Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

Transcription

1 American Journal of Economics and Sociology, Inc. The Wage Effects of Being Raised in the Catholic Religion: Does Religion Matter? Author(s): Bradley T. Ewing Source: The American Journal of Economics and Sociology, Vol. 59, No. 3 (Jul., 2000), pp Published by: American Journal of Economics and Sociology, Inc. Stable URL: Accessed: :11 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 American Journal of Economics and Sociology, Inc. is collaborating with JSTOR to digitize, preserve and extend access to The American Journal of Economics and Sociology

2 Focus ON ECONOMIC SOCIOLOGY The Wage Effects of Being Raised in the Catholic Religion Does Religion Matter? By BRADLEY T. EWING* ABSTRACT. This paper provides new empirical evidence about the existence of a Catholic wage premium. A simple allocation-of-time model provides two explanations for the observation that those persons raised in the Catholic religion earn more than their non-catholic counterparts. The Catholic religion may add to a person's stock of human capital and/or it may act as a signal of desirable labor market characteristics such as discipline, honesty, trustworthiness, and high motivation. I Introduction WHILE MUCH HAS BEEN WRITTEN about the effects of Catholic schooling educational attainment and, to a lesser extent, earnings, it is surprising that little research has explicitly addressed the effect that being raised in the Catholic religion has on future wages. This paper examines the wage effects of persons who were raised in the Catholic religion. The paper is motivated by the work of Hulsman (1999) who discusses a rational theory of the division of labor in the context of the Christian *Dr. Bradley T. Ewing is an Assistant Professor of Economics at Texas Tech University in Lubbock, Texas Bewing~ttacs.ttu.edu. His interests include labor economics, macroeconomics, and finance. Recently he has published in the Southern Economic Journal, Applied Economics, and the Review of Economics and Statistics. He would like to thank Professor Mark Yanochik and Professor Laurence Moss for valuable comments on this paper. American Journal of Economics and Sociology, Vol. 59, No. 3 (July, 2000). C) 2000 American Journal of Economics and Sociology, Inc.

3 420 American Journal of Economics and Sociology faith and the debate over the benefits of Catholic schooling (Sander 1996; Neal 1997). The effects of Catholic secondary schooling on high school and college graduation rates and future wages was examined by Neal (1997), who concluded that Catholic schools are not superior to public schools in general. His results indicate that there are dramatic positive effects of Catholic high school education for urban Blacks in terms of graduation rates and future earnings, and attributes this to the lower quality of urban public schools. The effects for urban Whites are much less dramatic, and no wage effects are found. Sander (1996) looked at the impact of Catholic grade school education on academic achievement. Examining the scores of non-hispanic Whites on tenth grade standardized tests, Sander determined that a positive Catholic grade school effect is driven by non-catholics who attend Catholic schools. Sander and Krautman (1995) studied the impact that Catholic schooling has on dropping out of high school and educational attainment, paying particular attention to the effect of selection into the Catholic school sector. Their evidence suggests that sophomores in Catholic high schools are more likely to graduate with their class than sophomores in non-catholic schools, and that Catholic high school seniors are more likely to acquire additional schooling than are non-catholic seniors, if selection is taken into account. Evans and Schwab (1995) compare the effectiveness of public and Catholic schools using information on high school completion and starting college as measures of school quality. They do not find evidence of selection bias and conclude that attending a Catholic high school raises the probability of finishing high school and beginning college. The common theme in all of the above mentioned papers is the study of the impact of Catholic schooling on either education or labor market attainment. A Catholic education may have positive benefits of the type that are typically rewarded in the labor market; however, it is not clear if actually being raised in the Catholic religion has any impact on realized future earnings. I examine this issue, and my results indicate that after controlling for standard human capital, institutional, and demographic variables that are also known to affect the wage a

