County-level Industry Job Shares and Alcohol Misuse/Alcohol Use Disorder among Working-Aged Individuals in the United States, 2015-2019

Abstract

This study employs an instrumental-variable approach to examine how county-level job shares influence alcohol use among working-aged individuals in the United States. Higher job shares in finance and other services are associated with greater risks of binge drinking and alcohol use disorder, whereas higher shares in construction and trade sectors appear to confer protective effects.

Short Summary

Higher county-level job shares in finance and other services are associated with greater binge drinking/AUD risk among working-aged individuals, whereas higher construction job shares appear protective overall, suggesting that greater availability of decent employment in the communities may offset potential alcohol-related risks from a higher proportion of workers in this high-risk sector.

Introduction

Alcohol misuse is a preventable public health issue, causing over 178,000 deaths annually in the United States (Centers for Disease Control and Prevention [CDC], 2024). As the most common form, binge drinking—consuming 5+ drinks (men) or 4+ drinks (women) on a single occasion—accounts for the majority of deaths and serious health risks from excessive alcohol consumption, including fatal motor vehicle accidents, chronic diseases (e.g., liver damage, cardiovascular diseases, and various cancers), and alcohol use disorder (AUD) (CDC, 2024). Moreover, binge drinking imposes significant economic and societal costs, amounting to an estimated annual cost of $171 billion for lost productivity and healthcare (Sacks et al., 2015). Despite these consequences, binge drinking remains prevalent especially among working-aged individuals, with past-year prevalence ranging from 14.3%-24.0% for binge drinking and 9.8%-23.4% for AUD across age (Bohm et al., 2021, Grant et al., 2017).

Among various multi-level determinants, the influence of local labor market contexts remains underexplored. In communities where high-risk industries—such as corporate/financial sectors or those with a disproportionately male workforce (McKetta et al., 2021, Roche et al., 2015)—are prevalent, a larger portion of the workforce is exposed to associated risk factors, increasing alcohol misuse/AUD risks. For instance, individuals in corporate/financial management roles report a higher risk of engaging in alcohol misuse, partly as a coping mechanism for high-pressure work environments and stress, as well as due to higher disposable incomes and greater opportunities for alcohol-involved social/professional activities (McKetta et al., 2021, Kahn et al., 2001). Similarly, workers in construction/extraction, arts/entertainments/recreation, and food preparation/serving sectors report elevated rates of binge drinking, driven by factors including self-medication for chronic pain, workplace alcohol exposure, and permissive drinking norms (Leadbeater et al., 2020, Roche et al., 2015). While research on AUD by industry is limited, recent studies suggest higher prevalence in construction, wholesale trade, arts/entertainment/recreation, and lower prevalence in manufacturing and education/health/social services (Park et al., 2024).

Local labor market contexts can also indirectly influence residents who are not employed in high-risk industries. These effects may arise from interactions with family members, friends, and neighbors who misuse alcohol, as well as from broader community characteristics that either promote or discourage alcohol use. Prior research has shown that alcohol use of spouses, romantic partners, and social networks significantly impact alcohol use (Rosenquist et al., 2010). Additionally, differences in earnings and employment stability across industries shape the socioeconomic conditions of communities and the overall send of hope and disorder, influencing alcohol misuse/AUD (Bryden et al., 2013).

Despite the potential roles of local labor market contexts, gaps remain in our understanding of their relationship with alcohol misuse/AUD. First, prior studies have mainly focused on alcohol use among individuals within specific occupations, leaving the influence of broader industry sectors on alcohol-related risks unclear. While occupational differences reflect specific jobs and tasks individuals perform that often exist in different industries, sectoral differences capture broader work environments and industry-wide alcohol norms (Thern et al., 2024). Second, few studies have specifically examined the relationship between work and alcohol misuse/AUD across industries (Roche et al., 2015) and most of these studies are based outside the U.S. Understanding these sectoral differences is crucial for identifying areas represented by high-risk industries, guiding resource allocation and developing geographically targeted prevention and intervention strategies. Lastly, many studies rely on descriptive or traditional regression approaches, which are prone to biases from various sources, such as omitted variables and reverse causality (Betz et al., 2018). When there exist unmeasured confounders that simultaneously correlate with both local industry job shares and alcohol-related outcomes, the estimates are likely to be biased.

