Wal-Mart Welfare?

:
Business Power and the Politics of Work Support in the American States

January 2007
 Abstract

States vary considerably in the extent to which they use antipoverty

programs to support workers and their families. The question is why.
Regression analyses show that income continues to be the single most
consistently important variable in determining work support policy
choices, although ideology and race play supporting roles. Business
political power has no effect on minimum wage increases and a
negative effect on Medicaid eligibility for working parents, which
suggests two things: active political lobbying is not the most effective
way for business to shape antipoverty policy, and ideological and
institutional factors are outweighing economic self-interest in the
business community.

2
 Since the 1970s, American social policy has undergone a

significant programmatic shift from providing support for the traditional

welfare population to helping low-income workers and their families .

This new emphasis on work support, through a new temporary cash

assistance program, health coverage, child care subsidies, food

stamps, earned income tax credits, and increased minimum wages, is

largely recognized as the inevitable result of public demand and a

bipartisan consensus among political elites that traditional “welfare”

(Aid to Families with Dependent Children, or AFDC) was an untenable

antipoverty strategy. The momentum for reform peaked in 1996 with

the overhaul of the welfare system, the various components of which

had far-reaching consequences for broader antipoverty policy.

The two policy areas that perhaps best represent this shift are

public health care (including Medicaid and the State Children’s Health

Insurance Program, or SCHIP) and the federal and state earned income

tax credits (EITCs). These programs, combined with both federal and

state minimum wage laws, constitute the bulk of support for low-

income workers and their families. But despite national agreement

among the public and elites that work support programs are an ideal

antipoverty tool, there is substantial state-level variation in both the

amount and kind of work support available for low-income workers.

3
Twenty states have consistently high or low work support (in some

cases very high or very low), while the other thirty states offer

moderate levels of work support through a range of complex patterns –

ten different patterns, in fact. The objective is to explain this variation.

Previous analyses have highlighted a wide range of variables

likely to affect state antipoverty policy choices, most notably state

wealth, ideology, and race. However, the emphasis in this analysis is

on an explanatory variable that is little discussed in the welfare policy

literature: employers. The guiding assumption is that recent policy

changes give employers new incentive for involvement in antipoverty

policy decisions. For many firms with low-wage employees, public

health care programs and wage subsidies allow them to externalize

some of their biggest – and growing – labor costs, while reducing

demand for higher wages, increasing productivity, and reducing

absenteeism. Minimum wages, on the other hand, directly impose

costs on employers and often force them to reduce employee hours or

raise prices. I call this new system of public benefits for low-wage

employees “Wal-Mart Welfare” for its implications for both workers and

their employers. However, not all businesses are affected equally, and

not all policies have the same effects, creating multiple sources of

variation that are ripe for analysis.

Through cross-sectional regression analysis, I find that there is a

politics of work support, though not in the sense that a uniform set of

4
factors affects all work support policies in the same way. State wealth

is the only factor shared by all five policies, regardless of whether the

policy requires state spending. Race – that is, the percent of the state

population that is black – has a negative effect on the minimum wage,

but does not have the expected negative effect on welfare-stigmatized

Medicaid. Party control affects only SCHIP eligibility levels and

minimum wages (which are also shaped by public liberalism, probably

because many wage increases occur through referenda).

Based on what we know of business political activity, we might

expect that employers would have a negative effect on minimum wage

increases and no effect on Medicaid. But the regression analyses show

precisely the opposite: business political power has a significant effect

on policy decisions in Medicaid (and to a lesser extent SCHIP), and no

effect on minimum wage policies. This suggests that direct lobbying is

not the most effective form of political power in this policy domain, and

that we must consider alternate forms of power.

Importantly, the effect of employers on Medicaid eligibility is

negative: in states where business is more politically powerful, fewer

working parents can access public health care. In some cases, this is

likely due to active business opposition to bills that fund program

expansion through direct costs on employers. But where the benefits

for employers outweigh the costs, the business community does not

seem to be shaping the policy process in a way that best serves its

5
self-interest. This is likely due to a combination of two things. First,

institutional and ideological forces overwhelm the ability of low-wage

firms to support government programs even when they are in their

material interest. Second, perhaps because of these constraints,

policymakers anticipate that business groups will oppose public health

care program expansions in the future, and make policy decisions

accordingly.

Background: The Shift to Work Support in the States

The four work support policies at hand represent very different

approaches to the ideal of supporting work through government policy,

varying in how they are financed, which families they benefit, and how

they are perceived. But they share two important characteristics: they

all offer support to low-income workers, and in all cases state

policymakers have a large amount of autonomy in policy design.

Medicaid and SCHIP

The shift toward supporting workers is perhaps least surprising

in Medicaid, the public health care program for the poor. Due to a

piecemeal American health care system that relies on public programs

for the very poor and employer provision for the middle and upper

classes, uninsurance is largely a problem of lower-income working

families. To address this, the federal government made gradual

changes in the 1980s, combined with provisions in the 1996 welfare

6
reforms, that opened Medicaid to more working families by severing

the link between Medicaid and cash assistance (programs in which

recipients historically enrolled simultaneously) and allowing states to

determine which of their residents were eligible. In recent years,

Medicaid expansions achieved through waivers have been the

preferred way for states to reduce the number of working uninsured .

An even more crucial federal policy change that has affected the

families of low-income workers was the 1997 enactment of the State

Children’s Health Insurance Program (SCHIP), a program similar to

Medicaid but targeted at children in families with higher earnings. At

the time, 10 million children were uninsured, 90 percent of whom lived

with a parent who worked . States had the ability to design their own

programs (within federal parameters), this time from scratch. Most

research indicates that it has been successful in helping to fill the

insurance gap .