4 Wage Effects of Being Raised Catholic 421 person receives, a substantial earnings premium exists for those person who are identified as being raised in the Catholic religion. Explanations for a Catholic wage premium are provided in the context of a simple allocation-of-time model, and the results and implications are discussed. II Data, Methodology, and Empirical Results THE DATA ARE FROM the National Longitudinal Surveys of Youth (NLSY) which has interviewed respondents annually from 1979 to present. The initial wave contained 12,686 individuals between the ages of 14 and 21. My sample consists of 3527 individuals who worked full-time for pay in the nonagricultural, nonpublic sector in the year prior to the 1990 wave. I am interested in the future effects of being raised in the Catholic religion. Workers are identified as being Catholic, Jewish, Protestant, of some other religion, or of no religion as described below. The 1979 wave asked the respondents "In what religion were you raised?" This definition of religious affiliation, or upbringing, is asked at a time when the respondents were still relatively young, between the ages of 14 and 21, but old enough to have developed a sense of identity and to have attached themselves to a particular faith, if any. As such, it does not reflect whether or not the person is a practicing Catholic, Jew, Protestant, etc. in 1990, the year under investigation. Unfortunately, that information does not exist in the NLSY. The variable Catholic is constructed such that it equals 1 if the respondent was raised in the Catholic religion and 0 otherwise. The variable Jewish is constructed such that it equals 1 if the respondent was raised Jewish and 0 otherwise. The variable Protestant is constructed such that it equals 1 if the respondent was raised as a Protestant and 0 otherwise. The variable Other Religion is constructed such that it equals 1 if the respondent was raised in another religion and 0 otherwise (e.g., Moslem).1 Finally, I define the binary variable No Religion as equal to 1 if the respondent reported having no religion and 0 otherwise. Thirty-two percent of the respondents were brought up Catholic,

5 422 American Journal of Economics and Sociology C a) - r - _ r Nvt ( 5 ) Z (n a E ~~~~~~~~ o ~~~~~~~~~~. C H a) N n N F i <H 0 r0 ' \ 'I' o- Cf9 V- U 0 o\ ( r n Cq Cr CZ, C0 o- 1 N 00or) 0 I,) t In q In 00N N 1-v I N rr O FS U ~ ~ ~ ~~ U *5H 0 Q)~~~~~~~a \C t n ", 0 O > 2o V)) 4-.4 = N 4 / oo N > 0 QN 00 v0 0 z~ ~~ -' ~ ~ ~ 4 ~ ~ * 14r\. X q~~~~~~~~~~~~~~~~~~c) a)0 i 0 ) (~~~~) ~ L a)o r t Q0?' EI(S? O O 4.,

6 Wage Effects of Being Raised Catholic 423 fifty-three percent Protestant, one percent Jewish, around ten percent in some other religion, and about four percent with no religion. Clearly, there is the potential for many to have been affected by religion. See Table 1 for selected descriptive statistics by religious upbringing. In this sample, Jewish respondents had the highest mean wage followed by Catholics, Other Religion, Protestants, and No Religion. This pattern holds for mean scores on the Armed Forces Qualifications Test (AFQT) also, while average years of education are highest for Jewish, then Catholic, Protestant, Other Religion, and No Religion. Note that the sample sizes for the five groups vary widely with 1138 Catholic, 1857 Protestant, 353 Other Religion, 142 No Religion, and only 37 Jewish. The empirical specification is: In WAGEj = a + 131(CATHOLIC)i + j2(jew7sh)i +f33(protestant) + 04(NO RELIGION)i + f35(actual Experience)i + f36(actual Experience2), + J7(Tenure)i + 08(Tenure2), + 09(Education)i + Pj0(AFQT)j + 011(Marital Status)i + f12(urban)i + fr13(number of Children)i + f14(establishment Size)i + 015(Local Unemployment Rate)i + I316(Black)i +f317(female)i + (Vector of Regional Dummies)p + (Vector of Industrial Dummies)iq + (Vector of Occupational Dummies)wo + Errori where CATHOLIC equals 1 for those raised in the Catholic religion and 0 otherwise, JEWISH equals 1 for those raised in the Jewish religion and 0 otherwise, PROTESTANT equals 1 for those raised in the Protestant religion and 0 otherwise, NO RELIGION equals 1 for those raised in no religion and 0 otherwise, and lnwage is the natural logarithm of the hourly rate of pay. Note that the control (or reference) group is people raised in some other religion (e.g., Moslem, etc.). I estimated the above model using ordinary least squares regression. Given the richness of the NLSY it is possible to construct a measure of work experience that represents actual weeks worked, less tenure at current firm. There are several reasons why a measure of ac-