To address these gaps, the present study leverages county-level job share data for 20 industry sectors, categorized using the North American Industry Classification System (NAICS), alongside binge drinking/AUD data from a nationally represented survey. By linking these datasets through geographic identifiers and employing instrumental-variable (IV) regression, this study mitigates potential biases arising from local conditions that simultaneously affect industrial composition of local economies and individuals’ alcohol use behaviors, offering insights into the ecological effects of the local labor market context on alcohol-related outcomes.

Methods

Data and Sample

This study utilized pooled data from the restricted-use National Survey on Drug Use and Health (NSDUH) for 2015-2019. Using a multistage area probability sampling of all 50 states and D.C., the NSDUH provides substance use and health-related information of a nationally representative sample of non-institutionalized U.S. civilians aged 12 and older (Center for Behavioral Health Statistics and Quality, 2020). By leveraging county of residence information, we linked the NSDUH data to county- and state-level contextual data on job shares by industry, substance supply factors and policies, and sociodemographic characteristics from multiple sources, including the US Census’ Longitudinal Employer-Household Dynamics Origin-Destination Employment Statistics, the Tax Data Center, the U.S. Drug Enforcement Administration’s National Forensic Laboratory Information System, CDC, and the U.S. Census’s American Community Surveys. Among the NSDUH’s 275,100 working-aged individuals (ages 15-64), the final analytic sample consisted of approximately 273,900, after excluding cases with missing data (0.4% of the total sample).

Measures

Alcohol use outcomes.

The outcome variables are binary measures of past-month binge drinking, and past-year AUD based on the DSM-IV.

Job shares by industry sectors.

The independent variable is the proportion of jobs in each 2-digit NAICS industry sector within county of residence. To capture the portion of industry-specific job share variations unrelated to unobserved local-level confounders, Bartik-style instruments—estimating the changes in industry job shares that would have occurred if local industries had changed at their national rates—were created (Bartik, 1991). These instruments were constructed by multiplying an industry’s job share (measured five years prior to the survey year) by one plus the national average percentage changes in that industry’s job share over the subsequent five years.

Controls.

Models were adjusted for individual characteristics and contextual factors, including county-level sociodemographic characteristics, state-level substance-related policies, and supply factors (see Table S1). Specifically, substance-related policies encompass not only an alcohol-specific policy but also policies on other drugs—particularly marijuana and opioids, two of the most prevalent substances often co-used with alcohol—allowing us to account for their indirect effects on alcohol use.

Analytic Strategy

The statistical analyses were conducted using Stata v. 18. First, we conducted IV probit analyses using Bartik-style instruments for the job share variables, employing the ivprobit command. The IV approach relies on an instrument that is strongly correlated with the independent variable—in this case, local industry job shares—but uncorrelated with the error term. By utilizing only the variation in the local industry job share variables that are not affected by unobserved confounders, the IV method helps mitigate bias from omitted variables. Using maximum likelihood estimation, we estimated the equations below for each outcome and industry separately to capture the effect of a given industry’s share of total employment within the county:

$$y_{ijs}=\beta w_{js}+{\gamma }_k\sum _{k=1}{x}_{ijsk}+u_{ijs}$$

$$w_{js}={\rho }_k\sum _{k=1}{x}_{ijsk}+\pi z_{js}+v_{ijs}$$

where $y_{ijs}$ is a binary indicator of binge drinking/AUD status for individual i living in county j in state s, $w_{js}$ is the industry’s job share of total employment, $x_{ijsk}$ is control variable k, $z_{js}$ is the Bartik-style instrument, and $u_{ijs}$ and $v_{ijs}$ are error terms assumed to follow a multivariate normal distribution, $N(0,\text{Σ}$). To evaluate the relevance of our instruments with the industry job share variables, we conducted F-tests for excluded instruments, which confirmed that our instruments possess strong explanatory power, consistent with Lee et al. (2022). Using the constructed IVs, we estimated two models: Model 1 included only individual-level controls, while Model 2 incorporated both individual- and contextual-level controls to assess the robustness of the estimates. To account for multiple hypothesis testing, p-values for the coefficients were adjusted using the Benjamini-Hochberg procedure (Benjamini et al., 2001). Then, we computed predicted probabilities of having binge drinking/AUD across the observed range of industry-specific job shares in the county of residence. All estimates accounted for the stratified cluster sampling design of the NSDUH for the generalizability of the findings.

