ally nds positive effects of Catholic schooling on educational attainment and student achievement.
1
However, Ludwig 1997 nds that many of these instruments are not exogenous, and Altonji, Elder, and Taber 1999 nd that functional form assump-
tions, rather than instruments, provide much of the explanatory power in these models.
The small literature on Catholic primary schooling suffers from similar problems with selection bias. Sander 1996 nds a positive and signicant effect of eight
years of Catholic primary school attendance on tenth-grade test scores. However, he does not control for Catholic high schooling in these regressions, so it is impossible to
determine whether Catholic grade school attendance or Catholic high school atten- dance drives these effects.
2
It is also unclear whether identication in his selection bias models comes from the instruments or the functional form assumptions. National
Assessment of Educational Progress NAEP data include Catholic primary schools but contain little student and school demographic data. Therefore, Lee and Stewart
1989 and Jones 1997 are only able to control for a few student characteristics and are unable to control for selection bias or prior achievement. Nonetheless, both
papers nd that Catholic primary school students outperform public primary school students in NAEP tests.
In this paper, I utilize a unique microdata set, the Prospects study of the Title 1 program, in order to analyze the effect of Catholic primary schools on test scores,
classroom behavior, and attendance. Because each Prospects school includes test score data for students in rst and fourth grade, I use the test scores from the students
in the rst-grade cohort as controls for selection bias for the students in the fourth- grade cohort under the assumption that ability and out-of-school factors have more
inuence than in-school factors in test scores for rst-grade students.
Using a value-added specication, I nd insignicant but generally positive effects of Catholic schooling on 1993 mathematics and reading test scores for students in
either cohort. Catholic schooling has no effect on classroom behavior. For the fourth- grade cohort, Catholic schooling is associated with a modest decrease in student
absences of less than one day per year. At the same time, Catholic school teachers are more likely than public school teachers to report that excessive absenteeism is
a problem.
II. Data
The data for this paper come from Prospects, the Congressionally mandated national study of the Chapter 1 program in the early 1990s. The program,
now called Title 1, provides additional resources to low-income and other at-risk students. The Prospects survey includes data on individual students in the rst,
fourth, and seventh grades in the spring of 1992. I use data from the rst- and fourth- grade cohorts, as the seventh grade cohort has no Catholic school counterpart. The
1. Recent work on the effects of Catholic high school attendance includes Evans and Schwab 1995; Sander and Krautmann 1995; Goldhaber 1996; Neal 1997; Figlio and Stone 1999; Grogger and
Neal 2000; Altonji, Elder, and Taber 2000; and Figlio and Ludwig 2000. 2. Sander 1996 nds that a single years-of-Catholic-schooling variable has no effect on test scores.
database for each cohort of Prospects is quite extensive. Students, parents, teachers, school administrators, principals, and district Chapter 1 coordinators for public
school students only completed questionnaires. Students in both cohorts took the fourth edition of the Comprehensive Test of
Basic Skills CTBS4 in reading and mathematics in the spring of 1992, with a followup test in the spring of 1993.
3
These tests are measured on a vertical scale, in order to allow for comparisons across cohorts and years. In other words, a score
of 500 for a second grade student in 1992 measures the same level of academic achievement as a score of 500 for a different student in another grade or during
another school year.
4
Each cohort contains over 10,000 public school students from approximately 200 schools, as well as more than 1,000 Catholic school students from 35 schools. Public
school districts and schools were carefully chosen to provide a representative sample of the schools and students receiving Chapter 1 funding, rather than to provide a
nationally representative sample of all primary schoolchildren.
5
The set of Catholic schools were chosen so that their overall demographic characteristics matched those
of the public schools. Within each school public or Catholic, if grade enrollment is under 150, all students in the grade are surveyed. Otherwise, every student in
four classrooms chosen at random is interviewed. Appendix Table A1 contains the descriptive statistics for student demographics.
6
The descriptive statistics by school type for the Prospects test score data are in Table 1. The table provides no clear evidence of positive or negative selection bias
on test scores. Catholic school students in Prospects actually have signicantly lower 1992 mathematics test scores for both cohorts. By 1993, Catholic and public
school students have approximately equal mathematics test scores, so the Catholic- public differences are insignicant for both cohorts. In reading, Catholic school stu-
dents in the rst-grade cohort have signicantly higher test scores in both years. For the fourth-grade cohort, Catholic and public school mean test scores are almost iden-
tical in both years.
I investigate the applicability of the data for representing low-income schools by comparing Prospects test scores with test scores from the voucher programs in
Cleveland, Dayton, New York, San Antonio, and Washington, D.C. Peterson, Greene, and Howell 1999; Peterson, Myers, and Howell 1999; and Howell et al.
2000. Both Catholic and public school students in the Prospects data have higher mathematics and reading national percentile scores than corresponding students in
the voucher programs. At the same time, all the Prospects test score means in Table 1 are below the national averages for the CTBS4 reported in CTB 1989.
3. Although nearly 40 percent of the students in the 1992 and 1993 data did not take the test in both years, sample attrition between 1992 and 1993, at least with respect to test scores, is similar in Catholic schools
and in public schools. Many of the students with missing test scores do have sufcient English language ability to take the test.
4. For more information on the test scores, see Jepsen 2000 as well as CTB 1989. 5. More information on the sampling technique for both the public and Catholic schools in Prospects is
available in Jepsen 2000, as well as in Bryant 1993 and Puma 1995. 6. All descriptive statistics and regressions are unweighted, as sampling weights are not available for the
entire sample.
Table 1 Descriptive Statistics for Test Scores, Behavior, and Absenteeism
Grade 1 Cohort Grade 4 Cohort
Public Catholic
Public Catholic
Reading 1992 mean test score
554.4 558.8
687.3 687.4
66.32 56.96
48.17 44.62
1993 mean test score 620.8
630.7 698.9
698.4 73.52
65.24 51.49
48.42 Mathematics
1992 mean test score 512.8
502.8 690.1
683.0 70.79
60.47 47.46
48.23 1993 mean test score
598.2 595.7
714.4 713.8
64.90 55.53
45.00 41.95
Observations 6,134
870 6,790
758 Behavior
Compliance scale 1.404
1.394 1.458
1.438 0.441
0.425 0.467
0.449 Motivation scale
1.715 1.678
1.762 1.744
0.584 0.569
0.592 0.585
Class participation 2.017
2.006 2.049
2.032 0.551
0.584 0.584
0.606 Absenteeism
Days missed 7.255
6.570 6.236
5.692 6.408
6.070 6.020
5.791 Excessive absenteeism
0.089 0.077
0.079 0.099
0.284 0.267
0.270 0.298
Observations 10,503
1,094 10,615
1,018
Notes: Standard deviations are in parentheses. Descriptive statistics for student demographics are in Appen- dix Table A1.
III. Econometric Approaches