The Politics of Minimum Income Protection in OECD Countries

Paper presented at the Annual Conference of the

Canadian Political Science Association

University of Calgary

Calgary, June 2nd, 2016

Alain Noël

alain.noel@umontreal.ca

Département de science politique

Université de Montréal

The Politics of Minimum Income Protection in OECD Countries *

“The RMI, I tell you, really it is crumbs, one cannot live with that. I don’t

know how such things can be done, you know? What more can I say? Me, I

didn’t even think things like this were possible, I didn’t even think they

existed.”

(Author’s translation; 61 year-old French woman, RMI beneficiary for

the last two years; quoted in Duvoux, 2009: 166)

Childless working-age adults who are deemed able to work and rely on social

assistance for an income are probably the poorest persons in advanced democracies. In

most countries, their disposable income falls well below the poverty line, sometimes

giving them not even half of what it takes to escape poverty. In the early days of the

welfare state, welfare incomes tended to be associated with some definition of needs, to

give persons an access to the basic necessities, but over time, this connection with

needs has receded. Welfare incomes have evolved haltingly and incrementally, as they

were or not adjusted for the evolution of consumption norms and for inflation, and as

successive governments worried more about work incentives than needs (Walker, 1993:

41-56; Van Mechelen and Marchal, 2013: 38-40). The general trend between 1990 and

2008 was steadily downward, with increasingly inadequate benefits (Nelson 2013). In

most countries, sanctions were also introduced to further reduce benefits for claimants

who failed to comply with behavioural rules, usually associated with activation measures

(Immervoll, 2009: 32).

We do not know much, however, about the politics of welfare incomes. We may

presume that it is congruent with the general politics of the welfare state, or with the left-

right politics of redistribution, but we do not really know. Like poverty, welfare incomes

have remained on the edge of welfare state research, because scholars focused on the

social insurance programs that covered the majority of citizens. They assumed that * Preliminary draft. Please quote with consideration. I thank Kenneth Nelson, for kindly sharing data, David Deault Picard and Audrey Roy for their help with the database, and Olivier Jacques for his methodological tips and suggestions.

2

these broad transfers defined the welfare state, and that over time they would make

social assistance, and poverty, increasingly marginal (Marx and Nelson, 2013: 7). In his

seminal book on welfare regimes, for instance, Gøsta Esping-Andersen paid little

attention to social assistance (1990). By his standards, these programs were all alike,

everywhere residual and means-tested.

This paper starts from the assumption that there is indeed a politics of social

assistance in advanced welfare states, and that this politics is an instance of the broader

conflict between the left and the right over the market, the state, and social justice. In

other words, the democratic class struggle over the welfare state reaches out to the

poorest, protecting them better in countries where solidarity among citizens is better

achieved.

To build this demonstration, the paper focuses not on social assistance per se,

but on minimum income protection (MIP), that is to say the disposable income a person

obtains when on social assistance, including the program’s benefits but also other cash

or in-kind benefits that may be allocated directly or though the tax system. Minimum

income protection data provide a more reliable comparative picture, because they

encompass all transfers aimed at social assistance recipients. The paper also

concentrates on single adults considered able to work, to better isolate minimum income

protection from other considerations, related in particular to family policy or national

approaches toward disability. Able-to-work single adults without income or assets — and

not eligible to unemployment insurance — are the arch-typical “undeserving” poor, the

least favoured of all social transfers beneficiaries. As such, they constitute the best case

to test a society’s commitment to redistribution.

Reliable comparative data on minimum income protection remain relatively

scarce. This paper relies mainly on the most encompassing and reliable source, the

Social Assistance and Minimum Income Protection Interim Dataset, or SaMip,

developed by Kenneth Nelson at the Swedish Institute for Social Research (SOFI)

(Nelson, 2013), with some references to the less complete and less convincing dataset

3

offered in the OECD Benefits and Wages database. Even though data are available for

the new EU members from Eastern Europe, this analysis covers only the eighteen

“classical’ welfare states for which we have complete series, assuming their longer

experience with democracy and higher level of economic development makes a

difference for minimum income protection.1 In their recent book on social policy in Latin

America, Evelyne Huber and John Stephens suggest it takes about twenty years of

democracy to influence durably income distribution (2012: 109). The period considered

here starts precisely at the moment Eastern European countries democratized, and it

runs for twenty years, from 1990 to 2010.

The first part of the paper reviews the literature on minimum income protection

and outlines the theoretical argument. The second part presents the data and

methodological approach. The last part exposes the results and discusses their

implications.

Explaining minimum income protection

Every advanced democracy has some form of minimum income protection (MIP).

