key: cord-0781360-r82obb90 authors: Leigh, J. Paul; Chakalov, Bozhidar title: Labor unions and health: A literature review of pathways and outcomes in the workplace date: 2021-08-02 journal: Prev Med Rep DOI: 10.1016/j.pmedr.2021.101502 sha: bf4a8d2015ebe1a4da2eec1ff1f185cac4729341 doc_id: 781360 cord_uid: r82obb90 Extensive economic research demonstrates correlations between unions with wages, income inequality, health insurance, discrimination, and other factors. Corresponding epidemiologic literature demonstrates correlations between income, income inequality, insurance, discrimination, and other factors with health. The first purpose of this narrative review is to link these literatures and identify 28 possible pathways whereby labor unions might affect the health of workers. This review is restricted to effects within workplaces; we do not consider unions' political activities. This review covers studies from the US, Europe, and Canada from 1980 through April 1, 2021. Pathways are grouped within five domains informed by the CDC 5-domain model of social determinants of health and the traditional 3-domain model of occupational medicine. Linked pathways include wages, inequality, excessive overtime, job satisfaction, employer-provided health insurance (EPHI), and discrimination. Second, we identify studies analyzing correlations between unions directly with health outcomes that do not require links. Outcomes include occupational injuries, sickness absence, and drug overdose deaths. Third, we offer judgments on the strength of pathways and outcomes --- labeled “consensus,” “likely,” “disputed” or “unknown” --- based on literature summaries. In our view, whereas there are four “consensus” pathways and outcomes and 16 “likely” pathways and outcomes for unions improving health, there are no “consensus” or “likely” pathways for harming health. The strongest “consensus” pathways and outcomes with salubrious associations include EPHI, OSHA inspections, dangerous working conditions, and injury deaths. Fourth, we identify research gaps and suggest methods for future studies. Unions are an underappreciated social determinant of health. Economic research is extensive pertaining to correlations between unions on the one hand and lower income inequality, higher wages, less overtime, less discrimination, more employer-provided health insurance and additional factors on the other. Epidemiologic research is extensive on correlations between health with the same or similar factors. It is somewhat surprising how separate these two literatures are. Only a few economic studies address how unions directly affect health, but these are limited to two outcomes: sickness absence and occupational injury. And whereas epidemiology has an entire subfield for occupational epidemiology, we found only a handful of studies (discussed below) addressing the direct associations between unionization and health. The authoritative text on occupational epidemiology no longer contains a chapter on unions (Levy et al., 2017) . The authoritative text on social epidemiology (Berkman et al., 2014) does not mention unions in its 39page index. The first purpose of this narrative review is to link the economic and epidemiologic literatures. Second, we review the few studies on direct correlations between unions and health outcomes. Third, we offer judgments on whether the links and correlations represent effects of unions on health. Fourth, we make suggestions for future research. Unions have been at the forefront of recent debates surrounding income inequality, stagnant wages, "deaths of despair," and protection and adequate pay for essential workers during the covid-19 pandemic. Income inequality in the US is now at historic levels not seen since the beginning of the Great Depression (Keshner, 2019) . Inflation-adjusted US wages have been stagnant or falling for middle-and low-wage workers for over 40 years (Mishel et al., 2015) . Epidemiologic studies find income inequality harms the population's health (Kawachi and Kennedy, 2002) . Kristol and Cohen (2017) estimate that 44% of the increase in wage inequality in US private sector jobs from 1988 to 2012 was attributed to the decline in unionization. "Deaths of despair" refer to the 20-year increase in deaths due to drug overdoses, alcoholism, and suicides primarily among middle-aged Americans (Case and Deaton, 2020) . These deaths have been cited as causes of the stunning recent annual drops in US life expectancy even before covid (Devitt, 2018) . The last time there were drops in US life expectancy was in 1918 during the Spanish Flu pandemic. "Deaths of despair" have been partially attributed to decades-long erosion of the numbers of unionized, well-paying, blue-collar jobs in the US (Case and Deaton, 2020) . A similar phenomenon is threatening the UK and the decline of unions is again being cited as one cause (Joyce and Xu, 2019) . Monnat (2016) and Blanchflower (2019) suggest that "deaths of despair" and falling wages since the Great Recession contributed to the election of Donald Trump and votes for Brexit. Many researchers and commentators view unions as playing pivotal roles in reducing income inequality, wage stagnation, "deaths of despair", and especially in improving treatment of essential workers (Case and Deaton, 2020; Kapos, 2020; Mishel et al., 2015; Nunn et al., 2017) . Many people view the 1950s as a time when America was "Great" for the American worker. The economy was growing rapidly, wages were increasing at all economic levels, and unemployment was low. It was also a time that private-sector labor unions were at their zenith of power in the US. Approximately 33% of the workforce was unionized in the 1950s and the vast majority was in private, not public unions (Mayer, 2014) . Private-sector unionization began a steady decline in the 1960s to arrive at the 2019 level of 6%, matching that from the 1910s in the US (Bureau of Labor Statistics, 2020). It is beyond the scope of this review to discuss the numerous factors causing these sharp declines in private-sector unions --declines in some manufacturing industries such as autos and steel with historically high unionization rates likely played roles-but one factor for which there is near unanimous agreement deserves a brief mention: politics. Countries such as France, Denmark, Sweden, Canada and Germany, among others, also experienced declines in manufacturing but experienced no or only modest declines in the percent of the workforce covered by union contracts; the political climate in these countries is cited as largely responsible (Rosenfeld, 2014) . There are unique examples from the US. The 1947 Taft-Hartley Act allowed states to pass socalled "right-to-work" (RTW) laws that permitted workers in unionized workplaces to opt-out of joining and paying dues, i.e. to free ride. RTW laws have had negative effects on unions' abilities to organize (Ellwood and Fine, 1987) . By 2017, 27 states had adopted these laws. As another example, President Reagan fired all striking air traffic controllers in 1981 thereby signaling that his administration did not support unions. Beginning in the 1980s, organized business began aggressively opposing unions (Rosenfeld, 2014) . But this 40-year decline is poised for change. Public approval for unions, 64% in 2019, is near a 50-year high, with the highest approval among people < 35 years old; and politicians have noticed (Jones, 2019) . Public-sector unions followed a different path. In 1959, Wisconsin became the first state to allow state workers to unionize; most other states subsequently followed. By the late 1970s, the percent of all public workers in unions rose to approximately 33%. Unlike private-sector unionization, public-sector unionization did not decline but rather has stabilized around 33-34% since the 1980s in the US. In 2019, approximately 7.1 million workers were in public-sector and 7.5 in privatesector unions (Bureau of Labor Statistics, 2020). Private-and publicsector unionization combined was 10.3% of the workforce in 2019. In 2019, for both private and public-sector unions combined, African Americans (11.2%) had higher union membership rates than white non-Hispanics (10.3%), Hispanics (9.1%) or Asians (8.9%) (Bureau of Labor Statistics, 2020) in the US. People in the 45-64 age bracket had the highest rates of all age brackets (12.7%) (Bureau of Labor Statistics, 2020). The rate for full-time workers (11.2%) was roughly double that for part-time workers (5.5%). Among all 16.4 million workers represented by unions, 14.6 million were members and 1.8 million were not members but were nevertheless covered by union contracts (Bureau of Labor Statistics, 2020). In 2017, women comprised approximately 47% of private-sector union members and 58% of public-sector members (Wolf and Schmitt, 2018) . Although we are not aware of statistics on occupations by gender by sector, data on educational attainment suggests that female union members are more likely to hold higher disproportionate shares of both low-and high-status occupations in the public than private sector. Approximately 14% of public-sector union members have less than a high school degree and 34% have more than a college degree; the percentages for private-sector union members are 9% and 11% (Wolf and Schmitt, 2018) . In part, this reflects public schools in which cafeteria workers and janitors are not required to have high school degrees, but teachers are required college degrees. We identify 28 possible pathways involving linked economic and epidemiologic literatures and seven health outcomes involving studies on the direct effects of unions on health. Each of these 28 economic and epidemiologic sets of studies as well as those for health outcomes studies merit their own literature review. This study, however, will not conduct 63 separate literature reviews (28 economic, 28 epidemiologic, and 7 direct). This review is restricted to effects within workplaces. We do not consider unions' political activities involving, for example, minimum wages, universal basic income, or universal health insurance. Unique literature search strategies were followed. Three leading labor economics texts (and references therein) formed the primary sources for the 28 economic pathways (Ehrenberg and Smith, 2015; Kaufman and Hotchkiss, 2003; McConnell et al., 2017) . Google scholar searches identified the prominent epidemiologic and public health studies for the 28 epidemiologic pathways as well as four of the seven direct effects (self-rated health, drug overdoses, mortality, food insecurity). Donado's (2015) literature review and subsequent studies citing Donado were used as the basis for our search involving the fifth and sixth direct effects: fatal and non-fatal occupational injuries. Brown and Sessions' (1996) literature review and subsequent studies citing Brown and Sessions formed the basis of our search involving the final direct effect: work absences. The appendix provides greater detail for the search strategies for injuries and absences. For all strategies, papers were excluded if they had no direct data-based estimate of effects of unions on pathways or pathways on health. Additional exclusions were for editorials, news stories, blogs, testimony, legal briefs, undergraduate papers, and thought pieces. All searches included 1980 through April 1, 2021. Most studies were from the US, but some were from Europe and Canada. No studies were drawn from economically developing nations. Whereas the subjects of this review -workers-are sometimes viewed only through the lens of occupational medicine, the effects of unions are farreaching and best viewed also through the lens of public health as shown in the model below. This review follows a straightforward format. First, we present a model for analyzing direct effects