Society For The Scientific Study of Religion, Wiley Journal For The Scientific Study of Religion

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Does Church Attendance Really Increase Schooling?

Author(s): Linda D. Loury


Source: Journal for the Scientific Study of Religion, Vol. 43, No. 1 (Mar., 2004), pp. 119-
127
Published by: Wiley on behalf of Society for the Scientific Study of Religion
Stable URL: https://www.jstor.org/stable/1387777
Accessed: 09-01-2019 06:55 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Society for the Scientific Study of Religion, Wiley are collaborating with JSTOR to
digitize, preserve and extend access to Journal for the Scientific Study of Religion

This content downloaded from 103.229.203.182 on Wed, 09 Jan 2019 06:55:55 UTC
All use subject to https://about.jstor.org/terms
Does Church Attendance Really
Increase Schooling?

LINDA D. LOURY

This article shows that religiosity during adolescence has a significant effect on total number of years of
attained. It differs from previous research by focusing on church attendance rather than on denominatio
controlling more completely for the effects of omitted-variables bias. Any estimated correlation between church
attendance and schooling without such controls may reflect unmeasured family, community, and individual char-
acteristics. The size of the effect for individuals who attended church 52 weeks per year compared to individuals
who do not attend at all is equivalent to over three years of parents' schooling. This finding implies that changes
in church attendance, either due to exogenous changes in attitudes or as an indirect effect of government or other
institutional activity, may have large spill-over effects on socioeconomic variables.

INTRODUCTION

The effects of religiosity on health and/or subjective well-being are well documented, but
relatively little work identifies its influence on education and other determinants of socioeconomic
status. Furthermore, much of this work has focused exclusively on the effects of denominational
differences on education without also examining effects of church attendance. The small number
of studies on how church attendance affects schooling may not provide an accurate view of
the relationship. These studies are generally restricted to limited populations and the estimated
effects may reflect unmeasured individual characteristics that increase both education and church
participation. The purpose of this article is to supplement previous work by providing consistent
estimates of the effects of church attendance during adolescence on total number of years of
schooling attained.

LITERATURE REVIEW AND EMPIRICAL MODEL

A variety of economic and sociological theories may explain any observed effects of church
attendance on schooling (Berger 1969; Jencks and Mayer 1990; Petersen and Roy 1985; Akerlof
1997; Cochran and Beeghley 1991). Reference group theories posit that, independent of fam-
ily characteristics, church peers may sway individuals toward more positive social behavior by
changing norms of desirable behavior.' Collective socialization theories imply that adults in the
church act as role models and sources of social control outside the family. Buffer theories focus
on stress that can be handled through the direct social support provided by other church members
and the set of meanings and structure inherent in religious beliefs.
The model that underlies the empirical work to test these effects can be represented as:

Y = PIXI +X2P2 +X3P3 +? IZ1I +2Z2 +8 (1)


where Y is schooling, XI is church attendance and/or religious denom
background characteristics such as mother's education, and X3 are other pr
characteristics. Z1 and Z2 are unobservable background variables that
schooling and church attendance. The Z1 are common among siblings a
siblings.

Linda D. Loury is Associate Professor Department of Economics, Tufts University, Medford, MA 02155. E-mail:
lindal oury@ tufts. edu

Journal for the Scientific Study of Religion 43:1 (2004) 119-127

This content downloaded from 103.229.203.182 on Wed, 09 Jan 2019 06:55:55 UTC
All use subject to https://about.jstor.org/terms
120 JOURNAL FOR THE SCIENTIFIC STUDY OF RELIGION

