36 pages
Le téléchargement nécessite un accès à la bibliothèque YouScribe
Tout savoir sur nos offres


  • cours - matière potentielle : representatives
  • cours - matière potentielle : the game
  • institutional changes
  • rg
  • economic welfare
  • fact of such behaviour of neoclassical schools representatives
  • victory
  • game
  • change
  • model
  • system



Publié par
Nombre de visites sur la page 20
Langue English
Signaler un problème

Thomas S. Dee
Swarthmore College
JULY 2003
I N T R O D U C T I O N CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education? CIRCLE W CIRCLE Working Paper 08: July 2003orking Paper 08: July 2003 Are There Civic Returns to Education?
“…since the achievement of American Independence, the universal and ever-repeated argument in
favor of Free Schools has been, that the general intelligence which they are capable of diffusing, and
which can be imparted by no other human instrumentality, is indispensable to a republican form of
Horace Mann (1846)
However, these correlations could actually be quite INTRODUCTION
misleading since both schooling and civic outcomes
are simultaneously infuenced by a wide variety of Economists typically justify the
inherently unobservable traits specifc to individuals government’s extensive and varied involvement
and the families and communities in which they were in the market for education by appealing to
reared. For example, individuals who grew up in distributional concerns and several types of market
cohesive families and communities that stressed civic failures. The most frequently discussed types of
responsibility may also be more likely to remain in market failure involve the positive externalities that
school. The plausible existence of such unobservables might be associated with schooling. For example,
implies that conventionally estimated correlations some have argued that education generates
may spuriously overstate the true civic returns to external social benefts by reducing the prevalence
education.2of crime and by promoting knowledge spillovers
This study attempts to construct less and technology diffusion in the workplace.1
ambiguous empirical evidence on this policy-relevant However, the externality that is arguably featured
issue by identifying the causal effects of additional most prominently in discussions about education
schooling on civic behaviors and knowledge. The involves civic behaviors and attitudes. Specifcally,
research designs adopted here essentially parallel the it is widely believed that education is an essential
extensive, empirical literature on the labor-market component of a stable democratic society because
returns to schooling (e.g., Angrist and Krueger 1999, it encourages citizens to participate in democratic
Card 1999). More specifcally, these inferences rely processes and prepares them to do so in an
critically on instrumental variables that generated informed and intelligent manner. The putative
possibly exogenous variation in individual levels of existence of such civic returns to education
schooling but that should otherwise be unrelated motivated the proliferation of common schools
to adult civic outcomes.3 First, using data from thin the early 19 century and early educational
the High School and Beyond (HS&B) longitudinal reformers like Horace Mann and continues to
study, I estimate the effects of college entrance on provide one of the most important justifcations
adult voter and volunteer participation by relying for the many public policies and institutions that
on the geographic proximity and density of junior promote access to all levels of education.
and community colleges as a teen. Then, using data An extensive, empirical literature in political
from the 1972-2000 General Social Surveys (GSS), science has documented a strong correlation
I estimate the effects of years of schooling on adult between educational attainment and various
voter participation, on group memberships and on civic behaviors. In particular, this literature has
attitudes towards free speech by relying on changes demonstrated that higher levels of schooling are
in teen exposure to child labor laws (Acemoglu and associated with substantive increases in voter
Angrist 2000). Using the GSS data, I also estimate turnout. Political scientists generally interpret
the effects of additional schooling on the frequency this literature as providing strong support for
of newspaper readership, an outcome that is closely the view that education is effective at promoting
related to measures of civic awareness. The results of the quantity and quality of civic participation.
these evaluations suggest that additional schooling,
www.civicyouth.org www.civicyouth.org
1 CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education? CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education?