of unions on the health of workers covered by collective bargaining contracts. Second, we review the economics literature pertaining to possible health pathways that are prevalent in unionized versus non-unionized workplaces. Third, we review the epidemiological literature pertaining to whether these likely pathways have health or behavior associations in working and other populations regardless of union status. Fourth, we consider the relatively few (mostly economic) studies on direct associations between union status and health outcomes. The paper closes with a summary, limitations, and suggestions for future research. Table 1 presents a model for union pathways with domains on the left side and pathways on the right. The model is informed by the literature on the Social Determinants of Health (SDoH) and Occupational Medicine. Healthy People 2020 posits five domains for the SDoH: 1) Economic Stability, 2) Education, 3) Health and Health Care, 4) Neighborhood and Built Environment, and 5) Social and Community Context (Centers for Disease Control, 2020). The leading textbook on Occupational Medicine posits three domains: Physical and Mechanical, Biological and Chemical, and Psychosocial (Levy et al., 2017) . For relevance in the union and workplace context, a subcategory of Neighborhood and Built Environment-pollution-can be combined with Physical, Mechanical, Biological and Chemical to form what may be labeled Environment and Work Organization. There is overlap between portions of Social and Community Context with Psychosocial which may be labeled simply Psychosocial. The pathways are gleaned from three leading labor economics texts as well as economics literature on differences between union and non-union workplaces for which there are corresponding epidemiologic studies. The first author has been teaching classes and researching subjects in labor economics and social epidemiology since 1980. Pathways are more fully described in the next section. The arrangement of pathways in the five domains is straightforward but some require comment. Excessive overtime, shift and graveyard work, workers and unemployment compensation, workplace flexibility for individual workers, non-standard and "gig" working conditions, piece-rate pay, and vacation leave are characteristics of the work environment and hence placed into Environment, but they also have economic and psychosocial dimensions. For example, excessive overtime can increase earnings and therefore be considered in Economic Stability. Shift and graveyard work can affect circadian rhythms which, in turn, can affect psychosocial health. In addition, whereas sub-categories for all three Occupational Medicine domains are well represented in the pathways, some subcategories for SDoH are not. Quality of housing and level of violence in the neighborhood are subcategories within Neighborhood and Built Environment. It is unlikely that workers within their workplaces are associated with housing or neighborhood violence in the same way that they are associated with, for example, wages or working conditions. The SDoH domains and subcategories are used as a guide, not the ultimate arbiter of which pathway to include in which domain. The pathways included in Table 1 are more numerous than any other list appearing in the literature of which we are aware. Future researchers will no doubt imagine others. It is likely that other imagined pathways will fall into one of the five domains, however, thereby underscoring the utility of the model. Table 1 describes pathways for workers at the workplace. This review will not extend to effects of unions on society at-large; they are simply too numerous. For example, teachers' unions might promote health classes for their students; police unions might support members who kill African Americans; national unions might support environmental laws; effects of labor unions on national levels of employment or housing availability -two important determinants of health-are largely unknown. Nevertheless, two broad effects will be addressed: economy-wide wage levels and income inequality. The first part of Table 2 , labeled I, applies to union versus non-union workplaces; part II applies to society at-large, combining union with non-union workplaces. The first column in Table 2 identifies pathways. Columns 2 and 3 pertain to predominantly economic literature on unions; columns 4 and 5 pertain to predominately epidemiologic literature regardless of union status. Column 2 provides references for the judgments. Column 3 provides our judgments regarding the findings in the literature. For some pathways, there is consensus. For example, all studies with which we are aware find unionized establishments have more OSHA inspections (pathway #11). For other pathways, such as "job security," (pathway #4) findings are disputed. We created a category, "likely", which indicates that judgment leans to one side. Finally, we use "unknown" for pathways for which there are two or fewer studies. We use these words to describe our judgments of findings, not the findings themselves. The reader should have high confidence that when we describe findings as "disputed" that there are at least two (probably many more) studies with inconsistent findings. Our "likely" judgment means that while there may be inconsistent findings, we believe the evidence leans in a consistent direction. In general, readers should have more confidence in "consensus", "disputed" and "unknown" judgments than "likely" ones. We sought to make Table 1 self-explanatory. Nevertheless, several pathways require additional comments. The lion's share of economic research has addressed private, not public unions. It is likely that all conclusions and judgments drawn in Table 1 also apply to public unions albeit to a lesser degree. For example, considering pathway #1, whereas private unions generate a 20% wage advantage over private