Two main conclusions can drawn be from existing empirical literature. First, there are large
and significant differences in average schooling across denominations that are moderated, to
a greater or lesser extent, by controlling for differences in family background (most recently,
Sander 1992; Lehrer 1999; Homola, Knudsen, and Marshall 1987; Stryker 1981). Second, church
attendance has an ambiguous effect on schooling.
Two recent studies that estimate the effects of church attendance are Ribar (1994) and Evans,
Oates, and Schwab (1992). Each used samples of women ages 14 to 21 in 1979 drawn from the
National Longitudinal Study of Youth (NLSY). Ribar found that both those who reported that
they attended religious services infrequently and those who attended often were more likely to
complete high school than those who were not at either extreme. In contrast, Evans, Oates, and
Schwab found that only those who attended religious services at least four times a month were
less likely to drop out of high school. In a third study, Freeman (1986) reported that church-going
inner-city African-American men ages 14-24 were more likely to attend school.
The problem with estimates from these studies is that church attendance may be correlated
with unobserved variables that actually account for the measured effects. For example, parents
who encourage or require their children to attend church may systematically differ from other
parents in how much they supervise their children's activities, in the amount of time they spend
with their children, or in a wide variety of other nonreligious attitudes and behaviors that are
the true determinants of children's schooling. In addition, children who typically attend church
may also differ from other children in their aspirations, aversion to criminality and destructive
behavior, and in other ways that are not affected by their religious commitment.
Often, proxies are used to capture the effects of such unobservables. Ribar (1994) includes
whether household members subscribed to a newspaper or magazine or held a library card as
measures of family encouragement to reading and schooling. Teachman (1987) and Krein and
Beller (1988) found that educational resources as measured by whether there was a specific
place to study in the home and whether there were reference books, a daily newspaper, or a
dictionary/encyclopedia in the home had a large and significant effect on schooling for females.
Adding such proxies will lessen the omitted-variables problem, but it may not completely
eliminate the bias. Large data sets have only a small fraction of the parental, individual, peer,
and neighborhood characteristics that could potentially alter schooling. As a result, most analyses
including the available subsets of proxies for unobserved explanatory factors will be subject to
criticisms of possible omitted-variables bias.
This article handles omitted-variables bias problems in three ways. First, in order to reduce
the effect from unobservables that are common among siblings (Z1), it includes the number of
older siblings by gender who attended college and the number of older siblings by gender who
were high school dropouts. These are good proxies for common family effects since the same Z1
unobservables determine schooling for all siblings in a given family.
Second, in order to reduce the effect from unobservables that differ between siblings (Z2), the
X2 used in this analysis include Armed Forces Qualifying Test (AFQT) scores and the individual's
own educational aspirations. Previous research suggests that the effects of many unobserved
characteristics that increase children's schooling are small holding these particular X2 constant.
Astone and McLanahan (1991) showed that variables that are typically unobserved-parental
practices as measured by education aspirations, mothers and fathers monitoring schoolwork, and
general supervision-had large and significant effects on dropping out of high school. However,
these effects generally dropped by 50 percent or more and often became insignificant when
children's educational aspirations, grades, school attendance, and attitude toward school were
added to the analysis. Similarly, Currie and Thomas (1999) found that both mother's AFQT
scores and mother's family background had large positive effects on her children's Peabody
Picture Vocabulary Test (PPVT) scores. However, when they compared PPVT scores of children
whose mothers are sisters (and thus shared the same family background while growing up), they
find a negative and insignificant relationship between mother's AFQT and the child's PPVT. One

This content downloaded from 103.229.203.182 on Wed, 09 Jan 2019 06:55:55 UTC
All use subject to https://about.jstor.org/terms
CHURCH ATTENDANCE AND SCHOOLING 121

interpretation of this result is that AFQT reflects family background differences in models that
use imperfect background proxies. Heckman (1995) and Neal and Johnson (1996) make similar
arguments.
Third, the article estimates the effects of church attendance on schooling using two-stage least
squares (2SLS). These estimates capture the effects of exogenous variation in church attendance
and not the influence of unobserved variables that are correlated with individual's schooling.