In their Handbook of Minimum Income Protection in Europe, Thomas Bahle, Vanessa

Hubl, and Michaela Pfeifer define MIP as “a social minimum based on a means test”

(2011: 13). The two elements are important in this definition: the reference to a social

minimum means that minimum income protection very much defines a social protection

floor for a category of citizens, an income below which nobody in a given situation

should fall; and the means test element specifies the manner in which this income is

attributed, not as a universal right or as a function of past contributions, but rather as a

last resort for those below a certain income (and usually with little or no assets, and no

family support). For retired persons, the MIP may take the form of a modest means-

tested top up, which complements other universal or insurance-based transfers. For

single mothers, it may coexist with universal or targeted family allowances. For 1 Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, the Netherlands, New Zealand, Norway, Spain, Sweden, Switzerland, the United Kingdom, and the United States.

4

unemployed, uninsured working-age adult, it basically corresponds to social assistance

and related transfers.

For many years, it was considered practically impossible to compare

systematically the protection offered by social assistance benefits across countries,

because social assistance regimes appeared as a maze of categorical programs with

arcane rules, abundant exceptions, local variations, and numerous in-kind advantages.

In a landmark report released in 1996, Tony Eardley and his coauthors used the family-

type approach to estimate in purchasing power parities the social assistance benefits

offered to various types of households by the different OECD countries. With data for

1992 only, they established important differences in minimum income protection across

countries, with Switzerland, the Netherlands, the Nordic countries and Australia standing

among the most generous (Eardley et al., 1996: 137). Their conclusions, however,

pointed to numerous variations in rules and regulations, which led them to classify

countries in a range of categories, each one containing only a few cases. It appeared

difficult to classify countries along a single monetary continuum. In any case, from the

standpoint of income, differences among cases did not seem to fit any clear institutional

or political pattern. The more recent Handbook of Minimum Income Protection in Europe

produced by Thomas Bahle, Vanessa Hubl, and Michaela Pfeifer updates these

conclusions and provides an elaborate, but complex, overview of social assistance in

Europe (2011).

The first break in this respect came with the early work of Kenneth Nelson, carried

in the context of a systematic effort to compare welfare states across different

dimensions, at the Swedish Institute for Social Research in Stockholm. As part of his

Ph.D. dissertation, Nelson paired the minimum income protection data of Eardley et al,

with the welfare state indicators gathered in Stockholm for the Social Citizenship

Indicators Program (SCIP). Using qualitative comparative analysis (QCA), fuzzy set

analysis, and ordinary least squares (OLS) regression, he found a positive relationship

between the protection offered to the middle class by social insurance programs and the

generosity of minimum income protection, confirming an important implication of the

5

power resources theory of the welfare state (Nelson, 2003: 125). There were limits,

however, to what could be concluded from a set of 18 country cases. In later works,

Nelson developed his own dataset on minimum income protection, covering at first the

18 classical welfare states for the years 1990-2002. Using time-series cross-sectional

(TSCS) analysis, he found that, as the conventional theory suggested, means-tested

benefits were more likely to be cut than social insurance programs in retrenchment

periods, especially where social insurance was well-entrenched (Nelson, 2007). As Paul

Pierson had suggested, the politics of retrenchment had its own logic; it was not simply

the politics of expansion in reverse (1994). In later works, Nelson continued the

development of his Social Assistance and Minimum Income Protection Interim Dataset

(SaMip) and focused on the question of convergence and divergence among European

or OECD countries. He found much diversity in minimum protection between 1990 and

2005. Even within Europe, there was no clear convergence in benefits, and no clear

correspondence between income levels and types of welfare states, except perhaps to

the extent that less encompassing welfare states also tended to be less generous for

social assistance recipients (Nelson, 2008: 114-5; Montanari, Nelson and Palme, 2008).

The arrival of new, less developed member states in recent years only increased

divergences with the European Union (Nelson, 2010). In recent years, Nelson also

documented the relationship between low levels of minimum income protection and

material deprivation (2012), and the inverse relationship between benefits adequacy and

a country’s spending on active labour market policies (2013: 394).

In a dissertation presented in 2009 at the University of Antwerp, Natascha Van

Mechelen more or less took up Nelson’s initial project of elucidating the socio-economic

determinants of minimum income protection. Using both fuzzy set analysis and time-

series cross-sectional analysis, she found no clear relationship between benefits

generosity and socio-economic conditions such as the government’s financial liabilities,

the unemployment rate, or the proportion of social assistance recipients. Only GDP per

capita seemed to have an influence, but mostly to hold down the poorest states, less

able to provide adequate incomes (2009: 93). Van Mechelen did not corroborate

Nelson’s findings on the relationship between encompassing social insurance programs

6

and minimum income protection adequacy, but found that strong trade unions were

favourable for social assistance recipients, a conclusion consistent with power resources

theory (164 and 189-91). She also established that countries combining national social

assistance rates with decentralized implementation tended to be more generous (270).

