the higher half of a population from the lower half.
d) The weighted average of possible values, with the weights being the probabilities of the re-
spective values.

Exercise 1.2 (Wages: mean vs. median.)

a) “The average monthly wage in a population is e 1,000.” Does the average wage relate to the
mean, median or mode of the population’s wage distribution?
b) Which is typically greater, the mean wage or the median wage? (Or, do most people earn
more than or less than the average wage?)

Exercise 1.3 (Calculating the expectations.)

a) Let x be a random variable (rv). Is it possible that Pr{x = Ex} = 0? (For instance, is it
possible that nobody actually earns the average wage?)
b) What is the expected value of a single die roll (for a six-sided die)?
c ) Consider a “loaded” die with uneven probabilities of the possible results, stated in Table 1.
What is the expected value of a roll with such a die?

Table 1: A loaded die

k 1 2 3 4 5 6
Pr{x = k} .1 .1 .1 .2 .2 .3

Exercise 1.4 (Variance and standard deviation.)

a) How would you describe the term variance of an rv to somebody who doesn’t know anything
about statistics?
b) Figure 1 shows distributions of two rvs. Which one has greater variance?
c ) Compare the variances of die-roll results for an ordinary six-sided die and the loaded die from
Table 1.

1
Introductory econometrics: exercises for tutorials Jan Zouhar

Figure 1: Two distributions, different variances

d ) Assume that height of a person is approximately normally distributed with a mean of 180 cm
and variance σ 2 . What percentage of the population falls within ±σ from the population
average (i.e., in the interval [180 − σ, 180 + σ])? And how about the ±2σ or ±3σ range? Draw
a plot that illustrates this.
e ) rv x has the following characteristics: Ex = 10, varx = 0. Is there anything more we can say
about x?

Exercise 1.5 (Calculations with expected value and variance.) Let x and y be independent rvs
with
Ex = 10, Ey = 5,
varx = 1, var y = 2.
Calculate:

a) E[4x]. f) var[4x].
b) E[4x + 5]. g) var[4x + 5].
c) E[x + y]. h) var[x + y].
d) E[x − y]. i) var[x − y].
e) E[4x − 3y + 5]. j) var[4x − 3y + 5].

Exercise 1.6 (Multiple dice.)

a) Imagine we roll two six-sided dice and add up the results. What are the possible outcomes?
What are their probabilities? Draw a plot of the probability function.
b) What is the variance of the rv from a? (Hint: the variance of a die roll is 3512 .)
c ) Suppose we take the arithmetic average of 10 die rolls. What is the expected value and variance
of the result?

Exercise 1.7 (Random sample and sample mean.) The population distribution of the number of
teeth (x) has a mean of 20 with a variance of 100. Assume we draw (at random) a sample of 10
people, measure the value of x for each one of them (thus obtaining values x1 , x2 , . . . , x10 ), and then
1
P10
calculate the arithmetic average x̄ = 10 i=1 xi . Due to random sampling, x̄ is a random variable.
a) What is the expected value of x̄? What is its variance?
b) (Law of large numbers.) Instead of 10 people, we take n now. What happens to E x̄ and var x̄
if we gradually raise n above all limits?
c ) (Central limit theorem.) Again, consider a sample of n people, only that now we study the
result of Pn
√ (xi − 20)
y = n(x̄ − 20) = i=1 √ .
n
As n grows, what happens to the distribution of y?

Exercise 1.8 (Visualizing the central limit theorem.) Download the CLT.zip archive from my
website and run the CLT.m file in Matlab. You should obtain a figure similar to Figure 2. It shows
probability distributions of standardized means (see below) of samples with different sizes, drawn
from the population stored in vector x in the Matlab code, line 4. Here, a standardized mean is the

2
Introductory econometrics: exercises for tutorials Jan Zouhar

√
expression n(x̄ − µ), where µ is the population average. Therefore, the figure shows the effect of
the central limit theorem (clt). The starting population is {1,2,3,4,5,6}, which means that each
observation is in fact a die roll. Notice how fast the distributions converge to the bell-shaped curve
of normal (or Gaussian) distribution. Try changing the population, perhaps repeating some of the
numbers, and see how the convergence process changes.