EMPIRICAL RESULTS

Data for this article come from the National Longitudinal Study of Youth (NLSY). The
NLSY panel is a nationally representative sample of 12,686 individuals ages 14-21 in 1979 who
were interviewed annually to determine a wide variety of information about schooling, work, and
other experiences. The sample for this study is limited to individuals ages 14-172 so that church
attendance and the other background variables are measured during adolescence prior to school
completion. Years of schooling for the individual and his or her siblings were measured as of 1993
when sample members were 28-35 years old. The individual data from the NLSY was merged
with county data on the total fraction of religious adherents (total adherents/total population)
and the fraction of adherents who belonged to the individual's denomination (denomination
adherents/total adherents) from the 1980 Survey of Churches and Church Membership collected
by the Glenmary Research Center (see Quinn et al. 1982 for details about this survey).
The sample includes Catholics, Protestants, and those who do not report a denomination.
It excludes members of non-Catholic or non-Protestant denominations. As indicated later, the
county variables used as instruments in the schooling equation are less likely to be valid for
smaller religious groups.3 The sample used here also excludes blacks because this data was not
available for most black churches. The means, standard deviations, and definitions of the variables
used in the analysis are listed in Table 1.
Column 1 of Table 2 lists the results of replicating past analyses. Looking at nonreligious
variables shows that while more siblings reduced schooling, parent's schooling, parent's white-
collar employment, and private school matriculation had large and significant (at least at the 10
percent level) positive effects on offspring schooling. Living in a family where someone had a
library card or received magazines had large and significant effects on schooling. This finding
does not mean that providing children with library cards and magazines will raise their schooling
by 0.3 and 0.5 years, respectively; instead, these coefficients should be interpreted (similarly to
Teachman 1987; Ribar 1994; Krein and Beller 1988) as capturing part of the effects of unobserved
family and individual characteristics that actually raise schooling.
The religious variables show that Episcopalians, Presbyterians, Methodists, Lutherans, and
Catholics average more years of schooling (in that order) than the left-out category (Baptists,
other Protestant denominations, and those who did not report a denomination).4 Furthermore,
each additional week of church attended raised schooling by about 0.016 years.5 As indicated
earlier, this estimate may combine the spurious relationship due to omitted-variables bias and the
causal relationship.
Column 2 shows the results of estimating Equation (1), which includes the individual's
1980 AFQT scores (measured by percentile rankings) and the individual's own 1979 educational
expectations as X2 variables and adds the number of college-educated and the number of high
school dropout older siblings by gender to the Z1 variables.6 According to Column 2, each
additional college-educated sibling raised years of schooling by about 0.2 and each additional
high school dropout older brother reduced schooling by about 0.09. One more year of expected
schooling raised education actually achieved by 0.36 years. The coefficient of the AFQT was
similarly large and significant at 0.03 higher years of schooling for every point on the percentile
ranking. Each additional week of church attendance now raised schooling by 0.010 years compared
to 0.0 16 in Column 1. In the absence of omitted-variables bias, this is a lower-bound estimate of

This content downloaded from 103.229.203.182 on Wed, 09 Jan 2019 06:55:55 UTC
All use subject to https://about.jstor.org/terms
122 JOURNAL FOR THE SCIENTIFIC STUDY OF RELIGION

TABLE 1
MEANS AND STANDARD DEVIATIONS OF
SELECTED VARIABLES

Years of schooling (1993) 13.2405


(2.4654)
Weeks attended church in previous year 25.9748
(22.4146)
Whether Episcopal 0.0270
(0.1620)
Whether Lutheran 0.0994
(0.2992)
Whether Methodist 0.0851
(0.2790)
Whether Presbyterian 0.0430
(0.2028)
Whether Catholic 0.3857
(0.4868)
County fraction of adherents 0.5722
(0.1600)
Denomination fraction of adherents 0.3560
(0.2514)
Number of observations 2,748

Note. Standard deviations are in parenthesis. All variables


were measured as of 1979 or 1980 except where indicated.

the effects of church attendance because church attendance also alters schooling through changing
AFQT scores and educational expectations. In results not shown here, the coefficients and standard
errors of church attendance in OLS regressions on AFQT scores and educational expectations
are 0.0536 (0.0244) and 0.0065 (0.0020), respectively, holding constant the other variables in
Column 2.7
Note that the magazine and library card coefficients fell from 0.5 and 0.3, respectively, to 0.03
and 0.07 when sibling schooling, AFQT scores, and educational expectations are included in the
analysis. Neither is now significant even at the 10 percent level. As indicated earlier, the primary
role of the library card and magazine variables was to act as proxies for unobservables. Reductions
in their coefficients would indicate that such unobservables no longer have a significant effect on
schooling, holding sibling schooling, AFQT scores, and education expectations constant.
Furthermore, the initial estimated effects of the other background variables (e.g., parents'
education and occupation, whether attended private school, and all the religious denominations)
include both the spurious relationship due to omitted variables and the causal relationship. These
estimated effects are now substantially reduced in Column 2 compared to Column 1. Several
of these variables (e.g., white-collar mother or father, whether Episcopal, Lutheran, or Catholic,
and whether attended a private school) are not significant even at the 10 percent level. This
suggests that both the causal relationship and any bias in the initial coefficients of these variables
in Column 1 are largely captured by including sibling schooling, AFQT scores, and educational
expectations. If the same omitted variables that account for these attenuated effects also influence
church attendance, then adding sibling schooling, AFQT scores, and educational expectations
would remove most of the bias associated in the estimated coefficient of church attendance.8
An alternative to adding proxy variables to reduce omitted-variables bias is 2SLS estimation.
The coefficient would, in this case, measure the effect of the exogenous or predetermined variation