All in all, Van Mechelen’s findings, like those of Nelson before her, appeared ambivalent:

minimum income protection was somehow tied to features of the welfare state but it

resisted the neat theories or classifications that prevailed in the study of social

protection.

Overall, the conclusions of Ivar Lødemel’s early study of Norway and Britain,

which suggest that the general configuration of the welfare state may well be a poor

predictor of social assistance arrangements, still seem to stand (1997). In a recent

collective book that brings together Stockholm and Antwerp, Ive Marx, Kenneth Nelson

and their coauthors explore further the question of minimum income protection, but

focus on general trends and do not address determinants (Marx and Nelson, 2013).

The objective of this paper is not to develop a full explanation of the determinants

of minimum income protection. Given the data we have, such an explanation, observes

Van Mechelen, is probably out of reach (281). We can nevertheless take advantage of

the passage of time to test again some of the core hypotheses about the political and

social determinants of social assistance incomes. This paper considers three such

hypotheses having to do with the weight of economic and financial constraints, the

influence of welfare state institutions, and the role of power resources (leaving aside,

among other questions, the complex matter of decentralization tackled by Van

Mechelen).

Consider, first, economic and financial constraints. Governments are always

sensitive to the macro-economic and budgetary context. In her exhaustive review of the

question, Van Mechelen identifies four variables associated with these constraints: the

evolution of national wealth, as measured by gross domestic product (GDP) per capita;

the size of the public debt as a proportion of GDP; the unemployment rate, as an

7

indicator of economic hardship and of the demand for income support; and the social

assistance rate, a direct measure of the demand for social assistance (2009: 78-90).

Van Mechelen finds some bivariate relationships, not all of them very strong, between

these variables and benefit generosity, but only the first variable remains significant

when combined with the others in a regression model. In this paper, we retain only the

first three, because reliable OECD data on social assistance rates are not available. In

any case, a study of Canadian provinces, where solid data are available, finds that the

unemployment and the social assistance rates are strongly correlated (Noël and Deault

Picard, 2015). In line with a very straightforward idea of economic and financial

constraints, GDP per capita should be positively correlated with the generosity of

benefits, and the size of the public debt as a proportion of GDP and the unemployment

rate should be negatively correlated with generosity. These relationships can be

considered together to test a first hypothesis, stating that economic and financial

constraints influence the generosity of minimum income protection.

The second hypothesis posits that welfare state institutions influence minimum

income protection. There are many ways to test this hypothesis. Nelson, for instance,

takes the importance of social insurance programs as an indicator of middle class

inclusion in the welfare state, and Van Mechelen considers the design and generosity of

various insurance programs. It is possible, now, to take a comprehensive measure of

this dimension, with the generosity index compiled by Lyle Scruggs and his collaborators

for the Comparative Welfare Entitlements Dataset (Scruggs, Jahn and Kuitto, 2014).

This generosity index was developed by Scruggs to update and improve upon Gøsta

Esping-Andersen’s decommodification index. It integrates a number of information on

social insurance programs, concerning eligibility rules, coverage, and replacement rates,

and provides a widely recognized measure of a country’s commitment to social

insurance (Van Kesbergen and Vis, 2014: 85).

This hypothesis about the influence of welfare state institutions derives from the

power resources argument whereby a welfare state that offers good protection to the

middle class will be more readily supported by the bulk of voters, who will accept to pay

8

taxes and redistribute because they also benefit (Korpi and Palme, 1998). This

argument suggests that encompassing welfare state will be more generous toward the

poor than welfare states where programs are targeted specifically at the poor. The effect

is direct, by increasing the legitimacy of social protection, and indirect, by allowing for a

larger redistributive budget. In addition to the generosity index, we should therefore also

measure the importance of public social spending as a percentage of GDP.

