Effect of religious attendance on the middle-aged worker’s wage in the United States: a possible causal connection

ABSTRACT

The current study explores the possibility of a causal connection between a worker’s religious attendance and wage. Numerous studies in the literature have already demonstrated a positive correlation between these two variables. For effective policy applications, however, it is necessary to find out whether there exists a causal connection between them. The study uses the US data and tests the hypothesis that religious attendance of middle-aged workers has a positive causal effect on their wages. Following an instrumental variable approach, the current study tests this hypothesis and demonstrates that religious attendance is very much likely to have a causal effect on middle-aged workers’ earnings. This has important policy implications and thus deserves further attention from researchers and policymakers.

KEYWORDS:

• Religious Attendance
• wage
• causal connection
• Instrumental variable

JEL CLASSIFICATION:

• J24
• J29
• J31
• Z12

Acknowledgments

The author wishes to thank the editor, three anonymous reviewers, Marco Del-Angel, Xiaohan Zhang, Sunil Sapra, seminar participants at California State University-Los Angeles and Sri Sathya Sai Institute of Higher Learning, India, and discussants of the Southern Economic Association Conference at Fort Lauderdale, Florida, for valuable suggestions on an earlier draft of this paper. The usual disclaimer applies.

Disclosure statements

No potential conflict of interest was reported by the author(s).

Data availability statement

For information, regarding the data and/or computer programs used for this study, please contact the author at [email protected].

Notes

¹ Note that the data on religious attendance is not available in all surveys of NLSY79. After the 1979 and 1982 surveys, the next survey in which this information is available is in the year 2000. Although a few other later surveys (2012, 2014, 2016 and 2018) include the information on this variable, the non-response rate on church attendance is so high that it is unreliable to be used for examining its relationship with wage. The number of respondents with valid response to the religiosity question in the 2014 survey is 293 only. It is only 128 in both 2016 and 2018 surveys even though the sample size with valid observations on all other variables in each survey is above 5000. Although there are a reasonable number of respondents with answer to the religiosity question in the 2012 survey, the relationship between church attendance and wage as shown in Appendix I is statistically insignificant.

² In fact, following a mediation analysis, Mohanty (2022) in a different context has demonstrated that church attendance by promoting psychological capital endowments is likely to enhance an individual’s earnings.

³ For older adults with shorter working career ahead of them, the present value of returns to their education is lower than that of younger adults. Moreover, older adults may have already completed their desired years of schooling and consequently their religious attendance may not have any significant effect on their school attendance. The positive productivity effect of religious attendance for this group of workers therefore is likely to be stronger than their negative part-time work status effect, leading to a net positive effect of religious attendance on their wages (Mohanty 2020).

⁴ Our study follows the line of research that associates religious attendance with positive psychological progress, leading to higher productivity (Goldsmith, Veum, and Darity 1997, 2000; Mohanty 2013), which in turn is likely to shift the production function upwards exactly the same way as the technical progress does in standard production theory (Becker 1993).

⁵ With rise in satellite families in twenty-first centuries in which wife and husband often live in different regions of a country, universal prevalence of assortative mating theory in marriage, although not impossible, remains questionable.

⁶ See Hill, Griffiths and Lim (Hill, William, and Lim 2011, 420–421) for a detailed textbook discussion on the surplus moment conditions and the procedure associated with the over-identification test.

⁷ Use of NC as an instrumental variable for some adult characteristics is not new in the literature. In different contexts, MacDonald and Shields (2001) and Ours and Jan (2004) have used ‘number or presence of children’ as an instrumental variable for adult alcohol consumption and drug use.

⁸ With the binary RASP as the dependent variable in equation (A1), the natural temptation is to estimate the first stage equation by probit. This, however, is likely to result in serious specification error. See Angrist and Krueger (2001, p. 80, footnote 10) for a detailed explanation.

⁹ The number of individuals in this category is quite small compared to the size of the sample.

¹⁰ We thank a reviewer for suggesting the inclusion of spousal characteristics in both first stage and second stage regression.

¹¹ $\mathit{F}=\frac{\left(380.44-223.21\right)/1}{223.21/3385}=2384.4$.

¹² $\mathit{F}=\frac{\left(380.44-223.19\right)/2}{223.19/3384}=1192.11$.

¹³ We thank a reviewer for suggesting Hansen–Sargan test to verify the validity of our instruments. Some statistical software, such as STATA, routinely report this statistic. Since the econometric software LIMDEP (Greene, 2016), which is used for estimating relevant equations in this study, does not report this statistic, we had to estimate it directly following the standard test procedure (see Appendix II).