Labor Unions and American Poverty

VanHeuvelen, Tom; Brady, David

Article — Published Version

ILR Review

Provided in Cooperation with:

WZB Berlin Social Science Center

Suggested Citation: VanHeuvelen, Tom; Brady, David (2022) : Labor Unions and American Poverty, ILR

Review, ISSN 2162-271X, Sage, Thousand Oaks, CA, Vol. 75, Iss. 4, pp. 891-917,

https://doi.org/10.1177/00197939211014855

This Version is available at:

https://hdl.handle.net/10419/240906

Standard-Nutzungsbedingungen:

Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen

Zwecken und zum Privatgebrauch gespeichert und kopiert werden.

Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle

Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich

machen, vertreiben oder anderweitig nutzen.

Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen

(insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten,

gelten abweichend von diesen Nutzungsbedingungen die in der dort

genannten Lizenz gewährten Nutzungsrechte.

Documents in EconStor may be saved and copied for your personal

and scholarly purposes.

You are not to copy documents for public or commercial purposes, to

exhibit the documents publicly, to make them publicly available on the

internet, or to distribute or otherwise use the documents in public.

If the documents have been made available under an Open Content

Licence (especially Creative Commons Licences), you may exercise

further usage rights as specified in the indicated licence.

LABOR UNIONS AND AMERICAN POVERTY

TOM VANHEUVELEN AND DAVID BRADY*

American poverty research largely neglects labor unions. The

authors use individual-level panel data, incorporate both household

union membership and state-level union density, and analyze both

working poverty and working-aged poverty (among households led

by 18- to 64-year-olds). They estimate three-way fixed effects (person,

year, and state) and fixed-effects individual slopes models on the

Panel Study of Income Dynamics (PSID), 1976–2015. They exploit

the higher quality income data in the Cross-National Equivalent

File—an extension of the PSID—to measure relative (\50% of

median in current year) and anchored (\50% of median in 1976)

poverty. Both union membership and state union density have statis-

tically and substantively significant negative relationships with relative

and anchored working and working-aged poverty. Household union

membership and state union density significantly negatively interact,

augmenting the poverty-reducing effects of each. Higher state union

density spills over to reduc e poverty among non-union households,

and there is no evidence that higher state union density worsens pov-

erty for non-union households or undermines employment.

y and large, American poverty research neglects labor unions.

Prominent public intellectual books on poverty fail to discuss unions

(e.g., Wilson 1996). High-profile edited volumes on poverty (e.g., Jencks

and Peterson 1991; Danziger and Haveman 2001) and O’Connor’s (2001)

influential history of American poverty scholarship do not mention unions.

Other prominent volumes contain only token mentions (e.g., Cancian and

Danziger 2009). For instance, Blank, Danziger, and Schoeni (2006: 374)

mention unions only in regards to unemployment insurance. To the best of

our knowledge, there have been no studies of how unions influence

*TOM VANHEUVELEN ( https://orcid.org/0000-0002-7504-8186) is an Assistant Professor at the

University of Minnesota. D

AVID BRADY is a Professor in the School of Public Policy at the University of

California, Riverside, and is a Fellow at the WZB Berlin Social Science Center.

For helpful comments, we thank Michaela Curran, Matthew Mahutga, UCR’s Political Economy

Working Paper Group, and the faculty and students at University of Illinois School of Labor and

Employment Relations. Data and replication files are available through the OpenICPSR PSID Repository

(workspace number: openicpsr-137301). The authors are listed reverse alphabetically, each contributed

equally. An Online Appendix is available at http://journals.sagepub.com/doi/suppl/10.1177/00197939

211014855. For additional information, please address correspondence to tvanheuv@umn.edu or

dbrady@ucr.edu.

EYWORDs: unionization, poverty, labor union, longitudinal, working poverty

ILR Review, 75(4), August 2022, pp. 891–917

DOI: 10.1177/00197939211014855. Ó The Author(s) 2021

Journal website: journals.sagepub.com/home/ilr

Article reuse guidelines: sagepub.com/journals-permissions

American poverty in Journal of Labor Economics, Journal of Human Resources,

Industrial and Labor Relations Review, Industrial Relations, or the ‘‘top five’’

Economics journals. Most Annual Review essays on American poverty in

anthropology (Morgen and Maskovsky 2003) and sociology (Small and

Newman 2001; Desmond and Western 2018) have zero mentions of unions.

The neglect of labor unions in American poverty research is surprising,

perhaps, because extensive literatures demonstrate the critical role they play

for outcomes closely related to poverty, such as wages, working conditions,

and equality. Scholars have long studied how labor unions are a key power

resource that provides workers influence in the workplace, mobilizes voters,

and allies with Leftist political parties to institutionalize egalitarianism

(Korpi 1983; Huber and Stephens 2001). Indeed, a few recent studies pro-

vide evidence that unionization is associated with lower poverty (Plasman

and Rycx 2001; Zuberi 2006; Lohmann 2009; Brady, Baker, and Finnigan

2013; Crettaz 2013; Rosenfeld and Laird 2016; Lohmann and Marx 2018).

Still, a striking disconnect remains between the American poverty literature

and research investigating union effects.

The present study explicitly builds on the few recent and relevant studies

while addressing their limitations. To the best of our knowledge, this is the

first study on unions and poverty to use individual-level panel data. Unlike

prior research, we examine both household union membership and state-

level union density (henceforth ‘‘state union density’’), and both working

poverty and working-aged poverty (among households led by 18- to 64-year-

olds). With the Panel Study of Income Dynamics (PSID) for 1976–2015, we

estimate three-way fixed effects (person, year, and state) and fixed-effects

individual slopes models (Wooldridge 2010). We exploit the Cross-National

Equivalent File’s—an extension of the PSID—higher quality income data to

measure relative (\50% of median in current year) and anchored (\50%

of median in 1976, adjusted for inflation) working and working-aged poverty.

We investigate three primary research questions: 1) Does household union

membership influence working and working-aged poverty? 2) Net of house-

hold union membership, does state union density influence working and

working-aged poverty? 3) By testing the interaction between union member-

ship and state union density, do the effects augment each other and do the

benefits of state union density spill over to non-union households?

The Case for Labor Unions

At the household and state levels as well as through their interactions, it is

plausible that unions reduce poverty. Extensive literatures link union

No essay on poverty in the United States has appeared in the Annual Review of Economics, and we are

not aware of any piece in the Journal of Economic Literature on unions and poverty. On balance, Lichter

(1997), Newman and Massengill (2006), and O’Connor (2000) have a few brief mentions of unioniza-

tion. But, none of these devote substantial attention to unionization as a principal cause of poverty.

Brady (2019) is the exception and identifies unions as a key measure of power resources.

892 ILR REVIEW

membership to higher wages and greater equality (Kalleberg, Wallace, and

Althauser 1981; Freeman and Medoff 1984; Card 1996; Rosenfeld and

Kleykamp 2012; Rosenfeld 2014; Rosenfeld, Denice, and Laird 2016; Kristal

and Cohen 2017). Many studies have demonstrated that union members

receive a wage premium compared to nearly identical non-members. Union

wage premia even exist for less-skilled workers (Eren 2009), those without a

high school education (Maxwell 2007), and those in precarious employ-

ment (Gomez and Lamb 2019). Union membership benefits also exist for a

wide variety of low-wage workers (Waddoups 2001; Appelbaum, Berg, Frost,

and Preuss 2003; Batt, Hunter, and Wilk 2003; Erickcek, Houseman, and

Kalleberg 2003; Firpo, Fortin, and Lemieux 2009). Because wages are the

dominant source of income for most working-aged households, union wage

premia suggest household union membership will reduce poverty.

Regarding state union density, evidence suggests a contextual effect of

firm- and industry-level unionization on poverty. Many studies have shown

that highly unionized contexts benefit both union members and non-

members (Kahn and Curme 1987; Leicht, Wallace, and Grant 1993;

Neumark and Wachter 1995; VanHeuvelen 2018). Such contextual effects

are attributable to a combination of threat and moral economy processes,

resulting in spillover effects onto non-union workers within a firm and

industry and into nearby non-union firms and industries.