The third hypothesis is drawn from the power resources theory and suggests that

the strength of leftist parties and of trade unions should have a positive impact on

minimum income protection. This theory presents the electoral conflict as a democratic

expression of class conflicts, where workers and their allies push for redistribution, and

are represented in this purpose by parties of the left, favourable to generous social

insurance and transfer programs. Trade unions share these objectives, but may be

ambivalent toward minimum income protection, which is addressed at non-workers. In

principle, unions should favour any measure that weakens market pressures on their

members or potential members. They belong, as David Brady puts it, to the “latent

coalition for egalitarianism” (2009: 102). In his work on insider-outsider politics, David

Rueda questions this assumption and suggests instead that trade unions primarily

defend insiders, well-protected workers with unionized jobs who protect their interests at

the expense of outsiders, persons at the margins of the labour market with little access

to the good jobs (2005). If Rueda is right, trade union density should not be a good

predictor of minimum income protection. Van Mechelen, however, finds some

relationship between union density and social assistance benefits, and so do Noël and

Deault Picard in their comparative study of Canadian provinces (Van Mechelen, 2009:

186; Noël and Deault Picard, 2015). The theory presented here points in the same

direction.

Table 1 sums up the expected effects of the variables considered with the three

hypotheses discussed above.

9

Table 1: Expected Effects on Minimum Income Protection

Independent Variables Expected Effect on Minimum Income Protection

GDP per capita +

Public Debt as % of GDP -

Unemployment Rate -

Welfare State Generosity +

Public Social Spending as % of GDP +

Strength of the Partisan Left +

Trade Union Density +

Data The study of minimum income protection has lagged behind that of social

insurance programs in part because social assistance appears less salient and less

central to the politics of the welfare state, and in part for lack of good, reliable

comparative data. Measuring social assistance benefits is notoriously difficult: these

benefits often mix standard and ad hoc transfers, they may or may not include in-kind

complements, they vary according to household type, and they are often determined

locally, within rather broad national parameters. In these circumstances, the best

approach consists in comparing the formal rules and transfers that apply, in specific

cities, to typical households. This is the model family approach (Van Mechelen, 2009:

35). The advantage of this approach is that it takes into consideration most benefits

obtained by households, without requiring access to extensive individual data. The main

disadvantage is that it is a formal, rules-based approach, which does not consider, for

instance, that for some targeted measures the real take up rate may well be low. Like

most measures of benefits, the model family approach also does not take into account

all in-kind benefits or services available through the welfare state. The most vexing

problem concerns the part of social assistance benefits that covers housing costs, which

is important in some countries. In real life, this component is adjusted to the rent actually

10

paid by beneficiaries, which creates important variations and implies that the model

family approach must assume a rent for a given household. To go around this problem,

the OECD posits a housing cost equivalent to 20 per cent of a country’s average

earnings. The problem with this solution is that it aligns the poor’s housing costs on the

norm for average families, which is not a realistic evaluation. The 20 per cent rule does

not even vary by family types, making it even more misleading for single person

households (Van Mechelen, 2009: 39). The OECD acknowledges this method generates

a “high but not unreasonably high” upper bound for welfare incomes, and it publishes as

well benefits without housing costs, as a lower bound (Immervoll, 2009: 12). The

problem is that, at least in some cases, this upper bound does appear unreasonable.

For the United Kingdom and a number of other countries, observe Jonathan Bradshaw

and Fran Bennett, this estimation “is wrong, and might seriously mislead policy makers”

(2009: 18). At the same time, using the lower bound would disregard the housing

component of benefits, which is important in some countries.

To solve this problem, Kenneth Nelson based his estimates of housing benefits

on the actual rent paid by households relying on social assistance, as established by

Eardley et al., who surveyed national informants to build their 1992 series. For

subsequent years, Nelson adjusted for rent inflation (Van Mechelen, 2009: 101; Nelson,

2013: 391). Nelson’s SaMip results remain estimates, but they nevertheless constitute

the best and most extensive dataset on minimum income protection. Because they are

based on assessments of national rules, they are more reliable for time-series than for

cross-sectional analysis, but they nevertheless constitute the best comparative tool

available (Van Mechelen, 2009: 100).

Once benefits are established, the next step consists in determining adequacy,

which is done by dividing minimum income benefits for a given household by the

country’s equivalised median income and then multiplying by 100 (Nelson, 2013: 391).

For the years between 1990 and the beginning of the 2000s, we relied on adequacy

estimates computed by Nelson and compiled in a file entitled SaMip 2.5 Beta Data (full)

(obtained from Nelson). For subsequent years, we relied on the SaMip benefits data

11

provided in the Social Policy Indicators (SPIN) database (http://www.sofi.su.se/spin/),

and followed the same procedure to establish adequacy, using OECD data for the

equivalised median disposable income (OECD, 2016).