both at the secondary and post-secondary levels, Interestingly, an economic perspective
had large and statistically signifcant effects on could also suggest alternative mechanisms by
voter participation. I also fnd that the additional which additional schooling might actually reduce
secondary schooling signifcantly increased the civic engagement. For example, by raising the
frequency of newspaper readership as well as the opportunity cost of an individual’s time, increased
amount of support for allowing most forms of possibly schooling could reduce the amount of time
controversial free speech. and attention allocated to civic activity. This
could be particularly relevant for volunteering,
which, unlike voting, can involve a substantial EDUCATION AND CIVIC ENGAGEMENT
commitment of time. However, education could

also reduce voter participation by promoting an
One of the fundamental mechanisms by which
awareness of voting as an essentially expressive
education has long been thought to generate civic
act with an infnitesimally small probability of
externalities involves improvements in the quality
infuencing actual policy.6 Nonetheless, the
of civic participation and awareness. Specifcally,
available empirical evidence seems to provide an
it is widely alleged that increases in education
emphatic confrmation of the conventional view
generate broad social benefts by allowing citizens
that education does promote civic engagement.
to make more informed evaluations of the complex,
Numerous studies over the last ffty years have
social, political and technological issues that might
demonstrated that higher levels of individual
be embedded in campaign literature, legislative
schooling are strongly associated with civic
initiatives and ballot referenda. However, the
behaviors and knowledge.7 For example, in a
contemporary literature among political scientists
widely repeated interpretation of this empirical
has also put a particular stress on the positive
evidence, Converse (1972) refers to educational
effects that schooling may have on the likelihood of
attainment as the “universal solvent” of political
civic participation, in particular, voter turnout (e.g.,
participation. Similarly, Putnam (2001) notes
Wolfnger and Rosenstone 1980). Education could
that “education is by far the strongest correlate
promote civic participation through at least two broad
that I have discovered of civic engagement in all
channels. First, schooling may reduce the effective
its forms” (emphasis mine). Also, in their earlier
costs of certain forms of civic participation. In
study of voting participation, Wolfnger and
particular, this is thought to occur because increased
Rosenstone (1980) suggest that their core fnding
cognitive ability makes it easier to process complex
is the “transcendent importance of education.”
political information, to make decisions and to
However, they also note that an individual’s level of
circumvent the various bureaucratic and technological
schooling could easily proxy for unobserved traits
impediments to civic participation.4 Second,
that also infuence civic behaviors (pages 19-20).
education may increase the perceived benefts
For example, they suggest that the types of family
of civic engagement by promoting “democratic
backgrounds that promote increased schooling
enlightenment” or, stated differently, by shaping
may also promote increased socialization into civic
individual preferences for civic activity. Similarly, it
activities like voting. Wolfnger and Rosenstone
is often alleged that education plays an important
(1980), like other researchers in this feld, have
public role by directly inculcating students with
attempted to control for the possible bias in the
other fundamental democratic and pluralistic values
estimated effect of education by introducing a
(e.g., support for free speech, for the separation of
few additional control variables (e.g., income and
church and state, etc.).5 However, it is also possible
occupational measures) into multiple regression
that additional schooling shapes civic preferences
models. The apparent robustness of the
indirectly through altering the composition of peer
correlations between education and civic outcomes
groups and shared social norms.
has led most researchers to conclude that
www.civicyouth.org www.civicyouth.org
2 CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education? CIRCLE W CIRCLE Working Paper 08: July 2003orking Paper 08: July 2003 Are There Civic Returns to Education?
