CHAPTER 15 Instrumental Variables Estimation and Two Stage Least Squares 543

exogenous, that is, uncorrelated with the error term of the structural equation; (2) it must
be partially correlated with the endogenous explanatory variable. Finding a variable with
these two properties is usually challenging.
The method of two stage least squares, which allows for more instrumental variables
than we have explanatory variables, is used routinely in the empirical social sciences.
When used properly, it can allow us to estimate ceteris paribus effects in the presence of
endogenous explanatory variables. This is true in cross-sectional, time series, and panel
data applications. But when instruments are poor—which means they are correlated with
the error term, only weakly correlated with the endogenous explanatory variable, or both—
then 2SLS can be worse than OLS.
When we have valid instrumental variables, we can test whether an explanatory vari-
able is endogenous, using the test in Section 15.5. In addition, though we can never test
whether all IVs are exogenous, we can test that at least some of them are—assuming that
we have more instruments than we need for consistent estimation (that is, the model is
overidentified). Heteroskedasticity and serial correlation can be tested for and dealt with
using methods similar to the case of models with exogenous explanatory variables.
In this chapter, we used omitted variables and measurement error to illustrate the
method of instrumental variables. IV methods are also indispensable for simultaneous
equations models, which we will cover in Chapter 16.

Key Terms
Endogenous Explanatory Instrumental Variable Overidentifying Restrictions
Variables Instrumental Variables (IV) Rank Condition
Errors-in-Variables Estimator Reduced Form Equation
Exclusion Restrictions Instrument Exogeneity Structural Equation
Exogenous Explanatory Instrument Relevance Two Stage Least Squares
Variables Natural Experiment (2SLS) Estimator
Exogenous Variables Omitted Variables Weak Instruments
Identification Order Condition

Problems
1 Consider a simple model to estimate the effect of personal computer (PC) ownership on
college grade point average for graduating seniors at a large public university:
GPA  0  1PC  u,
where PC is a binary variable indicating PC ownership.
(i) Why might PC ownership be correlated with u?
(ii) Explain why PC is likely to be related to parents’ annual income. Does this mean
parental income is a good IV for PC? Why or why not?
(iii) Suppose that, four years ago, the university gave grants to buy computers to roughly
one-half of the incoming students, and the students who received grants were
randomly chosen. Carefully explain how you would use this information to construct
an instrumental variable for PC.

Copyright 2012 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 544 PART 3 Advanced Topics

2 Suppose that you wish to estimate the effect of class attendance on student performance, as
in Example 6.3. A basic model is
stndfnl  0  1atndrte  2 priGPA  3 ACT  u,
where the variables are defined as in Chapter 6.
(i) Let dist be the distance from the students’ living quarters to the lecture hall. Do you
think dist is uncorrelated with u?
(ii) Assuming that dist and u are uncorrelated, what other assumption must dist satisfy to
be a valid IV for atndrte?
(iii) Suppose, as in equation (6.18), we add the interaction term priGPAatndrte:
stndfnl  0  1atndrte  2 priGPA  3 ACT  4 priGPAatndrte  u.
If atndrte is correlated with u, then, in general, so is priGPAatndrte. What might be a
good IV for priGPAatndrte? [Hint: If E(upriGPA, ACT, dist)  0, as happens when
priGPA, ACT, and dist are all exogenous, then any function of priGPA and dist is uncor-
related with u.]
3 Consider the simple regression model
y  0  1x  u
and let z be a binary instrumental variable for x. Use (15.10) to show that the IV estimator
ˆ1 can be written as

