Workplace safety regulations and worker well-being: the impact of mandatory helmet use on career length in the national hockey league

ABSTRACT

The long-term health problems resulting from professional careers in contact sports such as hockey and football have recently received lots of attention. The lawsuits initiated by NFL- and NHL-veterans suffering from repetitive concussions of the brain that eventually lead to chronic traumatic encephalopathy are two particularly prominent examples. The question, whether these long-term health problems could have been avoided, remains a highly contested issue. Using detailed career information on every single player drafted in the ten seasons before (1969 thru 1978) and the ten seasons after (1979 thru 1988) the implementation of mandatory helmets, we find that career length in professional hockey has increased by about two years, suggesting that adequate protective equipment contributes to player health and fitness. This positive effect, however, is somewhat reduced by specific forms of ‘offsetting behaviour’.

KEYWORDS:

• Workplace safety
• exogenous intervention
• worker well-being

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. In an already large (and still increasing) number of studies, sports medicine specialists have convincingly documented the (potentially) detrimental impact of recurrent concussions on hockey players’ long-term health (see e.g. Benson et al. 1999, Benson 2002, Benson et al. 2011, Gavett et al. 2011, Johnson 2011, Smith et al. 2011, Adams et al. 2015, Hutchinson et al. 2015).

2. Brandner et al. (2014) demonstrate that peak age is between 27 and 28 for forwards and 28 and 29 for defenders and that both types of skaters exhibit near-peak performance over a wide range going from about 24 to about 34.

3. A number of studies (such as e.g. Conybeare 1980, Crandall and Graham 1984, Traynor 1993, Keeler 1994, Peterson et al. 1995) find that mandatory seatbelts not only reduce traffic fatalities, but also lead to changes in driver behaviour (e.g. induce ‘reckless driving’). Dee (2009) as well as Mayrose (2008) confirm the existence of an offsetting effect of mandatory motorcycle helmets on traffic fatalities. Finally, Sobel and Nesbit (2007), looking at changes in driver behaviour in stock car racing, find offsetting behaviour in NASCAR in the sense that drivers do drive more recklessly in response to the increased safety of their automobiles. Total injuries, however, still decrease because this effect is not large enough to completely offset the direct impact of increased automobile safety.

4. The literature on risk compensation and risk homoeostasis is to a certain extent close to the concept of ‘sensation-seeking behavior’ (Zuckerman 2007; for a recent empirical application Frick 2020).

5. If anything, ticket demand in the NHL has increased over time, suggesting that the introduction of mandatory helmets has not made the sport less attractive to fans. Rockerbie (2012, 2016, 2017)) fails to find a significant (positive) impact of fighting on ticket sales and Haisken-DeNew and Vorrell (2008) find substantial returns not only to goal scoring, but also to fighting ability only for the first half of their observation period (the late 1990s), but not for the second half (the first years after the turn of the millennium). Taken together, these findings suggest that fans do not have a pronounced preference for violent play (any longer).

6. Thus, the majority of the players selected in the draft never played in the NHL but spent their entire careers in either one of the minor leagues in North America (such as the American Hockey League (AHL)) or in one of the professional leagues in Europe.

7. The data used to estimate the models presented below (the results of corresponding parametric survival models are available from the author upon request) was retrieved from www.hockeydb.com.

8. A closer look at the data reveals that 97% of the players are between 18 and 20 years old by the time they are drafted which, in turn, explains the low standard deviation.

9. This is not necessarily indicative of poorer player health, but suggests that with increasing roster size players may receive more time to recover from fatigue and injuries. Controlling for draft round, age, height, weight, position, year drafted and club dummies, the actual number of regular season games does not differ between the two regimes (the estimation results are, of course, available from the author upon request). However, the potential number of regular season games as well as the number of teams qualifying for the playoffs has increased during our period under investigation (see Figure A1 in the Appendix).

10. Whether this increase is due to more aggressive play or to stricter rule enforcement by the referee(s) remains to be tested.

11. The test statistics are Chi2 = 8.19, p < .01 for the complete and Chi2 = 7.86, p < .01 for the reduced sample.

12. In addition to a Cox hazard model, we have also estimated parametric hazard models. Parametric hazard models exploit the information about career length differently than Cox hazard models, and can be interpreted as regression models (Cleves et al. 2008). The specific form of parametric hazard models depends on the distribution of the dependent variable, i.e. career length. In our setting, the Akaike Information Criterion (Akaike 1974) indicates that a Gompertz model is to be preferred over its alternatives such as e.g. the log-logistic or Weibull models. The results of these models are, of course, available from the author upon request.

13. Estimating the models presented in (Tables 4(a,Tables 4b)) with the players’ body mass index instead of body height and weight leaves the coefficient(s) of interest completely unaffected. The detailed results are, of course, available from the author upon request.

14. The average annual player salary in the 1977–1978 season was 100,000 $ (Jones and Walsh (1988: 601), 200,000 $ in the 1989–1990 season (own calculation based on information on 520 players provided by Rod Fort on his website) and has increased to 2.9 million $ in the 2015–2016 season.

15. If we further restrict the data to two years around the introduction of the mandatory helmet, the coefficients again retain their sign, magnitude and statistical significance in the reduced sample with 315 observations (in the complete data set with 385 observations the coefficients also retain their magnitude, but lose their statistical significance).

16. The regression discontinuity design assumes that the trend of an outcome shifts following an exogenous policy intervention. Assuming that the environment before and after the policy change is the same, any observed difference – here in career length – can be attributed to the policy intervention.

17. As player salaries have increased more than the salaries in the players’ next best alternative, the opportunity costs of exiting professional hockey have increased too. In order to control for this development, we add year of entry dummies in models 4.1 and 4.2 (Table 4(b)). Although this reduces the magnitude of the helmet dummy, the coefficient retains its statistical significance.