Econ 410
Introductory Econometrics
Fall 2025
Problem set 1
Due September 26 (Friday) at 11:59pm
Submit through Canvas
Submission Instructions.
• For Stata questions, you must submit the Stata log file in the .pdf or .log format together
with your answers.
– See Section 7 of the discussion handout 1 for how to create a log file.
• Submit one pdf containing your written responses to all questions and another pdf or
log file for your Stata log. Alternatively, you may submit a single pdf that includes
both your written responses and Stata log.
– If you write your answers by hand, use a scanner app or platform to transform your written
work into pdf form.
– Files with the extensions .dta, .do, .smcl, .heic, and .gph cannot be opened on Canvas.
The questions cover topics discussed through September 15.
1. Let (X, Y ) be a pair of two random variables.
(i) Show that V ar(X) = E(X 2 ) − [E(X)]2 .
(ii) Show that Cov(X, Y ) = E(XY ) − E(X)E(Y ).
(iii) Suppose that E(Y |X) = α0 + α1 X, V ar(X) > 0, and Cov(X, Y ) = 0. Is α1 equal to zero?
Explain. (You may use the formula for α1 .)
2. Consider a simple linear regression model
Y = β0 + β1 X + U.
Suppose that we interpret this as a causal model. That is, we assume that the observable X and
the unobservable U determine Y . Suppose that
E(U |X) = α0 + α1 X,
where α0 and α1 are constants and α1 6= 0.
(i) Obtain the conditional expectation function E(Y |X) (i.e., write E(Y |X) in terms of α0 , α1 , β0 , β1 ,
and X).
Cov(X,Y )
(ii) Is β1 equal to V ar(X) ? Explain.
(iii) Consider an individual whose value of Y is determined by the above causal model. Suppose
that the X and U values of this individual are 5 and 2, respectively. What is the Y value of
this individual (in terms of α0 , α1 , β0 , β1 )?
(iv) Consider the same individual as in (iii). Suppose, hypothetically, that we move the X value
of this individual from 5 to 6, holding U fixed at 2. What is the effect of this change on the Y
value of this individual? (This is the causal effect of X on Y .)
1
3. [We use Stata for this question and for Question 4. Include your answers in the pdf of
your written responses, and also submit the Stata log file (in pdf/log format).] For this
question, we use the data set cps78_85.dta. This is a data set from the 1978 and 1985 Current
Population Surveys (CPS) and contains information on 1,084 people.
(i) Load the data set cps78_85.dta.
(ii) Use summarize to summarize the data.
(iii) This data set has “log-wage” rather than wage. Construct the wage using
gen wage=exp(lwage)
(iv) The variable year indicates the year that the data were taken. Produce an estimate of the
expectations of wages conditional on year. That is, the expected wages for 1978 and for 1985.
(One of the easiest ways to do this is to use the tabulate, summarize command.)
(v) female is a binary variable that takes one if female, and married is a binary variable that takes
one if married. Produce an estimate of the expectations of wages conditional on gender and
marital status. That is, the expected wages for married men, married women, unmarried men,
and unmarried women. (Again, you can use the tabulate, summarize command.)
4. For this question, we use the data set [Link]. This data set contains two variables lif eexp
(life expectancy at birth, measured in years) and alcohol (total alcohol consumption per capita,
measured in liters) of 182 countries and economies for 2018. Include the plots you obtain with
your answers in the pdf of your written responses.
(i) Load the data set [Link].
(ii) Use scatter to obtain a scatter plot of lif eexp (Y-axis) against alcohol (X-axis).
(iii) Use the commands mean and correlate (with option covariance) to compute the sample
means, variances, and covariance of lif eexp and alcohol.
(iv) Using the summary statistics computed in (iii) and the formula for the OLS estimator, compute
the OLS estimates β̂0 and β̂1 from a regression of lif eexp on alcohol (i.e., lif eexp is the
dependent variable and alcohol is the independent variable). You may compute the estimates
using a calculator; there is no need to use a Stata command.
(v) Use reg to run a regression of lif eexp on alcohol and check that the estimates in the Stata
output align with the estimates computed in (iv). Use twoway to draw the regression line
\
lif eexp = β̂0 + β̂1 alcohol on the top of the scatter plot.
(vi) Do you think this regression result suggests a causal relationship between alcohol consump-
tion and life expectancy (i.e., β̂1 measures the causal effect of alcohol consumption on life
expectancy)? Briefly explain why.