Figure 1 suggests the SaMip adequacy rates, which were used to order the

cases, are relatively reliable. These rates avoid the extreme values obtained with the

two OECD measures, and they tend to reduce the differences between countries, while

showing nevertheless a clear ordering, going from the United States at the bottom to

Norway at the top, Norway being the only country providing MIP above the European

Union at-risk-of-poverty threshold of 60% of the country’s median income. With the

OECD adequacy with housing benefits measure (oecdadqhg), the United Kingdom

appears much more generous than Norway, an unlikely outcome. With the OECD

adequacy without housing benefits (oecdadq), there is little difference between the

United Kingdom, Germany and Sweden. All in all the SaMip adequacy measure

(adequacy) appears more plausible.

12

Figure 1: Mean adequacy for OECD countries, 1990-2010, based on OECD MIP without housing benefits, OECD MIP with housing benefits, and SaMip benefits

Sources: OECD, Income Distribution and Poverty Database; SaMip.

Sources for the other variables considered here are more straightforward. GDP

per capita, public debt as a percentage of GDP, the unemployment rate, public social

spending as a percentage of GDP, and trade union density are taken from OECD

databases. The welfare state generosity index is obtained from the Comparative Welfare

Entitlements Dataset compiled by Lyle Scruggs and his collaborators (Scruggs, Jahn

and Kuitto, 2014). The strength of the partisan left is measured in two ways. First, a

score is given to represent the power of the left in the current year, based on the

proportion of left cabinet portfolios in the government. Second, this same score is

cumulated over the years between 1990 and 2010, to assess the cumulative power of

020

4060

80

Uni

ted

Stat

es

Can

ada

Spai

n

Fran

ce

Uni

ted

King

dom

Italy

Belg

ium

Aust

ria

New

Zea

land

Aust

ralia

Den

mar

k

Ger

man

y

Finl

and

Swed

en

Irela

nd

Switz

erla

nd

Net

herla

nds

Nor

way

mean of oecdadq mean of oecdadqhgmean of adequacy

13

the left. The source for these scores is Duane Swank’s Comparative Political Parties

Dataset (Swank, 2013).

The distribution of each variable was assessed to verify the normal distribution

assumption, and three variables were transformed: adequacy was squared (sqradeq),

pibhab was converted to the square root of pibhab (sqrtpibhab), and the unemployment

rate (txchom) was logged (ltxchom). There was no problem of collinearity between the

independent variables.

Results

Between 1990 and 2010, the adequacy of minimum income protection diminished

almost everywhere. On average, as can be seen in Figure 2, minimum incomes were

relatively stable until 1995, not too far below 50 per cent of the national median income.

They then decreased for ten years, to fall below 40 per cent by 2005. The average

adequacy then stabilized, slightly above the 40 per cent of median income level (Marx,

Nolan and Olivera, 2014: 27).

Figure 2: Average adequacy of minimum income protection, 18 OECD countries, 1990-2012

05

1015

2025

3035

4045

5055

60mean

1990 1995 2000 2005 2010year

14

This common downward trend did not necessarily imply a convergence among

countries. If anything, as Figure 3 on the evolution of the coefficient of variation

indicates, variation among countries increased over the years.

Figure 3: Coefficient of variation for the average adequacy of minimum income protection, 18 OECD countries, 1990-2012

Most countries, however, moved in the same direction and became less

generous. Figure 4 displays national trends over time. Starting and arriving points are

different, and the pace and timing of change vary, but most countries went from better to

worst adequacy, except perhaps Germany, Italy, and the Netherlands. Ireland also

stands out, with a pronounced U shaped curve of regression and recovery.

0.05

.1.15

.2.25

.3.35

.4.45

.5.55

.6cv

1990 1995 2000 2005 2010year

15

Figure 4: Adequacy of minimum income protection, 18 OECD countries, 1990-2012

Consider, now, the bivariate relationships between adequacy and the different

independent variables, when data for the eighteen countries are pooled over twenty

years.

020

4060

800

2040

6080

020

4060

800

2040

6080

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010

1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010

Australia Austria Belgium Canada Denmark

Finland France Germany Ireland Italy

Netherlands New Zealand Norway Spain Sweden

Switzerland United Kingdom United States

adeq

uacy

yearGraphs by cty

16

Table 2: Correlations between MIP adequacy and different independent variables, 18 OECD countries, 1990-2010