education does have a causal effect. For example, Gibson (2001) presents within-twin estimates,
in the most recent contribution to this literature, Nie which suggest that education actually reduces the
and Hillygus (2001) note that this orthodox view is probability of volunteering. In the next two sections,
“largely uncontested.” I present new empirical evidence on the effects of
However, the basic approach of introducing educational attainment on several civic outcomes. I
a few additional controls may not convincingly attempt to identify the causal effects of educational
resolve the question of whether the strong attainment by relying on instrumental variables that
correlations between education and civic generate plausibly exogenous changes in the levels of
outcomes actually refect the true causal effects. individual schooling.8
In particular, this could occur because so many
of the shared determinants of civic behavior and COLLEGE ATTENDANCE AND CIVIC PARTICIPATION
educational attainment are inherently diffcult High School and Beyond (HS&B)
for researchers to measure well. For example, The data for this section are drawn from High
as noted earlier, children who were raised in School and Beyond (HS&B), a major longitudinal
families or communities that stressed civic study conducted by the U.S. Department of
responsibility are almost certainly more likely to Education. This detailed study began with a cohort of
remain in school longer. This may occur in part high school sophomores in 1980. Follow-up interviews
because such families and communities are also of roughly 12,000 members of the sophomore
likely to impart values that encourage schooling. cohort occurred in 1984 when most respondents
However, it could also occur simply because civic- were 20 years old and again in 1992 when most
minded families and communities may do more respondents were 28 years old.9 In the 1992
to insure that their children attend well-funded, interview, respondents were asked four civic-related
high-quality schools. These plausible scenarios questions: whether they were currently registered
imply that the strong association between adult to vote (mean=.67), whether they had voted in a
civic outcomes and educational attainment may local, state or national election within the past year
refect, to an unknown degree, the confounding (mean=.36), whether they had voted in the 1988
infuence of unobserved family and community Presidential election (mean=.55) and whether they
traits. Alternatively, these correlations could had volunteered in the last month (mean=.37). The
also refect the confounding infuence of other, key measure of educational attainment examined
inherently unobservable individual traits like the here is college entrance defned as of the 1984
rate at which future outcomes are valued and the interview (mean=.54). This defnition of college
taste for altruism. Certainly, the recent trends in entrance is based on attendance at a junior college,
the United States (i.e., increases in educational a community college or a four-year college or
attainment not matched by increases in voter university and explicitly excludes those who only
turnout or political knowledge; Galston 2001) attended a vocational, trade, business or other
suggest that the association between education training school. While this is a somewhat narrow
and civic engagement could be specious. And at margin of educational attainment, the available
least two studies in the political science literature evidence indicates that it is also an increasingly
provide more formal evidence that such concerns important one. The rate of college enrollment among
about omitted variable biases may be empirically young adults has increased dramatically over the
relevant. Both Luskin (1990) and Cassel and last twenty years with roughly half of this increase
Lo (1997) present evidence that the apparent being absorbed by junior and community colleges
infuence of education on civic outcomes (political (Kane and Rouse 1999). And prior studies suggest
literacy and sophistication) may refect the spurious that modest persistence at two and four-year
infuence of other individual traits (e.g., intelligence colleges has benefcial labor-market consequences
and parents’ socio-economic status). Similarly,
www.civicyouth.org www.civicyouth.org
3 CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education? CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education?
even when it does not result in a degree (e.g., Kane proxy for the civic attitudes of the community in
and Rouse 1995). The HS&B respondents who had which the respondents grew up: the county-level
entered college by 1984 did generally remain in voter turnout in the 1980 Presidential election. The
college long enough to accumulate a relatively large second county-level variable is a measure of adult
amount of undergraduate credits.10 Furthermore, educational attainment in the respondent’s teen
the baseline evidence discussed below demonstrates community: the percent of adults aged 25 or older
that this measure of college entrance has a strong with high school degree. The third county-level
partial correlation with the probability of subsequent control, the population share aged 18-24, may
civic engagement. However, the choice of college be a relevant determinant of civic engagement
entrance as a measure of educational attainment and also infuence the competitiveness of post-
is also dictated by the availability of a plausible secondary institutions. The two state-level
instrument, the geographic availability of junior and variables refect infuential voter regulations
community colleges as a teen, which appears to defned as of 1992 (Knack 1995). One is a binary
have substantively infuenced the decision to attend indicator for whether the state had an active
college and to have been otherwise unrelated to civic policy of allowing voter registration by mail. The
engagement as an adult.11 second is the number of years the state had active
“motor-voter” regulations in place.13 The available
Baseline estimates evidence suggests that a years-based measure is
The validity of the geographic availability of the appropriate variable for identifying the early
junior and community colleges (hereafter referred effects of “motor-voter” policies because state
to as two-year colleges) as a basis for identifcation drivers licenses are renewed in cycles as long as
is a critical issue, which is discussed in some detail six years (Knack 1995).