Beyond the firm and industry, country-level union density similarly

reduces poverty. Power resources theory describes unions as key class-based

collective political actors shaping the distribution of economic resources

(Korpi 1983; Huber and Stephens 2001; Brady 2019). Unions bond the

working class and poor together, politically mobilize workers in elections,

exert pressure in workplaces and on governments, and ultimately result in a

more egalitarian income distribution. Accordingly, studies have demon-

strated that countries with higher unionization have significantly lower

working poverty (Zuberi 2006; Lohmann 2009; Crettaz 2013; Rosenfeld and

Laird 2016; Lohmann and Marx 2018). Although largely neglected by

American poverty research, this comparative literature shows that labor

unions are a key collective political actor driving lower poverty across rich

democracies (Plasman and Rycx 2001).

Similar to comparative analyses of rich democracies, US states can be

compared as polities where struggles and settlements occur over redistribu-

tion and inequality (Jacobs and Dirlam 2016; Bucci 2018). Indeed, in an era

of decentralized federalism, state polities could be increasingly important in

shaping poverty (DiGrazia and Dixon 2019; Hertel-Fernandez 2019).

Consistent with this account, Rosenfeld and Laird (2016) provided descrip-

tive correlations showing that states with higher density have lower working

and overall poverty. Brady and colleagues (2013) examined the relationship

between state union density and working poverty from 1991 to 2010. Using

In our PSID sample, 89% of ‘‘pre-fisc’’ total household income comes from labor income.

LABOR UNIONS AND AMERICAN POVERTY 893

multilevel models of individuals nested in states in 2010 and two-way fixed-

effects models of individuals nested in state-years (1991–2010), they found

that state union density reduces working poverty. They also found that

state union density has larger effects than states’ economic performance

and social policies, with effects comparable to standard individual-level

predictors of working poverty such as education and single motherhood.

Beyond the distinct effects of household union membership and state

union density, these two aspects should interact to augment the poverty-

reducing effects of each. Unions are a key component of a broader and

integrated complex of labor market institutions that govern and equalize

wages, constrain employers from paying very low wages, and protect

workers’ job security and physical safety (Blau and Kahn 2002; Koeniger,

Leonardi, and Nunziata 2007; Doellgast, Holtgrewe, and Deery 2009; Gautie´

and Schmidt 2009; Giesselmann 2014; Bucci 2018). In the United States,

which lacks centralized wage bargaining and corporatist governance, unions

are one of the few and most crucial labor market institutions (Rosenfeld

2014; Jacobs and Dirlam 2016; Bucci 2018). To the extent that unions are

an essential part of a broader institutional complex shaping equality, the

presence of both household union membership and state union density

should provide a particularly favorable situation for working families.

Moreover, as state union density reflects and amplifies an egalitarian institu-

tional context, it should spill over to reduce the poverty of non-union and

non-working households.

The Case for Skepticism

Despite the arguments above, at least four reasons suggest why unions

might fail to reduce poverty. First, union density is so low that unions may

have become ineffective, even irrelevant (DiGrazia and Dixon 2019; Hertel-

Fernandez 2019). As Rosenfeld (2014: 30) explained, ‘‘The private sector in

this country is now nearly union-free, to a degree not seen in a century.’’

The United States—and especially some states—exhibits cross-nationally

and historically exceptionally low union density (Rosenfeld 2014). Union

density has declined more rapidly among the less skilled, who are most vul-

nerable to poverty (Blank 2009). For such reasons, Autor (2011: 14) argued:

‘‘It appears unlikely their [unions’] role is paramount. . . . [Unions’] impact

is largely confined to manufacturing and public sector employment, neither

of which comprises a sufficiently large share of the aggregate economy.’’

Low union density also results in less variation across states than exists

across rich democracies (Hirsch and Macpherson 2003; Visser 2011). In

turn, there could be insufficient interstate heterogeneity in union density to

explain variation in poverty. Moreover, low and relatively invariant state

union density suggests that, although unions could theoretically affect US

poverty, other factors, such as economic performance and individual

characteristics, are likely to have far greater influence (Blank et al. 2006;

894 ILR REVIEW

Autor 2011). In total, analyses of recent data might reveal little to no rela-

tionship between labor unions and poverty.

Second, large underlying differences exist between those selecting into

unions and those not selecting into unions. Selection likely reflects unob-

served advantageous characteristics of union members, such as ambition and

social skills (Card, Lemieux, and Riddell 2004; Borjas 2015; VanHeuvelen

2018). Such unobserved characteristics are likely associated with poverty

for reasons independent of union membership. This unobserved hetero-

geneity and related selection bias are p lausibly even more notable among

the less skilled and those below or near the poverty line. Previous studies

on un ions and poverty have relied on cross-sectional data (e.g., Lohmann

2009; Brady et al. 2013), however. Therefore, panel data with techniques

to net out unobserved i ndividual characteristics might reveal no robust

union effect.

Third, even if unions benefit union households, there might be no bene-

ficial spillover effects for non-union households. Many studies argue that

unions only benefit workers in select industries or sectors where unions are

strong (Autor 2011). Those at the bottom of the income distribution are

unlikely to be unionized and may not benefit from state spillover effects.

While Brady and colleagues (2013) showed state union density reduces

household-level working poverty, they mostly could not control for house-

hold union membership and therefore could not establish such spillover

effects for non-union households.

Rather than a contextual spillover effect,

it is unclear if poverty-reducing effects of state union density are simply due

to compositional differences across states. Further, Brady and colleagues

(2013) found that the effects of state union density are much stronger for

working households closer to the median and insignificant for those in

deep poverty. Therefore, any benefits of unions may be narrowly restricted

to employed and less-poor union households. To accurately assess potential

spillover effects for non-union households, panel data with both household

and state union information, for working and non-working households, is

needed.

Fourth, unions could even have adverse spillover effects, thereby worsen-

ing poverty of non-members and disadvantaged groups. Directly, unions

and the policies they advocate for might only create rents for protected

insiders, and may even worsen the labor market for the truly disadvantaged

(for a discussion, see Rosenfeld and Kleykamp 2012). Some researchers the-

orize that unions have a crowding effect, in which union wage gains lead to

cuts in the quantity of union jobs (Kahn 1978; Neumark and Wachter

1995). Crowding then increases the supply of non-member workers,

depressing wages of non-members. While Brady and colleagues (2013)

They use Luxembourg Income Study (LIS) data from the March Current Population Survey (CPS),

but the LIS removes union membership data. Although Brady and colleagues (2013) conducted sensitiv-

ity analyses with the underlying CPS data from the smaller outgoing rotation group including union

information, they can only approximate the higher quality LIS income measures.

LABOR UNIONS AND AMERICAN POVERTY 895

analyzed working households only, a comprehensive test for direct adverse

spillover effects must include non-working households.

Indirectly, by raising wages among the employed, unions could increase

labor costs, cause labor market rigidity, and discourage hiring (Blau and

Kahn 2002; Magnani and Prentice 2010). Similar to well-known arguments

about adverse effects of minimum wages, higher wages and labor costs

could force firms to reduce employment (Kahn and Morimune 1977;

Walsworth 2010). Such reduced employment would worsen poverty because

employment is the most salient individual-level predictor of poverty

(Rainwater and Smeeding 2004; Brady, Finnigan, and Hu

bgen 2017; Brady

2019). Therefore, it is essential to include non-working households in the

sample and to test for state union density effects on employment as well.

To recapitulate, previous research on unions suggests that the beneficial

impacts on workers in the middle and bottom of the distribution should

also apply for poverty. Direct mechanisms of union household residence and

contextual effects of state union density might each have distinct negative

associations with poverty, and these effects might even interact. However,

there are several reasons to remain skeptical. Any poverty differences may be

attributable to variation in observed or unobserved characteristics across

individuals or across states. Meanwhile, restricting focus to employed workers

might miss negative spillover effects in which individuals are crowded out of

gainful employment.

Data and Methods

We use individual-level data from the PSID and the Cross-National

Equivalent File (CNEF), which we merge with state-level data (described

below). This data set has critical advantages over the LIS-CPS data used by

Brady and colleagues (2013). Primarily, the CPS is cross-sectional,

meaning

they could not control for the unobserved characteristics that select

individuals into unions. Furthermore, they only examined from 1990 to

2010. By contrast, our study using the PSID spans a longer and more varying

time period.

The CNEF, which is a supplement to the PSID, provides higher quality

standardized measures of income incorporating taxes, tax credits, and

transfers (Frick et al. 2007). The PSID is the longest-running panel survey

in the United States, with the initial survey wave administered in 1968. With

weights, the economic characteristics of the PSID—including wages and

inequality, and all but the most extreme high and low family incomes—are

similar to the data used to construct official poverty statistics, the CPS

(Gouskova and Schoeni 2007; Heathcote, Perri, and Violante 2010;

VanHeuvelen 2018).

Although the CPS outgoing rotation group can be treated as longitudinal, respondents are in the

panel for only one year, an insufficient time span for our purposes.

896 ILR REVIEW

We use the PSID-CNEF waves 1976 and 1979 to 2015 because 1976 and

1979 were the first PSID waves with information on spouse’s union member-

ship. The analyses end with 2015 as it is the final year of available CNEF

data, which we need for the income measures. Our data use three PSID

samples: the Survey Research Center, the Survey of Economic Opportunity,

and the 1997 Immigrant samples. We drop the 1990 Latino sample

(VanHeuvelen 2018).

Individuals, the unit of analysis, are nested in households, which are

nested in states and years. We construct two samples corresponding to work-

ing households and working-aged households: 1) individuals in households

with at least one employed working-aged adult (18–64 years) and 2)

individuals in households with a working-aged adult household head. We

include the 50 US states and the District of Columbia, which is treated as a

state.

Dependent Variable

Following the overwhelming majority of international poverty research

(Rainwater and Smeeding 2004; Brady et al. 2013; Smeeding 2016; Brady

et al. 2017), we operationalize poverty as those residing in households with

less than 50% of the median equivalized disposable household income (ref-

erence = not poor). Thus, poverty is a household-level variable. A house-

hold pools its expenses and resources, so if the household is poor, all

members are poor. We measure household income with the CNEF house-

hold ‘‘post-fisc’’ income variable. Unlike the official US poverty measure

(OPM), our measure of income comprehensively incorporates taxes and

tax credits (e.g., the Earned Income Tax Credit) and cash and near-cash

(e.g., the Supplemental Nutritional Assistance Program) transfers.

Thus,

we intentionally avoid the OPM because of its well-documented and serious

validity and reliability problems (Rainwater and Smeeding 2004; Brady et al.

2013; Smeeding 2016). Following prevailing international standards on

income measurement (Duncan and Petersen 2001; Rainwater and

Smeeding 2004; Brady et al. 2013; Brady et al. 2017), we equivalize income

for household size by dividing by the square root of household members.

The poverty threshold is calculated yearly using all individuals regardless of

the household head’s age or the employment status of any household mem-

ber. The sample is reduced to employed or working-aged households only

after calculating the threshold.

Using the current year’s median, we analyze this standard relative poverty

measure described above. This measure is the most widely accepted defini-

tion in the international poverty literature. We supplement that measure

with anchored poverty. Anchored poverty sets the threshold for poverty in the

first year of analysis (1976) and uses that threshold across years, adjusting

The CNEF employs the National Bureau of Economic Research’s TAXSIM model.

LABOR UNIONS AND AMERICAN POVERTY 897

only for inflation (Chen and Corak 2008; Brady et al. 2013; Smeeding

2016). Anchored poverty is a well-established approximation of ‘‘absolute’’

poverty as it applies the same threshold over time, even when medians rise

and fall. Whereas relative poverty is less responsive to the business cycle and

economic development, anchored poverty should mechanically decline as

the typical household experiences rising affluence since 1976.

Household and State Union Measures

We measure labor unions at the household and state levels. First, union

household membership is a binary measure of living in a union household or

not, where either or both the household head and spouse are union

members.

Second, we measure state-level union density among non-

agricultural workers age 16 and older, collected from the CPS by Hirsch

and Macpherson (2003).

Union membership for household heads is avail-

able from 1970 onward; however, spouses were asked about union member-

ship only in 1976 and from 1979 onward. Fortunately, union membership

in the PSID tracks closely to union membership in the CPS (VanHeuvelen

2018). Online Appendix Figure A.1 displays the variation in state union

density over time.

Other Independent Variables

We adjust for a standard set of variables that may confound the association

between unions and poverty (Rainwater and Smeeding 2004; Blank et al.

2006; Lohmann 2009; Brady et al. 2013; Brady et al. 2017). We include two

sets of controls, based on our samples of working households and working-

aged households. For working households, we identify the household lead

earner, defined as the highest earner, with ties broken by age (i.e., not nec-

essarily the head). Household age distribution includes lead’s age (under

25, 25–34, 35–54, and 55 or older), the number of household members under 18,

the number of household members over 64, and a binary measure of whether the

household contains a child under age 5. With a couple as the reference, we

include binary measures for single mother, single father, female-head no child,

and male-head no child households. With white as the reference, we include

indicators for Black and other lead earners.

With less than high school

degree as the reference, we include binary measures for whether the lead

Union information for other household members is not available. Heads and spouses make up more

than 90% of employed individuals in our sample. Although our measure probably slightly

underestimates union households, we are skeptical that our main results are significantly biased by this

limitation.

An alternative measure, state union coverage, produces similar results. This outcome is expected

because of the similar levels and trends of membership and coverage over this period.

While race and education are usually time-invariant among employed adults, lead earners can vary

across surveys, and individuals can transition across households. Thus, race and education as household

properties have some time variation for individuals. For the working-aged sample, heads also vary over

time.

898 ILR REVIEW

earner has a high school degree, some college, college degree,andgraduate education.

Following VanHeuvelen’s (2018) harmonization of Census industry codes

in the PSID, we include dummies for 18 industries of the lead earner.

Further, we include indicators for 13 occupations of the lead earner.

For working-aged households, we assign household characteristics based

on the head rather than the lead earner because approximately 8% of sam-

ple households have no one employed. This approach applies to age, race,

and education. We omit the industry and occupation indicators and instead

include indicators of whether no one is employed in the household and multi-

ple earners in the household (reference = one earner) (Brady et al. 2017).

The working-aged poverty models retain the controls for age distribution

and family structure.

For both samples, we adjust for several state characteristics: 1) GDP per

capita, in thousands of real 2000 dollars,2)employment rate of the population,3)

GDP growth, and 4) the natural log of population. Data are collected from the

US Bureau of Economic Analysis (2020). Descriptive statistics are included

in Online Appendix Tables A.1 and A.2.

Estimation Techniques

We begin by estimating three-way fixed-effects linear probability regression

models:

ist

= UnionHH

ist

b + UnionSt

b + x

ist

b + year

g + state

p + a

+ 

ist

ð1Þ

Individuals, i, are nested in states, s, which are nested in years, t. The out-

come y indicates whether an individual is poor in survey wave t.

Individual

fixed effects, a

, remove time-invariant unobserved person-level heteroge-

neity, while year contrasts, g, remove shared period-specific shocks, and

state fixed effects, p, remove time-invariant state-level characteristics and

transform state-level variables to within-state deviations. The variable x is the

set of observed household- and state-level characteristics included beyond

household union membership (UnionHH

ist

b) and state union density

(UnionSt

b).

Compared to previous research, these models provide a more rigorous

test of the association between unions and poverty. Most critically, individ-

ual fixed effects remove time-invariant individual unobserved heterogeneity,

which addresses concerns of selection into unions discussed above. Thus,

our main results indicate the association between change in union member-

ship, at both the household and state levels, and the change in the probabil-

ity of an individuals’ poverty status.

Because we are primarily interested in average marginal effects, we use linear probability models,

which provide similar results to average marginal effects from logistic regression models. We also esti-

mated conditional logistic regression models and found similar results.

We use robust standard errors clustered at the individual level.

LABOR UNIONS AND AMERICAN POVERTY 899

State fixed effects allow us to better identify the influence of state union den-

sity on poverty by measuring within-state deviations. There are many reasons

why states with high union density, such as California, differ from states with

low union density, such as Mississippi. We can more directly test the influence

of union density by measuring its change within states after net ting out stable

unobserved between-state characteristics. Moreover, including both household

and state union measures more rigorously assesses whether previous findings

relied on compositional differences in states’ union membership.