ˆ1  ( y-1  y-0)/(x-1  x-0),
where y-0 and x-0 are the sample averages of yi and xi over the part of the sample with zi  0, and
where y-1 and x-1 are the sample averages of yi and xi over the part of the sample with zi  1. This
estimator, known as a grouping estimator, was first suggested by Wald (1940).
4 Suppose that, for a given state in the United States, you wish to use annual time series data
to estimate the effect of the state-level minimum wage on the employment of those 18 to
25 years old (EMP). A simple model is
gEMPt  0  1gMINt  2gPOPt  3gGSPt  4gGDPt  ut,
where MINt is the minimum wage, in real dollars, POPt is the population from 18 to
25 years old, GSPt is gross state product, and GDPt is U.S. gross domestic product. The
g prefix indicates the growth rate from year t  1 to year t, which would typically be
approximated by the difference in the logs.
(i) If we are worried that the state chooses its minimum wage partly based on unob-
served (to us) factors that affect youth employment, what is the problem with OLS
estimation?
(ii) Let USMINt be the U.S. minimum wage, which is also measured in real terms. Do
you think gUSMINt is uncorrelated with ut?
(iii) By law, any state’s minimum wage must be at least as large as the U.S. minimum.
Explain why this makes gUSMINt a potential IV candidate for gMINt.
5 Refer to equations (15.19) and (15.20). Assume that u  x, so that the population varia-
tion in the error term is the same as it is in x. Suppose that the instrumental variable, z, is
slightly correlated with u: Corr(z,u)  .1. Suppose also that z and x have a somewhat stron-
ger correlation: Corr(z,x)  .2.
(i) What is the asymptotic bias in the IV estimator?
(ii) How much correlation would have to exist between x and u before OLS has more
asymptotic bias than 2SLS?

Copyright 2012 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 CHAPTER 15 Instrumental Variables Estimation and Two Stage Least Squares 545

6 (i) In the model with one endogenous explanatory variable, one exogenous explanatory
variable, and one extra exogenous variable, take the reduced form for y2 (15.26), and
plug it into the structural equation (15.22). This gives the reduced form for y1:
y1  0  1z1  2z2  v1.
Find the j in terms of the j and the j.
(ii) Find the reduced form error, v1, in terms of u1, v2, and the parameters.
(iii) How would you consistently estimate the aj?
7 The following is a simple model to measure the effect of a school choice program on stan-
dardized test performance [see Rouse (1998) for motivation and Computer Exercise C11
for an analysis of a subset of Rouse’s data]:
score  0  1choice  2 faminc  u1,

where score is the score on a statewide test, choice is a binary variable indicating whether
a student attended a choice school in the last year, and faminc is family income. The IV for
choice is grant, the dollar amount granted to students to use for tuition at choice schools.
The grant amount differed by family income level, which is why we control for faminc in
the equation.
(i) Even with faminc in the equation, why might choice be correlated with u1?
(ii) If within each income class, the grant amounts were assigned randomly, is grant
uncorrelated with u1?
(iii) Write the reduced form equation for choice. What is needed for grant to be partially
correlated with choice?
(iv) Write the reduced form equation for score. Explain why this is useful. (Hint: How do
you interpret the coefficient on grant?)
8 Suppose you want to test whether girls who attend a girls’ high school do better in math
than girls who attend coed schools. You have a random sample of senior high school girls
from a state in the United States, and score is the score on a standardized math test. Let
girlhs be a dummy variable indicating whether a student attends a girls’ high school.
(i) What other factors would you control for in the equation? (You should be able to rea-
sonably collect data on these factors.)
(ii) Write an equation relating score to girlhs and the other factors you listed in part (i).
(iii) Suppose that parental support and motivation are unmeasured factors in the error
term in part (ii). Are these likely to be correlated with girlhs? Explain.
(iv) Discuss the assumptions needed for the number of girls’ high schools within a
20-mile radius of a girl’s home to be a valid IV for girlhs.
(v) Suppose that, when you estimate the reduced form for girlshs, you find that the
coefficient on numghs (the number of girls’ high schools within a 20-mile radius) is
negative and statistically significant. Would you feel comfortable proceeding with IV
estimation where numghs is used as an IV for girlshs? Explain.
9 Suppose that, in equation (15.8), you do not have a good instrumental variable candidate for
skipped. But you have two other pieces of information on students: combined SAT score
and cumulative GPA prior to the semester. What would you do instead of IV estimation?
10 In a recent article, Evans and Schwab (1995) studied the effects of attending a Catholic
high school on the probability of attending college. For concreteness, let college be a
binary variable equal to unity if a student attends college, and zero otherwise. Let CathHS