below. However, before turning to an assessment These models are somewhat unusual
of the relevant instrumental variables, it is useful in comparison to the prior literature since they
to establish an empirical baseline by estimating condition on detailed individual and community-
the effects of college entrance on subsequent civic level socioeconomic variables defned as of each
behaviors in specifcations that assume the absence respondent’s teen years. Furthermore, HS&B’s
of omitted variable biases. Table 1 presents the clustered sampling design also makes it possible
estimated marginal effects from single-equation to control for the possibly confounding infuence
probits in which the four measures of civic behavior of unobserved community traits through the
are the dependent variables. The frst specifcation introduction of school fxed effects. The key results
(column (1)) conditions on 10 variables representing from these evaluations, which are presented in
basic demographic information on age, race, ethnicity, Table 1, uniformly suggest that college entrance
gender and religious affliation, 18 other variables had positive and statistically signifcant effects on
that refect family income, family composition and civic participation. Interestingly, the magnitudes
parental education as defned during the 1980 of these estimated marginal effects are also
interview and a single variable refecting each quite robust to the introduction of the additional
respondent’s 1980 composite score on reading, controls, including fxed effects for the 961 base-
mathematics and vocabulary tests.12 The subsequent year schools. These estimated effects are also
models introduce school-level controls (i.e., miles quite large, implying that a relatively modest
to the nearest 4-year college and urbanicity fxed increase in educational attainment has a sizable
effects), state and county-level controls based on the infuence on subsequent civic participation. For
location of the base-year school, fxed effects for the example, these estimates imply that college
Census division of the base-year school and, fnally, entrance increased voter registration by
fxed effects for each of the 961 base-year schools. approximately 12 percentage points, an increase
One of the county-level variables is a well-measured of nearly 18 percent in the mean probability
www.civicyouth.org www.civicyouth.org
4 CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education? CIRCLE W CIRCLE Working Paper 08: July 2003orking Paper 08: July 2003 Are There Civic Returns to Education?
of being registered. Similarly, these estimates results in Table 1 involves the quality of the self-
imply that college entrance increases the mean reported data on civic engagement. It is well-known
probability of voting in the last year, voting in the that survey respondents often overstate their
1988 Presidential election and volunteering by 26 participation. Furthermore, studies that compare self-
percent, 28 percent and 15 percent, respectively. reported voting with validated measures often fnd
However, the central concern with the that more highly educated people are particularly
results in Table 1 is that the strong partial likely to overstate their voter participation (e.g.,
correlations between college entrance and civic Silver, Anderson and Abramson 1986). The basic
behaviors may refect the confounding infuence explanation for this phenomenon is that additional
of unobserved determinants of both schooling education may change peer norms and create a sense
and civic engagement. One straightforward way of obligation that leads more educated respondents
to assess the possible empirical relevance of this to overstate their actual civic engagement more
concern is to examine the partial correlations frequently than those with lower educational
between college entrance and measures of civic attainment. This possibility implies that the apparent
attitudes and knowledge that preceded attendance effects of post-secondary schooling on adult voter
in college. I rely on two such measures based participation identifed here (e.g., Table 1) could
on data from the sophomore-year survey. One refect, to an unknown degree, education-specifc
is a standardized test score on questions related patterns of over-reporting.