7 424 American Journal of Economics and Sociology tual experience is preferred to using potential work experience (usually defined as age-education-6). Potential experience may understate the returns to experience because it treats time not working the same as time working. This is particularly troublesome when estimating wages of persons who are more likely to have intermittent labor force participation.2 The use of both actual experience and tenure at the current firm, and their squares, as control variables should capture the total work experience of the respondent. There are several reasons to include the Armed Forces Qualifications Test as an independent variable in the model. First, it may proxy for unobserved ability (Blackburn and Neumark 1992). Secondly, Maxwell (1994) has successfully argued that AFQT proxies for quality of schooling received. AFQT is included as an explanatory variable in addition to years of education (as measured by number of years of schooling completed). In this respect, elements of both school quality and quantity are incorporated into the study. Additionally, vectors of industry and occupation controls are included, which presumably capture much of the heterogeneity in monitoring technology not captured by establishment size. Other variables include controls for marital status, number of children, race, gender, and local unemployment rate, as well as residence in region of country, and urban area.3 I conducted Ramsey's (1969) regression specification error test (RESET) for omitted variables. Basically, this test estimates w = xb + zt + u and then tests t = 0 via a standard F statistic, where powers of the fitted values are used for z. The null hypothesis is that the model has no omitted variables and the F-statistic was 0.14 with a probability value of Thus, I do not reject the null hypothesis of no omitted variables. Tests for heteroscedasticity were conducted using the Cook- Weisberg (1983) method which tests t = 0 in var(e) = c2exp(zt). The null hypothesis is constant variance. The estimated chi-squared statistic was significant at less than the 1 percent level (X2(1) = 9.13). The model was estimated using White's (1980) method to produce robust standard errors. The summary results of the regression are presented in Table 2.4 Given my main interest of investigating the Catholic-non-Catholic

8 Wage Effects of Being Raised Catholic 425 Table 2 Summary of Regression Results Dependent Variable is InWAGE Estimated coefficient Catholic (0.040) Jewish (0.173) Protestant (0.647) No Religion (0.959) Education (0.000) AFQT (0.000) F-statistic (0.000) [30,3496] Adjusted R-squared Sample size 3527 Note: Actual probability values are in the parentheses. wage differential, I limit the discussion to the estimated Catholic, Jewish, Protestant, and No Religion term coefficients from the model. I find the coefficient on Catholic is positive and significant at the 4 percent level. The general indication is that the effect of having been raised in the Catholic religion translates into about a 6 percent wage premium. None of the other religion variables were found to be significantly different from zero in the wage regression.5 However, the coefficient on Jewish was found to be positive, but still insignificant at the 10 percent level (p-value = 0.17). The lack of significance may be the result of the small number of Jewish respondents in the survey. It is interesting to note that while the sign of the coefficient on Jewish is