Results

Table 1 presents changes in z-scores associated with a 1-percentage point higher job share of each industry sector at the county level. For binge drinking, when controlled for both individual and contextual controls (Model 2), the strongest positive associations were observed in Other Services (Coeff=0.019, p<0.001), followed by Arts/Entertainment/Recreation (Coeff=0.013, p<0.01) and Finance/Insurance (Coeff=0.010, p<0.001), whereas the strongest negative associations were found in Construction (Coeff =−0.013, p<0.001) and A/WM/R services (Coeff=−0.013, p<0.001). For AUD, the strongest positive associations were seen in Real Estate/Rental/Leasing (Coeff=0.033, p<0.01), followed by Other Services (Coeff=0.030, p<0.001), Information (Coeff=0.021, p<0.001) and Finance/Insurance (Coeff=0.016, p<0.001) while the strongest negative associations were observed in Wholesale (Coeff=−0.021, p<0.001), Construction (Coeff=−0.020, p<0.001), and Retail (Coeff=−0.018, p<0.001).

Table 1.

Effects of Industry-Specific Job Proportions in US Counties on Working-Aged Individuals (Ages 15-64)’ Alcohol Misuse and Use Disorder, NSDUH 2015-2019: Instrumental Variable Probit Regression Analyses

	Binge Drinking				Alcohol Use Disorder
	Model 1		Model 2		Model 1		Model 2
Industry Categories	Coeff	SE	Coeff	SE	Coeff	SE	Coeff	SE
Agriculture/Forestry/Fishing/Hunting	−0.001	0.003	0.008*	0.003	0.009**	0.003	0.011**	0.004
Mining/Quarrying/Extractions	−0.001	0.002	0.004	0.002	−0.001	0.003	0.003	0.003
Utilities	−0.006	0.012	0.023	0.012	−0.022	0.015	−0.011	0.015
Construction	−0.019 ***	0.003	−0.013 ***	0.004	−0.023 ***	0.004	−0.020 ***	0.005
Manufacturing	−0.005 ***	0.001	−0.004 ***	0.001	−0.007 ***	0.001	−0.005 ***	0.001
Wholesale Trade	−0.013 ***	0.004	0.001	0.004	−0.031 ***	0.005	−0.021 ***	0.006
Retail Trade	−0.016 ***	0.003	−0.010 **	0.004	−0.024 ***	0.004	−0.018 ***	0.005
Transportation/Warehousing	−0.014 ***	0.003	−0.009 **	0.003	−0.022 ***	0.004	−0.014 **	0.005
Information	0.007	0.004	0.011*	0.005	0.024***	0.005	0.021***	0.006
Finance/Insurance	0.014***	0.003	0.010***	0.003	0.012***	0.004	0.016***	0.004
Real estate/rental/leasing	0.016	0.009	0.015	0.009	0.037**	0.012	0.033**	0.012
Professional/Scientific/Technical Services	0.003	0.002	0.001	0.002	0.007**	0.002	0.006*	0.002
Management	0.016***	0.004	0.009*	0.005	0.006	0.006	0.007	0.006
A/WM/R Services	−0.011 ***	0.003	−0.013 ***	0.004	−0.011 *	0.004	−0.007	0.005
Educational services	0.005***	0.004	−0.001	0.002	−0.001	0.002	−0.003	0.002
Health care/social assistance	0.005***	0.001	0.001	0.002	0.003	0.002	0.003	0.002
Arts/Entertainment/Recreation	0.022***	0.004	0.013**	0.005	0.024***	0.006	0.013*	0.006
Accommodation/Food Services	0.007***	0.001	0.006***	0.001	0.008***	0.002	0.006***	0.002
Other Services	0.028***	0.004	0.019***	0.005	0.039***	0.005	0.030***	0.006
Public administration	0.003	0.002	−0.001	0.002	0.005*	0.002	0.001	0.002

Notes. NSDUH = National Survey on Drug Use and Health. A/WM/R = Administrative/Waste Management/Remediation. The total sample size is rounded to the nearest 100 in accordance with SAMHSA’s data confidentiality agreements. Each row represents a separate regression model estimating the impact of the respective industry’s job proportions on individuals’ binge drinking and alcohol use disorder. Model 1 controls for individual characteristics (age, sex, race/ethnicity, marital status, number of children in household, educational attainment, employment status, household income, metropolitan status of residence, and overall health status). Model 2 controls additionally for county-level sociodemographic (% male, % individuals aged 65+, total population size, % black, % single-parent families, % individuals aged 25+ without a high school diploma or equivalent, % individuals living in poverty, % unemployed) and substance-related (beer tax rate, recreational marijuana legalization status, prescription drug monitoring law status) characteristics. All analyses were adjusted with sampling weights to account for the NSDUH’s complex sampling design. Coefficients for control variables are not provided to conserve space. The coefficients in bold indicate statistical significance after adjusting p-values using the Benjamini-Hochberg procedure to account for multiple hypothesis testing.