to civics. The other is the student’s response to This issue cannot be addressed defnitively in
a question about the importance of correcting this context since HS&B did not validate self-reported
social and economic inequality (1=not important, voting. However, the available evidence suggests that
2=somewhat important and 3=very important). this is not particularly problematic. First of all, the
Each of these variables is highly predictive of each HS&B respondents had comparatively little incentive
measure of future civic engagement. For example, to over-report since the survey instrument focused
a 10 percent increase in the sophomore-year almost exclusively on labor-market and educational
civics test score is associated with a statistically experiences, not political values and participation.
signifcant 7 percent increase in the mean The November voter supplements to the Current
probability of voting within the last year. Similarly, Population Surveys shared this feature and the
a one-unit increase in the ordered attitudinal aggregate voter-participation rates implied by those
measure is associated with a statistically signifcant self reports are relatively close to the actual rates
10 percent increase in the mean probability of (Teixeira 1992, Appendix A). Furthermore, the voter-
voting. In auxiliary regressions where these registration rate implied by the HS&B responses
sophomore-year measures are the dependent (67 percent) is similar to the contemporaneous
variables, the estimated effects of college entrance CPS-reported rate for 25-34 year olds (61 percent,
are positive and statistically signifcant. However, U.S. Census Bureau 1996). And the percent of
since the dependent variables in these models HS&B respondents who reported voting in the past
preceded college entrance, these results cannot year (36 percent) is actually lower than the CPS-
plausibly refect causal effects. Instead, these reported turnout rate for 25-34 year olds in the
results suggest the existence of individual-level 1992 Presidential election (53 percent).14 However,
unobservables that may have a positive covariance further comparisons with the CPS data suggest that
with both educational attainment and adult civic the HS&B respondents’ 1992 recall of having voted in
engagement. This stylized evidence underscores 1988 may be more biased. In the 1988 CPS survey,
the need to rely on instrumental variables in approximately 38 percent of 21-24 year olds reported
estimating the effects of college attendance on civic voting in the Presidential election while the 1992
outcomes. HS&B survey suggests that 55 percent of respondents
Another fundamental concern with the did. So, a caveat about this particular variable is
www.civicyouth.org www.civicyouth.org
5 CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education? CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education?
outcomes. Rouse (1995) also presents evidence appropriate.
that the availability of two-year colleges increases A second indication that there are not
educational attainment for those on the margin a potentially confounding reporting biases in
of attending college (a “democratization” effect) models based on the HS&B data is that estimates
but actually reduces it among those who would based on actual voter turnout suggest that
have otherwise attended a four-year college (a educational attainment has similarly sized effects.
“diversion” effect). I also fnd some support for Specifcally, county-level regressions based on
a modest “diversion” effect (i.e., the proximity 1980 data from the 516 counties represented in
of two-year colleges reducing the probability of HS&B suggest that completing a year or more
completing a bachelor’s degree) but rely on the of college increased voter turnout by at least 12
stronger “democratization” effect as a source of percentage points. Third, it should be noted that,
identifying information. even if schooling did increase over-reporting,
Specifcally, I rely on two measures of the that would necessarily imply that schooling has
local availability of two-year colleges. One is the a type of structural effect (i.e., instilling a sense
distance in miles from each respondent’s high of civic obligation) that should also generate true
school to the nearest two-year college (as reported increases in civic engagement. In other words,
by a high-school offcial as part of the HS&B school though these evaluations would not identify the
survey). The second is a count of the number true effect of schooling on civic participation, the
of two-year colleges within each respondent’s very existence of such reporting biases would
county in 1983 (mean=2.4).16 These measures suggest that schooling had some of its intended
of the availability of two-year colleges are clearly civic consequences.