9 426 American Journal of Economics and Sociology positive, the 95 percent confidence interval is from -0.5 to.30, while these bands are from -.08 to.07 and from -.06 to.04 for No Religion and Protestant, respectively. The corresponding band for Catholic is from.00 to.11. I interpret these confidence intervals as consistent with the evidence reported earlier of a positive Catholic wage premium, and believe they also suggest the possibility of a Jewish wage premium that has not been picked up in this sample.6 Note also that the estimated coefficients on Education and AFQT are positive and significant. In the next section, I provide two explanations that are consistent with the existence of a Catholic wage premium. III Explanations for a Catholic Wage Premium I NOW DISCUSS how being raised in the Catholic religion may affect the f ture wages of workers in the context of Alchian and Demsetz's team production model and Spence's signaling model, both within an allocation-of-time framework (Alchian and Demsetz 1972; Spence 1973). Having observed that Catholics in the sample earn more than non- Catholics, I develop a framework that offers insights into the wage- Catholic relationship. The model's predictions are consistent with Catholics earning more and suggest that there is more to the wage-catholic relationship than mere correlation between a Catholic upbringing and accumulation of other income-generating human capital. Participation in Catholic religion may affect wages in two ways. First, the Catholic religion may enhance productivity by teaching discipline or by fostering other characteristics that are rewarded in the labor market. Second, firms may recognize that Catholics have certain desirable traits that are generally unobservable, such as drive, motivation, honesty, trustworthiness, or strong work ethic, which may indicate a lower likelihood of shirking. Thus, there may be a signaling aspect to having been raised in the Catholic religion. The development of a simple variant of the team production model of Alchian and Demsetz, which purports that agents may gain traits from Catholic religion participation that increase their future earning power, may shed some light on the empirical findings (Alchian and Demsetz 1972). These traits may be described as enhancements to

10 Wage Effects of Being Raised Catholic 427 self-control, perseverance, and discipline. Consequently, Catholics may work better in a variety of production situations than other workers, thereby increasing workplace efficiency. Let us consider the standard allocation-of-time model extended to the case where an agent can divide his or her time between three activities: leisure, Catholic religion, and education (Becker 1965). Suppose that a person's time is allocated among these three alternatives with the goal of maximizing their lifetime utility. I allow time choices to be made by the parents of a person or in conjunction with them (i.e., the choice of being raised Catholic may be decided for the individual by the parents). However, once the decision has been made, the standard-allocation-of-time framework applies since it is as if the person had made the decision to be raised Catholic or not. Parents will make the decision to raise their children in the Catholic religion if they believe it is in their child's best interest; in other words, it can be treated as a utility maximizing choice. With the above in mind, suppose that both education and the Catholic religion add to the stock of human capital, whereas leisure does not. Below I also discuss the implication of relaxing the assumption that Catholic religion adds directly to human capital. Let Tbe the total time available to an individual, L be the amount of time devoted to leisure, E the amount of time devoted to education, and C the amount of time devoted to Catholicism. Lifetime earnings, appropriately discounted, are given by Y. In the Alchian and Demsetz (1972) framework, the ability to generate income depends positively on both time devoted to academics and time devoted to Catholic religion participation, such that Y = g(e, C). The individual's time is allocated to maximize utility: U(L, Y, C). For ease of exposition I assume a specific functional form of UQ) that is separable in its arguments and linear in consumption. That is, I assume that Uij = 0 for i 0 j, Uji < O for i=1, 3, and Uji = O for i=2. A utility maximizing individual solves the following problem by choosing the amount of time to devote to education and Catholic participation. max [ a f (L) + g(e, C) + h (C)] subject to C 0- (1) where standard concavity assumptions apply, a non-negativity constraint on C is imposed, and L = T - E - C. Note that individuals can

11 428 AmericanJournal of Economics and Sociology differ in two dimensions, a and,3, the relative costs of leisure and a Catholic upbringing, respectively. Further, I assume g(e, C) is separable such that g12 = 0. The following first order conditions define the optimal choices of E and C, respectively: -a f' + g1 = 0 (2) -ap' D h' + gc =? (3) pc is the Lagrange multiplier for the non-negativity con Equation (2) shows that an individual acquires additional education in order to generate income until the marginal gain from doing so just equals the marginal cost of lost leisure. Note that (3) has the positive term g2 If C did not enter g(o), that is, if Catholic did not add to the stock of human capital, then this term would be zero. Thus, from (3) we see that more people will participate in the Catholic religion when it adds to the stock of human capital than when it does not. Additionally, for any given E, C affects the income generating potential of an individual. The agent who participates in Catholic religion will balance the total gain from additional units of time devoted to Catholic activities with the leisure cost. The heterogeneity in the weights attached to the functions describing the utility from time devoted to leisure and to Catholic religion, a and I, implies different choices of E and C. Comparative static results are derived below and reveal how the optimal allocation of time choices change in response to changes in the parameters. Total differentiation of the first order conditions, (2) and (3), with respect to E, C, a, and I, gives the following system of equations: ( a f' + g11 ) de + ( 0 ) dc = ( f' )da (4) (0 ) de + ( a f' + g22+ 1 h" ) dc = (f' ) da (5) ( a f' + g1l ) de + (0) dc = ( 0) d (6) (0 ) de + ( a f + g22 + h"') dc = ( -h' ) do (7) Rearranging these reveals how the optimal allocation of time choices respond to changes in the parameters a and P. Comparative static results are shown in (8) to (11).

