Marital status and productivity: evidence from personnel data.

• Author: Mehay, Stephen L.

1. Introduction

   Although the existence of a wage premium for married men has been documented extensively in the labor economics literature, the source of the premium has remained controversial. One hypothesis attributes the wage differential to economies of scale and specialization within the household, which provides incentives for additional investment in specific human capital by married men (Becker 1981). Alternative hypotheses are that the marriage premium stems from employer favoritism or that more productive or higher earning men are selected into marriage. The extensive empirical literature that has attempted to test the competing explanations has produced conflicting conclusions. However, one research approach to assess the effect of marital status that seldom has been implemented in the prior literature is to use data on the realized productivity of employees of a single firm.

   This study uses administrative data on male employees of a large, hierarchical organization to explore the effect of marital status on selected job performance measures. The personnel database covers male U.S. Naval officers and contains relatively detailed information on career performance, including promotion outcomes and annual performance reviews. The article seeks to shed light on the existence and magnitude of job-productivity differentials between married and single males and to assess competing explanations for marriage-related performance differentials.

   Navy personnel data are especially fertile for analyzing the effect of marital status on employee job productivity. Several performance measures are available for each individual, and the indicators capture performance throughout an employee's tenure rather than at a single career point. The design also controls for key aspects of the organization's internal labor market--career ladders, job assignment, and promotion policies, for example--that normally are unobserved in national surveys. In addition to reducing firm heterogeneity, the data also minimize individual heterogeneity by controlling for differences in jobs, occupations, and labor supply, all of which may differ by marital status.

   The article finds that married men receive significantly higher supervisor performance ratings than single men during their early career with the organization and, for those who stay, are more likely to be promoted at the up-or-out review. Although controlling for selection arising from quit decisions reduces the estimated marriage effect, it remains robust for most subgroups. In addition, supplementary tests suggest that selectivity in the decision to marry explains only a small portion of the unadjusted marriage premium for the largest group of (line) officers and none of the premium for staff officers.

2. Background

   Most prior studies attempt to assess the relative strength of the selection versus the household specialization explanation of the marriage premium. Korenman and Neumark (1991) obtain an 11% wage premium for married men in cross-sectional estimates but only a 6% premium in fixed-effects estimates. They find that the premium increases with each year of marriage and that less than 20% of the marriage premium is associated with fixed unobservable characteristics that are positively correlated with both marriage and wages. In sharp contrast, Cornwell and Rupert (1997) find that marriage tenure has no effect on wages. Moreover, they find that current earnings of single men who will marry in the future exceed those of currently married men, suggesting that married men possess characteristics valued in both marriage and labor markets.

   Several studies attempt to test Becket's hypothesis that intrahousehold specialization provides additional incentives for married men to invest in specific human capital. Becker's theory implies that fewer opportunities for specialization exist in households in which the wife works. In support of this prediction, Chun and Lee (2001) find that the wage premium of married men decreases with the number of labor market hours worked by wives. Loh (1996), on the other hand, shows that married men whose wives work receive a larger wage premium than men whose wives do not work, holding marriage duration constant. Jacobsen and Rayack (1996) find that, when marital status is treated as endogenous, the earnings differential between husbands with and without working wives disappears. Hotchkiss and Moore (1999) verify Jacobsen and Rayack's conclusion for nonmanagers but not for managers. They find that wives' labor supply is exogenous for managers and those with working spouses earn lower wages than those with nonworking wives. Hersch and Stratton (2000) find that, when direct indicators of household specialization are used in place of wives' market work hours, the male marriage premium is not affected. They conclude that household specialization cannot be the source of the marriage premium.

   Korenman and Neumark (1991) provide the only prior study to use administrative data from a single firm. They find that married male employees in the firm earn 25% more than single men. They attempt to determine whether the higher job grades occupied by married men are explained by their longer tenure or by their higher productivity. Their results show that married new hires are less likely to quit and more likely to be promoted, leading them to conclude that the promotion differential and the resulting difference in job grades are due to consistently higher performance ratings received by married men.

   The conflicting results in prior wage studies may stem in part from unmeasured firm heterogeneity. (1) Key aspects of a firm's personnel policies and internal labor market typically are not controlled in studies that rely on national surveys. Although Korenman and Neumark (1991) overcome these weaknesses by using internal firm data, they are not able to account for nonrandom selection among employees to stay to important career points. (2) Also, their promotion analysis may be subject to measurement error because they cannot identify whether workers are qualified for promotion consideration. In contrast, our data minimize the effects of unmeasured firm heterogeneity, allow us to account for selection due to quit behavior, and restrict the analysis of promotion outcomes to employees who are eligible for promotion.

3. Personnel Data and Job Performance Models

   Our personnel data capture cohorts that began their careers after college graduation as officers in professional, technical, and managerial positions in the U.S. Navy between 1977 and 1985. The database approximates a longitudinal file by building employment histories for men observed initially at the grade 3 promotion point at approximately three years of service. Each cohort member is tracked through his first 10 years of service or until he separates. Data are available for 26,385 line and 4283 staff personnel, who range in age from their early 20s at entry to their early 30s at the later career stage. (3)

   The Navy's personnel system displays features of a classic internal labor market. The hierarchy consists of 10 grades and all new hires begin employment in the entry grade (grade 1). Upon graduation from college, officers choose to enter staff or operational (line) career fields. Staff officers have a relatively short formal training period and fill mostly administrative and support jobs, such as in the supply corps. Line officers receive more extensive formal training and fill jobs in aviation, on ships, or on submarines. New entrants in each career field receive the same amount of formal training and incur a service obligation of either four or five years, depending on commissioning program. There is no lateral entry and all promotions are from within the organization.

   An important feature of the personnel system is rank-order promotion tournaments (see Asch and Warner 2001). Individuals spend two years in grade 1 and two years in grade 2 and promotion during this four years is nearly automatic. Those who stay beyond the initial service obligation spend six years in grade 3 and undergo an up-or-out review for promotion to grade 4 at the 10-year point. Promotion targets for each major occupational specialty are based on vacancies at the next highest grade. Officers reviewed for promotion are evaluated on the basis of previous jobs, skill qualifications, demonstrated leadership and achievements, and prior annual performance ratings. Those who are promoted effectively are granted tenure and are guaranteed at least a 20-year career and a pension.

   In our job performance models, an individual's relative performance is specified as a function of his stock of accumulated human capital (Barrel 1995). In the case of professional and managerial employees, cognitive skills are assumed to depend on college background, including grades and major. Affective skills also have been identified as important contributors to job performance of a firm's employees (Wise 1975). Affective skills consist of work-related attributes such as perseverance, self-discipline, leadership, initiative, and the ability to cooperate, traits that are especially valuable in the military's team production environment. Performance models are estimated separately for the line and staff occupational groups, each of which has unique training requirements, career paths, qualification standards, and promotion opportunities. For example, career progression differs in each field due to the sequence and types of jobs and applicable qualification standards.

   We construct three separate performance indicators. (4) The first two are based on annual performance ratings from each individual's supervisor. The evaluation consists of 25 graded items and other items such as a "recommendation for accelerated promotion" (RAP). Although the distribution of grades on the 25 items tends to be concentrated at the upper end of the scale, the incidence of a...