related but they also appear to have had plausibly
distinct effects on educational attainment.17 For Measures of college availability as
example, inferences based on these data suggest instruments
that the proximity of base-year high schools to The partial correlations reported in Table
a 2-year college increased college attendance in 1 are consistent with the prior empirical studies
the late teens and early twenties, had no effect of civic participation. However, a more convincing
on later spells of college attendance and, overall, strategy for assessing whether the estimates in
may have diverted students away from eventually Table 1 refect the causal effects of attending
completing a bachelor’s degree. In contrast, the college is to exploit instrumental variables
number of two-year colleges within a county that generate plausibly exogenous variation in
appears to have generated more sustained spells this measure of educational attainment. The
of college attendance throughout young adulthood fundamental requirements of such instrument
and to have increased the probability of ultimately are that they actually infuence educational
completing a bachelor’s degree.attainment and that they are uncorrelated
Since the identifcation strategy with the unobserved determinants of civic
implemented here exploits the cross-sectional engagement. A recent study of the labor-market
variation in the availability of two-year colleges, returns to schooling by Card (1995) suggests
the key sources of this variation should be noted. that the geographic availability of colleges may
While every state has two-year colleges, their provide valid instruments for schooling.15 The
geographic distribution across the United States basic motivation for such instruments is that the
is somewhat uneven. For example, several states proximity of colleges as a teen should substantially
in the West and Southwest (e.g., California, reduce the costs of attending college (particularly
Washington, Texas and Arizona) and in the upper for students from disadvantaged backgrounds)
Midwest (e.g., Illinois, Michigan) have relatively but should otherwise have no effects on adult
extensive systems of public community colleges.
www.civicyouth.org www.civicyouth.org
6 CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education? CIRCLE W CIRCLE Working Paper 08: July 2003orking Paper 08: July 2003 Are There Civic Returns to Education?
Medsker and Tillery (1971) note that this instruments and sophomore-year measures of civic
distribution refects the dramatic growth in new attitudes and knowledge that are strongly correlated
two-year colleges that occurred in the middle with future civic participation (i.e., scores on a civics
of the last century (i.e., in decades prior to the test and attitudes towards correcting inequality).
HS&B study). They also note that growth of And, third, I assess how the effects of these
two-year colleges was shaped by the interaction instruments vary across students from advantaged
of state-specifc enabling legislation and several and disadvantaged backgrounds. To the extent that
sources of enrollment pressure (e.g., the G.I. the estimated effects of these instruments truly
Bill, the baby boom and population migration). refect variation in the costs of attending college,
However, it should be noted that the states of these effects should be concentrated among
New York, Pennsylvania, Ohio and Florida also students from disadvantaged backgrounds (Card
have a large number of two-year colleges, with 1995, Kling 2001). The results of all of these ad-
a particularly large share of them being older, hoc specifcation checks suggest that the proximity
private junior colleges. to 4-year colleges may be an invalid instrument. In
I also considered, but rejected, the particular, nearness to 4-year colleges is associated
idea of using proximity to four-year colleges as with sharp increases in the probability of graduating
an instrument. Specifcally, a central concern from high school as well as signifcant increases in
with any instrument based on the geographic sophomore-year civics knowledge. These results
availability of colleges is that it might be fawed do not constitute a defnitive case against this
because it is associated with the unobserved particular measure as an instrument for educational
determinants of both educational attainment attainment. Nonetheless, all of the models for
and civic behavior. In particular, the unobserved educational attainment and civic outcomes reported
traits of communities near colleges (e.g., high here condition on this measure.20
socioeconomic status) could simultaneously In Table 2, I present the estimated marginal
encourage both higher educational attainment effects of the availability of two-year colleges on the
and increased civic participation. Furthermore, probability of entering college. The results in the
the availability of colleges may promote a top panel suggest that both measures of availability
youth-oriented and politically aware culture have plausibly signed and statistically signifcant
that promotes the civic engagement of teens effects on college entrance and that these estimates
independently of its effects on educational are relatively robust to the introduction of additional
attainment.18 I assess the empirical relevance controls. Specifcally, the results from Model (4)
of these concerns in a number of ways. For suggest that a location 100 miles further away
example, I discuss the robustness of the from a two-year college reduces the probability of
key results to the introduction of the school, college entrance by 7.3 percentage points. Similarly,
county and state-level controls. However, I also these results suggest that an additional two-year
provide three other types of ad-hoc empirical institution within county is associated with a 0.6
evidence on the validity of these instruments. percentage point increase in the probability of
First, I examine their effects on different levels entering college. Since recent studies (Bound et
of educational attainment and base-year test al. 1995, Stock and Staiger 1997) have illustrated
scores. If the estimated effects of college the biases that might be generated by relying on
availability truly refect the costs of attending relatively “weak” instruments, I also tested the joint
college and not the infuence of omitted signifcance of these two variables. The extremely
variables, these instruments are likely to have low p-values associated with these tests suggest
little or no effects on these other measures that those concerns are not relevant in in this
of educational achievement.19 Second, I application.