In both Medicaid and SCHIP, the new state autonomy has led to

dramatic variation across states in terms of who can access public

health care. This is partly because state health policy decisions often

involve choosing between programmatic goals. Workers and their

families constitute only one needy group, and compete with other

categories of recipients such as the elderly, the disabled, and

unemployed single mothers. And because federal benefit standards

are extremely low, states can choose how to allocate their resources;

7
for example, whether to cover mandated groups at high benefit levels,

or to expand access and spread resources more thinly across a broader

range of groups. The goal is to explain the conditions under which

states will expand health programs to give access to workers and their

families.

State Earned Income Tax Credits

The earned income tax credit (EITC) differs from public health

programs in that it is more targeted (only workers with children

benefit), it is funded through the tax code (which tends to hide costs),

and federal and state programs are separate (states can choose to

enact their own EITCs, and fund them entirely themselves). The

federal credit, which had modest beginnings with its quiet enactment

in 1975, is today the largest, fastest growing non-elderly antipoverty

program . The effect on low-income workers is substantial and direct.

The EITC is basically a wage subsidy, paid for by taxpayers, which adds

as much as $1 to $2 to a worker’s hourly earnings . It also acts as an

incentive to work, and research has shown that it has significantly

increased the marginal return to work for very low income families .1

Because of its popularity and its bipartisan support, states began

to follow the federal model, starting with Rhode Island in 1986. Like so
1
The effect is strongest on the labor force participation rates of single mothers, who
are most likely to find jobs in the low-wage service industry .

8
many other antipoverty programs, state EITCs were affected by the

1996 welfare reforms, which allowed states to devote part of their

TANF allocations toward refundable state earned income tax credits.

Today, twenty states and the District of Columbia have their own

earned income credits, all of which are linked to the federal credit but

funded exclusively by states. Their generosity varies widely. The

politics behind earned income credits, particularly at the state level,

have gone largely unexamined.

State Minimum Wages

Despite the fact that minimum wages are not generally thought

of as welfare policies, they are unquestionably a core element of

government support for low-wage workers. They differ most notably

from EITCs and public health programs in that government spends

nothing when minimum wages are enacted; the costs are borne by

employers and consumers. Minimum wages are also unique in that

states took the lead in creating them, beginning with Massachusetts in

1912 and followed by 14 other states over the next decade. (A federal

wage floor wasn’t enacted until after the Great Depression in 1938.)

From 1997 to 2007, the federal minimum languished at $5.15,

eventually holding its lowest real value in over fifty years. Over the

course of those ten years, inflation increased 26 percent, dramatically

reducing the purchasing power of the minimum wage over time .

9
 States responded to the declining value of the federal minimum

wage by raising their own minimums. 29 states had set minimum

wages above the federal level as of January 2007 (before the most

recent federal increase). At $7.93, Washington currently has the

highest. Minimum wage increases are highly salient among the public

(one Pew Research study found that 83 percent of respondents support

an increase ), and many successful minimum wage increases occur

through ballot initiatives (including six increases in the November 2006

election cycle). Few scholars have examined the politics of the

minimum wage, and even fewer have considered state minimum

wages. How are they different from the politics of tax credits and

public health programs?

Business Political Power and Antipoverty Policy

This paper is largely motivated by the speculation that the shift

toward greater support for workers will affect the politics behind

antipoverty policy choices. Most importantly, the emphasis on work

support will draw different groups into the policy debate. I am

particularly interested in employers, who have traditionally been

absent from most aspects of antipoverty policymaking outside of the

minimum wage, but now have a material interest in the other choices

states are making. The question is what role do businesses and their

political organizations play in state antipoverty choices – does their

relative power or their deliberate lobbying efforts shape which states

10
will prioritize workers over other low-income residents, or what policies

will be chosen?

In order to hypothesize that business will affect the policy

debate, it is not necessary to establish that the business community

has been actively involved on these issues. It is necessary only to

establish that the business community is politically engaged generally

and does have the potential to influence the political process when it

comes to relevant antipoverty policies, and that these four work

support policies are relevant. In fact, business can influence policy

outcomes without even forming a clear preference on the particular

issue at hand.

This is an important point because business rarely actively

involves itself in most social policy, both for institutional and

ideological reasons. The most powerful American employer

organizations are massive umbrella associations that represent

extremely heterogeneous businesses with divergent interests and

political preferences . These groups tend to take positions only on non-

contentious issues with unambiguous, short-range effects on the

business community (Martin 2000) – characteristics rare in antipoverty

policies. In addition, these organizations have a general anti-

government ideological bias that dominates preference formation .

This semi-consistent conservative ideology has been beneficial, as

business gets much of its power and access from its alliance with the

11
Republican party . Support for social program expansion would violate

the tenants on which many of these organizations are based, and

would likely antagonize their political allies.

Without direct involvement, how can employers affect outcomes?

As long as they are involved in the political process and have the

power to affect political decisions generally, they can affect any of

these antipoverty policies indirectly in two ways. First, by taking

positions on related issues, employers can substantially affect the

likelihood of a bill’s passage. As issues that are controversial or require

additional funds, work support policies are often packaged with other

policies when they are presented to a state legislature. For example,

Medicaid and SCHIP expansions are expensive, and bills will often be

combined with financing mechanisms such as employer mandates or

cigarette tax increases. Business groups will frequently help block

passage of the bill based on their opposition to these funding sources,

regardless of how they feel about Medicaid expansion itself. (This is

precisely what happened in recent years in Maryland.) Business can

also have a positive effect because of its view of the package as a

whole. For example, minimum wage increases are almost always

combined with tax breaks or other benefits for small business owners

that offset – or sometimes more than offset – their increases in labor

costs. Regardless of how vehemently local business groups have

opposed the minimum wage increase itself, these benefits cause

12
employers to back away. Given that employers are often the primary

opposition to such bills, by not opposing the bill employers then

indirectly aid in its passage.