We next estimate linear probability regression models interacting with

household union membership and state union density, which allows us to

formally assess whether state union density effects are concentrated among

union households, whether state union density effects spill over to non-

union households, and whether household union membership and state

union density augment each other’s effects.

Although three-way fixed-effects models improve upon previous research

and address many of the reasons for skepticism discussed above, they neverthe-

less have limitations. Such models rely on certain strong assumptions that might

be unreasonable when applied to the study of unions and poverty. Thus, to

scrutinize the robustness of our main results, we consider an extension of a

fixed-effects method that allows for the relaxation of these assumptions.

We estimate fixed-effects individual slopes (FEIS) linear probability

regression models. These models, popularized and detailed by Wooldridge

(2010: 377–81), can be written as:

= a

year

+ x

b + state

p + period

u + a

+ 

ð2Þ

The difference between Equations (1) and (2) is the treatment of time.

Year contrasts are replaced with an individual-specific linear year coeffi-

cient.

FEIS models adjust not only for time-invariant individual-level het-

erogeneity in the probability of working poverty but also individual-specific

time trajectories in the probability of working poverty (Ludwig and Bru

derl

2018). To partially account for broadly shared poverty trends, we include a

categorical variable, period

, which measures period contrasts in the business

cycle.

Additional details of the FEIS are included in the Online Appendix.

Results

Descriptive Patterns

We present descriptive statistics of key variables in Table 1. Across the entire

sample, approximately 10.1% of person-years fall into anchored poverty,

and 16.1% fall into relative working poverty. As expected, we observe

For coefficients of interest, FEIS is equivalent to including time-by-individual interactions in the

regression model.

Because of fewer observations in earlier years, we combine 1976–1992 (1), 1993–1999 (2), 2000–

2007 (3), and 2008–2015 (4).

900 ILR REVIEW

substantively and statistically significant differences in poverty rates across

union and non-union households, as well as between states with high and

low union density rates, defined as those in the top or bottom third of the

entire sample’s state union density rates. Across working and working-aged

relative and anchored poverty, we observe that union households have pov-

erty rates between 9 (anchored working-aged) and 15 (relative working-

aged) percentage points lower than non-union households, with unionized

households consistently having poverty rates of only 3 to 4%. Similarly,

highly unionized state-years have poverty rates between 3 (anchored-all)

and 7 (relative-all) percentage points lower than state-years with lower

union density.

Figure 1 shows trends in poverty. Figure 2 shows trends separately by

household union membership and state union density. Figure 3 visualizes

the differences across groups from Figure 2. Although many patterns in

these figures are notable, we highlight three.

First, we observe consistent differences in poverty rates, of 5 to 10 percent-

age points, between union and non-union households over time. Poverty

rates among union households remain at low values across time, meaning

that non-union households largely drive changes in poverty over time.

Second, we observe consistently lower working-aged poverty rates among

states with high union density. Anchored and relative working-aged poverty

are consistently about 5 percentage points lower in highly unionized states.

Third, we observe convergence across states of working relative and

anchored poverty rates in recent years, driven primarily by working poverty

rates in low union density states converging with the low poverty rates in

high union density states. This result partly reflects convergence in union

density across states between 1976 and 2015 and the modest decline in

working poverty in the United States since the 1990s (Brady et al. 2013).

For relative working poverty, the difference between states shrinks from

approximately 15 percentage points in 1980 to approximately 3 percentage

Table 1. Descriptive Statistics

Household State union density

Mean Non-union Union Difference Low High Difference

Relative poverty, all 0.164 0.189 0.051 –0.138* 0.200 0.129 –0.071*

Relative poverty, working 0.130 0.149 0.049 –0.100* 0.158 0.099 –0.059*

Anchored poverty, all 0.131 0.152 0.036 –0.116* 0.155 0.111 –0.043*

Anchored poverty, working 0.098 0.113 0.034 –0.079* 0.114 0.082 –0.031*

Union membership 0.160

State union density 16.00

Source: Panel Study of Income Dynamics (PSID), 1976–2015.

Notes: ‘‘Low’’ and ‘‘High’’ state union density are defined as state-years in the bottom and top third of

the distribution of state union density among the whole sample.

*p \ 0.001, two-tailed test.

LABOR UNIONS AND AMERICAN POVERTY 901

points in the most recent waves. For anchored working poverty, we see no

significant difference in the last year of data, 2015. However, this last point

is an exception to the otherwise consistent set of findings. At least descrip-

tively, unionized households and states have lower rates of each dimension

of poverty.

Of course, differences in poverty could result from compositional

differences of states and households, as well as unobserved characteristics of

individuals resulting in unequal sorting across union dimensions. We there-

fore turn to fixed-effects regression models to assess the robustness of the

association between labor unions and poverty.

Regression Analyses

Table 2 presents results from three-way fixed-effects regression models. For

both relative and anchored poverty and both working and working-aged

households, we present three models. The first includes only our two union

measures and individual, state, and year fixed effects. The second adds

individual-, household-, and state-level controls. The third includes an inter-

action between household and state union density.

Across Table 2, several conclusions can be drawn. First, residence in a

union household clearly and significantly reduces all poverty outcomes.

These coefficients are robust across all eight of the first two models. While

adding controls in the second models lowers the magnitude of coefficients,

we find that union household membership reduces the probability of pov-

erty by between 0.04 and 0.06 compared to non-union households (p \

0.001 in all models with controls, two-tailed tests). Critically, all models

include individual fixed effects. Results thus do not reflect differences in

Figure 1. Poverty over Time

Data source: Panel Study of Income Dynamics (PSID), 1976–2015.

Notes: Dotted lines are 95% confidence intervals.

902 ILR REVIEW

Figure 2. Poverty over Time, by Household and State Union Density

Data source: Panel Study of Income Dynamics (PSID), 1976–2015.

Notes: Dotted lines are 95% confidence intervals. ‘‘Low’’ and ‘‘High’’ state union density are defined as state-years in the bottom and top third of the distribution of state

union density among the whole sample.

Figure 3. Difference in Poverty across Household and State Union Density

Data source: Panel Study of Income Dynamics (PSID), 1976–2015.

Notes: Dotted lines are 95% confidence intervals. ‘‘Low’’ and ‘‘High’’ state union density are defined as state-years in the bottom and top third of the distribution of state

union density among the whole sample.

the probability of poverty across union and non-union households, but

rather change in the probability of poverty for an individual when a

household changes its union membership. A ssociations are thus net of

time-invariant characteristics that differ between union and non-union

households.

Second, we find significant and negative associations between state union

density and all four poverty outcomes. For three of four outcomes, state

union density becomes statistically significant only when control variables

are included. Nevertheless, declines in state union density, or the inverse of

presented coefficients, correspond with an increase in the probability of all

four poverty outcomes (p \ 0.01). Because the models include state fixed

effects, coefficients represent the association between working poverty and

change in state union density within states over time. Notably, results are

net of household union membership. Thus, these state-level results do not

simply reflect individual compositional differences in the PSID across times

Table 2. Fixed-Effects Linear Probability Models, Poverty on Unions and Controls

Relative poverty Anchored poverty

Working-aged households

(1) (2) (3) (4) (5) (6)

Union household –0.100*** –0.059*** –0.051*** –0.087*** –0.049*** –0.024***

(0.003) (0.003) (0.007) (0.003) (0.003) (0.006)

State union density –0.001 –0.002*** –0.002*** –0.001 –0.002** –0.001*

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)

Household

3 State union density –0.001 –0.001***

(0.001) (0.001)

Working poverty households

(7) (8) (9) (10) (11) (12)

Union household –0.066*** –0.041*** –0.028*** –0.052*** –0.031*** –0.009

(0.003) (0.003) (0.007) (0.003) (0.002) (0.006)

State union density –0.002** –0.002*** –0.002*** –0.001

–0.002*** –0.002**

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)

Household

3 State union density –0.001* –0.001***

(0.001) (0.001)

Controls? No Yes Yes No Yes Yes

Notes: Robust clustered standard errors in parentheses. All models include individual, year, and state

fixed effects. Standard errors clustered at individual level. Working-aged sample controls include

household head age, number of household members under age 18, number of household members

over age 64, whether the household contains a child under age 5, household composition, household

head race, household head education, whether none, one, or two or more household members are

employed. Working poverty sample controls include lead earner age, number of household members

under age 18, number of household members over age 64, whether the household contains a child

under age 5, household composition, lead earner race, lead earner education, lead earner industry,

lead earner occupation. All models include state controls: GDP per capita, employment rate per

population, GDP growth, and natural log of population.

Samples: 381,112.

Samples: 324,391.

p \ 0.10; *p \ 0.05; **p \ 0.01; ***p \ 0.001, two-tailed test.

LABOR UNIONS AND AMERICAN POVERTY 905

and places. Rather, this is a state-level effect of union density net of house-

hold membership.

Third, the interaction effects between household and state union density

are statistically significant for three of four outcomes (relative working-aged

is the exception). Results mostly suggest that state union density and house-

hold membership augment the effects of each other. Moreover, the main

effect of state union density is significantly negative in all four models.

Overall, results suggest that state union density reduces poverty for non-

union households. In addition, there is no evidence that state union density

has adverse spillover effects, as negative associations are found for both

working and working-aged poverty.

Although the models in Table 2 improve on previous studies of unioniza-

tion and poverty, they nevertheless rely on potentially strict assumptions for

fixed effects. We relax these assumptions by fitting FEIS models, as

presented in Table 3. Model sequencing remains the same as in Table 2.

Our first two conclusions, that household membership and state union

density independently reduce poverty, are clearly replicated in FEIS models.

Losing household union membership increases the probability of poverty

for all four outcomes, and the decline of state union density increases the

risk of all as well. Thus, both state and household union effects are

Table 3. Fixed-Effects Individual Slopes Linear Probability Models, Poverty

on Unions and Controls

Relative poverty Anchored poverty

Working-aged

(1) (2) (3) (4) (5) (6)

Union household –0.093*** –0.056*** –0.035*** –0.080*** –0.045*** –0.014*

(0.004) (0.003) (0.008) (0.003) (0.003) (0.007)

State union density –0.002** –0.002*** –0.002** –0.001

–0.002*** –0.002**

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)

Household

3 State union density –0.001* –0.001***

(0.001) (0.001)

Working poverty

(7) (8) (9) (10) (11) (12)

Union household –0.063*** –0.037*** –0.024*** –0.048*** –0.027*** –0.011

(0.003) (0.003) (0.007) (0.003) (0.003) (0.006)

State union density –0.001** –0.002** –0.002* –0.001 –0.001* –0.001*

(0.001) (0.001) (0.001) (0.001) (0.000) (0.000)

Household

3 State union density –0.001* –0.001**

(0.001) (0.000)

Controls? No Yes Yes No Yes Yes

Notes: Robust clustered standard errors in parentheses. See Table 2 for discussion of controls.

Samples: 379,076.

Samples: 321,654.

p \ 0.10; *p \ 0.05; **p \ 0.01; ***p \ 0.001, two-tailed test.

906 ILR REVIEW

detectable net of not only stable individual characteristics but also idiosyn-

cratic individual poverty trajectories over time.

When we include individual slopes, we detect even more clearly consis-

tent, significant interactions between household and state union density.

Across all poverty outcomes, state union density has a significant and nega-

tive association for non-union households with a steeper impact for union

households. As shown in Figures 2 and 3, union and non-union households

have substantially dissimilar trajectories in poverty over time, as have states

with different union densities. It is uncertain whether assuming uniform

time trajectories is appropriate. Partly for this reason, we can more clearly

detect variation of union effects across states and household membership

when we relax this assumption.

Figure 4 plots the predicted probabilities of our four poverty dimensions

across union and non-union households, and across state union density

(based on Table 3). The dashed line for union households clearly shows

that living in a union household has its most protective benefits when one

Figure 4. Predicted Poverty Levels by State and Household Union Density

Data source: Panel Study of Income Dynamics (PSID), 1976–2015.

Notes: Predictions from models 3, 6, 9, and 12 in Table 3. Dotted lines are 95% confidence intervals.

LABOR UNIONS AND AMERICAN POVERTY 907

also lives in a state with high union density, with the probability of poverty

ranging between 0.03 and 0.07 among high union density state-years.

Similarly, non-union households benefit from higher state union density.

We observe lower predicted probabilities of poverty among non-union

households among high union density states across all poverty outcomes.

Figure 5 displays the marginal effects of household union membership

across levels of state unionization. The marginal effects of poverty decline

between 2 and 5 percentage points from low to high state unionization. For

example, union households have approximately 0.02 lower probability of

being in anchored poverty compared to non-union households among low

unionized states, and have approximately 0.06 lower probability among the

highest unionized states. The interaction reveals that there is more similar-

ity in the risk of poverty between union and non-union households among

less unionized states. This between-group equality comes at the cost of an

overall higher risk of poverty, however. Although there is a greater between-

group difference between union and non-union households in highly

Figure 5. Household and Standardized State-Level Associations with Poverty, Fixed-Effects

Individual Slopes Models

Notes: Bars indicate 95% confidence intervals. Coefficients based on models 2, 5, 8, and 11 in Table 3.

We show household marginal effects in Online Appendix Figure A.2.

908 ILR REVIEW

unionized contexts, this inequality occurs within a context of an overall

lower risk of experiencing poverty in the first place.

Before proceeding, it is important to clarify how the magnitude of state

union density and household union membership compares to our other

predictors of poverty. Figure 5 presents x-standardized coefficients of all

state-level variables in our second models of Table 3: union density, GDP

per capita, employment, GDP growth, and logged population.

It also

compares household union membership against the individual-level

coefficients for the ‘‘big four’’ risk factors of poverty (Brady et al. 2017): sin-

gle motherhood, low education (i.e., less than high school), unemployment,

and young household head/lead earner (age 25 or younger).

Focusing first on state-level effects, Figure 5 shows that state union density

has a comparable magnitude to GDP per capita. GDP per capita tracks rising

economic development and affluence, and hence gauges long-term economic

growth. Both state-level characteristics have significant and negative effects:

The decline of union membership associates with higher poverty, whereas ris-

ing GDP per capita associates with lower poverty. The absolute magnitudes of

the two variables are similar. A standard deviation increase in state union den-

sity decreases poverty by between 0.01 and 0.018, whereas a standard deviation

increase in GDP per capita decreases poverty by between 0.011 and 0.016. Put

differently, state union density has about the same influence on poverty

change as does long-term state economic growth. Further, state union density

has larger and more robust standardized coefficients than employment rates,

GDP growth (i.e., short-term economic growth), and logged population.

Next, focusing on household-level effects, union membership consistently

has effects smaller than the big four risks. For working poverty, household

union membership effects are between 25% and 50% the magnitude of

other household variables; for working-aged poverty, union effects are

between 15% and 75% the magnitude of other key household explanations.

It is reasonable that household changes in employment or single mother-

hood status have substantively larger effects on poverty than household

union membership. It is notable, however, that the magnitudes of house-

hold union membership effects are almost as large as the contrast of lacking

a high school degree, compared to having a college degree, or having a

young household lead earner or young household head. Altogether, the

results in Figure 5 suggest that while unions alone are insufficient to explain

poverty trajectories, the decline of unions is a substantively significant con-

tributor to American poverty trajectories.

Supplementary Analyses

Thus far, we have presented evidence that both household union member-

ship and state union density reduce poverty. Recall, however, that

We present a table of standardized coefficients in Online Appendix Table A.4.

LABOR UNIONS AND AMERICAN POVERTY 909

employment is consistently found to be the most important predictor of

working (Brady et al. 2013) and working-aged poverty (Rainwater and

Smeeding 2004; Brady et al. 2017). Being an employed household or having

multiple earners (e.g., note effects of unemployment in Figure 5) have

much larger effects on poverty than do unions, raising concerns that unions

might have adverse spillover effects for those outside, or marginally

attached to, the labor market. As we explained, researchers have long been

concerned that higher state union density leads to fewer employment

opportunities or lower wages for non-members. It is therefore valuable to

assess whether state union density undermines employment.