This content downloaded from 103.229.203.182 on Wed, 09 Jan 2019 06:55:55 UTC
All use subject to https://about.jstor.org/terms
CHURCH ATTENDANCE AND SCHOOLING 123

TABLE 2
ESTIMATES OF THE EFFECTS OF SELECTED VARIABLES ON YEARS
OF SCHOOLING

OLS with
OLS Proxies 2SLS

Weeks attended church in previous year 0.0156 0.0103 0.0157


(0.0020) (0.0017) (0.0075)
Mother's years of schooling 0.1562 0.0489 0.0477
(0.0214) (0.0178) (0.0179)
Mother-white-collar worker 0.3115 0.1557 0.1584
(0.1019) (0.0817) (0.0821)
Father's years of schooling 0.1268 0.0314 0.0307
(0.0184) (0.0151) (0.0152)
Father-white-collar worker 0.4719 0.1261 0.1012
(0.1343) (0.1047) (0.1067)
Father-craftsman -0.3246 -0.1062 -0.1296
(0.1090) (0.0888) (0.0905)
Age 0.0009 -0.0765 -0.0631
(0.0400) (0.0334) (0.0368)
Female-headed family 0.1184 0.1543 0.1845
(0.1540) (0.1252) (0.1273)
Number of college-educated older sisters (1993) 0.2525 0.2483
(0.0597) (0.0602)
Number of high school dropout older sisters (1993) -0.0492 -0.0119
(0.0865) (0.0865)
Number of college-educated older brothers (1993) 0.2403 0.2476
(0.0625) (0.0632)
Number of high school dropout older brothers (1993) -0.2040 -0.1898
(0.0645) (0.0684)
Number of older siblings (1993) -0.0831 -0.0740 -0.0795
(0.0229) (0.0224) (0.0235)
Number of younger siblings (1993) -0.2115 -0.1148 -0.1115
(0.0376) (0.0283) (0.0286)
Whether female 0.0617 0.0490 0.0292
(0.0868) (0.0698) (0.0736)
Whether Episcopal 0.7440 0.2667 0.2688
(0.3320) (0.2736) (0.2729)
Whether Lutheran 0.1959 -0.0344 -0.0804
(0.1463) (0.1187) (0.1348)
Whether Methodist 0.7199 0.5026 0.4578
(0.1648) (0.1280) (0.1394)
Whether Presbyterian 0.7712 0.4881 0.4779
(0.2391) (0.1942) (0.1958)
Whether Catholic 0.2699 0.0225 -0.0425
(0.1020) (0.0827) (0.1198)
Whether Baptist, other Protestant, or no denomination specified
Whether family had magazines 0.4854 0.0314 0.0342
(0.1002) (0.0787) (0.0809)

(Continues)

This content downloaded from 103.229.203.182 on Wed, 09 Jan 2019 06:55:55 UTC
All use subject to https://about.jstor.org/terms
124 JOURNAL FOR THE SCIENTIFIC STUDY OF RELIGION

TABLE 2
(Continued)

OLS with
OLS Proxies 2SLS

Whether family had library card 0.3438 0.1050 0.1106


(0.0973) (0.0777) (0.0788)
Attended private school 0.3709 0.0624 0.0198
(0.2085) (0.1649) (0.1887)
Years of schooling expects to complete 0.3494 0.3439
(0.0242) (0.0249)
AFQT score 0.0299 0.0301
(0.0018) (0.0019)
Constant 8.7940 6.8145 6.6288
(0.6647) (0.6470) (0.7073)
R2 0.3388 0.5752 0.5726

Note. Standard errors are in parentheses. All variables were measured as of 1979 or 1980 except where
indicated. Other variables included in the analysis were dummy variables for don't know father's schooling,
don't know mother's schooling, don't know years of schooling expects to complete, and don't know AFQT
score.