^*p<0.05,

^**p<0.01,

^***p<0.001

Figure 1 illustrates the predicted probabilities of binge drinking/AUD across observed job shares for select industries with significant associations. For binge drinking, notable increases were found in Other Services (from 25.5% in no jobs to 35.6% with a 48% job share) and decreases in Construction (from 29.7% with no jobs to 17.8% with a 30% job share). For AUD, notable increases were seen in Real Estate/Rental/Leasing (from 5.2% with no jobs to 10.9% with a 12% job share) and Other Services (from 4.7% with no jobs to 14.6% with a 21% job share). Conversely, notable decreases were observed within Wholesale Trade (from 7.1% with no jobs to 1.8% with a 30% job share) and Construction (from 6.9% with no jobs to 1.8% with a 30% job share).

Figure 1.

Predicted Probabilities of Binge Drinking and AUD among Working-aged Individuals (Ages 15-64), NSDUH 2015-2019: County-Level Job Shares by Industry Sector

Notes. NSDUH = National Survey on Drug Use and Health. A/WM/R = Administrative/Waste Management/Remediation. Predicted probabilities of binge drinking and alcohol use disorder were estimated using the significant coefficient estimates from the IV probit models in Table 1. The job proportion ranges for each industry reflect the actual job proportions observed across U.S. counties. For the estimated probabilities, only coefficients for significant industries greater than 0.01 were presented to enhance the readability of the line figure.

Discussion

Our findings indicate that county-level job shares in certain industries are significantly associated with the risk of working-aged individuals reporting binge drinking/AUD. Specifically, higher job shares in financial and other service sector—primarily composed of occupations such as automotive service technicians/mechanics—were linked to greater risks of binge drinking/AUD, aligning with prior research (Schweitzer et al., 2000, Mandell et al., 1992).

In contrast, individuals in counties with higher construction or trade job shares generally exhibited lower risks of binge drinking/AUD. Understanding the protective effects of the construction sector on residents’ alcohol use warrant a nuanced perspective, given evidence that construction workers report higher alcohol consumption. This finding may reflect broader ecological benefits of a thriving construction industry that extend beyond construction workers. While a higher share of construction workers may be associated with increased alcohol-related risk, the overall protective effects likely stem from benefits to residents employed in supply chain and related service industries, who experience job and income growth, as well as from county-wide gains such as infrastructure development and increased government revenues (Pheng et al., 2019, U.S. Bureau of Labor Statistics, 2024). Ecological-level benefits have likewise been documented in drug mortality research, which identified protective effects of growth in the construction sector on Black residents during the 2010s (Oh et al., 2025) and of low-skill job availability on overall drug mortality (Betz et al., 2018), though further evidence is needed to assess the broader effects of the construction sector on overall drug mortality.

These results should be interpreted with caution. As our results capture the overall relationship between a county’s industry-specific job shares and individuals’ alcohol behaviors, it does not distinguish between direct effects (from having more workers employed in a given sector) and indirect effects (e.g., potential spillover effects on other residents). Further research incorporating individual-level information on employment industry may help to disentangle these effects. Additionally, while our use of IV regression approach helps reduce bias from county-specific unobserved confounders, potential biases due to selection and spillover effects from neighboring counties cannot be ruled out (Canan et al., 2017).

Despite these limitations, this study offers new insights into how local labor market contexts—a key determinant of alcohol-related outcomes among working-age individuals—may influence binge drinking/AUD. Our findings of higher risks of binge drinking/AUD risks among individuals in counties with higher job shares in Finance and Other Services indicate the need for effective workplace and community-based interventions to mitigate the risks (Oh, 2023, Porthé et al., 2021). Also, further research is warranted to elucidate the potential protective effects of having higher shares of construction and trade sector jobs by investigating the role of economic vitality and other protective benefits in reducing binge drinking/AUD risks at the population level.

Data Availability:

The National Survey on Drug Use and Health (NSDUH) data used in this study are not publicly available due to restrictions under the data contract with the Substance Abuse and Mental Health Services Administration (SAMHSA). Information on contextual data sources is provided in the Supplemental Material.