Variables Coefficient P value Observations

GDP per capita - 0.1702 * 0.0018 335 Public Debt - 0.3389 * 0.0000 237

Unemployment Rate - 0.0884 0.1173 315 WS Generosity 0.5472 * 0.0000 313

Social Expenditures 0.2958 * 0.0000 335 Left 0.1791 * 0.0011 327

Left Cumulative 0.0614 0.2683 327 Union Density 0.4620 * 0.0000 335

As expected, there is a negative relationship between adequacy and what is

perhaps the main economic constraint, public debt as a percentage of GDP. The

relationship with the unemployment rate is also negative, but not significant. More

surprisingly, the relationship between GDP per capita and MIP benefits is not positive,

but negative. Being wealthy does not encourage solidarity with the poorest; in fact, the

contrary seems true. The second hypothesis, on welfare state institutions, seems

validated: high levels of public social spending and generous social insurance programs

are correlated with more adequate minimum incomes. The results are also consistent

with the third, power resources, hypothesis. The current strength of the partisan left and

a strong union movement are associated with more generous social assistance

revenues. The cumulative power of the left, however, is not significant.

Pooling all cases, however, confounds differences in space and in time and may

hide, in particular, important evolutions over time. To evaluate this possibility, and to

better understand how the different variables interact, it is useful to test a cross-sectional

regression model, for different years. For this model, we will consider the variables that

were found significant in Table 2, namely GDP per capita, public debt, welfare state

17

generosity, social expenditures, left power, and union density. Table 3 presents the

results for four years between 1995 and 2010, using robust ordinary least square

regression.

Table 3: Cross-sectional analysis of the determinants of MIP adequacy, 18 OECD countries, 1995, 2000, 2005 and 2010

(1) (2) (3) (4) Variables 1995 2000 2005 2010 sqrtgdp -1.437 21.44 10.26 -5.720 (29.44) (23.42) (17.96) (21.18) debtgdp -17.86** -15.63* -16.54** -35.22*** (5.460) (8.512) (5.235) (9.353) totgen 107.0** 71.48 85.20 135.6* (41.49) (48.22) (43.67) (67.20) socex -75.66 27.64 17.04 -107.4 (48.73) (40.30) (37.34) (140.1) leftpower -0.690 2.768 -4.084 -21.13* (4.823) (4.010) (3.073) (10.55) uniond 24.61* 7.618 10.19 -13.90 (12.83) (8.771) (7.298) (13.63) Constant 1,285 -3,997 -2,816 4,960 (4,293) (3,621) (4,178) (6,201) Observations 15 17 12 15 R-squared 0.734 0.676 0.859 0.558

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

In the four cross-sectional analysis, public debt as a percentage of GDP acts as a

significant constraint on benefit generosity. The welfare state generosity index, on the

other hand, seems to lift welfare incomes, at least in 1995 and 2010. As for union

density, it operates in the expected direction, but only in 1995. The power of trade

unions appears to fade in the 2000s. Finally, the power of the partisan left is only

significant in 2010, but it then works in the opposite direction, as a negative factor.

These results, of course, are very fragile because we have six independent variables for

18

a very small number of cases. The best strategy, in the circumstances, is to use a time-

series cross-sectional model, to pool country cases and years in a panel analysis.

There are vigorous debates among welfare state scholars, and more broadly

among scholars using time-series cross-sectional models, on the merits of using lagged

dependent variables and fixed effects in such models. From a purely statistical point of

view, there is no straightforward answer to this question, because each approach has

advantages and disadvantages. The choice must be made in light of the theoretical

argument and the data at hand, to determine which avenue appears less likely to

misestimate the model. Because we have a similar set of cases (a small N with a

relatively small number of years) and a comparable theoretical argument about the

economic, institutional and political determinants of change, we follow the logic adopted

by Evelyne Huber and John D. Stephens in their recent book on social policy and

inequality in Latin America (2012). Like many comparative politics scholars, Huber and

Stephens use Nathaniel Beck and Jonathan Katz’s panel-corrected standard errors

procedure (PCSE; Beck and Katz, 1995). Unlike these authors, however, and following

Christopher Achen (2000), they do not include a lagged dependent variable, because it

risks inappropriately suppressing the effect of important independent variables (2012:

135-36). They use instead an AR(1) or Prais-Winsten correction, which corrects for first-

order auto-regression without misestimating other variables. As for country fixed effects,

Huber and Stephens agree with Beck and Katz (2001) and Thomas Plümper, Vera

Troeger and Philip Manow (2005) that their inclusion eliminates the effect of largely time-

invariant variables, precisely the variables that we suspect explain country differences in

institutional and political explanations (2012: 136). Fixed effects basically reduce the

analysis to an explanation of within case variations. If a theory predicts variations across

cases caused by largely time-invariant factors, such as the characteristics of the welfare

state or trade union density, fixed effects are inappropriate. In such a case, argue

Plümper, Troeger and Manow, “allowing for a mild bias resulting from omitted variables

is less harmful than running a fixed effects specification” (2005: 334).