First, if state union density did more harm than good at the bottom of

the labor market, we would find a positive effect on the poverty of non-

members among higher union density states. However, Figure 4 shows state

union density reduces all four poverty outcomes among non-union

households. Thus, there is no evidence of adverse spillover effects from

state union density for the poverty of non-union households. Second, per-

haps controlling for employment attenuates (or is a post-treatment control

for) the coefficient of state union density and conceals the adverse spillover

effects. Models 1, 4, 7, and 10 of Tables 2 and 3, however, show the effects

of state union density before controlling for employment. Although the

coefficients for state union density are less robustly significant before

controls are added, no ‘‘reduced form’’ model shows a positive coefficient

for poverty. Third, that the coefficients for state union density are consis-

tently negative among both working and working-aged households

undermines claims of adverse spillover effects. Even if state union density

worsens the employment of some households, the net effect across the sam-

ple is to reduce poverty.

Going further, we test if state union density undermines being employed

among working-aged households or having multiple earners among employed

households. These models mimic the fixed-effectsmodelsinTable2.Results

(included in the Online Appendix) show that across model specifications,

state union density is not significantly associated with whet her a household is

employed among the sample of working-aged households. In fact, state union

density’s coefficient is positively signed when controls are included. Similarly,

state union density is not significantly associated with whether a household

has multiple earners among the sample of working households. In total, we

find no evidence of adverse spillover effects on non-union households or

those marginally attached to the labor market.

Beyond spillover effects, we considered several potential concerns regard-

ing modeling decisions. First, we consider how our results were sensitive to

various clustering strategies. Although clustering at the incorrect level could

deflate standard errors, each approach has strengths and weaknesses. While

some researchers suggest clustering at the highest level possible, others

argue that it is better to cluster at the level of data collection or treatment

(Abadie, Athey, Imbens, and Wooldridge 2017). We replicated our main

910 ILR REVIEW

results while clustering at the household and state levels. State-level cluster-

ing reveals slightly weaker and modestly less robust coefficients for state

union density in the fixed-effects models. However, even when clustering at

the state level, state union density FEIS coefficients remain largely

unchanged. Thus, overall, our decision to cluster at the individual level does

not appear to change our conclusions.

Second, cross-level interactions with state-level within-unit deviations

might be biased without special consideration. Giesselmann and Schmidt-

Catran (2019) showed that such interactions in standard fixed-effects

models can be problematic, as they still retain a partial mix of between-unit

and within-unit effects. Thus, our interaction results may not appropriately

identify the more robust results stemming from within-state union density

changes. Following their advice, we interacted household union member-

ship with state and year fixed effects. We also replicated our interaction

models after including an additional interaction between household union

membership and individual fixed effects. These specifications did not alter

our main results, which we interpret as evidence that the results are not

unduly rooted in misspecification of within-unit deviations.

Conclusion

This study investigates the relationship between labor unions and poverty.

We measure unions as household union membership, state union density,

and their interaction. Distinctively, we assess the spillover effects of state

union density on non-union households. We analyze individual-level panel

data from the PSID between 1976 and 2015 using three-way fixed effects

and FEIS models. To the best of our knowledge, this is the first study of the

consequences for labor unions on poverty that uses individual-level panel

data. Because we use the CNEF’s ‘‘post-fisc’’ income measures, our study

also has more valid and reliable measures of poverty than studies based on

the OPM. Relatedly, we verify our results across both working and working-

aged poverty, and both relative and anchored measures.

We ask three research questions. First, does household union member-

ship influence working and working-aged poverty? The descriptive evidence

shows that all four dimensions of poverty are significantly lower in union

versus non-union households. These results are not wholly based in

characteristics that vary between union and non-union households. We find

robust evidence that entering a union household has a consistent negative

effect on the probability of all forms of poverty. Although the magnitudes

are smaller than those of the big four risk factors of poverty (Brady et al.

2017), household union membership has substantively meaningful effects.

Second, net of household characteristics, does state union density influ-

ence working and working-aged poverty? We show poverty is significantly

lower in states with high union density. Although not quite as robust as

household union membership, we find significant associations in 21 of 24

LABOR UNIONS AND AMERICAN POVERTY 911

reported models. Furthermore, state union density has a significant negative

coefficient in all eight fixed-effects and FEIS models for the four dimensions

of poverty when including individual- and state-level controls. These results

demonstrate that state union density has a poverty-reducing contextual

effect net of the compositional effects of states’ union membership. The

coefficients for state union density are also substantively meaningful com-

pared to standard predictors of poverty. State union density has coefficients

comparable to those of GDP per capita (i.e., affluence and long-term eco-

nomic development), and only state union density and GDP per capita have

consistent associations across poverty dimensions.

Perhaps more salient is that state union density reduces poverty for both

union and non-union households. Hence, state union density has a poverty-

reducing contextual spillover effect for non-union households specifically.

Higher state union density reduces poverty even for those who are not

members and for the working-aged population as a whole. Because we ana-

lyze the working-aged and not just working households, these results are

particularly relevant for the argument that state union density has a contex-

tual spillover anti-poverty effect.

Third, do any poverty-reducing effects of household union membership

and state union density significantly interact to offset or augment one

another? The interactions discussed above show not only that state union

density has effects for both union and non-union households but also that

household union membership is particularly beneficial in a context of high

state union density. There are distinct, non-redundant effects to both house-

hold union memberships and state union density. Although high state

union density benefits non-union households, union households are partic-

ularly unlikely to be poor in states with high union density.

This study also provides evidence that undermines each major reason for

skepticism about labor unions. Even though unions are exceptionally weak

in the contemporary United States, we continue to find significant effects.

This is the case even though our analyses includes three recent time points

(2011, 2013, 2015) of very low union density—all after the last time point

(i.e., 2010) observed by Brady and colleagues (2013). Further, this study

exploits panel data to remove the stable unobserved characteristics that

select people into union membership. Along with a rich set of time-varying

controls, the use of fixed effects and FEIS models should reduce concerns

about unobserved heterogeneity and selection. Further, we find beneficial

spillover effects for non-union households and the broader working-aged

population. Finally, analyses yield no evidence of adverse spillover effects

for the bottom of the income distribution, especially those marginally

attached to the labor market. Higher state union densities do not appear to

marginalize people into unemployment, fewer workers per household, or

poverty. Beyond these points, we underline that the present study advances

beyond past research’s focus on state union density and working poverty. By

examining both household union membership and working-aged poverty,

912 ILR REVIEW

this study substantially deepens and expands the evidentiary base for argu-

ing that unions reduce poverty.

Initially, it may seem counterintuitive that household and state union

measures have such robust associations with poverty, given the relative sta-

bility of poverty rates during the period of our study compared to steep

union decline. This seeming contradiction is best understood through a

cross-national comparison, in which the United States has an unusually high

poverty rate compared to other high-income countries (Brady et al. 2017).

The decline of labor unions and their protective social, political, and eco-

nomic consequences has allowed for US poverty rates to remain stable at a

high level despite four decades of economic growth, rising educational

attainment, female labor force participation and multi-earner households,

declining rates of young heads of households, and several other poverty-

reducing trends (Brady et al. 2017). A counterfactual world in which US

labor union membership caught up to Western Europe, rather than the

experienced inverse, would potentially have pushed American relative pov-

erty much nearer to Western European levels.

Our study has limitations that can motivate future research. Although we

demonstrate that state union density has broad effects on poverty reduction,

meaningful heterogeneity may be present across subgroups. Given that

labor union membership tends to stabilize marital patterns (Schneider and

Reich 2014), does union density have similar effects across single- and two-

parent households? Our focus on individual trajectories may not fully

account for the substantial geographical reorganization of industrial loca-

tion over the time period of study. Manufacturing employment shifted from

the Northeast to the South over the period of our study. To what extent do

the poverty-reducing effects of union density track with state-level shifts in

industrial composition (i.e., net of household-level industry and occupation

of employment, which we control for)? Further, how does union density

affect poverty among senior and child populations specifically, two groups

at especially high risks of poverty? While all these questions are beyond the

scope of the current article, they would help further establish when, and

how, union density reduces poverty.

