Historical religious concentrations and the effects of Catholic schoolingIntroductionDataNELS:88 and ECLS-KHistorical Catholic share dataHistorical Catholic shares, the supply of Catholic schools, and Catholic school attendanceEvidence for the exogeneity of historical Catholic sharesReduced-form relationships between Catholic shares and outcomesEstimates of the causal effects of Catholic high school attendanceConclusions and directions for future researchData appendixReferencesHistorical religious concentrations and the effects of Catholic schoolingqDanny Cohen-Zadaa,*, Todd ElderbaBen-Gurion University, Department of Economics, P.O. Box 653, Beer-Sheva 84105, IsraelbMichigan State University, Department of Economics, 110 Marshall-Adams Hall, East Lansing, MI 48824-1038, USAarticle infoArticle history:Received 18 December 2008Revised 15 April 2009Available online 7 May 2009JEL classification:I21Keywords:Catholic schoolsTreatment effectabstractThe causal effects of Catholic schooling on student outcomes have proven challenging to estimate, withseveral previous studies using the proportion of a geographic unit’s population which is Catholic as apotentially exogenous source of variation in the availability of Catholic high schools. We propose anew approach which instead relies on the historical distribution of religious preferences. Specifically,we find that county-level Catholic shares measured at the end of the 19th century are far more stronglyassociated with Catholic school attendance than are current Catholic shares. Using several strategies, weshow that historical Catholic shares are likely to be exogenous to student outcomes conditional on thecurrent distribution of religion. Estimates based on this identification strategy point to smaller Catholicschooling effects than those implied by OLS, in contrast to instrumental variables estimates from previ-ous studies.Ó 2009 Elsevier Inc. All rights reserved.1. IntroductionNumerous studies have attempted to quantify the causal effectsof Catholic school attendance on student outcomes.1Acknowledg-ing that selection of students into Catholic schools is non-randombut lacking experimental data, researchers have typically relied oninstrumental variables (IV) strategies to disentangle causal effectsfrom spurious correlations due to sorting. Several creative instru-ments have been proposed, including those based on a student’sown religion, the religious composition of the local population (bothdue to Evans and Schwab (1995)), and the local availability of Cath-olic schools (Neal, 1997).Although these IV strategies were all plausibly valid for identi-fying Catholic schooling effects, Altonji et al. (2005a) recently pro-vided several indications that the proposed exclusion restrictionsfail in practice. First, the instruments are strongly related to stu-dent outcomes among eighth graders attending public schools.Since public eighth graders almost never attend Catholic highschools, a reduced-form relationship between an instrument andoutcomes in this subsample suggests that the instrument directlyaffects outcomes and is therefore not excludable. Second, 2SLS esti-mates have typically yielded implausibly large Catholic school ef-fects, much larger than the corresponding OLS estimates. Third,the proposed instruments are strongly associated with observabledeterminants of outcomes, and Altonji et al. (2005a) argue that thisselection on observables implies substantial selection on unob-servables as well. These authors conclude that the prospects forfinding valid exclusion restrictions in this setting are poor, so theydevelop new methods that allow for the estimation of bounds ontreatment effects in the absence of valid instruments (Altonjiet al., 2005b).2In this paper, we propose a new strategy for identifying the ef-fects of Catholic schooling. We use student-level data from the Na-tional Educational Longitudinal Study of 1988 (NELS:88) and theEarly Childhood Longitudinal Study – Kindergarten Cohort (ECLS-K), together with county-level data from several sources on theCatholic share in the population at different points in time, to showthat the fraction of Catholics in the county population in 1890 (theearliest date that this measure is available) can serve as a poten-tially useful instrument for Catholic school attendance.3First, wefind that the 1890 local Catholic share is substantially more powerfulthan the current Catholic share in explaining the current supply ofCatholic schools, many of which were established in the early 20thcentury. Consequently, it is also a stronger predictor of current0094-1190/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.doi:10.1016/j.jue.2009.04.002qWe are grateful to Chris Jepsen, William Sander and Gary Solon for helpfulcomments and suggestions. Valuable comments from the editor and anonymousreferees significantly improved the paper. We thankfully acknowledge financialassistance from the National Institute of Child Health and Human Development. Thecontent is solely the responsibility of the authors and does not necessarily representthe view of the National Institute of Child Health and Human Development or theNational Institutes of Health.* Corresponding author. Fax: +972 8 6472941.E-mail addresses: [email protected] (D. Cohen-Zada), [email protected] (T. Elder).1Coleman et al. (1982), Murnane (1984), Tyler (1994), Evans and Schwab (1995),Sander and Krautmann (1995), Sander (1996), Neal (1997), Vella (1999), Grogger andNeal, 2000, Jepsen (2003), and Dee (2005) are prominent examples.2Jepsen and Montgomery (2009) find that measures of the local availability ofcommunity colleges are exogenous to educational attainment, providing reason foroptimism that location-based instruments can be useful in some circumstances.3Cohen-Zada (2009) similarly found county-level historical Catholic populationshares in 1890 to be a valid instrument for private school competition.Journal of Urban Economics 66 (2009) 65–74Contents lists available at ScienceDirectJournal of Urban Economicswww.elsevier.com/locate/jueCatholic school attendance. Second, a host of evidence suggests thathistorical Catholic shares are much more likely to be exogenous tostudent outcomes than are current Catholic shares. For example, fol-lowing the logic of Altonji et al. (2005a), the 1990 Catholic share hasa significantly positive effect on college attendance and on 12thgrade math test scores in a sample of public eighth graders,