19

Table 4 presents the results of a time-series cross-sectional analysis of MIP

adequacy, using two alternative model for the power of the partisan left, a one-year left

power indicator as used in Table 3, and a cumulative power measure, which was

dropped in Table 3 for the sake of simplicity, because it was found not significant in

pooled correlations.

Table 4: Panel corrected standard errors analysis of the determinants of MIP

adequacy, 18 OECD countries, 1990-2010

(1) (2) Variables Model 1 Model 2 sqrtgdp -6.071*** -3.253 (2.297) (2.524) debtgdp -6.190** -6.158** (2.434) (2.672) totgen 55.30*** 53.34*** (13.59) (13.88) socex 28.33* 43.03*** (15.63) (16.22) leftpower 1.121 (0.923) leftcum -42.32*** (9.254) uniond 12.75*** 13.85*** (2.044) (2.292) Constant 558.3 -0.935 (575.1) (687.9) Observations 216 216 R-squared 0.608 0.587 Number of cty 17 17

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

As was observed with pooled correlations and cross-sectional analyses,

economic constraints are negatively correlated to adequacy. Public debt as a proportion

of GDP, in particular, is significant in two models. The negative sign of GDP per capita

20

appears more puzzling, but it is steady across statistical tests. National wealth does not

favour compassion for the poor. Welfare state generosity and high social expenditures,

on the other hand, do facilitate adequacy, as observed in previous tests. The power

resource argument, however, is more uncertain. Trade union density is a good predictor

of minimum income protection, but the power of the left in a given year is not. More

surprisingly, the cumulative power of the left is negatively associated with adequacy.

This counter-intuitive finding probably reflects important reductions in benefits in the last

ten years, in countries like Norway, Sweden and Spain (see Figure 4). It nevertheless

challenges in part the power resource argument.

Various models, presented in an online appendix, were run to test the robustness

of these results, including robust regression with fixed effects, robust regression with a

lagged dependent variable and random effects, and a first differences model. As

expected, the introduction of a lagged dependent variable or of fixed effects neutralized

the impact of largely time-invariant variables such as union density, but the welfare state

impact, measured through the generosity index or through social expenditures,

remained significant.

Going back to the models presented in Table 4, we can assess the impact of

each variable by estimating the standardized coefficients. Figure 5 presents these

coefficients for the significant variables of model 2, which includes a variable for the

cumulative power of the left.

21

Figure 5: Standardized coefficients for the panel corrected standard errors analysis of the determinants of MIP adequacy, 18 OECD countries, 1990-2010

The most important influence on adequacy comes from welfare state

characteristics. For a standard deviation increase in the generosity index, MIP adequacy

is expected to go up by about 0.34 standard deviations; for a rise of one standard

deviation in social spending, there is an effect on adequacy of about 0.19 standard

deviations. Then comes union density, with a standardized coefficient of 0.27, and, on

the negative side, public debt as a percentage of GDP (-0.16) and the cumulative power

of the left (-0.14).

Substantively, these findings suggest that institutional welfare arrangements

matter for minimum income protection. The logic at play seems more consistent with the

paradox of redistribution identified by Walter Korpi and Joakim Palme (1998) than with

Lødemel’s own welfare paradox, which pointed to a disconnection between welfare

institutions and social assistance (1997). Generous and encompassing social insurance

programs seem indeed to foster middle class support for a sizable redistribution budget,

and thus favour the poor. These findings also corroborate Nelson’s early results (2003),

-.2 0 .2 .4Standardized Coefficient

WS Generosity

Union Density

Social Spending

Cum. Power of Left

Public Debt

22

which were not confirmed by Van Mechelen (2009). For social assistance recipients, it

appears preferable to live in a welfare state committed to spend and redistribute.

The evidence in this case is more mixed, but one’s country need not even be

among the wealthiest. If anything, for reasons that cannot be elucidated here, a high

GDP per capita tends to discourage more than encourage minimum income protection.

Ideally, however, the country should not face excessive financial constraints. Contrary to

the results obtained by Van Mechelen, public debt as a proportion of GDP has a

significant and negative impact on minimum income protection, and this seems true

across periods, as suggest the cross-sectional results presented in Table 3. The idea

that financial constraints place the welfare state under stress seems vindicated

(Hemerijck, 2013; Van Kersbergen and Vis, 2014; Marchal, Marx and Van Mechelen,

2014).

Finally, our findings are in part consistent, and in part inconsistent, with the power

resources argument. Trade union density is indeed a strong correlate of minimum

income protection, which is what the classical argument about power resources predicts.