Future research should also interrogate the causal ordering. For exam-

ple, we cannot fully disentangle whether labor unions themselves reduce

poverty or whether labor unions are associated with good jobs available to

those in the middle and bottom of the income distribution. Also, unions

might reduce employment churn and downward pressures on wages in

years spent unemployed, or union membership might follow after attain-

ment of high-wage employment when one can pay union dues. Labor

unions could be closely associated with establishing the social conditions

necessary for quality employment in spaces other than the top of the labor

market, but more research should test if unions have the direct effect on

poverty that our results suggest. Although these mechanisms are critical for

developing a precise theoretical understanding of how labor unions

LABOR UNIONS AND AMERICAN POVERTY 913

alleviate poverty, our study still demonstrates that declining labor unions in

the US have clear practical implications for poverty.

Although Brady and colleagues (2013) provided evidence that part of

union density’s effect on poverty works through the mechanism of social

policy, we are forced to leave that question to future research. In contrast to

Brady and colleagues’ time period of 1991 to 2010, it is more difficult to

compare social policy generosity across state-years in our longer and more

recent time period (1976 to 2015). While their analysis of Aid to Families

with Dependent Children/Temporary Assistance for Needy Families plus

Supplemental Nutrition Assistance Program (AFDC/TANF+SNAP) and

unemployment insurance (UI) maximum benefit levels is informative,

updated analyses would require greater consideration of coverage, eligibil-

ity, and access. Future research could ideally examine the decline of welfare

programs that were previously more prominent, such as General Assistance,

as well as the emergence of higher state minimum wages, earned income

tax credits, and the expansion of Medicaid through the Affordable Care

Act (as well as the shift from programs such as TANF to SNAP and

Supplemental Security Income (SSI)).

We conclude by encouraging American poverty researchers to incorpo-

rate labor unions into the study of poverty. As shown in the introduction,

the neglect of labor unions in American poverty research is pervasive. This

situation is unfortunate because the omission of unions from analyses of

poverty could arguably be a substantial omitted variable bias. Even if the

focus of an analysis is far from unions, we conjecture that unions should still

be accounted for in models of poverty. To the extent studies are interested

in causes of poverty that are in any way related to labor unions (e.g.,

employment), it is essential for American poverty scholars to incorporate

labor unions into their analyses. The omission is even more unfortunate

given growing interest in political theories of poverty generally (Brady

2019), and given the extensive related literatures on unions and wages, jobs,

and social equality. Scholars are increasingly recognizing that poverty is the

result of politics, and that power resources and institutions exert tremen-

dous influence on poverty. Unions are one of the most important power

resources and institutions and are pivotal to the politics of social policies.

Moreover, it is well understood that declining unionization is substantially

shaped by politics at the federal, state, and local levels (Rosenfeld 2014;

DiGrazia and Dixon 2019; Hertel-Fernandez 2019). Therefore, to under-

stand poverty, we must understand unionization as a key aspect of the poli-

tics of poverty and as a key aspect of the broader political processes that

ultimately shape the distribution of economic resources in society.

References

Abadie, Alberto, Susan Athey, Guido W. Imbens, and Jeffrey Wooldridge. 2017. When should

you adjust standard errors for clustering? NBER Working Paper No. 24003. Cambridge,

MA: National Bureau of Economic Research.

914 ILR REVIEW

Appelbaum, Eileen, Peter Berg, Ann Frost, and Gil Preuss. 2003. The effects of work

restructuring on low-wage, low-skilled workers in US hospitals. In Eileen Appelbaum,

Annette Bernhardt, and Richard J. Murnane (Eds.), Low-Wage America: How Employers Are

Reshaping Opportunity in the Workplace, pp. 77–117. New York: Russell Sage Foundation.

Autor, David. 2011. The polarization of job opportunities in the US labor market:

Implications for employment and earnings. Community Investments 23(2): 11–16.

Batt, Rosemary, Larry W. Hunter, and Steffanie Wilk. 2003. How and when does manage-

ment matter? Job quality and career opportunities for call center workers. In Eileen

Appelbaum, Annette Bernhardt, and Richard J. Murnane (Eds.), Low-Wage America: How

Employers Are Reshaping Opportunity in the Workplace, pp. 270–313. New York: Russell Sage

Foundation.

Blank, Rebecca M. 2009. Economic change and the structure of opportunity for less-skilled

workers. In Maria Cancian and Sheldon H. Danziger (Eds.), Changing Poverty, Changing

Policies, pp. 63–91. New York: Russell Sage Foundation.

Blank, Rebecca M., Sheldon H. Danziger, and Robert F. Schoeni (Eds.). 2006. Working and

Poor: How Economic and Policy Changes Are Affecting Low-Wage Workers. New York: Russell

Sage Foundation.

Blau, Francine D., and Lawrence M. Kahn. 2002. At Home and Abroad: U.S. Labor Market Perfor-

mance in International Perspective. New York: Russell Sage Foundation.

Borjas, George. 2015. Labor Economics, 7th ed. Boston: McGraw-Hill.

Brady, David. 2019. Theories of the causes of poverty. Annual Review of Sociology 45: 155–75.

Brady, David, Regina S. Baker, and Ryan Finnigan. 2013. When unionization disappears:

State-level unionization and working poverty in the US. American Sociological Review 78(5):

872–96.

Brady, David, Ryan Finnigan, and Sabine Hu

bgen. 2017. Rethinking the risks of poverty: A

framework for analyzing and comparing prevalences and penalties. American Journal of

Sociology 123(3): 740–86.

Bucci, Laura C. 2018. Organized labor’s check on rising economic inequality in the US

states. State Politics & Policy Quarterly 18(2): 148–73.

Cancian, Maria, and Sheldon H. Danziger (Eds.). 2009. Changing Poverty, Changing Policies.

New York: Russell Sage Foundation.

Card, David. 1996. The effect of unions on the structure of wages: A longitudinal analysis.

Econometrica 64(4): 957–79.

Card, David, Thomas Lemieux, and W. Craig Riddell. 2004. Unions and wage inequality. Jour-

nal of Labor Research 25: 519–59.

Chen, Wen-Hao, and Miles Corak. 2008. Child poverty and changes in child poverty. Demog-

raphy 45: 537–53.

Crettaz, Eric. 2013. A state-of-the-art review of working poverty in advanced economies: Theo-

retical models, measurement issues and risk groups. Journal of European Social Policy 23(4):

347–62.

Danziger, Sheldon H., and Robert H. Haveman (Eds.). 2001. Understanding Poverty. New York

and Cambridge, MA: Russell Sage Foundation and Harvard University Press.

Desmond, Matthew, and Bruce Western. 2018. Poverty in America: New directions and

debates. Annual Review of Sociology 44: 305–18.

DiGrazia, Joseph, and Marc Dixon. 2019. The conservative upsurge and labor policy in the

states. Work & Occupations. Published Online September 24.

Doellgast, Virginia, Ursula Holtgrewe, and Stephen Deery. 2009. The effects of national

institutions and collective bargaining arrangements on job quality in front-line service

workplaces. Industrial and Labor Relations Review 62(4): 489–509.

Duncan, Greg J., and Eric Petersen. 2001. The long and short of asking questions about

income, wealth, and labor supply. Social Science Research 30(2): 248–63.

Eren, Ozkan. 2009. Does membership pay off for covered workers? A distributional analysis

of the free rider problem. Industrial and Labor Relations Review 62(3): 367–80.

Erickcek, George A., Susan N. Houseman, and Arne L. Kalleberg. 2003. The effects of tem-

porary services and contracting out on low-skilled workers: Evidence from auto suppliers,

hospitals, and public schools. In Eileen Appelbaum, Annette Bernhardt, and Richard J.

LABOR UNIONS AND AMERICAN POVERTY 915

Murnane (Eds.), Low-Wage America: How Employers Are Reshaping Opportunity in the Work-

place, pp. 368–403. New York: Russell Sage Foundation.

Firpo, Sergio, Nicole M. Fortin, and Thomas Lemieux. 2009. Unconditional quantile regres-

sion. Econometrica 77(3): 953–73.

Freeman, Richard B., and James L. Medoff. 1984. What Do Unions Do? New York: Basic Books.

Frick, Joachim R., Stephen P. Jenkins, Dean R. Lillard, Oliver Lipps, and Mark Wooden.

2007. The Cross-National Equivalent File (CNEF) and its member country household

panel studies. Schmollers Jahrbuch 127: 627–54.