in church attendance on schooling. Potential candidates for instruments include variables that
affect church attendance but do not alter schooling choices. The instruments used for this 2SLS
analysis are whether the individual resided in the South, the fraction of the county's population
who were church adherents (total adherents/total population), and the fraction of the adherents
in the county who belonged to the individual's denomination (total denomination adherents/total
adherents).9 The choice of the regional variable is based on previous research. Chalfant and
Heller (1991), Kanagy, Firebaugh, and Nelsen (1994), and Stump (1986) found significant church
attendance differences by region. Furthermore, a review of intergenerational mobility by Haveman
and Wolfe (1995) showed that region does not have significant effects on years of schooling for
whites. The county adherents fraction was used because it measures local attitudes toward church
attendance. Similarly, the denomination fraction measures the ease with which individuals may
find similar believers and places of worship for their own denomination.
Column 3 of Table 2 shows that the 2SLS coefficient estimate of church attendance equals
0.016. This result suggests that the estimated coefficient of church attendance in Column 2 may,
in fact, be biased downward due to measurement error. Church attendance figures are generally
regarded as noisy indicators of church participation (see, e.g., Chaves and Cavendish 1994; Smith
1998; Presser and Stinson 1998). The effects of such measurement error on the estimates of the
effects of church attendance in the multivariate context are ambiguous. However, in the one-
variable case, measurement error results in downward bias (see Greene 1993).
The 2SLS estimates are unbiased only if the proposed instruments are correlated with church
attendance but uncorrelated with total years of schooling ceteris paribus. The first assumption
is valid since, in analysis not reported here, each of the proposed instruments has a large and
significant effect on church attendance. The second assumption is likely to be valid in most cases.
Religious adherents are measured at the county rather than at the neighborhood level. Among
residential moves within state between 1975 and 1980, over 70 percent occurred within the same
county (U.S. Bureau of the Census 1983:2). This suggests that dissatisfaction with neighborhood
characteristics is often resolved by choosing a different neighborhood in the same county.
The second assumption may not be valid for members of smaller religious groups who
may tend to choose their residence to find concentrations of similar believers. If, for example,

This content downloaded from 103.229.203.182 on Wed, 09 Jan 2019 06:55:55 UTC
All use subject to https://about.jstor.org/terms
CHURCH ATTENDANCE AND SCHOOLING 125

there were related geographic variations in school quality, the fraction of the adherents in the
county who belonged to the individual's denomination could be correlated with educational
attainment. County-level residential choices based on religious concerns are unlikely to be a
problem for this analysis since it includes only Catholics, Protestants, and those who did not
report a denomination.'0
The church attendance coefficient from Column 2 of Table 2 implies that the 19 percent of
the sample who never attended church in 1979 would have 0.52 fewer years of schooling than
the 37 percent who attended at least once per week ceteris paribus (0.010 * 52 = 0.52). This
is equivalent to over three years of mothers' schooling or over four years of fathers' schooling
(see Table 2, Column 1). Separate analyses by gender and other ways of specifying the church
attendance variable yield similar results. The coefficients and standard errors from running sep-
arate regressions for men and women were 0.012 (0.002) and 0.009 (0.002), respectively. The
coefficients of dummy variables for attended religious services at least once a week, attended
about two to three times per month, and attended once per month or infrequently were 0.529
(0.105), 0.269 (0.135), and 0.038 (0.100). Never attended is the left-out variable.
Finding significant church attendance effects is also not sensitive to how educational attain-
ment is measured. Church attendance increases both the likelihood that individuals will complete
high school and the likelihood that they will attend college. Using the same explanatory variables
as in Column 2 of Table 2, the logit coefficient and standard errors for church attendance in
analyses of high school graduation and college attendance were 0.0 12 (0.003) and 0.0 10 (0.003),
respectively. The high school graduation results differ from Ribar (1994). He reports higher
schooling both for those who attended church often and those who attended infrequently. Since
his older sample members were ages 18 to 24, Ribar may be measuring church attendance for
college-educated individuals at the point in their lifecycle when it is especially low. College at-
tendance coincides with relatively low church attendance among individuals in their late teens
and early 20s (see Chaves 1991; Hout and Greeley 1987).

DIsCUSSION

The results of this article imply that church attendance raises an important component of
socioeconomic status in addition to its effects on health and subjective measures of well-being.
Although this observation is independently noteworthy, it is particularly salient for evaluating
programs to fund religious organizations that provide social services to the disadvantaged. This
research suggests that any measured direct effects of faith-based programs may understate their
overall impact if such programs alter church going among children of program participants. It
also suggests that the religious character of such programs could have an independent role in
promoting their success.