Copyright 2012 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 546 PART 3 Advanced Topics

be a binary variable equal to one if the student attends a Catholic high school. A linear
probability model is
college  0  1CathHS  other factors  u,
where the other factors include gender, race, family income, and parental education.
(i) Why might CathHS be correlated with u?
(ii) Evans and Schwab have data on a standardized test score taken when each student
was a sophomore. What can be done with this variable to improve the ceteris paribus
estimate of attending a Catholic high school?
(iii) Let CathRel be a binary variable equal to one if the student is Catholic. Discuss
the two requirements needed for this to be a valid IV for CathHS in the preceding
equation. Which of these can be tested?
(iv) Not surprisingly, being Catholic has a significant positive effect on attending a
Catholic high school. Do you think CathRel is a convincing instrument for CathHS?
11 Consider a simple time series model where the explanatory variable has classical
measurement error:
yt  0  1xt*  ut [15.58]
xt  xt*  et,
where ut has zero mean and is uncorrelated with x *t and et. We observe yt and xt only.
Assume that et has zero mean and is uncorrelated with x*t and that x*t also has a zero mean
(this last assumption is only to simplify the algebra).
(i) Write xt*  xt  et and plug this into (15.58). Show that the error term in the new
equation, say, vt, is negatively correlated with xt if 1  0. What does this imply
about the OLS estimator of 1 from the regression of yt on xt?
(ii) In addition to the previous assumptions, assume that ut and et are uncorrelated with
all past values of x*t and et; in particular, with xt*1 and et1. Show that E(xt1vt)  0,
where vt is the error term in the model from part (i).
(iii) Are xt and xt1 likely to be correlated? Explain.
(iv) What do parts (ii) and (iii) suggest as a useful strategy for consistently estimating
0 and 1?

Computer Exercises
C1 Use the data in WAGE2.RAW for this exercise.
(i) In Example 15.2, if sibs is used as an instrument for educ, the IV estimate of the
return to education is .122. To convince yourself that using sibs as an IV for educ
is not the same as just plugging sibs in for educ and running an OLS regression,
run the regression of log(wage) on sibs and explain your findings.
(ii) The variable brthord is birth order (brthord is one for a first-born child, two for a
second-born child, and so on). Explain why educ and brthord might be negatively
correlated. Regress educ on brthord to determine whether there is a statistically
significant negative correlation.
(iii) Use brthord as an IV for educ in equation (15.1). Report and interpret the results.
(iv) Now, suppose that we include number of siblings as an explanatory variable in the
wage equation; this controls for family background, to some extent:
log(wage)  0  1educ  2sibs  u.

Copyright 2012 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 CHAPTER 15 Instrumental Variables Estimation and Two Stage Least Squares 547

Suppose that we want to use brthord as an IV for educ, assuming that sibs is exog-
enous. The reduced form for educ is
educ  0  1sibs  2brthord  v.
State and test the identification assumption.
(v) Estimate the equation from part (iv) using brthord as an IV for educ (and sibs as
its own IV). Comment on the standard errors for  ˆeduc and 
ˆsibs.
(vi) Using the fitted values from part (iv), educ, compute the correlation between 
 educ
and sibs. Use this result to explain your findings from part (v).
C2 The data in FERTIL2.RAW include, for women in Botswana during 1988, informa-
tion on number of children, years of education, age, and religious and economic status
variables.
(i) Estimate the model
children  0  1educ  2age  3age2  u
by OLS, and interpret the estimates. In particular, holding age fixed, what is the
estimated effect of another year of education on fertility? If 100 women receive
another year of education, how many fewer children are they expected to have?
(ii) The variable frsthalf is a dummy variable equal to one if the woman was born dur-
ing the first six months of the year. Assuming that frsthalf is uncorrelated with the
error term from part (i), show that frsthalf is a reasonable IV candidate for educ.
(Hint: You need to do a regression.)
(iii) Estimate the model from part (i) by using frsthalf as an IV for educ. Compare the
estimated effect of education with the OLS estimate from part (i).
(iv) Add the binary variables electric, tv, and bicycle to the model and assume these
are exogenous. Estimate the equation by OLS and 2SLS and compare the esti-
mated coefficients on educ. Interpret the coefficient on tv and explain why televi-
sion ownership has a negative effect on fertility.
C3 Use the data in CARD.RAW for this exercise.
(i) The equation we estimated in Example 15.4 can be written as
log(wage)  0  1educ  2exper  …  u,
where the other explanatory variables are listed in Table 15.1. In order for IV to
be consistent, the IV for educ, nearc4, must be uncorrelated with u. Could nearc4
be correlated with things in the error term, such as unobserved ability? Explain.
(ii) For a subsample of the men in the data set, an IQ score is available. Regress IQ on
nearc4 to check whether average IQ scores vary by whether the man grew up near
a four-year college. What do you conclude?
(iii) Now, regress IQ on nearc4, smsa66, and the 1966 regional dummy variables
reg662, …, reg669. Are IQ and nearc4 related after the geographic dummy vari-
ables have been partialled out? Reconcile this with your findings from part (ii).
(iv) From parts (ii) and (iii), what do you conclude about the importance of controlling
for smsa66 and the 1966 regional dummies in the log(wage) equation?
C4 Use the data in INTDEF.RAW for this exercise. A simple equation relating the three-
month T-bill rate to the inflation rate (constructed from the Consumer Price Index) is
i3t  0  1inft  ut.