Second, business can have tremendous indirect influence on

policy through the actions of lawmakers. There is a great deal of

evidence that policymakers anticipate business reactions to policies,

limiting the menu of viable options to programs that business would

potentially support (or at least not oppose) . Even if business is not

involved in the policy debate, policymakers can estimate the business

effect, or read the signals sent by the business community, and make

reasonable guesses about their responses to proposed policy changes.

When policymakers anticipate business reactions, they are

looking for two potential types of response. One is simply future

political activity. Politicians do not want to risk business opposition,

particularly if business organizations are politically powerful in their

state. But another possibility is that business could disinvest or

relocate to another jurisdiction. This type of business influence is

known as structural power, and relies on the crucial economic role

played by employers and investors . This is particularly relevant at the

state level, where differing economic and tax policies make the threat

of capital flight sufficiently credible to constrain the policy choices

available to politicians.

13
 All of these scenarios require that employers be politically

powerful – and there is evidence that business does indeed occupy a

“privileged position” in American politics, either from instrumental

power (power rooted in disproportionate resources and active

involvement in political affairs) or structural power (power rooted in the

crucial role of business in the American economy, which constrains the

policy choices available to politicians) . Moreover, it appears that

employers have greater influence in state policy than federal .

However, this influence is inherently limited both because of

heterogeneity in the corporate community that hinders action and

systemic features that allow other groups – and voters – to have voice .

As for welfare policy itself, there are some indications that

business has had the capacity to shape outcomes in the past, although

most discussions focus on national policy, and they are generally

limited to the minimum wage, Social Security, or unemployment

insurance, since these are the policies with which employers have

historically been involved . Research on state welfare policymaking

has little to say about business and antipoverty policy. In fact, these

studies have tended to rehash the same variables in study after study,

focusing almost exclusively on TANF and using expenditures as the

default dependent variables, and paying little attention to interest

groups at all. The only known study to empirically test the effect of

business power on a state means-tested antipoverty policy found that

14
business lobbyists can negatively affect time limits and generosity in

TANF . This finding suggests that the failure to examine business in

this arena has been an important oversight.

If employers have the potential to affect relevant antipoverty

policies, the final task is to establish that work support programs are

indeed relevant for employers today. The policy with the clearest

effect on business – and the policy with which business has historically

been the most involved – is the minimum wage. Employers have

largely opposed the minimum wage ,2 believing that they will have to

offset increases in wages with one of several equally undesirable

options: raise prices, reduce work hours for current employees, or hire

fewer employees in the future. In addition to its economic costs,

businesses also resent the minimum wage on an ideological level: it

represents unacceptable government intrusion into private labor

contracts .

Other work support policies, such as Medicaid, SCHIP, and the

EITC, have less obvious but nearly as important effects on employers.

Of course, one way that these programs are relevant to business is

that they are funded with taxes, which can be seen as a cost (though

one spread across all taxpayers rather than directly imposed on

employers). However, for a number of reasons these programs also

potentially reduce labor costs. First, access to Medicaid and earned

income credits has been shown to reduce worker demand for increased
2
However, see Swenson for an opposing view.

15
wages and benefits from their employers, particularly at low-wage

firms . Even if workers continue to demand greater compensation,

they may have weakened bargaining power due to an increased labor

supply. Second, employees can opt out of expensive employer benefits

when public health care programs are available, a phenomenon known

as “crowd-out” .3 Third, greater financial security and physical health

for both employees and their families would be expected to increase

productivity due to reduced absenteeism and fewer preventable

catastrophic illnesses . Fourth, government benefits reduce risk for

employers, decreasing volatility of labor costs and protecting

employers from swings in the cost of health care. Finally, government

health care programs could reduce employers’ labor costs by lowering

health care costs overall. By covering more uninsured workers and

their families, antipoverty programs help eliminate the uncompensated

care that is at least partly to blame for skyrocketing costs .

The fact that these programs can potentially lower employer

labor costs becomes even more important when we consider that they

are often alternatives for policies that have similar effects on workers

3
The question that arises is whether employers’ labor costs will decrease if they
reduce spending on benefits, or whether these costs will be shifted to wages.
Business is likely to reap at least short-term gains because an individual employee
who opts out of employer-based insurance will not see a corresponding increase in
wages. That is, total compensation for the employee will decrease. In the long term,
however, most economic models assume that market forces are likely to bring total
compensation back to equilibrium levels, with no change in labor costs regardless of
the distribution between wages and benefits. These economic assumptions are not
universally accepted , and most importantly do not resonate with business managers,
who believe quite strongly that their labor costs will decrease if they spend less on
health care .

16
but raise labor costs for employers. For example, the EITC is often

championed as an alternative to minimum wage increases, and

Medicaid expansions can substitute for employer health mandates

(which many states have considered imposing). In an environment

where some relief for low-income workers seems to be a political

inevitability, interest groups that might have otherwise stayed out of

the debate have an incentive to fight for more preferable policies.

Putting it all together: theory and hypotheses

This paper does not claim that business will dominate the

agenda in work support policymaking, nor that it will be the driving

force behind policy decisions. Rather, interest groups will be one of

several factors that shape state policymaking. Evidence from TANF

(and, occasionally, Medicaid spending) suggests three others that are

also likely to be highly relevant: state wealth, partisanship, and race.

The hypotheses in this paper are built around a key expectation:

the factors that affect state policy decisions will vary depending on a

policy’s characteristics. The one exception is state wealth, which will

likely affect all antipoverty policies. States with higher median

incomes have a greater tax base from which to finance programs.4 For

example, more prosperous states have been found to have higher

4
It is conceivable that the condition of state budgets would be more likely to affect
Medicaid, SCHIP, and the EITC, given that states would be more likely to expand
programs in times of surplus regardless of median income. However, this would be
nearly impossible to measure in a cross-sectional analysis because budgets can
change dramatically over time and even from year to year. State income is a more
reliable measure because it does not change substantially over time.

17
welfare benefits and public health coverage for workers .5 But state

wealth will also likely affect the minimum wage, which requires no

government funding. States with higher median incomes have fewer

workers earning very low wages, which means there will be less

opposition to an increase.

Aside from wealth, however, policy features determine which

factors will come into play during policy battles. The four policies in

this analysis share many characteristics – primarily, they all exclusively

benefit the poor and near poor. However, even within this category of

policies there are three key distinctions that affect policy choices:

whether the policy is theoretically controversial; whether it is salient

for the public, political parties, or competing interest groups; and

whether it is associated with welfare and poverty.

The first consideration is whether the policy is theoretically

controversial. One policy at hand is actually quite uncontroversial: the

EITC. Almost everyone – including conservative economists – agrees

that this is an ideal way to provide incentives for the poor to get and

retain jobs. In this situation, the only factor that should matter is

whether the state can afford it. A policy that almost everyone supports

will not face opposition from either party (unless the size is extreme),

5
The effect of income on health programs is complicated somewhat by the role of
federal matching rates for Medicaid and SCHIP, which give poorer states more federal
money. This could act as an incentive to expand programs, and personal interviews
have revealed (at least anecdotally) that these states feel some desire to expand
programs to avoid leaving money on the table. However, it is unlikely that this
incentive overrides the capacity for states to fund their shares.

18
nor will interest groups have much of an effect. Therefore, I expect

that the only factor to shape state EITC decisions will be state wealth.

The remaining three policies are all theoretically controversial –

there is substantial disagreement on how to go about increasing the

earnings of or providing health care for the poor and their families.

Other factors then come into play. First is whether the policy is salient

among the public, political parties, and competing interest groups.

When policies are salient to competing interest groups, this may

counter the effects of business influence . The policy where this is

most relevant is the minimum wage, where organized labor has

actively supported increases. At the federal level, union campaign

contributions were found to be the most significant variable that

affected whether legislators voted for minimum wage increases .

Organized labor has also been shown to positively affect general state

spending , but there has been little evidence that they have had any

effect on public health care programs or EITCs. Of course, the power of

unions has been waning for some time, which reduces the likelihood

that they will have any substantial mitigating effect on business

influence. With respect to health policies, the primary opposition

would be progressive health groups that advocate particular expensive

Medicaid and SCHIP expansions that employers would oppose.

Policies that are salient to the public and parties will be less open

to interest group influence. Issues that are easily comprehensible and

19
publicly salient are more likely to generate political rewards (or

punishment) from voters, and thus policymakers are more likely to

cater to public opinion on those issues . (Conversely, highly complex

issues with low public salience will be more open to interest group

influence. Particular features of Medicaid fit into this category.)

Therefore, when policies are publicly salient, factors other than

interest group power will be more likely to drive outcomes. Ideology,

channeled through party control of government, is likely to be a

significant determinant.6 Studies have shown that greater Democratic

composition of state legislatures leads to more generous TANF ,

Medicaid , and unemployment benefits. Democrats have also proven

more willing to tax residents . As liberalism increases (which is related

to partisan affiliation but measures intra- as well as inter-party

differences), Medicaid spending increases and TANF programs become

less restrictive .

We can expect that partisan affiliation will likely affect both

Medicaid and the minimum wage, with Democrats historically in favor

of expansions and increases. It is also likely that public liberalism (and

not just legislative party control) will affect minimum wages directly,

given that in many states increases can be enacted through referenda.

SCHIP has historically been less ideological than these two policies

6
Ideology among the electorate has also been shown to affect policy choices, both in
general and with respect to welfare policies , but with the exception of the minimum
wage (when it is passed by referendum) the effect is indirect. That is, policy reflects
the preferences of the public because electorates choose representatives that mirror
their own ideological preferences.

20
(and was highly bipartisan at its inception), but debates over eligibility

expansion often ultimately fall along party lines.

Another consideration that will determine which factors affect

policy decisions on an issue that is theoretically controversial is

whether it is connected to welfare. Cash assistance programs that are

traditionally thought of as “welfare,” such as AFDC and TANF, have

been closely tied with race; many people believe that welfare

disproportionately benefits blacks and Latinos, which negatively affects

program popularity . This relationship continues in state policymaking,

and a number of studies have found a negative relationship between

race (measured as both the total percentage of blacks in the

population and the proportion of AFDC caseloads made up of blacks

and Latinos) and welfare generosity . Therefore, it is reasonable to

expect that other antipoverty programs that are associated with

welfare will have similar racial effects. Medicaid is the most likely to

have this stigma, as historically only welfare recipients could enroll in

the program, and it is expected that race will still negatively affect the

generosity of Medicaid programs. The minimum wage, which is the

most explicitly tied to poverty, might also connect with race.

Based on the above discussion, the following are the main

hypotheses:

21
• EITC – The policy is theoretically uncontroversial, so only state

wealth will have a positive effect on the existence and generosity

of a state’s earned income credit.

• Minimum wage – State wealth, Democratic party control in state

government, and public liberalism will have positive effects on a

state’s minimum wage. The proportion of black state residents

will have a negative effect. It is unlikely that business will affect

outcomes due to the policy’s public salience and simplicity.

• SCHIP – State wealth and Democratic party control will have

positive effects on SCHIP eligibility levels. Business political

power may have little effect on outcomes due to its public

salience.

• Medicaid – State wealth will have a positive effect on Medicaid

eligibility levels, while the effect of the percent black in the

population will be negative due to its association with welfare.

Business political power will affect Medicaid eligibility indirectly,

though it is unclear whether this effect will be positive

(policymakers anticipating that employers would support this

over more costly solutions to uninsurance) or negative

(employers opposing funding mechanisms).

Data Measures and Results

22
 The dependent variables measure the extent to which workers

benefit from each policy. For Medicaid and SCHIP, eligibility thresholds

are used as indicators of work support. This appears as a percent of

the federal poverty level (FPL) under which household income must fall

in order for a child or working parent to qualify. In Medicaid, all states

have had to decide whether to keep the minimum federal eligibility

level, set as part of the welfare reforms in 1996, or to apply for a

waiver to increase the threshold. Most states have increased their

limits, but there is substantial variation.7 In SCHIP, all states designed

their own programs in the late 1990s following the federal allocation of

funds. The recommended level was 200 percent of the federal poverty

level, but in 2005, 15 states had thresholds higher than 200 percent,

while 10 states were below 200 percent.8

State EITCs have never been quantified in a way that makes

them comparable. I created a composite measure, the first part of

which is based on whether the state has an EITC. If so, a state receives

additional points for refundability (i.e., allowing residents who do not

have tax liabilities due to low earnings to claim the credit) and level of

generosity (which ranges from 5 to 50 percent of the federal EITC).

7
2005 eligibility limits ranged from 18 percent of the poverty level (Alabama) to 275
percent (Minnesota). The average threshold was 86 percent.
8
The average SCHIP threshold was around 226 percent, ranging from 140 (North
Dakota) to 400 percent (Massachusetts).

23
This scale ranges from 0 to 5, with 0 indicating no state credit and 5

indicating the most generous policy.9

Finally, the minimum wage is a straightforward variable

indicating a state’s legal minimum. The five states with no minimum

wage policy were assigned a value of $0.10 As of January 2007, 29

states had federal minimums above the federal level, while 15 states

have enacted legislation setting their minimum to the federal level.

Only one state, Kansas, had a state minimum wage below the federal.

The average minimum wage was $5.58.

The regression analyses examine the effect of six independent

variables on the work support scores for each policy.11 Wealth is

measured by state per capita income (in thousands).12 For

partisanship, a score of Democratic party control of state government

was computed that takes into account both houses of the legislature

and the governorship.13 The inclusion of the governor is critical, since

9
The mean score is 1.42. Eight states have no income tax and were assigned the
mean value. Results did not change when these states were dropped from the
analysis.
10
This coding decision is supported by state legislative data (collected by author)
showing that many of these states have rejected bills that would set the state
minimum wage to that of the federal level. In these states, workers not covered by
the federal minimum are not subject to a wage floor. Thus, there is a substantive
difference between these states and those that have enacted state minimum wages
equal to the federal level.
11
Most independent variables are calculated as the average of three data points:
1986 (the year the first state EITC was enacted), 1996, and 2004. Specific years
used vary depending on data availability. The averages are taken to account for
change over time and lags in time between the phenomena measured and policy
changes. They also minimize one-year spikes in the data.
12
Average of 1986, 1996, and 2005, all adjusted to 2005 dollars. Data from the
Bureau of Economic Analysis, Regional Economic Accounts. Mean: $29,373; standard
deviation: $4,151.
13
This is a 4-point scale created by Robert D. Brown that takes into account partisan
majorities in each house of the legislature and in the governor’s office. Higher scores

24
executive leadership can determine the fate of many of these issues.

Race was measured as the percent of black residents in the state

population.14 Interest group power is measured in two ways: the

proportion of campaign contributions from business and from labor,

and the percent of large firms that are in low-wage service industries.

Multiple other variables, including political culture, unionization rates,

the percent of workers in manufacturing, uninsurance rates, the cost of

a state EITC, legislative professionalism, poverty rates, and state tax

rates have been tested in previous iterations and have been found to

have no effect (or they are strongly correlated with other independent

variables). Concerns about multicollinearity in the six key independent

variables are minimal.

The interest group variables merit some explanation. Interest

group power is incredibly difficult to measure.15 This analysis uses two

indicate more Democratic control. Averages scores over multiple election years
(every two years from 1986 to 2004). Mean: 2.63; standard deviation: 0.76.
14
Average of 1990, 1996, and 2004 percentages. Data from US Census Population
Estimates Program and the American Community Survey. Mean: 10.22%; standard
deviation: 9.57%. The percent Hispanic and the percent of total nonwhite residents
were also tested. The percent Hispanic had no effect on the results, and the percent
nonwhite either had no effect or the effect was similar to the effect of the percent
black, indicating that the black population was the relevant measure.
15
Several other measures of business power were considered and rejected due to
problems with theory or data. Perhaps the most commonly used measure is Gray
and Lowery’s database of registered state lobbyists. These data were tested but
ultimately rejected due to concerns about data quality (largely because state
lobbying registration rules vary substantially from state to state). These data also do
not improve much upon the campaign contribution data because they suffer from the
same problem as most measures of business power: they are potentially
endogenous. As a measure of structural power, a variable was constructed that
measured the proportion of a state’s revenue that came from corporate taxation, but
this measure was deemed problematic because tax policies are likely to be
dependent on many of the same factors that shape antipoverty policies. Another
way to operationalize the potential power of low-wage employers is the percentage of
the labor force working in low-wage service industry jobs. This measure was
computed but ultimately rejected because it says little about centers of business

25
measures. First, state election campaign contributions are employed

to capture instrumental power (i.e., power that comes from direct

political activity).16 The assumption is that in states where business or

labor interests contribute a high proportion of campaign funds, these

groups will have more power in determining later policy choices.

Admittedly, this is a crude measure because it is possible that groups

contribute to campaigns not because they are strong, but because

they are weak and want to increase their power.

The second business measure attempts to capture the

dominance of low-wage service sector employers in the economy by

measuring the proportion of large firms (those with more than 1000

employees) that come from these industries.17 This accounts for the

fact that low-wage service industry employers are much more likely to

care about antipoverty policy because they will be most affected. This

measure most likely captures potential power. In states where large

low-wage firms are more dominant in the economy, rational

power (these employees could be scattered throughout the state in various small
businesses, which may not have any political power as a group). The variable also
had no effect on the results when included in the model. Finally, the percent of a
state’s employees who work at Wal-Mart was tested but had no effect.
16
Average percentage from 1996 and 2004 (or nearest year available). The business
percentage includes general business; finance, insurance, and real estate;
agriculture; transportation; construction; health; and communications and
electronics. Two separate measures of contributions from low-wage industries and
from business associations were tested and rejected for their extremely small
percentages and the lack of any independent effects. Percentages were calculated
from the Institute on Money in State Politics. Business mean: 28.64%; business
standard deviation: 8.98%. Labor mean: 4.96%; labor standard deviation: 2.96%.
17
Includes two industry categories: retail trade, and accommodation and food
services. From the average of 1998 and 2004. From County Business Patterns data.
Mean: 3.97%; standard deviation: 9.77%.

26
policymakers would be more likely to take their actual or potential

preferences into account.

All models are cross-sectional, and are tested using standard

OLS. Model 1 (table 2) examines the variables that affect Medicaid

eligibility levels (i.e., Medicaid work support). There are two extremely

clear relationships. The most significant (at the 1 percent level of

statistical significance) is a strong positive association between per

capita income and Medicaid eligibility levels. This is also substantively

significant: a $10,000 increase in median income (which would be the

equivalent of moving from Arkansas to Rhode Island) would increase

the Medicaid eligibility threshold by 46.2 percentage points (a

meaningful increase when the average threshold is only 86 percent of

the poverty level). The second finding is that business campaign

contributions are negatively associated with Medicaid eligibility

(p=.03). This striking finding holds through multiple variations of the

analysis, and is substantively meaningful. A ten percent increase in

business contributions can be expected to decrease Medicaid eligibility

levels by 19 percentage points.

Model 2 (table 3) looks at SCHIP. Three measured variables have

statistically significant effects. Two of these were extremely significant

(p<.01): both median income and Democratic party control have

strong positive effects on SCHIP eligibility levels. A $10,000 increase in

median income would raise SCHIP eligibility by 86.1 percentage points,

27
well over one standard deviation. The effect of party is slightly less

substantial; if one branch of state government were to switch party

control, eligibility levels would be expected to increase by 27

percentage points.

Model 3 (table 4), which considers the factors that affect state

EITCs, has somewhat weak explanatory power, with an adjusted r2

value of 0.20. The results here are clear: only income has a significant

effect, and its relationship to EITCs is quite strongly positive (significant

at the .01 percent level). (This finding held through multiple variations

of the model.) Its effect is also very substantial: a $10,000 increase in

median income would raise the EITC score by more than one standard

deviation. A second model substituted estimated EITC costs for

income, hypothesizing that higher cost would lead to a lower likelihood

of enacting the credit.18 Indeed, program cost was negatively

associated with EITC score, at the .01 percent level. However, in this

model there was a second significant variable: race. The percent

black, surprisingly, had a positive impact on EITC score (p=.03).

The six key variables are regressed on state minimum wages in

model 4 (table 5). This model has good explanatory power (the

adjusted r2 value is 0.51), and two variables are shown to be highly

statistically significant, with p-values below the 0.01 percent threshold.

18
The correlation between income and EITC cost (which is primarily a measure of
poverty in the state) is -0.62, excluding the possibility of keeping both variables in
the model. EITC cost is per resident, calculated based on 2003 federal EITC claims
from state residents. From “A Hand Up,” table 8 .

28
The most important of these is race: the percent black in the

population has an extremely negative impact on state minimum

wages. A ten percent increase in the black population would lower a

state’s minimum wage by $1.27. (This result holds when including a

dummy variable for southern states.) Income is also strongly related

to minimum wage levels, with a $2.60 increase in the minimum wage

for every $1000 in median income. These results were unchanged

when substituting state poverty for income.19 (Both could not be used

because of multicolinearity.) Interest groups have no statistically

significant effect on minimum wage levels; the coefficient for business

campaign contributions is actually negative.

Discussion

The findings are expected to be imprecise (but efficient and

unbiased) due to the small sample size and the inherent difficulties in

measuring power. The goal of the analysis is to isolate key

relationships and to determine whether theoretical expectations are

borne out in the data – and to consider potential explanations when

they are not.

Is there a politics of work support? There are two ways to answer

this question. First, table 1 demonstrates how the four work support

policies are related to one another. Twenty states can be considered

19
This is a measure of the percent of the state population with income below 200
percent of the federal poverty level. Data from 2004-5. Ranges from 23 to 48
percent.

29
either high or low work support states. That is, they demonstrate clear

patterns in their choices of either supporting or not supporting low-

income workers. That these states follow one of these two patterns

reveals that all four policies follow broadly similar political logics. To

some extent, states have characteristics that lead them either to use

government policy to support workers, or to rely more on markets.

Moreover, the states that fall into these categories would surprise no

one. The nine “high work support states,” which have above average

Medicaid eligibility thresholds, SCHIP thresholds at or above 200

percent, earned income credits, and minimum wages above the federal

level, are all northeastern (with the exceptions of Minnesota and

Illinois), historically liberal, and quite wealthy. Eleven states are “low

work support states,” meaning they have below average Medicaid

eligibility thresholds, SCHIP thresholds at 200 percent or lower, no

earned income credits, and no separate minimum wages. Six of the

low work support states are southern, five are western, and all are

relatively poor and conservative. Many of these are the same states

that have offered either high or low levels of traditional welfare

support.

However, 30 other states don’t follow these patterns at all. They

demonstrate some commitment to helping low-income workers and

their families, but choose only a selection of the policies to accomplish

this goal. These states are less predictable, and represent ten different

30
patterns. The most common pattern is for states to have a higher

minimum wage than the federal level, while remaining at lower levels

on the other three policies. In this category, all but Michigan are

thought to be politically conservative, and many of these states are

poor. The second most common pattern, shared by six states, is to

provide work support only through an EITC. This appears to be a

Midwestern phenomenon (with the exception of Virginia). The other

moderate work support states fall into eight different patterns, none of

which are followed by more than three states. This implies that 16

states are too idiosyncratic to generalize without more sophisticated

analysis.