CONCLUSION

This article shows that church attendance during adolescence significantly increases total
years of schooling that individuals obtain. It implies that variation in church attendance, either
due to exogenous changes in attitudes, demographic characteristics, or government or other in-
stitutional activity, may have spill-over effects on socioeconomic variables in addition to the
extensively documented effects on subjective well-being.

ACKNOWLEDGMENTS

The author would like to acknowledge the helpful suggestions of an anonymous referee and
funding from the Ford Foundation.

This content downloaded from 103.229.203.182 on Wed, 09 Jan 2019 06:55:55 UTC
All use subject to https://about.jstor.org/terms
126 JOURNAL FOR THE SCIENTIFIC STUDY OF RELIGION

NOTES

1. In conservative Protestant denominations, higher education in secular institutions is often regarded as anti-God and
anti-moral (see LaHaye 1977). However, Sherkat and Darnell (1999) show that concerns about these dangers might
be regarded as small for children sufficiently inoculated from worldly influences. Church attendance is a potential
way for inoculating children.
2. Restricting the sample to those ages 14-17 means that church attendance and AFQT are measured prior to schooling's
completion. This would, therefore, reduce any potential simultaneity problems (Sander 1992).
3. The total number of white sample members ages 14-17 is 4,086. Nine-hundred-seventy-eight were omitted because
of missing data for own and sibling schooling as of 1993, five were omitted because they did not report church
attendance as of 1979, 355 were omitted because they belonged to a non-Catholic or Protestant denomination. The
analysis is weighted due to initial nonrandom sample selection and nonrandom attrition.
4. Separate dummy variables for Baptists and other Protestants were not included in the analysis since their coefficients
were not significantly different from zero. For example, in Column 1 of Table 2, the coefficients and standard errors
for Baptists and other Protestants would have been -0.0353 (0.1396) and -0.0771 (0.2031).
5. Individuals were coded as 52 if they attended religious services at least once a week, as 30 for two to three times per
month, as 12 for once per month, as 6 for infrequently, and as 0 if they did not attend at all.
6. The coefficient for the number of all older siblings is then the effect of older siblings who have exactly 12 years of
schooling.
7. Any simultaneity due to church attendance depending on AFQT scores and educational expectations is likely to
be small. Most recent analyses of determinants of church attendance or choice of denomination find either a small
(sometimes negative) or no relationship between socioeconomic status and measures of religiosity for adults (e.g.,
Condran and Tamney 1985; Alston and McIntosh 1979).
8. Since AFQT scores may depend on years of schooling, age was included in the analysis. According to Column 2, age
is, in fact, negatively related to years of schooling, holding AFQT constant. In addition, note that the results do not
change substantially when sample members are restricted to ages 14-16 where this problem would be less important.
In separate regressions, the church attendance coefficient is 0.010 for ages 14-16 and 0.009 for age 17.
9. The actual variable used here equals the fraction Methodist if the individual was Methodist, the fraction Baptist
if the individual was Baptist, and so forth for Lutherans, Presbyterians, Catholics, and Episcopalians. Those from
other denominations or no denomination were coded as the fraction of the county's adherents who belonged to
religions other than those just listed. This variable would reflect prevalence of religious beliefs outside the mainstream
groups.
10. Adding individuals from non-Catholic or Protestant denominations into the analysis has little effect on the results.

REFERENCES

Akerlof, G. 1997. Social distance and social decisions. Econometrica 65:1005-27.