12 Wage Effects of Being Raised Catholic 429 de/ da < 0 (8) de/ do =0 (9) dc/da '0 (10) dc /d-o (O1) Note that dc / da = 0 an These results indicate that persons who value leisure less will devote more time to education and therefore earn more in the future. This is true even if the person optimally chooses not to participate in the Catholic religion. It is also true that those with a high value of leisure (or perhaps those from a family with a high value of leisure) will not participate as much in Catholic activities. Furthermore, the weight attached to Catholic participation does not affect time allocated to education. Not surprisingly, those that do not gain much utility from Catholic religion will spend less time participating in Catholic activities. Note that if Catholic participation does not add to the stock of human capital, then comparative static reasoning suggests that if one adequately controls for education in the wage regression, a positive wage-catholic relationship should not be observed. Thus, a finding that Catholics earn a wage premium is suggestive of some type of human capital enhancement that goes beyond mere correlation of being raised Catholic and being educated. This discussion underlies the importance of estimating the wage model while controlling for education. Asymmetric information about the value a worker places on leisure also provides an explanation for a wage effect of being raised Catholic. The idea is based on Spence's signaling/screening model. In an efficiency wage context, firms value workers who place less value on leisure because they may be less likely to shirk. If firms have perfect information on a, the value of leisure, then there is no need for the firm to use a signal. The model suggests that a correlation between high a type persons (i.e., those with a low value of leisure) and wages exists, and there is a correlation between high a type persons and being raised in the Catholic religion. Now suppose that firms do not observe a, but that they do observe Catholic participation. In this case firms will use C as a signal for a. Consequently, workers with a low

13 430 American Journal of Economics and Sociology value for leisure find it less costly to acquire the signal. Efficiency wage firms would prefer to hire Catholics because they are perceived as more productive. If employers regard Catholics as better workers than non-catholics then we would expect some type of premium paid to Catholics. From the comparative static results we know that higher a type individuals value leisure less than other individual types. People with a higher a will choose to devote more time to Catholic activity and, if worker productivity depends on E and a, then Catholic participation signals potential employers that the worker does not value leisure highly. Catholic acts as a signal on a. Efficiency wage firms would prefer to hire persons raised in the Catholic religion in order to avoid the low a type workers, the Catholic religion indicating that the potential worker is less likely to shirk because he or she does not value leisure highly. IV Concluding Remarks I EXAMINED THE RELATIONSHIP between wages and being raised in the Catholic religion and found that Catholics earn more than their non-catholic counterparts. Two explanations were given that are consistent with a Catholic-wage premium. Being raised in the Catholic religion (1) may add to the stock of human capital and (2) may signal employers that a person has some desirable labor market characteristics such as discipline, honesty, trustworthiness, motivation, and a low value for leisure. The results of this study have policy implications. For example, these findings suggest that religion may be an important factor in determining future economic well-being. Thus, policy-makers should think twice about promoting programs and policies that may discourage or inhibit religious participation. Notes 1. It is not possible to determine which other religion the respondent was actually raised in, as this is simply coded as "other" religion in the NLSY data set. 2. Stratton (1995) examined the effect that timing of work interruptions has on the degree of human capital depreciation. 3. A factor that is not controlled for in this study but may have an effect on