non-union workplaces, public unions generate a 10% advantage (McConnell et al., 2017) . When the same conclusions and judgments cannot be drawn for public unions, we will so indicate. Wages and within-firm wage inequality are the first and second pathways and there is consensus: unions increase wages -especially for low-wage workers-and reduce within-firm inequality compared to non-union workplaces. A leading labor economics textbook estimates a 15% union wage for private-sector and government-sector unions combined (McConnell et al., 2017) . This wage advantage represents a transfer from capital (business owners) to labor (Mishel, 2012) . Mishel (2012) estimates that unionized workers are 53.9 % more likely to have employer-provided pensions (pathway #3). Pathway #5 involves discrimination for which there are two forms: employment and wages. Historians disagree on the extent of union discrimination against Blacks for the first 60 years of the twentieth century. Some unions and affiliates (e.g. AFL) explicitly excluded Blacks but others (e.g. CIO) welcomed them (Hill, 1996) . In addition, some unions, such as the Brotherhood of Sleeping Car Porters, were exclusively Black. A recent analysis finds that in the two decades following World War II, Blacks were over-represented in unions and enjoyed a greater union wage advantage than whites (Farber et al., 2020) . Explicit discrimination against other minorities, particularly Asians, and women also occurred within some unions in these 60 years but, again, there is no consensus on the extent of that discrimination (Hill, 1996) . Beginning in the 1970s, most studies find less discrimination against either women or minorities than in the labor market at-large with respect to employment (Leonard, 1985) . In 2016, 65% of persons covered by either private or public-sector union contracts were either women or minorities (Bivens et al., 2017) . Beginning in the 1970s, while there is no consensus, numerous studies indicate less wage discrimination in either private-or public-sector unions (Bivens et al. 2017) . Within the public sector, unions raise wages for women more than for men (Freeman and Leonard, 1987) . For both public-and private-sector unions combined, Mishel (2012) estimates union wage premiums are higher for Blacks (17%) and Hispanics (23%) than whites (11%). (2015) p. 487 Kaufman and Hotchkiss (2003) p. 648 Leigh (1981) and (2015) Likely unions promote social support at work Park et al. (2004) Likely social support at work improves health 25. Fairness, justice at work See Kochan (1979) Likely decrease in wage and/or income inequality improves population health Even though formal education (pathway #7) is a powerful SDoH, research on the effects of unions on educational attainment is sparse and ambiguous (Blanchflower, 2016; Ewer, 2000) . For example, Blanchflower (2016) finds educational attainment is negatively correlated with private sector membership but positively correlated with public sector membership. There is consensus that unionized workplaces are more likely to have employer-provided health insurance (EPHI) in the US (pathway #8). One estimate is that unionized workplaces have insurance coverage rates that are 18.3 percentage points higher than those for nonunionized workplaces (Bivens et al., 2017) . Buchmueller et al. (2002) find that de-unionization explains about a third of the decline in EPHI coverage between 1983 and 1997 in the US. Mishel (2012) estimates that unionized workers are 3.4% more likely to have paid sick leave (pathway #9). We distinguish between exposure to dangerous working conditions (pathway #12) versus health and injury outcomes resulting from exposures (Table 3) . Exposure can be assessed, for example, with questions to workers such as "does your job ever expose you to…." followed by possibilities including, for example, dangerous chemicals, viruses, bacteria, radiation, fire, electricity, or air pollution (Leigh, 1982) . The consensus is that unionized workplaces are more hazardous than nonunionized ones. But there is a question regarding assessing blame: do unions create hazards or vice versa? A leading labor economics text suggests unions are more likely to form in hazardous workplaces and once formed, hazardous conditions are reduced, at least in the private sector; few analyses have addressed the public sector (Kaufman and Hotchkiss, 2003) . A thorough discussion of this issue appears in the analysis of Table 3 . The consensus is that unions increase the likelihood of OSHA inspections (pathway #13) and, given workers experience injuries or unemployment, unions increase the likelihood of receipt of workers compensation and unemployment compensation benefits (pathway #14). Union members may feel less threatened than non-union members by possible employer retaliation resulting from contacting OSHA or filing workers' compensation claims (Weil, 1991; Hirsch et al., 1997) . Both pathways #17 and #18 include shiftwork. Pathway #17 applies to the total amount of shiftwork while #18 applies to worker control over whether to engage in shiftwork. Individual worker-controlled flexibility (#18) includes, for example, the ability to: work at home; have compressed workweeks part of the year; temporarily change start and quit times; alter the pace of work; choose shifts; require predictable hours (Cotti et al. 2013; Duncan and Stafford 1980; Kaufman and Hotchkiss, 2003; Keune, 2013) . Flexibility has implications for family health as it would allow workers more time to take care of sick family members. Whereas unions may decrease the availability of working from home or ability to alter the pace of work, they enlarge the capacity to choose which shift to work and to require predictable work hours, particularly for members with seniority. The effects of unions on individual worker-controlled flexibility are therefore disputed. Jobs with non-standard work arrangements have also been referred to as alternative, precarious, contingent, gig, freelance, or independent contract (#19). There is no agreed-upon definition, but these jobs are typically temporary, do not have an explicit or implicit contract for ongoing employment, and shift some of the risk of business onto workers (Howard, 2017) . The Bureau of Labor Statistics estimates that nearly 14% of the workforce held contingent or alternative jobs in 2017 (Bureau of Labor Statistics, 2018). These jobs are disproportionately nonunion (OECD, 2019). Unions have been at the forefront of keeping standard jobs from becoming gig jobs as well as helping to change the legal classification of gig work from independent contractor to employee. (Tronsor, 2018; CBS News, 2021) . European unions might have success in their attempts to include gig work in their sectoral bargaining arrangements (Doherty and Franca, 2020) . A systematic review with meta-analysis finds unions correlate with low job satisfaction (pathway #22) but does not find unions cause low satisfaction (Laroche, 2016) . One explanation is that union members are encouraged to "speak up", to express any displeasure with working conditions to management; another is that dissatisfied workers are more likely to join unions. Further evidence for this "speaking up" hypothesis pertains to quit rates which can be viewed as the strongest expression of job dissatisfaction. Most studies find unionization lowers quit rates (McConnell et al., 2017) . Pathway #23 pertains to "job strain" and/or "job control" (Schnall et al., 1994) . We are unaware of economic studies comparing union and non-union workers on these dimensions and only one epidemiologic study (Gillen et al. 2002) which, incidentally, finds no union/non-union differences. Social support at work includes mentoring, cooperative spirit, willingness to share resources, companionship, emotional support, and other factors among co-workers and, when appropriate, supervisors (Park et al., 2004) . We are unaware of empirical studies addressing union/non-union differences for social support at work (#24). Nevertheless, it is likely that unions promote this pathway given unions encourage members to attend meetings, voice grievances with one another, and solidarity (Hagedorn et al., 2016) Fairness and justice (#25) encompass many dimensions and these can be in conflict (Fuller and Hester, 2001) . For example, does fairness Table 3 Direct associations between unions and health outcomes. Reeves (2021) Unknown. Unions decrease mortality in 1 study 35. Food insecurity Reeves et al. (2021) Unknown. Unions decrease food insecurity in 1 study dictate that workers be promoted based on seniority or productivity or some combination of the two? Unions give considerable weight to seniority (McConnell et al., 2017) . On the other hand, unions generally have grievance and arbitration procedures that encourage workers to voice complaints and sometimes change work practices. A (dated) survey finds that 83% of American workers believe unions protect them from "unfair practices" by management (Kochan, 1979) . Self-esteem, respect, and stigma (#26) receive enormous attention in the organizational psychology literature but we are unaware of studies exploring union/non-union differences (Pierce and Gardner, 2004) . One survey of steelworkers finds union membership satisfies socio-emotional needs such as "approval, self-esteem, affiliation, and respect" (Fuller and Hester, 2001) . On the other hand, it could be that unions carry a stigma given some historical connections to organized crime (Gibney et al., 2018) . Pathways #27 and #28 pertain to society at-large. Unions may have effects on wages, benefits, and unemployment throughout the economy depending on effects on non-union workplaces (#27). There are conflicting theories (McConnell et al., 2017) . One theory holds that unions raise wages "too high", forcing employers to cut their unionized workforces. These laid-off workers will flood the labor market in non-union sectors, leading to declines in wages and benefits and increasing unemployment in those sectors. But there are also threat effects according to a different theory: non-union firms might increase wages and benefits to discourage any threat posed by their workers possibly deciding to unionize (Fortin et al., 2021) . Finally, there is consensus that unions reduce economy-wide wage and income inequality (#28). Western and Rosenfeld (2011) find de-unionization from 1973 to 2007 explains from 20% to 33% of rising wage inequality in the US. Raphael (2011) finds unionization and inequality associations across 21 OECD countries. Epidemiologists and some economists have investigated the pathways in Table 2 independent of union status. Column 4 provides prominent studies and texts and column 5, our judgments. Again, many entries in Table 2 are self-explanatory and do not require additional comment. Several hypotheses surround higher wages (pathway #1). First, higher wages can improve access to health care as workers are more able to afford it. Women with low incomes will feed and clothe their children before spending on themselves (Elliot et al., 2015) . On the other hand, higher wages might allow workers to buy more cigarettes, drugs, or alcohol (Leigh et al., 2019) . We are unaware of studies on the effects of pensions on health (#3). Studies find Social Security Supplemental Income benefits (theoretically like pensions) decrease disability in the elderly (Arno et al., 2011) . Substantial literature indicates increases in income, especially for low-income people, improves health . For example, Davis et al. (2018) query cross-sections of Britons to determine minimum levels of income required to meet basic material needs for food, clothing, shelter, and so on; Gibson et al. (2020) review 27 studies on interventions similar to Universal Basic Incomes and find some health benefits. Discrimination (#5) harms psychological health and increases unhealthy behaviors (e.g. smoking) of women and especially minorities; effects on physiological health are "inconsistent and weak" (Krieger, 2014) . Epidemiologic studies typically do not separate employment from wage discrimination. When compared to "no insurance", EPHI (pathway #8), improves health of workers and their families (O'Brien, 2003) . If the US moves to universal coverage, however, this comparison may become moot. We are not aware of studies that compare health outcomes among similar workers with EPHI versus, for example, Medicaid or individual private insurance. But for the foreseeable future, many employed people in the US will likely not have insurance. Regarding other fringe benefits, Asfaw et al. (2017) find sick leave (pathway #9) reduces flu-related absences because the flu is less likely to spread and Rossin (2011) finds maternity leave (pathway #10) improves child health. Studies have identified disproportionate shares of workplace hazards and injuries within non-standard jobs (#19) (Howard, 2017) . Apouey and Stabile (2019) find that non-standard employment is associated with good mental health due to the job control and flexibility. Piece work and incentive pay (#20) likely harm health (DeVaro and Heywood, 2017) . The first economist to suggest harm was Adam Smith: "Workmen …. when they are liberally paid by the piece, are very apt to overwork themselves, and to ruin their health and constitution in a few years" (DeVaro and Heywood, 2017) . Economy-wide increases in either wage or income inequality (#28) likely harm population health. We are not aware of epidemiologic studies addressing only wage inequality (as there are in economics). A plethora of epidemiologic studies, however, have addressed income inequality. Ross et al.(2000) find inequality increases mortality in the US but has no effect in Canada. But Kawachi et al. (2014) and Pickett and Wilkinson (2009) find reduced inequality improves health across states and nations. Moreover, there is a wealth of epidemiologic literature addressing health effects of income reaching back decades (Kitagawa and Hauser, 1973) Table 3 presents pathways, judgments, and studies on the direct associations of unions with various measures of health. There are far fewer of these studies than appear in Table 2 . These studies differ from those in Table 2 in that the dependent variable measures health and the key independent variable measures unions; no links are required between economic and epidemiologic studies. The first two dependent variables in the first two rows summarize literature reviews in the Appendix. All studies recognize the possibility of reverse causality: workplace hazards might result in more unions because unions might be more likely to form in workplaces that have significant hazards. Some studies attempt to remove reverse causality with instrumental variables and/or longitudinal data (Donado, 2015) so that researchers can test whether unions reduce the number of injuries from existing high levels. An additional complicating factor is that unions likely help workers apply for and receive workers' compensation benefits (Hirsch et al., 1997) . Our assessment of the literature in the Appendix is that unions decrease fatal injuries (pathway #29) but findings for non-fatal injuries (#30) are (table 3) *Note: Reported absence is not an outcome; only actual sickness or injury leading to absence is an outcome. disputed. There is consensus for the second dependent variable: unions increase reported sickness absence (#31). There is no consensus, however, for explaining the correlation. Most researchers suggest that rather than indicating unions cause sickness, unions encourage workers to take more sick days when they are truly sick. Union workers may not feel as threatened with employer retaliation as non-union workers for taking days off. Finally, seven unique studies in the bottom rows of Table 3 pertain to unions improving self-rated physiological and psychological health, drug overdoses, mortality, and food insecurity. Table 4 provides a summary of findings. We created five categories for assessments. For a pathway to qualify within either the "Consensus unions improve health" or "Consensus unions harm health" there must have been consensus either in both columns 2 and 4 of Table 2 or within Table 3 . For the two "likely" categories, qualification was broader: either the pathways in both columns 2 and 4 of Table 2 were "likely" or one was "likely" and the other was "consensus"; or the pathway in Table 3 is "likely." If any pathway garnered a "disputed" or "unknown" judgment in either column 2 or 4 of Table 2 or Table 3 , that pathway was classified as "disputed and unknown" in summary Table 4 . Findings in Table 4 reveal that whereas there are four consensus pathways and outcomes and 16 likely pathways and outcomes for unions improving health, there are no consensus or likely pathways for harming health. We cannot conclude, however, that unions improve health overall because there are 15 disputed and unknown pathways and outcomes and any of these may have powerful harmful effects. This study has limitations. First, readers may not agree with our judgments regarding summaries of findings, but they at least have a place to begin to form their own or construct studies to test these pathways. Yet setting aside judgments, we identify 35 pathways and outcomes and cite relevant studies for each. Virtually all "union and health" studies with which we are aware have identified no more than three pathways. Malinowski et al. (2015) and Hagedorn et al. (2016) identify more but they do not link their pathways to the economics literature, nor do they identify as many as in this study. Second, with the exception of effects on economy-wide wages and inequality, we do not include any other possible economy-wide effects such as possible effects on the quality of products (e.g. unionized nurses providing better cardiovascular care (Ash and Seago, 2015) or union support for political public health initiatives such as Obamacare or effects on broader Social Determinants of Health such as housing. With four exceptions -employer-provided health insurance (EPHI), paid family leave, individual worker-controlled flexibility and wages-we do not address effects on families. Finally, this review primarily focuses on private-sector unions because, apart from wages, there is little research on possible pathways for public-sector unions. There are lessons from studies in Tables 2 and 3 for future research. First, most research has been conducted within the private sector. When public-and private-sector sector unions are analyzed, they are frequently separated. Second, comparison groups must be constructed with an eye to the structure of the relevant labor market. For example, within the private-sector, blue-collar rather than white-collar workers are much more likely to be unionized. The comparison group for privatesector unions, therefore, should be non-unionized blue-collar, not whitecollar, workers. Third, there are major gaps in research pertaining to, for example, union/non-union differences in education, job strain, and justice, and effects of pensions on health. Fourth, whatever the health dependent variable might be there is the possibility of reverse causality e.g. hazardous conditions may lead workers to form unions. To address this possibility researchers might use longitudinal data containing people who have joined or left unions over time or instrumental variables or propensity scores. We first link predominately economic with predominately epidemiologic literatures to identify 28 job-related pathways whereby unions might influence the health of workers. Pathways include wages, wage inequality, and discrimination. Second, we report on studies with direct associations between unions and health including occupational injuries and absence from work. We cannot conclude that there is consensus that unions improve or harm overall worker health; we nevertheless find considerably more salubrious than harmful pathways and outcomes. Unions can also have effects outside workplaces; for example, they can help galvanize political support for public health legislation or minimum wages. But such effects are myriad and beyond the scope of this review. Unions are underappreciated institutions for affecting not only worker health, but the health of workers' families and the public at-large. Trade unions, absenteeism, and exit-voice The effects of self and temporary employment on mental health: The role of the gig economy in the UK INSEAD Working Paper No. 2019/23/EPS Social security and mortality: The role of income support policies and population health in the United States Vacation -still an issue for worker protection? An empirical study of vacation and recuperation Potential economic benefits of paid sick leave in reducing absenteeism related to the spread of influenza-like illness The effect of registered nurses' unions on heart-attack mortality Job insecurity and health: The moderating role of workplace control Precarious employment: Understanding an emerging social determinant of health Trade union membership, tenure and the level of job insecurity 2014. Social Epidemiology How today's unions help working people. Economic Policy Institute A Cross-Country Study of Union Membership. IZA Discussion Paper #2016 Not Working: Where Have All the Good Jobs Gone? Union coverage and non-standard work in Britain The economics of absence: Theory and evidence Union effects on health insurance provision and coverage in the United States The effect of unions on the receipt of unemployment insurance benefits Bureau of Labor Statistics, 2020. Economic News Release. Union Members Summary Workplace flexibility, selfreported health, and health care utilization Inequality at work: the effect of peer salaries on job satisfaction CBS News. Unions and gig drivers sue to overturn proposition 22 Chapter 11: The role of training in promotion workplace safety and health Employers steal billions from workers' paychecks each year: Survey data show millions of workers are paid less than the minimum wage, at significant cost to taxpayers and state economies Workplace flexibilities, job satisfaction and union membership in the US workforce Health effects of unemployment benefit program generosity Employment discrimination, segregation, and health A Minimum Income Standard for the UK Drug use, de-unionization and drug death rates Health and the wage rate: Cause, effect, both, or neither? New evidence on an old question Performance pay and work-related health problems CDC data show U.S. life expectancy continues to decline Suicides, drug overdose deaths named as key contributors Unfairness and health: Evidence from the Whitehall II Study Solving the 'Gig-saw'? Collective Rights and Platform Work Why do union workers have more nonfatal occupational injuries? Welfare regimes, labour policies and unhealthy psychosocial working conditions: a comparative study with 9917 older employees from 12 European countries Worker health is good for the economy: Union density and psychosocial safety climate as determinants of country differences in worker health and productivity in 31 countries Do union members receive compensating wage differentials? Join the union and be safe: The effects of unionization on Occupational Safety and Health in the European Union Does the Union Make Us Strong? Labor-Union Membership, Self-Rated Health, and Mental Illness: A Parametric G-Formula Approach Solidarity and disparity: Declining labor union density and changing racial and educational mortality inequities in the United States Modern Labor Economics: Theory and Public Policy Being a good mom: Low-income, black single mothers