examine the partial correlations between the
www.civicyouth.org www.civicyouth.org
7 CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education? CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education?
backgrounds. However, the results in the bottom In the bottom panel of Table 2, I provide some
panel of Table 3 also suggest that, for students with ad-hoc evidence on the validity of these instruments
poorly educated parents, the number of two-year by estimating their unique effects on students from
colleges within county had strong “democratization” advantaged and disadvantaged backgrounds. Card
effects that increased the probability of entering (1995) suggests that, if the interpretation of college
college as well as the probability of obtaining a availability as an independent measure of the costs
bachelor’s degree.22 However, the more general of attending college is a valid one, the effects of
and important result from Table 3 is that the these instruments should be concentrated among
effects associated with the availability of two-year students from disadvantaged backgrounds. Following
colleges are highly concentrated on the margin of Card (1995), I assess the existence of such response
attending college. This evidence is consistent with heterogeneity by interacting the availability measures
the maintained assumption that these measures with indicators for high and low parental education.21
refect plausibly exogenous variation in the costs of The results in the bottom panel of Table 2 indicate
college entrance and not other unobserved traits that the effects of the availability of two-year colleges
of these communities. In particular, because these are highly concentrated among students with
instruments have such a narrowly focused effect poorly educated parents. Furthermore, the p-values
on this single margin of educational attainment, it reported in the bottom panel of Table 2 indicate
suggests that they do not proxy for the unobserved that the interactions of low parental education with
determinants of future civic engagement.the two instruments are, jointly, highly signifcant
However, in Table 4, I present further determinants of college entrance. In contrast, the
evidence on the validity of these exclusion estimated effect of the proximity of two-year colleges
restrictions. As noted earlier, the base-year survey is statistically insignifcant for students with highly
of HS&B sophomores contained two variables educated parents (though it has the same sign).
that appear to refect each student’s latent civic Similarly, the number of two-year colleges within a
engagement and knowledge well: a standardized county has a smaller effect for students with highly
test score on questions related to civics and an educated parents.
attitudinal question about the importance of In Table 3, I present evidence how the
correcting social and economic inequality (1=not availability of two-year colleges infuenced different
important, 2=somewhat important and 3=very measures of educational achievement. These
important). These latent indicators, which are estimates are based on specifcations that include
highly predictive of future civic engagement, all the individual, family, school, county and state-
provide a potentially plausible basis for evaluating level controls and division fxed effects. The OLS
the validity of the instruments. Specifcally, if estimates from models of base-year test scores
the measures of college availability have an indicate that the availability of two-year colleges
association with the unobserved determinants of has small and statistically insignifcant effects.