Astone, N. M. and S. McLanahan. 1991. Family structure, parental practices, and high school completion. American
Sociological Review 56:309-20.
Berger, P. 1969. The sacred canopy. New York: Anchor Books.
Chalfant, H. P. and P. Heller. 1991. Rural/urban versus regional differences in religiosity. Review of Religious Research
33:76-86.
Chaves, M. 1991. Family structure and Protestant church attendance: The sociological basis of cohort and age effects.
Journalfor the Scientific Study of Religion 30:501-14.
Chaves, M. and J. Cavendish. 1994. More evidence on U.S. Catholic church attendance. Journalfor the Scientific Study
of Religion 33:376-81
Cochran, J. and L. Beeghley. 1991. The influence of religion on attitudes toward nonmarital sexuality: A preliminary
assessment of reference group theory. Journalfor the Scientific Study of Religion 30:45-62
Condran, J. and J. Tamney. 1985. Religious nones: 1957 to 1982. Sociological Analysis 46:415-23.
Currie, J. and D. Thomas. 1999. The intergenerational transmission of intelligence: Down the slippery slopes of The bell
curve. Industrial Relations 38:297-330.
Evans, W., W. Oates, and R. Schwab. 1992. Measuring peer group effects: A study of teenage behavior. Journal of Polit
Economy 100(5):966-91.
Freeman, R. 1986. Who escapes? The relation of churchgoing and other background factors to the socioeconomic per-
formance of black male youths from inner-city poverty tracts. In The black youth employment crisis, edited by R.
Freeman and H. Holzer, pp. 353-76. Chicago, IL: University of Chicago Press.
Greene, W. 1993. Econometric analysis. New York: Macmillan Publishing.
Haveman, R. and B. Wolfe. 1995. The determinants of children's attainments: A review of methods and findings. Journal
of Economic Literature 30: 1829-78.
Heckman, J. 1995. Lessons from the bell curve. Journal of Political Economy 103: 1091-1120.

This content downloaded from 103.229.203.182 on Wed, 09 Jan 2019 06:55:55 UTC
All use subject to https://about.jstor.org/terms
CHURCH ATTENDANCE AND SCHOOLING 127

Homola, M., D. Knudsen, and H. Marshall. 1987. Religion and socio-economic achievement. Journalfor the Scientific
Study of Religion 26:201-17.
Hout, M. and A. Greeley. 1987. The center doesn't hold: Church attendance in the United States, 1940-1984. American
Sociological Review 52:325-45.
Kanagy, C., G. Firebaugh, and H. Nelsen. 1994. The narrowing regional gap in church attendance in the United States.
Rural Sociology 59:515-24.
Jencks, C. and S. Mayer. 1990. The social consequences of growing up in a poor neighborhood. In Inner-city poverty in
the United States, edited by L. E. Lynn and M. McGeary, pp. 111-86. Washington, DC: National Academy Press.
Krein, S. and A. Beller. 1988. Educational attainment of children from single-parent families: Differences by exposure,
gender, and race. Demography 25:221-34.
LaHaye, B. 1977. How to develop your child's temperament. Eugene, OR: Harvest House.
Lehrer, E. 1999. Religion as a determinant of educational attainment: An economic perspective. Social Science Research
28:358-79.
Neal, D. and W. Johnson. 1996. The role of pre-market factors in black-white wage differences. Journal of Political
Economy 104:869-95.
Petersen, L. and A. Roy. 1985. Religiosity, anxiety, and meaning and purpose: Religions' consequences for psychological
well-being. Review of Religious Research 27:49-62.
Presser, S. and L. Stinson. 1998. Data collection mode and social desirability bias in self-reported religious attendance.
American Sociological Review 63:137-45.
Quinn, B., H. Anderson, M. Bradley, P. Goetting, and P. Shriver. 1982. Churches and church membership in the United
States 1980. Atlanta, GA: Glenmary Research Center.
Ribar, D. 1994. Teenage fertility and high school completion. Review of Economics and Statistics 76:413-24.
Sander, W. 1992. The effects of ethnicity and religion on educational attainment. Economics of Education Review 1 1:1 19-
35.
Sherkat, D. and A. Darnell. 1999. The effect of parents' fundamentalism on children's educational attainment: Examining
differences by gender and children's fundamentalism. Journalfor the Scientific Study of Religion 38:23-35.
Smith, T. 1998. A review of church attendance measures. American Sociological Review 63:131-36.
Stump, R. 1986. Regional variations in the determinants of religious participation. Review of Religious Research 27:208-
25.
Stryker, R. 1981. Relio-ethnic effects on attainment in the early career. American Sociological Review 46:212-31.
Teachman, J. 1987. Family background, educational resources, and educational attainment. American Sociological Review
52:548-57.
U.S. Bureau of the Census. 1983. County and city data book, 1983 Washington, DC: U.S. Government Printing Office.

This content downloaded from 103.229.203.182 on Wed, 09 Jan 2019 06:55:55 UTC
All use subject to https://about.jstor.org/terms

You might also like