negotiate intensive mothering The impact of right-to-work laws on union organizing Trade unions and vocational education and training: Questions of strategy and identity. Labor and Industry: A The relationship between job satisfaction and health: a meta-analysis Unions and inequality over the twentieth century: New evidence from survey data. Princeton University Working Paper Union maids: Unions and the female workforce What Do Unions Do? Basic Books Labor market institutions and the distribution of wages: The role of spillover effects A closer look at the relationship between justice perceptions and union participation Unions, incentive systems, and job design Public perceptions of unions, members, and joining intentions The public health effects of interventions similar to basic income: a scoping review Perceived safety climate, job demands, and coworker support among union and nonunion injured construction workers The effects of workplace flexibility on health behaviors: A cross-sectional and longitudinal analysis Socioeconomic status and health The role of labor unions in creating working conditions that promote public health Upstream solutions: Does the supplemental security income program reduce disability in the elderly? The problem of race in American Labor History Workers compensation recipiency in union and nonunion workplaces Nonstandard work arrangements and worker safety and health Association of the healthpromoting workplace with trade unionism and other industrial factors Dignity and health: A review As Labor Day turns 125, union approval near 50-year high Women, working families, and unions Inequalities in the twenty-first century: introducing the IFS Deaton Review. Institute for Fiscal Studies Unions tap into burst of worker angst over coronavirus The Economics of Labor Markets The Health of Nations: Why Inequality Is Harmful to Your Health Chapter 4. Income Inequality The determinants of worksite health promotion behavior America's 1% hasn't had this much wealth since just before the Great Depression MarketWatch Trade union responses to precarious work in seven European countries Differential Mortality in the United States: A Study in Socioeconomic Epidemiology Job strain as a risk factor for coronary heart disease: a collaborative meta-analysis of individual participant data How American workers view labor unions Chapter 13: Returning to work after occupational injury Chapter 3: Discrimination and health inequities The causes of rising wage inequality: the race between institutions and technology Work health promotion, job well-being, and sickness absences-a systematic review and meta-analysis A meta-analysis of the union-job satisfaction relationship The effects of union membership on absence from work due to illness Are unionized blue-collar jobs more hazardous than non-unionized blue-collar jobs? Minimum wages and health: A literature review The effect of unions on the employment of blacks, Hispanics, and women Industrial and Labor Relations Occupational and Environmental Health The effect of workplace inspection on worker safety Social class myopia: The case of psychology and labor unions Labor unions: A public health institution Union membership trends in the United States Contemporary Labor Economics Wage theft as a neglected public health problem: An overview and case study from San Francisco's Chinatown district Unions, inequality, and faltering middle-class wages. Economic Policy Institute. Issue Brief #342 Wage stagnation in nine charts Deaths of despair and support for Trump in the 2016 election New evidence on unions and layoff rates Unions as social capital: Renewal through a return to the logic of mutual aid? The shift in private sector union participation: Explanation and effects. The Hamilton Project. Economic Analysis Employers' benefits from workers' health insurance Non-standard workers are underrepresented by trade unions: Actual and adjusted ratio of trade union density among non-standard workers relative to standard workers (%), latest available year. OECD Employment Outlook 2019: The Future of Work Effects of social support at work on depression and organizational productivity What do unions do for mothers? Paid maternity leave use and the multifaceted roles of labor unions The Spirit Level. Why Greater Equality Makes Societies Stronger Self-esteem within the work and organizational context: A review of the organization-based self-esteem literature The political economy of health promotion: part 2, national provision of the prerequisites of health The health effects of wage setting institutions: How collective bargaining improves health but not because it reduces inequality. Sociology of Health and Illness Wage-Setting policies, employment, and food insecurity: A multilevel analysis of 492 078 people in 139 countries Bringing you more than the weekend: Union membership and self-rated health in the United States What unions no longer do Enforcing OSHA: The role of labor unions Relation between income inequality and mortality in Canada and in the United States: Cross sectional assessment using census data and vital statistics The effects of maternity leave on children's birth and infant health outcomes in the United States Job strain and cardiovascular disease The public health effect of economic crises and alternative policy responses in Europe: An empirical analysis. The Lancet Overtime pay, overtime hours, and labor unions Unions for workers in the gig economy: Time for a new labor movement Shift and night work and long working hours-a systematic review of safety implications Unions, norms, and the rise in US wage inequality union-workers-in-state-and-local-government-key-facts-about-the-sector-for-follow ers-of-janus-v-afscme-council-31 This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Supplementary data to this article can be found online at https://doi. org/10.1016/j.pmedr.2021.101502.