Copyright 2012 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 548 PART 3 Advanced Topics

(i) Estimate this equation by OLS, omitting the first time period for later compari-
sons. Report the results in the usual form.
(ii) Some economists feel that the Consumer Price Index mismeasures the true rate of
inflation, so that the OLS from part (i) suffers from measurement error bias. Re-
estimate the equation from part (i), using inft1 as an IV for inft. How does the IV
estimate of 1 compare with the OLS estimate?
(iii) Now, first difference the equation:
i3t  0  1inft  ut.
Estimate this by OLS and compare the estimate of 1 with the previous estimates.
(iv) Can you use inft1 as an IV for inft in the differenced equation in part (iii)? Ex-
plain. (Hint: Are inft and inft1 sufficiently correlated?)
C5 Use the data in CARD.RAW for this exercise.
(i) In Table 15.1, the difference between the IV and OLS estimates of the return to
education is economically important. Obtain the reduced form residuals, vˆ2, from
the reduced form regression educ on nearc4, exper, exper2, black, smsa, south,
smsa66, reg662, ..., reg669—see Table15.1. Use these to test whether educ is
exogenous; that is, determine if the difference between OLS and IV is statistically
significant.
(ii) Estimate the equation by 2SLS, adding nearc2 as an instrument. Does the coeffi-
cient on educ change much?
(iii) Test the single overidentifying restriction from part (ii).
C6 Use the data in MURDER.RAW for this exercise. The variable mrdrte is the murder
rate, that is, the number of murders per 100,000 people. The variable exec is the total
number of prisoners executed for the current and prior two years; unem is the state un-
employment rate.
(i) How many states executed at least one prisoner in 1991, 1992, or 1993? Which
state had the most executions?
(ii) Using the two years 1990 and 1993, do a pooled regression of mrdrte on d93,
exec, and unem. What do you make of the coefficient on exec?
(iii) Using the changes from 1990 to 1993 only (for a total of 51 observations), esti-
mate the equation
mrdrte  0  1exec  2unem  u
by OLS and report the results in the usual form. Now, does capital punishment ap-
pear to have a deterrent effect?
(iv) The change in executions may be at least partly related to changes in the expected
murder rate, so that exec is correlated with u in part (iii). It might be reasonable
to assume that exec1 is uncorrelated with u. (After all, exec1 depends on ex-
ecutions that occurred three or more years ago.) Regress exec on exec1 to see
if they are sufficiently correlated; interpret the coefficient on exec1.
(v) Reestimate the equation from part (iii), using exec1 as an IV for exec. Assume
that unem is exogenous. How do your conclusions change from part (iii)?
C7 Use the data in PHILLIPS.RAW for this exercise.
(i) In Example 11.5, we estimated an expectations augmented Phillips curve of the form
inft  0  1unemt  et,

