Physical activity guidelines for Americans and their employment decisions

ABSTRACT

The Physical Activity Guidelines recommend Americans improve their health through sufficient aerobic exercise. Given the required energy and time commitment of the Guidelines, we consider its unintended consequence to market work in the US. Using 2011 Behavioural Risk Factor Surveillance System (BRFSS) data, we find adhering to the Guidelines actually raises the chance of employment by roughly 10 percentage points, not primarily because of health but likely greater socio-emotional skills.

KEYWORDS:

• Physical activity
• employment
• noncognitive skills
• bivariate probit

JEL CLASSIFICATION:

• C35
• I12
• J22
• J24

acknowledgment

We thank Christopher Patillo for research assistance through the Undergraduate Research and Creative Activity program and Ariel Belasen for constructive comments and proof-reading manuscript drafts.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ See Lechner and Downward (2017) for a review of the literature and a comparison of methodologies.

² Only the 2011 survey has publicly available county identifiers and questions assessing PA duration and frequency.

³ This model consistently estimates the overall (not local) average treatment effect and outperforms competing methods in simulation exercises based on treatment and outcome probabilities like ours (Bhattacharya, Goldman, and McCaffrey 2006).

⁴ We use all states (except Alaska because county identifiers unavailable) and the District of Columbia.

⁵ Weather data sources are CDC Wonder and the National Centres for Environmental Information.

⁶ We use establishments classified by NAICS as: Sports and Recreation Instruction, Golf Courses and Country Clubs, Skiing Facilities, Marinas, Fitness and Recreational Sports Centres, and Bowling Centres. Huang and Humphreys (2012) computed establishment density similarly in their application.

⁷ A potentially key variable unavailable in the 2011 BRFSS is hours of caregiving, which makes it costly to engage in PA and labour market activity. While we control for several variables related to caregiving hours, omitting this variable could still bias our estimates if it makes the exclusion restriction (that county fitness/sports supply is uncorrelated with ${ϵ}_2$) untenable. To provide some suggestive evidence against this possibility, we use caregiving data from the 2009–2010 BRFSS available for four states (Connecticut, Illinois, Louisiana, New Hampshire) and the District of Columbia (DC), where about 20% of adults are caregivers and nearly 30% of which provide care for at least 20 hours (the mean) – comparable to CDC reports based on 2015–2016 BRFSS for 38 states and DC. We find the simple correlation between county fitness/sports supply and county-average caregiving hours is –0.147 ($p=0.065$), but approaches zero (–0.045, $p=0.598$) after partialling out our county controls and state fixed effects.

⁸ A weighted proportion of about 20% of adults are at or near retirement (55–64 years old) and about 62% of them are employed. We find their PAG effect is 0.226 ($s.e.=0.079)$. This is not surprising if PAG greatly improves their fitness for work. In fact, 55–64-year-olds have more health problems (especially, multiple chronic ones) than younger adults, which implies adhering to PAG can substantially improve health outcomes (see Sari 2014).

⁹ Interestingly, accounting for health factors decreases the PAG effect for prime-working age adults from 0.090 to 0.072 and for older working-age adults from 0.226 to 0.092, albeit with much loss in precision. Thus, health factors appear to largely explain why PAG increase employment prospects for 55–64-year-olds, but not for all working-age adults.

¹⁰ See https://www.kornferry.com/press/new-research-shows-women-are-better-at-using-soft-skills-crucial-for-effective-leadership, last accessed 22 February 2020.

¹¹ We would have liked to provide estimates by occupation type, as noncognitive skills are more relevant in some occupations, but such data are unavailable.