The second way to identify the politics of work support is to tease

out more specific patterns through regression analysis. The clearest

finding from the models is that income matters substantially across all

four policies—it is, as the theory predicts, the only variable that is

universally important. State wealth has a positive relationship with

work support policies no matter how it is conceptualized, whether as

state median income or a measure of poverty such as the cost of

enacting an EITC or the percentage of low-income residents. States

are simply constrained by their available resources.

The second important finding is that partisanship matters only in

policies with particular characteristics. I hypothesized that party

control would affect the policies that were theoretical controversial and

31
salient to the public and political parties, a category that includes

SCHIP and minimum wages. The only policy for which partisanship

mattered significantly at the 5 percent level or better was SCHIP;

Democratically controlled state governments will have more expansive

programs. Partisanship had only a weak effect on minimum wages, but

the effect was stronger with public liberalism, reflecting the fact that

the minimum wage can be enacted in the legislature or at the ballot

box. In other words, public opinion matters substantially more for this

policy than for the other three, where public opinion is always filtered

through the agents.

The third clear finding from the analyses is that race does matter.

The percent black in a state’s population had a negative effect on

minimum wages, as expected. Whether this is due to racial bias or

something else is unclear. However, we can be sure that race is not a

proxy for liberal public opinion. First, and most importantly, percent

black and public liberalism are actually negatively correlated (r=-0.4).

Second, when public liberalism is controlled for in the model, the

negative relationship between the black population and minimum

wages remains.

The most surprising finding related to race is that when

controlling for poverty in the EITC model (as measured by the

estimated cost of enacting a state EITC), race has a positive effect.

That is, states with higher proportions of black residents are more likely

32
to have EITCs. It is unclear exactly how to account for this finding (and

it is important to note that race is not at all significant in the primary

model). At the very least, this indicates that the politics of the EITC

differ substantially from those of the other policies.

However, race did not have a statistically significant effect on

Medicaid eligibility levels (although the sign was in the expected

negative direction), despite Medicaid’s historical connection with

welfare. This may indicate that Medicaid has been effectively delinked

from AFDC and TANF, or simply that eligibility levels are not the policy

feature that racial biases affect. SCHIP was also unaffected by race,

although I did not expect to find a relationship since the program is

symbolically removed from welfare, and the “deservingness” of

children likely overrides negative racial stereotypes.

There are four primary conclusions about business power when it

comes to work support policies. First, the economic dominance of low-

wage service sector employers in a state has no relationship with work

support policies. It may be that these large businesses are not

sufficiently politically active to affect lawmakers’ decisions, or that

their political activity is channeled through campaign contributions

(which means their influence is captured in that variable). It may also

be that this measure simply isn’t capturing an important element of

business power.

33
 Second, as expected, business has no influence on EITCs.

Policies that are not theoretically controversial are likely to only be

shaped by the state’s ability to fund them. This is true regardless of

the material benefit a state EITC may have for low-wage employers. If

it is possible for interest groups to influence this category of policies at

the margins, perhaps by raising the profile or priority of the policy in

the competition for limited funds, they are not doing so in this case.

Third, business power, as measured by the proportion of state

campaign contributions coming from business entities, does have a

statistically significant negative effect on Medicaid (and a statistically

weaker negative effect on SCHIP). In states where business

contributes more to state elections, ceteris paribus, public health

programs have lower eligibility thresholds. In other words, business

political power is having the effect that is the opposite of what would

be in most employers’ self interest.

How can we explain this effect? There is little evidence that the

business community has engaged in the debate over public health care

at the state level, so the effect must not be due to direct business

persuasion on these issues. However, it does seem that some indirect

influence is taking place. Most likely, politicians see the business

community as relevant to the public health care debate, and are

anticipating the reactions of business leaders to their policy decisions.

However, these politicians are not anticipating that employers will

34
support these programs, despite their direct material benefit – they are

anticipating employer opposition. Perhaps this is an assumption based

on experience: certainly there are a few key principles that the

business community (or at least the handful of prominent business

associations) has consistently espoused: low taxes and small

government chief among them. These ideological constraints,

combined with the institutional incentives for prominent business

organizations to remain closely allied with the Republican party,

probably account for a great deal. They prevent employers from

actively lobbying in favor of big government programs, even when

these programs are in their material self interest, and they prevent

policymakers from anticipating such support in the future.

It is important to note that business power has affected Medicaid

eligibility levels despite the salience of the policy for political parties.

Most policymaking theories predict that interest groups can have little

impact on such salient policies because legislators must prioritize

public and partisan preferences. We can speculate that particular

policy features, such as eligibility levels for working parents, are simply

not as salient as the broad contours of the program. This finding is

likely a sign that on a policy with only moderate salience to the public,

business can exercise substantial power if policymakers anticipate its

preferences.

35
 The fourth conclusion about business power is that direct, noisy

lobbying on an issue does not guarantee influence. Campaign

contributions do not have an effect on the minimum wage, the only

policy in which employers have actively engaged – and, in fact, the

coefficient is positive, despite hard lobbying in opposition. Two things

could be drawn from this. First, on extremely simple and salient

policies, particularly those on which the public can vote directly,

interest groups are likely to be drowned out. Second, active lobbying is

not necessarily the way that business exercises its influence. From

other evidence in the analysis, it seems that the most important source

of influence comes indirectly through policymakers, who anticipate the

reactions of employers. On the minimum wage, perhaps legislators

interpret signals from the business community that indicate that

business leaders put on a good show of opposition but don’t actually

care enough to punish policymakers later, since few of them actually

pay minimum wages. Interviews at the state level support this

interpretation of the data.

The politics of work support are clearly quite complex, and more

work remains to be done. This analysis has specified a theory for

determining which factors will shape which work support policies.

