Religious service attendance and labour force status: evidence from survey data using count data methods

Abstract

I undertake count data estimation with data from the National Longitudinal Survey of Youth 1979 cohort and the Health and Retirement Study to evaluate the relationship between time spent out of the labour force and the frequency of religious service attendance for individuals of working age. I also examine whether being out of the labour force is correlated with the frequency of religious service attendance.

 Results using Poisson fixed-effect and negative binomial estimation suggest that men under age 50 appear to attend religious services less frequently when out of the labour force. I ascribe this finding to younger men’s religious service attendance being related to having work or the pursuit of work. Men between ages 50 and 65 attend religious services less frequently when out of the labour force, which I attribute to serious health problems in later age forcing labour market exiting and reduced frequency of religious service attendance. Women between ages 50 and 65 attend religious services more frequently when out of the labour force, which I ascribe to having more time to pursue religious activity in addition to women’s established proclivity to religious commitment.

Keywords:

• labour
• religion

JEL Classification:

• J00
• C29

Acknowledgements

I would like to thank David Mustard, Ronald Warren Jr., Christopher Cornwell and the UGA Department of Economics for helpful feedback throughout the development of this article.

Notes

¹ Other reasons for dropping out of the labour force such as moving to a new state exist. Studies such as Smith-Lovin and Tickamyer (1978), Pencavel (1986), Killingsworth and Heckman (1986), Currie and Madrian (1999), Altonji and Blank (1999) and Clark et al. (1979) substantiate the specific reasons I mention for dropping out of the labour force with the exception of high nonwage income and retirement. Retired individuals by definition are out of the labour force while I argue that high nonwage income earners, on average, are more likely to drop out of the labour force than individuals without high nonwage incomes.

² Studies such as Brown (2009), Sacerdote and Glaeser (2001), Lehrer (2004), Becker and Hofmeister (2001), Johnson et al. (1991) and Musick et al. (2004) provide evidence substantiating this claim with the exception of retirement. Because individuals tend to retire at a later age and age and the frequency of religious service attendance are positively correlated as in Iannaccone (1998), I contend that there is a positive relationship between retirement and the frequency of religious service attendance.

³ To my knowledge, no study exists exploring the frequency of religious service attendance as a determinant of labour force status or time spent out of the labour force.

⁴ I discuss these controls further in Section III.

⁵ The following examples are measures that were pursued as instrumental variables but were not employed due to data restrictions, missing data, inconsistent measurement across both data sets or incomplete data: job search activities and intentions, educational loans or financial aid, participation in government programs, extent spouse worked in the previous year, local cost of living changes, local or state tax law changes, changes to local or state labour laws, size of birth cohort, size of graduation cohort when education is completed, local workforce diversity, number of siblings, density of religious organizations in the local or state area and reason for leaving job.

⁶ The examples of instruments are my own except for number of siblings, which comes from Smith-Lovin and Tickamyer (1978).

⁷ The panel is left unbalanced because balancing would result in an additional loss of 6272 observations.

⁸ For the conversion, I set ‘not at all’ equal to 0, ‘infrequently’ equal to 6, ‘once per month’ equal to 12, ‘2–3 times per month’ equal to 30, ‘once per week’ equal to 52 and ‘greater than once per week’ equal to 78. Varying this conversion where appropriate does not significantly impact the results of this study.

⁹ Data on time spent out of the labour force is originally in number of weeks in the past calendar year. I convert the measure to number of months in the past calendar year to be consistent with the Health and Retirement Study.

¹⁰ I convert income to real 2008 dollars to match the real income data I obtain from the Health and Retirement Study.

¹¹ I replace income values of ‘0’ with ‘1’ to use the log of income.

¹² According to Cameron and Trivedi (2010), overdispersion refers to when the conditional variance of the dependent variable is greater than the conditional mean. Because the conditional mean and variance do not usually vary substantially from the unconditional mean and variance, I evaluate the unconditional mean and unconditional variance of the dependent variable for evidence of overdispersion as in Cameron and Trivedi (2010). Overdispersion is a violation of the equidispersion property of the Poisson distribution where the conditional variance and conditional mean of the dependent variable are assumed to be equal. To address overdispersion, I utilize pooled negative binomial and fixed-effect Poisson estimation with clustered-robust SEs. I discuss this further in Section III and Appendix 2.

¹³ See Appendix 1 for details on the data source.

¹⁴ For the conversion, I set ‘not at all’ equal to 0, ‘1 or more times a year’ equal to 9, ‘2–3 times a month’ equal to 30, ‘once per week’ equal to 52 and ‘greater than once per week’ equal to 78. Varying this conversion where appropriate does not significantly impact the results of this study.

¹⁵ Appendix 1 provides the details on the out of labour force since last job designation and the construction of months out of the labour force.

¹⁶ I replace income values of ‘0’ with ‘1’ to use the log of income.

¹⁷ A similar measure of overall health does not exist in the NLSY79 for the years 1979, 1982 and 2000.

¹⁸ See Appendix 2 for a longer discussion of why I use pooled negative binomial and Poisson fixed effects to address overdispersion and why I use Poisson fixed effects instead of negative binomial fixed effects in the traditional sense or the Hausman et al. (1984) sense.

¹⁹ Results in the first and third columns of Table 3, which I consider less reliable because they do not control for student status or unobservable heterogeneity, do show a negative relationship between time spent out of the labour force and religious service and no relationship between being out of the labour force and religious service attendance. The results imply just being out of the labour force is not associated with changes in religious service attendance. Instead, it is the length of time spent out of the labour force that matters when it comes to changes in religious service attendance.

²⁰ To interpret coefficients for a dummy variable in a count data estimated model, coefficients must be adjusted, according to Cameron and Trivedi (2010, p. 343–346). To adjust a dummy variable coefficient, , the following operation is performed: (exp()−1). For example, in the first column of Table 3, the adjustment for the coefficient for months out of the labour force for men is (exp(0.02)−1) = 0.02, which is 2% in percentage form.

²¹ As Greene (2004) argues, I employ random-effect ordered probit because fixed-effect ordered probit yields inconsistent coefficient estimates. Additionally, a consensus for an adequate estimator for fixed-effect ordered logit does not exist, according to Baetschmann et al. (2011) and Dickerson et al. (2011).

²² See Miller and Hoffman (1995), Thompson Jr. (1991) and Collett and Lizardo (2009) for a few of many examples.