Figure 2: Die-roll standardized means (clt demonstration)

Exercise 1.9 (Correlation & covariance.)

a) Would you say that wages and education are positively correlated, negatively correlated or
uncorrelated? And how about wages a person’s height?
b) Find an example of negatively correlated economic variables.
c ) If we know that two rvs are negatively correlated, what does it tell us about their covariance?
d ) What are the possible values of a covariance of two rvs?
e ) Let x and y be independent. Is it possible that cov(x, y) = 0.58? Why?
f ) We know that cov(x, y) = 0. Do x and y have to be independent? (If not, find an example of
rvs that are uncorrelated despite not being independent.)
g ) What are the possible values of a correlation coefficient of two rvs?
h) Which of the following can happen:
1) corr(x, y) = −1.56.
2) corr(x, y) = 0.28, cov(x, y) = 0.
3) corr(x, y) = 0.28, cov(x, y) = −0.5.
4) corr(x, y) = 0.28, cov(x, y) = 0.5.
Why? What is the relationship between the covariance and correlation coefficient of two rvs?

Exercise 1.10 (Sharpening your eyes.) In your web browser, type in the address
http://www.ruf.rice.edu/~lane/stat_sim/reg_by_eye/.
Follow the instructions on the website: first press the begin button in the upper-left corner, then
look at the scatterplot of two rvs (we’ll denote them x and y here), and guess which of the five
suggested numbers for corr(x, y) is correct. To see the correct value, click on the Show r button.
Repeat the procedure (using the New Data button) until you manage to guess the right answer three
times in a row.

3
Introductory econometrics: exercises for tutorials Jan Zouhar

Exercise 1.11 (Conditional expectations.)

a) What is the average monthly wage in your country, expressed in e? (Make a rough guess.)
b) Imagine you meet a person on the street and he/she tells you he/she had only finished elemen-
tary school before getting employed. Does this information change your expectation about the
person’s wage?
c ) Estimate the following: E[wage|educ = 9], E[wage|educ = 13], E[wage|educ = 18].1
d ) Based on c, try to approximate E[wage|educ] using a linear relationship

E[wage|educ] = β0 + β1 educ.

e ) Based on d, what is the expected difference in wages for two people with a gap of 1 year
in their education? In other words, what is the population average of ∆wage
∆educ ? (Or: what is
∆E[wage|educ]
∆educ ?)

Exercise 1.12 (Calculations with conditional expectations.)

a) Let x and y be independent rvs, Ey = 12.5. Find E[y |x].
b) Let x and y be rvs, E[y |x] = 2 + 5x. Find E[4y + 3xy + x2 |x] and E[4y + 3xy + x2 |x = 5].

Exercise 1.13 (Conditional expectations II.) Suppose that at a large university, college grade point
average, GPA, and SAT score, SAT , are related by the conditional expectation

E[GPA|SAT ] = .70 + .002 SAT .

a) Find the expected GPA when SAT = 800. Find E[GPA|SAT = 1,400]. Comment on the
difference.
b) If the average SAT in the university is 1,100, what is the average GPA?

Exercise 1.14 (Conditional variance.) Do you think the variance of wages varies among groups of
people with different levels of education? E.g., is there a difference between var[wage|educ = 9] and
var[wage|educ = 18]?

Tutorial 2: Simple regression

Exercise 2.1 (Gretl practice.) Import data from the MS Excel file simplereg.xls into Gretl (use
the drag-and-drop trick). (This is a fictitious dataset, the numbers don’t have any real meaning.)
a) Regress y on x; i.e., estimate (using Model → Ordinary least squares) the model

E[y |x] = β0 + β1 x.

b) Write down the estimated regression function. (Note: once we’ve estimated a model, we
typically write ŷ = β̂0 + β̂1 x.)
c ) From the Gretl output, read the value of R2 . What does it tell us about the model?
d ) Find (in Gretl or by calculation) the values of SST, SSR and SSE.
e ) Draw the scatterplot of xi ’s and yi ’s with the estimated regression line in it (Graphs → Fitted,
actual plot → Against x).
f ) Save the residuals (ûi ) from the estimated model as a new variable (Save → Residuals). Next,
find the sample mean of û (View → Summary Statistics) and sample correlation between û
and x (View → Correlation Matrix). Is the result unexpected, or can we generalize it to other
regression models? Explain why.

Exercise 2.2 (Campaign expenditures.) Load the data voting.gdt. The data describe election
outcomes and campaign expenditures for 173 two-party races (A,B) for the U.S. House of Repre-
sentatives in 1988.
1 Here educ is expressed in years, i.e. 9 years typically represent elementary education and 18 years a master’s degree.

4
Introductory econometrics: exercises for tutorials Jan Zouhar

a) Consider the following regression model:

E[voteA|expenA] = β0 + β1 expenA.

Does it make sense to use a model like this to describe the relationship between campaign
expenditures and the eventual vote share? How would you interpret β0 and β1 ? In a two-
party race, do you think it makes sense to look at the campaign expenditures of party A
alone?
b) Consider the following regression model:

E[voteA|shareA] = β0 + β1 shareA

where shareA is A’s percentage share in the total campaign expenditures (“total” meaning the
sum across all parties). Generate shareA in Gretl (use Add → Define new variable), estimate
the model and interpret the estimates.
c ) Find a story for the association between voteA and shareA supporting each of the three cau-
sation schemes.

Exercise 2.3 (Constant elasticity model.) Load the data ceosal1.gdt (the ceos’ salaries data
from Lecture 2). This time, we relate salary to sales.
a) Consider the following population regression model:

log(salary) = β0 + β1 log(sales) + u.

Can you express the elasticity of salary with respect to sales in terms of the regression coeffi-
cients β0 and β1 ?
b) Generate log(salary) and log(sales) in Gretl (use Add → Logs of selected variables) and estimate
the regression model.
c ) Regress salary on sales without logarithms and look at the R2 ’s in both models. Does a
comparison of the two R2 ’s tell us something meaningful?

Exercise 2.4 (Monte Carlo.) In the lectures, we study the linear regression model (lrm) using
analytical means. There’s another way to study the linear regression model, and that is using a
computer simulation. Consider a population model

y = β0 + β1 x + u, β0 = 5, β1 = 10 (1)

and a random sample consisting of 15 observations. Carry out the following simulation in MS Excel.
You can use the MonteCarlo.xls file if you like.
a) Generate x and u values at random and write them down in two columns. Use the RAND-
BETWEEN function to do this (the function generates a random integer within the specified
bounds). You can pick any range for x; however, note that in a clrm, Eu has to be zero.
Therefore, the lower and upper bounds for u have to be opposite numbers; i.e., use RANDBE-
TWEEN(−umax , umax ).
b) Create columns with both y and E[y |x]; these will be calculated based on (1).
c ) Draw a scatterplot of y vs. x and include the prf in it (i.e., the line E[y |x] = β0 + β1 x). Press
F9 repeatedly to see how the random sampling procedure looks like. What does u represent
in the plot?
d ) Calculate the ols-estimates β̂0 and β̂1 using INTERCEPT and SLOPE functions. Next, calculate
the values of ŷ and û, and add the srf line (i.e., the line ŷ = β̂0 + β̂1 x) into your plot. Press F9
repeatedly to see how accurately the ols procedure estimates β0 and β1 . Which one is more
accurate (on average), β̂0 or β̂1 ?
e ) Press F9 ten times, write down the results for β̂0 and β̂1 , and then make the arithmetic average
of the 10 trials for both β̂0 and β̂1 . What results would you expect if we did 1000 trials instead
of 10?

5
Introductory econometrics: exercises for tutorials Jan Zouhar

f ) Open the MonteCarlo2.xls file. The experiment from e is automated here, only with 1000
trials instead of 10. The 1000 values of β̂0 and β̂1 are shown in columns W and AC. In the same
columns, you can see the mean and (sample) standard deviation of the 1000 trials. Compare
the standard deviations for β̂0 and β̂1 . Does the difference in the two values reflect your
conclusions about the accuracy of the estimates?
g ) The histogram plots on the right show the frequencies of β̂0 (green) and β̂1 (blue) values
among the 1000 trials in the experiment. These plots tell us something about the shape of the
distribution of rvs β̂0 and β̂1 . Does the shape of the plots remind you of one of the well-known
distributions?

Exercise 2.5 (Conditional variance of β̂1 using Monte Carlo simulation.) In the lectures, we
derived a formula for conditional variance of β̂1 given x by analytical means. In this exercise, you’ll
verify the result using Monte Carlo simulation. In principle, this will be the same simulation as in
the previous exercise, only that now we study the conditional distribution of β̂1 . In order to do this,
we need to select particular values of the explanatory variable and keep these values fixed in the
repeated samples (i.e., only the values of u are sampled each time, x remains the same). Proceed
as follows:
a) Open the MonteCarlo3.xls file.
b) Select specific values for x; e.g., replace the formulas in the green column I with odd numbers
going from 1 to 29.
c ) Look at the sample variance of the resulting 1000 trials for β̂1 in cell X1006. Compare this
number to the analytic result, which says

σ2
var[β̂1 |x] = .
s2x
P15
Note that s2x = i=1 (xi − x̄)2 and σ 2 is the variance of u. We’re using RANDBETWEEN to
generate u, which means u has discrete uniform distribution, the variance of which is

(b − a − 1)2 − 1
,
12
where a and b are the lower and upper bound, respectively. (Fill the formulas for σ 2, s2x
and var[β̂1 |x] in cells Q5, Q6 and Q7, respectively.) Press F9 repeatedly and comment on the
difference between the analytical and simulation results.

Exercise 2.6 (Factors affecting β̂1 variance.) We’ll continue working with the MonteCarlo3.xls
file. From the lectures, you know that . . .
i) . . . the less variance in the disturbances,
ii) . . . the more variance in the explanatory variable,
the more accurate estimates we obtain. Verify this using the simulation model.
a) Try changing the range of x-values (e.g., pick the numbers 10–24 or 0–70) and see how the
variance of the estimates varies.
b) Try changing the umax value and watch the resulting change in the variance of the estimates.

Tutorial 3: Multiple regression I

Exercise 3.1 (Sleep vs. work). The following model is a simplified version of the multiple regression
model used by Biddle and Hamermesh (1990) to study the tradeoff between time spent sleeping and
working and to look at other factors affecting sleep:

sleep = β0 + β1 totwrk + β2 educ + β3 age + u,

where sleep and totwrk (total work) are measured in minutes per week and educ and age are measured
in years.

6
Introductory econometrics: exercises for tutorials Jan Zouhar

a) If adults trade off sleep for work, what is the sign of β1 ?

b) What signs do you think β2 and β3 will have?
c) Using the data in sleep.gdt, estimate the model and write out the results in equation form.
d) If someone works five more hours per week, by how many minutes is sleep predicted to fall?
Is this a large tradeoff?
e ) Discuss the sign and magnitude of the estimated coefficient on educ.
f ) Would you say totwrk, educ, and age explain much of the variation in sleep? What other
factors might affect the time spent sleeping? Are these likely to be correlated with totwrk?

Exercise 3.2 (Housing prices and pollution). The following equation describes the median housing
price in a community in terms of amount of pollution (nox for nitrous oxide) and the average number
of rooms in houses in the community (rooms):

log(price) = β0 + β1 log(nox) + β2 rooms + u. (2)

a) What are the probable signs of β1 and β2 ? What is the interpretation of β1 ? Explain.
b) Why might nox (or the log of nox) and rooms be negatively correlated? If this is the case,
does the simple regression of log(price) on log(nox) produce an upward or downward biased
estimator of β1 ?
c ) Using the data in houses.gdt, estimate (2) and the following model:

log(price) = β0 + β1 log(nox) + u.

Is the relationship between the simple and multiple regression estimates of the elasticity of
price with respect to nox what you would have predicted, given your answer in part b? Does
this mean that β̂1 from (2) is definitely closer to the true elasticity than β̂1 from the simple
regression model?

Exercise 3.3 (Building up an econometric model). You were asked to carry out empirical research
in order to quantify the so-called returns to schooling, i.e., the effect of additional education on a
person’s wage. In lecture 1, we discussed the steps to be carried out in empirical analysis:

Step 1: Formulate the question of interest.

Step 2: Find a suitable economic model.
Step 3: Turn it into an econometric model.
Step 4: Obtain suitable data.
Step 5: Use econometric methods to estimate the econometric model.
Step 6: If needed, use hypothesis tests to answer the question of interest.

Step 1 has already been made for you: the question of interest is stated above. Your task here is to
discuss steps 2, 3 and 4 in detail.
a) Put up a list of all thinkable factors that shape a person’s wage.
b) Try to argue the causal link from each of the factors to a person’s wage from the standpoint
of an economic theory. Arguments such as “we all know that clever people earn more money”
do not count here.
c ) Explain how you would quantify the factors you identified. First, decide whether a factor is
directly measurable, or whether we need to find a suitable proxy variable. Second, explain
what units you’d use use for quantification.
d ) Write down the econometric model you would use in order to estimate the effect of wages on
education. Is it necessary to include all the factors (or their proxies) in the regression model?
Are some of them more important than others?
e ) Is it possible to drop one of the variables you included in the model without violating the
E[u|educ] = 0 assumption?
f ) Based on your economic argumentation from b, what values of the β parameters do you expect?
(For each jth variable, give at least the expected sign of βj .)

7
Introductory econometrics: exercises for tutorials Jan Zouhar

g ) Imagine you’ve collected the data you need and saved them into an MS Excel file. Sketch a
structure of the MS Excel file (think up arbitrary data for the first two observations).

Tutorial 4: Multiple regression II

Exercise 4.1 (Partialling out). Using the 526 observations on workers in wage.gdt, regress the log
of wage (hourly wage in $) on educ (years of education), exper (years of labor market experience),
and tenure (years with the current employer).
a) Formulate the population regression model you are estimating.
b) Write down the estimated equation and interpret the values of the estimated coefficients.
c ) Confirm the partialling out interpretation of the ols estimates by explicitly doing the par-
tialling out. This first requires regressing educ on exper and tenure, and saving the residuals,
r̂1 . Then, regress log(wage) on r̂1 . Compare the coefficient on r̂1 with the coefficient on educ
in the regression from b.

Exercise 4.2 (Working with categories). In a study relating college grade point average to time
spent in various activities, you distribute a survey to several students. The students are asked how
many hours they spend each week in four activities: studying, sleeping, working, and leisure. Any
activity is put into one of the four categories, so that for each student the sum of hours in the four
activities must be 168.
a) Consider the model

GPA = β0 + β1 study + β2 work + β3 leisure + β4 sleep + u.

Interpret the coefficient β1 . Does it make sense to hold work, leisure, and sleep fixed, while
changing study?
b) Explain why this model violates Assumption MLR.3.
c ) How could you reformulate the model so that its parameters have a useful interpretation and
it satisfies Assumption MLR.3?

Exercise 4.3 (“Harmless” multicollinearity). Suppose you postulate a model explaining final exam
score in terms of class attendance. Thus, the dependent variable is final exam score, and the key
explanatory variable is number of classes attended. To control for student abilities and efforts
outside the classroom, you include among the explanatory variables cumulative GPA, SAT score,
and measures of high school performance. Someone says, “You cannot hope to learn anything from
this exercise because cumulative GPA, SAT score, and high school performance are likely to be
highly collinear.” What should be your response?

Exercise 4.4 (Variable selection). The causal links between variables y, x, z1 , z2 , z3 and z4 are
shown in Figure 3. Your task is to quantify the causal effect of x on y. Which of the variables will
you include in your equation?

y z1 z2 z3

z4 x

Figure 3: Causal links between variables y, x, z1 , z2 , z3 and z4

Exercise 4.5 (Used cars). The file used_cars_original.xls contains data on used Škodas that
I collected in January 2004. At that time, I owned an old Škoda Felicia and was thinking about
selling it; I didn’t sell it in the end, and later in 2009, it suddenly caught fire when my dad was
driving it (accidentaly, this happened very close to our university premises), see

8
Introductory econometrics: exercises for tutorials Jan Zouhar

http://www.pozary.cz/clanek/20375-obrazem-u-bulhara-horela-skodovka/.
When you open the file in MS Excel, you’ll notice that the data format is not suitable for econometric
work (why?). Compare used_cars_original.xls with used_cars.xls and notice the way the
qualitative variables were encoded into dummies.
a) Regress price on km and age; interpret all the estimated coefficients. Is there any meaningful
interpretation of the intercept? Do you find its level reasonable?
b) Regress price on km and year now. Compare the results with your previous model. How would
you interpret the intercept in this case?
c ) In the data, I created the variable age from the year of manufacture in the following way:
age = 2004 − year. Notice the impact this relationship had on the coefficients you estimated
in a and b.
d ) Regress price on all available explanatory variables. Why were some of the variables omitted
by Gretl? Interpret the coefficients for the dummy variables that were retained.
e ) Try and find the best regression model for the price of a used Škoda. Consider (and estimate)
various function forms.
f ) How much value does a used car lose (on average) with each additional kilometre? Discuss the
various function shapes you used in e.
g ) What price would you ask (in 2004) for Škoda Felicia, which has 100.000 km on the clock, the
engine 1.9D and was manufactured in 1998?
h) Would be a version with petrol engine be cheaper? By how much?
i ) What is the price difference between used Octavias and Felicias? What data will you use to
find this out?
j) Find out whether the extra charge for the combi version varies for Octavias and Felicias.
k ) Find out whether the extra charge for a diesel engine varies for Octavias and Felicias.
l ) Find out whether the average value loss per km varies for Octavias and Felicias.

Tutorial 5: Hypothesis testing

Exercise 5.1 (Theory check). Which of the following can cause the usual ols t statistics to be
invalid (that is, not to have t distributions under H0 )?
a) Heteroskedasticity.
b) A sample correlation coefficient of .95 between two independent variables that are in the model.
c ) Omitting an important explanatory variable.

Exercise 5.2 (Practical vs. statistical significance). Consider an equation to explain salaries of
ceos in terms of annual firm sales, return on equity (roe, in percent form), and return on the firm’s
stock (ros, in percent form):
log(salary) = β0 + β1 log(sales) + β2 roe + β3 ros + u.
a) In terms of the model parameters, state the null hypothesis that, after controlling for sales and
roe, ros has no effect on ceo salary. State the alternative that better stock market performance
increases a ceo’s salary.
b) Esitmate the equation using the data in ceosal1.gdt. By what percent is salary predicted to
increase, if ros increases by 50 points? Does ros have a practically large effect on salary?
c ) Test the null hypothesis that ros has no effect on salary, against the alternative that ros has
a positive effect. Carry out the test at the 10% significance level.
d ) Would you include ros in a final model explaining ceo compensation in terms of firm perfor-
mance? Explain.

Exercise 5.3 (Individual vs. joint significance). Using the data in sleep.gdt, estimate
sleep = β0 + β1 totwrk + β2 educ + β3 age + u
and report the results in equation form.

9
Introductory econometrics: exercises for tutorials Jan Zouhar

a) Is either educ or age individually significant at the 5% level against a two-sided alternative?
Show your work.
b) Drop educ and age from the equation and report the results in equation form. Are educ and
age jointly significant in the original equation at the 5% level? Justify your answer.
c ) Does including educ and age in the model greatly affect the estimated tradeoff between sleeping
and working?
d ) Suppose that the sleep equation contains heteroskedasticity. What does this mean about the
tests computed in parts a and b?

Exercise 5.4 (Using confidence intervals to test hypotheses). The variables in GPA.gdt include
college grade point average (colGPA), high school GPA (hsGPA), achievement test score (ACT ),
and the average number of lectures missed per week (skipped) for a sample of 141 students from a
large university; both college and high school GPAs are on a four-point scale. Estimate the following
equation, which can be used to study the effects of skipping class on college GPA:

colGPA = β0 + β1 hsGPA + β2 ACT + β3 skipped + u.

a) Using the standard normal approximation, find the 95% confidence interval for hsGPA.
b) Can you reject the hypothesis H0 : hsGPA = .4 against the two-sided alternative at the 5%
level?
c ) Can you reject the hypothesis H0 : hsGPA = 1 against the two-sided alternative at the 5%
level?

Exercise 5.5 (Linear restrictions). Use the data in wages2.gdt for this exercise.
a) Consider the standard wage equation

log(wage) = β0 + β1 educ + β2 exper + β3 tenure + u.

State the null hypothesis that another year of general workforce experience has the same effect
on log(wage) as another year of tenure with the current employer.
b) Test the null hypothesis in part a against a two-sided alternative, at the 5% significance level,
by constructing a 95% confidence interval. What do you conclude?

10