Perhaps more importantly, it has identified a heretofore overlooked role

for employers. Based on the literature and contemporary observation,

we might expect business to be shaping minimum wage laws (which

36
they publicly oppose) and not public health care programs (which they

generally ignore). The results of this analysis indicate that the truth

may be precisely the opposite of these expectations. The answer may

be in a broader definition of power, which takes into account the role of

indirect influence in policy decisions.

References

Table 1. State Policy Choices

Medicaid SCHIP EITC Minimum
STATE Eligibility Eligibility Score Wage
Low Work
Support
States
ALABAMA 26% 200 0 $5.15
IDAHO 43% 185 0 $5.15
KENTUCKY 66% 200 0 $5.15
LOUISIANA 20% 200 0 $5.15
MISSISSIPPI 33% 200 0 $5.15
NORTH DAKOTA 65% 140 0 $5.15
SOUTH DAKOTA 58% 200 0 $5.15
TENNESSEE 80% 185 $5.15
TEXAS 29% 200 $5.15
UTAH 49% 200 0 $5.15
WYOMING 57% 200 $5.15

37
 High Work
Support
States
DELAWARE 107% 200 3 $6.65
ILLINOIS 192% 400 2 $6.50
MAINE 207% 200 1 $7.00
MASSACHUSET
TS 133% 400 3 $7.50
MINNESOTA 275% 280 5 $6.15
NEW JERSEY 115% 350 4 $7.15
NEW YORK 150% 250 5 $7.15
RHODE ISLAND 192% 250 4 $7.40
VERMONT 192% 300 5 $7.25
Moderate
Work Support
States
ALASKA 81% 175 $7.15
ARIZONA 200% 200 0 $6.75
ARKANSAS 18% 200 0 $6.25
CALIFORNIA 107% 250 0 $7.50
COLORADO 67% 200 3 $6.85
CONNECTICUT 157% 300 0 $7.65
FLORIDA 58% 200 $6.67
GEORGIA 55% 235 0 $5.15
HAWAII 100% 200 0 $7.25
INDIANA 27% 200 2 $5.15
IOWA 77% 200 1 $5.15
KANSAS 36% 200 3 $2.65
MARYLAND 38% 300 5 $6.15
MICHIGAN 61% 200 0 $7.15
MISSOURI 40% 300 0 $6.50
MONTANA 62% 150 0 $6.15
NEBRASKA 58% 185 2 $5.15
NEVADA 86% 200 $6.15
NEW
HAMPSHIRE 56% 300 $5.15
NEW MEXICO 65% 235 0 $5.15
NORTH
CAROLINA 54% 200 0 $6.15
OHIO 90% 200 0 $6.85
OKLAHOMA 43% 185 2 $5.15
OREGON 100% 185 2 $7.80
PENNSYLVANIA 61% 235 0 $7.15

38
SOUTH
CAROLINA 97% 185 0 $5.15
VIRGINIA 31% 200 3 $5.15
WASHINGTON 79% 250 $7.93
WEST VIRGINIA 36% 200 0 $5.85
WISCONSIN 192% 185 5 $6.50

Numerical values in the EITC column are on a scale of 0 to 5, with

5 being the most generous. Zero indicates no EITC. States with no
state income tax are left blank.

Table 2: Determinants of Medicaid eligibility levels

Model 2: Medicaid Work Support
OLS N=50
Adjusted R2 = 0.226
Expected
Direction Coef. (s.e.)
Per capita income (in
thousands) + 4.618 (1.850)*
Democratic control + 14.201 (10.929)
Percent black - -1.547 (0.956)
Business campaign
contributions +/- -1.908 (0.855)*

39
Labor campaign
contributions + 2.512 (2.930)
Large low-wage
employers +/- 0.241 (0.788)
Constant -29.481 (66.475)
Note: The significance of business campaign contributions
and large low-wage employers are tested against a two-sided
alternative. All other variables are tested against a one-sided
alternative.
* Indicates a statistically significant result (p<.05).

Table 3: Determinants of SCHIP

eligibility levels
Model 2: SCHIP Work Support
OLS N=50
Adjusted R2 = 0.444
Expected
Direction Coef. (s.e.)
Per capita income (in
thousands) + 8.612 (1.499)*
Democratic control + 27.099 (8.859)*
Percent black +/- -0.759 (0.775)
Business campaign
contributions +/- -1.314 (0.693)
Labor campaign
contributions + -2.397 (2.376)
Large low-wage
employers +/- 0.024 (0.638)
Constant -40.672 (53.891)
Note: The significance of percent black, business campaign
contributions, and large low-wage employers are tested
against a two-sided alternative. All other variables are tested
against a one-sided alternative.
* Indicates a statistically significant result (p<.05).

Table 4: Determinants of state EITC existence and generosity

Model 3: State EITCs
OLS N=50
Adjusted R2 = 0.197
Expected
Direction Coef. (s.e.)

40
Per capita income (in
thousands) + 0.215 (0.053)*
Democratic control +/- 0.173 (0.316)
Percent black +/- -0.006 (0.028)
Business campaign
contributions + -0.009 (0.025)
Labor campaign
contributions + -0.011 (0.085)
Large low-wage
employers + -0.009 (0.023)
Constant -5.013 (1.924)
Note: The significance of Democratic control and percent
black are tested against a two-sided alternative. All other
variables are tested against a one-sided alternative.
* Indicates a statistically significant
result (p<.05).

Table 5: Determinants of state minimum wage levels

Model 4: State Minimum Wages
OLS N=50
Adjusted R2 = 0.509
Expected
Direction Coef. (s.e.)
Per capita income (in
thousands) + 0.261 (0.053)*
Democratic control + 0.491 (0.315)
Percent Black - -0.139 (0.026)*
Business campaign
contributions - 0.016 (0.025)
Labor campaign
contributions + -0.021 (0.084)
Large low-wage
employers - -0.015 (0.023)
Constant -2.577 (1.916)
Note: All variables are tested against a one-sided alternative.
* Indicates a statistically significant
result (p<.05).

41