Copyright 2012 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 CHAPTER 15 Instrumental Variables Estimation and Two Stage Least Squares 549

where inft  inft  inft1. In estimating this equation by OLS, we assumed that
the supply shock, et, was uncorrelated with unemt. If this is false, what can be said
about the OLS estimator of 1?
(ii) Suppose that et is unpredictable given all past information: E(etinft1, unemt1, …) ฀0.
Explain why this makes unemt1 a good IV candidate for unemt.
(iii) Regress unemt on unemt1. Are unemt and unemt1 significantly correlated?
(iv) Estimate the expectations augmented Phillips curve by IV. Report the results in
the usual form and compare them with the OLS estimates from Example 11.5.
C8 Use the data in 401KSUBS.RAW for this exercise. The equation of interest is a linear
probability model:
pira  0  1p401k  2inc  3inc2  4age  5age2  u.
The goal is to test whether there is a tradeoff between participating in a 401(k) plan
and having an individual retirement account (IRA). Therefore, we want to estimate 1.
(i) Estimate the equation by OLS and discuss the estimated effect of p401k.
(ii) For the purposes of estimating the ceteris paribus tradeoff between participation
in two different types of retirement savings plans, what might be a problem with
ordinary least squares?
(iii) The variable e401k is a binary variable equal to one if a worker is eligible to par-
ticipate in a 401(k) plan. Explain what is required for e401k to be a valid IV for
p401k. Do these assumptions seem reasonable?
(iv) Estimate the reduced form for p401k and verify that e401k has significant partial
correlation with p401k. Since the reduced form is also a linear probability model,
use a heteroskedasticity-robust standard error.
(v) Now, estimate the structural equation by IV and compare the estimate of 1 with the
OLS estimate. Again, you should obtain heteroskedasticity-robust standard errors.
(vi) Test the null hypothesis that p401k is in fact exogenous, using a heteroskedasticity-
robust test.
C9 The purpose of this exercise is to compare the estimates and standard errors obtained by
correctly using 2SLS with those obtained using inappropriate procedures. Use the data
file WAGE2.RAW.
(i) Use a 2SLS routine to estimate the equation
log(wage)  0  1educ  2exper  3tenure  4black  u,
where sibs is the IV for educ. Report the results in the usual form.
(ii) Now, manually carry out 2SLS. That is, first regress educi on sibsi, experi, tenurei,
and blacki and obtain the fitted values,  educi, i  1, ..., n. Then, run the second
stage regression log(wagei) on  educi, experi, tenurei, and blacki, i  1, ..., n. Verify
ˆj are identical to those obtained from part (i), but that the standard errors
that the 
are somewhat different. The standard errors obtained from the second stage regres-
sion when manually carrying out 2SLS are generally inappropriate.
(iii) Now, use the following two-step procedure, which generally yields inconsistent
parameter estimates of the j, and not just inconsistent standard errors. In step
one, regress educi on sibsi only and obtain the fitted values, say  educi. (Note that
this is an incorrect first stage regression.) Then, in the second step, run the regres-
sion of log(wagei) on  educi, experi, tenurei, and blacki, i  1, …, n. How does the
estimate from this incorrect, two-step procedure compare with the correct 2SLS
estimate of the return to education?
Copyright 2012 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 550 PART 3 Advanced Topics

C10 Use the data in HTV.RAW for this exercise.