14 Wage Effects of Being Raised Catholic 431 wages nonetheless is whether the respondent was native-born or foreignborn. Recent evidence suggests that during the 1980s there was an increase in the level of schooling among European/Canadian and Asian immigrants to the U.S. but not among Hispanic immigrants (Cohen et. al. 1997) In a related study (Duleep and Regets 1997) it was suggested that there has been a decline in the entry earnings of immigrants over the period after controlling for education level. This decline is primarily attributed to the skills of immigrants not being transferable to the U.S. labor market. The rich set of control variables included in the wage model of this study, particularly AFQT, Education, Occupation, and Industry, should help to mitigate any problem associated with immigration. 4. Full regression results available on request. 5. The lack of a significant coefficient on the Protestant term is not consistent with what one would expect based on the sociological model of Max Weber. The Weber model contends that Protestants have certain ethical values (e.g., self-discipline, personal honesty, creative innovation, etc.) and that they work hard because it is they and they alone that bear responsibility for their salvation. Based on the Weber model one would expect to actually find a positive coefficient. See Jones (1997) for a review of the Weber model. 6. Further tests of hypotheses regarding the estimated coefficients on the religion variables were also conducted. In particular, F-tests rejected the null hypothesis that the estimated coefficient on Catholic was equal to the estimated coefficient on Protestant (F = 19.02, p-value = 0.00) and also that the estimated coefficient on Catholic was equal to the estimated coefficient on No Religion (F = 3.07, p-value = 0.08). However, I was unable to reject the null hypothesis that the estimated coefficient on Catholic was equal to the estimated coefficient on Jewish (F=0.58, p-value = 0.45). Although this latter finding should be interpreted with caution because of the small number of Jewish respondents in the sample, it points to a possible avenue for further research into the possible labor market effects of being raised in a Jewish household. References Alchian, Armen and Harold Demsetz. (1972). "Production, Information Costs, and Economic Organization." American Economic Review, vol. 62 (5): Becker, Gary. (1965). "The Theory of the Allocation of Time." EconomicsJournal 75: Blackburn, McKinley and David Neumark. (1992). "Unobserved Ability, Efficiency Wages, and Interindustry Wage Differentials." Quarterly Journal of Economics 107: Cohen, Y., T. Zach, and B. R. Chiswick. (1997). "The Educational Attainment of Immigrants: Changes Over Time." Quarterly Review of Economics and Finance 37:

15 432 American Journal of Economics and Sociology Cook, R. D. and S. Weisberg. (1983). "Diagnostics for Heteroskedasticity in Regression." Biometrika 70: Duleep H.O. and M. C. Regets. (1997)." The Decline in Immigrant Entry Earnings: Less Transferable Skills or Lower Ability?" Quarterly Review of Economics and Finance 37: Evans, William and Robert Schwab. (1995). "Finishing High School and Starting College: Do Catholic Schools Make a Difference?" QuarterlyJournal of Economics 110: Hulsman, Jorg G. (1999). "Discursive Rationality and the Division of Labour: How Cooperation Emerges." American Journal of Economics and Sociology 58: Jones, Harold B. (1997). "The Protestant Ethic: Weber's Model and the Empirical Literature." Human Relations 50: Lindberg, Assar and Dennis J. Snower. (1987). "Efficiency Wages Versus Insiders and Outsiders." European Economic Review 31: Maxwell, Nan. (1994). "The Effect on Black-White Wage Differences of Differences in the Quantity and Quality of Education." Industrial and LaborRelations Review 47: Neal, Derek. (1997). "The Effects of Catholic Secondary Schooling on Educational Achievement." Journal of Labor Economics 15: Ramsey, J. B. (1969). "Tests for Specification Errors in Classical Linear Least Squares Regression Analysis." Journal of the Royal Statistical Society Series B 31: Sander, William. (1996). "Catholic Grade Schools and Academic Achievement." Journal of Human Resources 31: Sander, William and Anthony Krautmann. (1995). "Catholic Schools, Dropout Rates, and Educational Achievement." Economic Inquiry 23: Spence, Michael. (1973). "Job Market Signaling." Quarterly Journal of Economics 87: Stratton, Leslie. (1995). "The Effect Interruptions in Work Experience Have on Wages." Southern Economic Journal 61: White, Halbert. (1980). "A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity." Econometrica 48: