The Estimation of Compensating Wage Differentials: Lessons From the Deadliest Catch

ABSTRACT

I use longitudinal survey data from commercial fishing deckhands in the Alaskan Bering Sea to provide new insights on empirical methods commonly used to estimate compensating wage differentials and the value of statistical life (VSL). The unique setting exploits intertemporal variation in fatality rates and wages within worker-vessel pairs caused by a combination of weather patterns and policy changes, allowing identification of parameters and biases that it has only been possible to speculate about in more general settings. I show that estimation strategies common in the literature produce biased estimates in this setting, and decompose the bias components due to latent worker, establishment, and job-match heterogeneity. The estimates also remove the confounding effects of endogenous job mobility and dynamic labor market search, narrowing a conceptual gap between search-based hedonic wage theory and its empirical applications. I find that workers’ marginal aversion to fatal risk falls as risk levels rise, which suggests complementarities in the benefits of public safety policies. Supplementary materials for this article are available online.

KEY WORDS:

• Applications and case studies
• Exploratory data analysis
• Model selection/variable selection.

ACKNOWLEDGMENTS

The author is grateful for dissertation guidance from George Jakubson, Lawrence Kahn, Nicholas Kiefer, and Kosali Simon, for helpful comments from John Abowd, David Card, Raj Chetty, David DeAngelis, Jed DeVaro, Brian Dillon, Francesca Molinari, Jöerg Stoye, Nicolas Treich, and seminar/conference participants at Boston College, Cal State East Bay, Collegio Carlo Alberto, the Congressional Budget Office, Cornell, Harvard, Ohio State, Penn State, Rice, SOLE, Stanford, and Toulouse, and to Jennifer Lincoln (NIOSH), Devin Lucas (NIOSH), and Cathy Tide (ADFG) for generously sharing data.

Notes

1 See, for example, Card, Heining, and Kline (2013), Krueger and Summers (1988), Murphy and Topel (1987, 1990), Gibbons and Katz (1992), Abowd, Kramarz, and Margolis (1999), and Taber and Vejlin (2016).

2 See Gibbons and Katz (1992) for empirical evidence suggesting that endogenous job mobility biases estimates of compensating differentials.

3 See Bonhomme and Jolivet (2009), Dey and Flinn (2008), and Lavetti and Schmutte (2018) for discussions of identification of compensating wage differentials under frictional search.

4 The fisheries studied are not unique for having spot labor markets. Beaudry and DiNardo (1991) found insignificant and wrong-sided evidence of implicit contracts in the forestry and fishing industry using CPS data, in contrast to most other industries.

5 Source: CDC. State-Specific Smoking-Attributable Mortality and Years of Potential Life Lost — United States 2000–2004. MMWR 2009;58:29–33.

6 See Greenstone et al. (2014).

7 Such as Kniesner, Viscusi, and Ziliak (2010).

8 This assumption is implicit in nearly all research on the value of statistical life (VSL) that describe “The VSL” as a fixed value.

9 Of course, policies that consider only efficiency could have potentially inequitable implications. For example, if willingness to pay for safety is increasing in income, a pure efficiency argument would imply spending more public resources per life saved in high income areas. Although the White House’s Office of Management and Budget requires federal agencies to use estimates of the VSL in calculating the benefits of significant safety policies (See Federal Register Vol. 67, No. 60, March 28, 2002, p. 15044 and Executive Order 12866) federal policies tend to use uniform VSLs rather than adjusting for demographics, where efficiency conditional on this equitable restriction occurs when the marginal cost of safety equals the average marginal willingness to pay. However, the choice of VSLs differs substantially across policies, providing scope for policies to consider the average characteristics of the affected populations, including the baseline mortality rate of the population.

10 See, for example, Hwang et al. (1998), Sullivan and To (2014), Dey and Flinn (2008), and Bonhomme and Jolivet (2009).

11 The model is also related to the hedonic search models studied by Sullivan and To (2014), Dey and Flinn (2008), and Bonhomme and Jolivet (2009).

12 Specifically, sufficient assumptions are that indirect utility is additively separable in log wages and amenities, as in Hwang et al. (1998), and log unit labor costs are additively separable in wages and the cost to the firm of providing amenities.

13 In 1998, the Coast Guard implemented a mandatory Dockside Stability and Safety Compliance Check Examination in BSAI fisheries. The examination reviewed the engineering designs of each vessel to assure that they were sufficiently stable given the fishing gear on deck and had proper lifesaving equipment on board before they were permitted to begin fishing. In January 2005, the F/V Big Valley skipped the mandatory inspection and left port with far more crab pots on deck than the stability requirement would have allowed, and subsequently sank killing five of the six crew members. This event contributed to a 2006 review and revision of the examination policy to improve its effectiveness.

14 Worker earnings are actually equal to a share of revenue net of certain limited costs such as food and fuel, so the inferred revenue is this net amount.

15 The sampling frame for this survey was derived from an Alaska Department of Fish and Game (ADFG) database of commercial fishing licenses sold, which provided the name, mailing address, and location of license purchase for every commercial fishermen in the state since 1988. However, the mailing addresses in the database were several years old, and many deckhands had moved and the surveys were not forwarded.

16 A copy of the survey instrument is available at http://www.kurtlavetti.com/research.

17 There are no deckhands in the sample that were not U.S. residents at the time of survey.

18 This is a weighted average, weighted by job-spell duration measured in days.

19 Historical weather data come from NOAA weather buoy station #46035, located in the Bering Sea, which provides hourly weather data dating back to 1985.

20 Although the AKM model is more common in the labor economics literature, I report the match effects specification because it produces virtually identical results, requires weaker assumptions, is easier to conceptualize than a two-way fixed effects model.

21 See Arnould and Nichols (1983), Olson (1981), Dorsey and Walzer (1983), for example.

22 Note that this conclusion does not violate the convexity of preferences. An example of convex preferences consistent with this finding would be if indifference curves in Figure 2 become flatter at high levels of risk for a given wage.

23 Dickstein and Morales (2013) called this type of error “expectational error,” and showed that it creates attenuation bias very similar to measurement error.

24 Note that weather is not used as an instrumental variable. The observed fatality rate is not endogenous, nor is weather excludable from the wage equation.

25 Deckhands typically arrive to help prepare a vessel about 2 weeks prior to departure, and the contract terms are generally set in advance of arrival since transportation to the vessel is often expensive.

26 See Lincoln and Conway (1999).

27 The ISS is a modified measure of the Abbreviated Injury Score that takes into account the severity of multiple simultaneous injuries to different regions of the body.

28 Hospital costs are based on up to three hospitalization spells per injury, and data on the source of payments is available for up to two of those spells.

Additional information

Funding

The author is also grateful for financial support from the National Science Foundation (SES-0851605), the Graduate School at Cornell University, and the Robert Wood Johnson Scholars in Health Policy Program at UC Berkeley.