(i) Run a simple OLS regression of log(wage) on educ. Without controlling for other
factors, what is the 95% confidence interval for the return to another year of
education?
(ii) The variable ctuit, in thousands of dollars, is the change in college tuition facing
students from age 17 to age 18. Show that educ and ctuit are essentially uncorre-
lated. What does this say about ctuit as a possible IV for educ in a simple regres-
sion analysis?
(iii) Now, add to the simple regression model in part (i) a quadratic in experience and a
full set of regional dummy variables for current residence and residence at age 18.
Also include the urban indicators for current and age 18 residences. What is the
estimated return to a year of education?
(iv) Again using ctuit as a potential IV for educ, estimate the reduced form for educ.
[Naturally, the reduced form for educ now includes the explanatory variables in
part (iii).] Show that ctuit is now statistically significant in the reduced form for
educ.
(v) Estimate the model from part (iii) by IV, using ctuit as an IV for educ. How does
the confidence interval for the return to education compare with the OLS CI from
part (iii)?
(vi) Do you think the IV procedure from part (v) is convincing?
C11 The data set in VOUCHER.DTA, which is a subset of the data used in Rouse (1998),
can be used to estimate the effect of school choice on academic achievement. Atten-
dance at a choice school was paid for by a voucher, which was determined by a lottery
among those who applied. The data subset was chosen so that any student in the sample
has a valid 1994 math test score (the last year available in Rouse’s sample). Unfortu-
nately, as pointed out by Rouse, many students have missing test scores, possibly due to
attrition (that is, leaving the Milwaukee public school district). These data include stu-
dents who applied to the voucher program and were accepted, students who applied and
were not accepted, and students who did not apply. Therefore, even though the vouchers
were chosen by lottery among those who applied, we do not necessarily have a random
sample from a population where being selected for a voucher has been randomly deter-
mined. (An important consideration is that students who never applied to the program
may be systematically different from those who did—and in ways that we cannot know
based on the data.)
Rouse (1998) uses panel data methods of the kind we discussed in Chapter 14 to
allow student fixed effects; she also uses instrumental variables methods. This problem
asks you to do a cross-sectional analysis where winning the lottery for a voucher acts as
an instrumental variable for attending a choice school. Actually, because we have multi-
ple years of data on each student, we construct two variables. The first, choiceyrs, is the
number of years from 1991 to 1994 that a student attended a choice school; this variable
ranges from zero to four. The variable selectyrs indicates the number of years a student
was selected for a voucher. If the student applied for the program in 1990 and received
a voucher then selectyrs = 4; if he or she applied in 1991 and received a voucher then
selectyrs = 3; and so on. The outcome of interest is mnce, the student’s percentile score
on a math test administered in 1994.
(i) Of the 990 students in the sample, how many were never awarded a voucher?
How many had a voucher available for four years? How many students actually
attended a choice school for four years?
Copyright 2012 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
 CHAPTER 15 Instrumental Variables Estimation and Two Stage Least Squares 551

(ii) Run a simple regression of choiceyrs on selectyrs. Are these variables related in
the direction you expected? How strong is the relationship? Is selectyrs a sensible
IV candidate for choiceyrs?
(iii) Run a simple regression of mnce on choiceyrs. What do you find? Is this what you
expected? What happens if you add the variables black, hispanic, and female?
(iv) Why might choiceyrs be endogenous in an equation such as
mnce  0  1choiceyrs  2black  3hispanic  4female  u1?
(v) Estimate the equation in part (iv) by instrumental variables, using selectyrs as the
IV for choiceyrs. Does using IV produce a positive effect of attending a choice
school? What do you make of the coefficients on the other explanatory vairables?
(vi) To control for the possibility that prior achievement affects participating in the
lottery (as well as predicting attrition), add mnce90—the math score in 1990—to
the equation in part (iv). Estimate the equation by OLS and IV, and compare the
results for 1. For the IV estimate, how much is each year in a choice school worth
on the math percentile score? Is this a practically large effect?
(vii) Why is the analysis from part (vi) not entirely convincing? [Hint: Compared with
part (v), what happens to the number of observations, and why?]
(viii) The variables choiceyrs1, choiceyrs2, and so on are dummy variables indicating
the different number of years a student could have been in a choice school (from
1991 to 1994). The dummy variables selectyrs1, selectyrs2, and so on have a
similar definition, but for being selected from the lottery. Estimate the equation
mnce  0  1 choiceyrs1  2choiceyrs2  3choiceyrs3  4 choiceyrs4
 5 black  6hispanic  7female
by IV, using as instruments the four selectyrs dummy variables. (As before,
the variables black, hispanic, and female act as their own IVs.) Describe your
findings. Do they make sense?

APPENDIX 15A

15A.1 Assumptions for Two Stage Least Squares

This appendix covers the assumptions under which 2SLS has desirable large sample
properties. We first state the assumptions for cross-sectional applications under random
sampling. Then, we discuss what needs to be added for them to apply to time series and
panel data.

15A.2 Assumption 2SLS.1 (Linear in Parameters)

The model in the population can be written as

y  0  1x1  2x2 ฀…  kxk  u,

where 0, 1, …, k are the unknown parameters (constants) of interest, and u is an
unobserved random error or random disturbance term. The instrumental variables are
denoted as zj.

It is worth emphasizing that Assumption 2SLS.1 is virtually identical to MLR.1 (with

the minor exception that 2SLS.1 mentions the notation for the instrumental variables, zj).

Copyright 2012 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has
deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.