future civic engagement, we would expect them Similarly, the results from probit models suggest that
to be correlated with these observed measures the availability of two-year colleges has small and
as well. In Table 4, I present the key results from statistically insignifcant effects on the probability of
auxiliary regressions in which these sophomore-graduating from high school and on the probability
year traits are the dependent variables. These of obtaining an associate’s degree. These results also
results are based on models that include all of suggest that the geographic proximity of two-year
the prior controls (e.g., Model (4) in Tables 1 and colleges led to relatively small and weakly signifcant
2). The estimates in Table 4 uniformly suggest reductions in the probability of obtaining a bachelor’s
that availability of two-year colleges, both degree: a “diversion” effect that appears to be
generally and for students with poorly educated concentrated among students from disadvantaged
www.civicyouth.org www.civicyouth.org
8 CIRCLE Working Paper 08: July 2003 Are There Civic Returns to Education? CIRCLE W CIRCLE Working Paper 08: July 2003orking Paper 08: July 2003 Are There Civic Returns to Education?
probability of volunteering but uniformly large and parents, has a small and statistically insignifcant
positive effects on each of the three measures of association with sophomore-year civics knowledge
voter participation. Specifcally, these estimates and with community attitudes. Some studies also
indicate that college entrance increases voter assess instrument validity and possible biases by
participation by roughly 17 to 22 percentage considering the sign of the relationship between
points. These results are clearly consistent with the candidate instruments and observed determinants
conventional claims that educational attainment of the outcomes under study (e.g., Altonji et al.
is a critical determinant of civic engagement. In 2002). The mixed signs of the estimates reported
fact, with respect to voter registration and having in Table 4 do not provide consistent evidence for
voted in the last 12 months, these estimated effects particular violations of exclusion restrictions. In a
are noticeably larger than those based on partial similar vein, I also examined the partial correlations
correlations (Table 1). The sampling variation between these instruments and the 1980 county-
associated with these estimates suggests that level voter turnout. Interestingly, the results
these differences should not be overemphasized. indicated that communities with better access to
Nonetheless, it is also worth noting at least two-year colleges had lower voter turnout rates.
three reasons that the true effects of educational These negative relationships suggest that, if there
attainment might exceed the estimates based on are violations of the exclusion restrictions, they
partial correlations (Table 1). First, as frequently may impart a negative bias, which would not be
noted in the literature on wages and schooling, fundamentally confounding for most of the results
this could refect an attenuation bias driven by presented below.
measurement error in reported schooling. Second,
these estimates could indicate that the civic returns RESULTS
associated with college entrance are particularly
large for the non-random subset of individuals The results in Tables 2, 3 and 4 are
whose post-secondary attainments were infuenced consistent with the maintained assumption that the
by the instruments (e.g., those from disadvantaged geographic availability of two-year colleges provides
backgrounds, Imbens and Angrist 1994).24 And, a potentially valid source of identifcation. The
third, a downward bias in conventional estimates availability of two-year colleges is associated with a
could also refect the infuence of unobserved ability signifcant increase in college attendance but smaller
on both schooling decisions and time allocated to and statistically insignifcant changes in base-year
civic endeavors. test scores and in other measures of educational
However, a fourth possibility with very attainment. These increases are plausibly
different implications is that the size of these concentrated among students with poorly educated
estimates refects undiagnosed violations of the parents. And these measures are unrelated to
maintained exclusion restrictions. One indication sophomore-year indicators of civic attitudes (e.g.,
that this is not so is that the results from Table 5 civics knowledge and attitudes towards inequality).
are quite similar across models, which incrementally In Table 5, I present the key results from bivariate
introduce the school, county and state-level probits in which the adult civic behaviors are the
controls. However, another way to assess this dependent variable of interest and college entrance
concern is to use as instruments the interaction of is an endogenous regressor (Wooldridge 2002).23
low parental education and the measures of two-The excluded instruments in Model (1) are miles
year college availability. Specifcally, in such models, to the nearest two-year college and the number
the interaction of high parental education and the of two-year colleges within county. These results
measures of two-year college availability can then of these models suggest that college entrance has
be included as controls in the outcome equations small and imprecisely estimated effects on the
(e.g., Card 1995). This approach to identifcation
www.civicyouth.org www.civicyouth.org