Gautie´, Je´roˆme, and John Schmitt (Eds.). 2009. Low-Wage Work in the Wealthy World. New York:

Russell Sage Foundation.

Giesselmann, Marco. 2014. The impact of labour market reform policies on insiders’ and

outsiders’ low-wage risk. European Sociological Review 30(5): 549–61.

Giesselmann, Marco, and Alexander W. Schmidt-Catran. 2019. Getting the within estimator

of cross-level interactions in multilevel models with pooled cross-sections: Why country

dummies (sometimes) do not do the job. Sociological Methodology 49(1): 190–219.

Gomez, Rafael, and Danielle Lamb. 2019. Unions and non-standard work: Union representa-

tion and wage premiums across non-standard work arrangements in Canada, 1997–2014.

ILR Review 72(4): 1009–35.

Gouskova, Elena, and Robert F. Schoeni. 2007. Comparing estimates of family income in the

Panel Study of Income Dynamics and the March Current Population Survey, 1968–2005.

Technical Series Paper #07-01. University of Michigan.

Heathcote, Jonathan, Fabrizio Perri, and Giovanni L. Violante. 2010. Unequal we stand: An

empirical analysis of economic inequality in the United States, 1967–2006. Review of Eco-

nomic Dynamics 13(1): 15–51.

Hertel-Fernandez, Alex. 2019. State Capture: How Conservative Activists, Big Businesses, and Wealthy

Donors Reshaped the American States — and the Nation. New York: Oxford University Press.

Hirsch, Barry T., and David A. Macpherson. 2003. Union membership and coverage database

from the Current Population Survey: Note. Industrial and Labor Relations Review 56(2): 349–54.

Huber, Evelyne, and John D. Stephens. 2001. Development and Crisis of the Welfare State: Parties

and Policies in Global Markets. Chicago: University of Chicago Press.

Jacobs, David, and Jonathan C. Dirlam. 2016. Politics and economic stratification: Power

resources and income inequality in the United States. American Journal of Sociology 122(2):

469–500.

Jencks, Christopher, and Paul E. Peterson (Eds.). 1991. The Urban Underclass. Washington,

DC: Brookings Institution.

Kahn, Lawrence M. 1978. The effect of unions on the earnings of nonunion workers. Indus-

trial and Labor Relations Review 31(2): 205–16.

Kahn, Lawrence M., and Michael Curme. 1987. Unions and nonunion wage dispersion.

Review of Economics and Statistics 69(4): 600–607.

Kahn, Lawrence M., and Kimio Morimune. 1977. Unions and employment stability: A

sequential logit approach. International Economic Review 20(1): 217–35.

Kalleberg, Arne L., Michael Wallace, and Robert P. Althauser. 1981. Economic segmentation,

worker power, and income inequality. American Journal of Sociology 87(3): 651–83.

Koeniger, Winfried, Marco Leonardi, and Luca Nunziata. 2007. Labor market institutions

and wage inequality. Industrial and Labor Relations Review 60(3): 340–56.

Korpi, Walter. 1983. The Democratic Class Struggle. Boston: Routledge.

Kristal, Tali, and Yinon Cohen. 2017. The causes of rising wage inequality: The race between

institutions and technology. Socio-Economic Review 15(1): 187–212.

Leicht, Kevin T., Michael Wallace, and Don Sherman Grant. 1993. Union presence, class

and individual earnings inequality. Work and Occupations 20(4): 429–51.

Lichter, Daniel T. 1997. Poverty and inequality among children. Annual Review of Sociology 23:

121–45.

Lohmann, Henning. 2009. Welfare states, labor market institutions and the working poor: A

comparative analysis of 20 European countries. European Sociological Review 25(4):

489–504.

916 ILR REVIEW

Lohmann, Henning, and Ive Marx (Eds.). 2018. Handbook of Research on In-Work Poverty.

Northampton, MA: Edward Elgar.

Ludwig, Volker, and Josef Bru

derl. 2018. Is there a male marital wage premium? New evi-

dence from the United States. American Sociological Review 83(4): 744–80.

Magnani, Elisabetta, and David Prentice. 2010. Did reducing unionization create more flexi-

ble American industries? Industrial and Labor Relations Review 63(4): 662–80.

Maxwell, Nan L. 2007. Wage differentials, skills, and institutions in low-skill jobs. Industrial

and Labor Relations Review 61(3): 394–409.

Morgen, Sandra, and Jeff Maskovsky. 2003. The anthropology of welfare ‘‘reform’’: New

perspectives on U.S. urban poverty in the post-welfare era. Annual Review of Anthropology

32: 315–38.

Neumark, David, and Michael L. Wachter. 1995. Union effects on nonunion wages: Evidence

from panel data on industries and cities. Industrial and Labor Relations Review 49(1): 20–38.

Newman, Katherine S., and Rebekah Peeples Massengill. 2006. The texture of hardship:

Qualitative sociology of poverty, 1995–2005. Annual Review of Sociology 32: 423–46.

O’Connor, Alice. 2001. Poverty Knowledge: Social Science, Social Policy, and the Poor in Twentieth-

Century U.S. History. Princeton, NJ: Princeton University Press.

———. 2000. Poverty research and policy for the post-welfare era. Annual Review of Sociology

26: 547–62.

Plasman, Robert, and Francxois Rycx. 2001. Collective bargaining and poverty: A cross-

national perspective. European Journal of Industrial Relations 7(2): 175–202.

Rainwater, Lee, and Timothy M. Smeeding. 2004. Poor Kids in a Rich Country: America’s Chil-

dren in Comparative Perspective. New York: Russell Sage Foundation.

Rosenfeld, Jake. 2014. What Unions No Longer Do. Cambridge, MA: Harvard University Press.

Rosenfeld, Jake, Patrick Denice, and Jennifer Laird. 2016. Union decline lowers wages of

nonunion workers. Technical Report, Economic Policy Institute.

Rosenfeld, Jake, and Meredith Kleykamp. 2012. Organized labor and racial wage inequality

in the United States. American Journal of Sociology 117(5): 1460–502.

Rosenfeld, Jake, and Jennifer Laird. 2016. Unions and poverty. In David Brady and Linda M.

Burton (Eds.), Oxford Handbook of the Social Science of Poverty, pp. 800–819. Oxford, UK:

Oxford University Press.

Schneider, Daniel, and Adam Reich. 2014. Marrying ain’t hard when you got a union card?

Labor union membership and first marriage. Social Problems 61(4): 625–43.

Small, Mario Luis, and Katherine Newman. 2001. Urban poverty after The Truly Disadvan-

taged: The rediscovery of the family, the neighborhood, and culture. Annual Review of Soci-

ology 27: 23–45.

Smeeding, Timothy. 2016. Poverty measurement. In David Brady and Linda M. Burton

(Eds.), Oxford Handbook of the Social Science of Poverty, pp. 21–46. Oxford, UK: Oxford Uni-

versity Press.

US Bureau of Economic Analysis. 2020. Regional economic accounts: Interactive data tables.

Accessed at https://www.bea.gov/data/economic-accounts/regional.

VanHeuvelen, Tom. 2018. Moral economies or hidden talents? A longitudinal analysis of

union decline and wage inequality, 1973–2015. Social Forces 97(2): 495–530.

Visser, Jelle. 2011. Database on institutional characteristics of trade unions, wage setting, state

intervention and social pacts, 1960–2010, version 3.0. Amsterdam Institute for Advanced

Labour Studies (AIAS), University of Amsterdam.

Waddoups, C. Jeffrey. 2001. Unionism and poverty-level wages in the service sector: The case

of Nevada’s hotel-casino industry. Applied Economic Letters 8(3): 163–67.

Walsworth, Scott. 2010. Unions and employment growth: The Canadian experience. Indus-

trial Relations 49(1): 142–56.

Wilson, William Julius. 1996. When Work Disappears: The World of the New Urban Poor. New York:

Norton.

Wooldridge, Jeffrey M. 2010. Econometric Analysis of Cross Section and Panel Data

. Cambridge,

MA: MIT Press.

Zuberi, Dan. 2006. Differences That Matter: Social Policy and the Working Poor in the United States

and Canada. Ithaca, NY: ILR Press, an imprint of Cornell University Press.

LABOR UNIONS AND AMERICAN POVERTY 917