Some may find this finding surprising, given that the labour movement is often

suspected of working more for insiders than for outsiders, social assistance beneficiaries

being the outsiders par excellence (Rueda, 2005). Our results, however, are in continuity

with the earlier findings of Van Mechelen. Strong trade unions do seem to consolidate

what Brady calls the “latent coalition for egalitarianism” (2009: 102). Parties of the left,

however, have a more surprising, negative impact. Considering the downward trends of

minimum income protection in many countries governed by the left in the 2000s, one

may think Bea Cantillon is right to suspect the general turn toward social investment

policies, accepted by both the left and the right, works against the poorest, those more

distant from the labour market, who are unable to take advantage of new programs

meant to facilitate labour market integration and make work pay (2014). In a recent

article, Nelson concurs with this interpretation, and suggests public expenditures on

active labour market policies (ALMP) have been negatively correlated with benefits

adequacy (2013: 393-4). In this perspective, governments of the left, usually more

23

committed to invest in ALMP, may have been engaged in a trade-off between activation

and minimum income protection. Such a trade-off, however, is far from evident. Figure 6

presents the relationship between expenditures on active labour market policy as a

percentage of GDP and our adequacy measure, for 2009 (ALMP data are from Brady,

Huber and Stephens, 2014).

Figure 6: Relationship between the adequacy of minimum income protection and expenditures on active labour market policy, 15 OECD countries, 2009

This relationship is not strong, it is only for a year, and it is presented here without

probing other possible factors, notably partisan variables. The possibility of a trade-off

between activation and adequacy, however, appears far from obvious. In the late 1990s

and early 2000s, the leftist politics of the Third Way often combined an emphasis on

activation with a commitment to redistribute (Huo, 2009; Larocque and Noël, 2014). In

any case, this elusive trade-off between activation and adequacy cannot explain, by

itself, the puzzle raised by the negative impact we found for left cumulative power. On

this question, more research needs to be done.

Austria

Belgium

Canada

Denmark

FinlandFrance Germany

Ireland

Italy

Netherlands

Norway

Spain

Sweden

United Kingdom

United States

0.5

11.

5al

mp_

pub

0 20 40 60 80adequacy

R = 0.49 (0.06)

24

Conclusion

As the effective social floor prevailing in advanced democracies, minimum income

protection constitutes the rock-bottom foundation of citizenship rights, and it is a

distinctive test of a welfare state commitment to social justice (Bahle, Hubl, and Pfeifer,

2011: 2; Kenworthy, 2011: 4). Minimum incomes, however, have been neglected by

welfare state scholars, because they appear marginal, are difficult to assess, and poorly

documented. Even the OECD maintains a sketchy representation of the amounts

involved, and only for a few recent years. Building on the work of Kenneth Nelson, who

developed the Social Assistance and Minimum Income Protection Interim Dataset

(SaMip), we estimated adequacy levels for the 18 “classical” welfare states, and

developed a time-series cross-sectional analysis to account for the economic,

institutional and political determinants of these levels.

The first conclusion that can be drawn from this study is that almost everywhere

in the last two decades, minimum incomes as a proportion of median disposable income

have been going down. This trend was not the result of a convergence among countries,

became variations across cases remained important, but most countries did go from

more to less redistributive.

The second conclusion is that this widespread downward trend was not entirely a

consequence of economic difficulties. Sure enough, governments that had a higher

public debt as a proportion of GDP appeared more likely to let social assistance benefits

decline, but the poorest countries were not more prone to retrenchment. If anything, it

was in the countries that had the highest GDP per capita that redistribution toward the

poor fared worst. Theoretically, this observation is not easy to explain, but it is not very

robust empirically either.

Third, and most importantly, as went the welfare state so went minimum income

protection. There is a clear association between the decommodifying character of social

insurance programs as measured by Scruggs’ generosity index and public social

spending as a percentage of GDP, on the one hand, and social assistance benefits on

25

the other hand. This conclusion is consistent with Korpi and Palme’s paradox of

redistribution and with Nelson’s early findings, and it appears robust across our different

tests. Even when we consider more demanding statistical procedures, with lagged

dependent variables or fixed effects, which tend to erase the effect of political variables

such as union density, welfare state characteristics remain a significant determinant of

minimum income protection. When measured with standardized coefficients, in our main

models, the welfare state context produces the most important impacts on the variation

in minimum income protection.

Fourth, politics matter. Trade union density, in particular, has a positive influence

on minimum incomes. Trade unions work first for their members, and as such they

naturally favour insiders, workers who are employed in good, steady jobs. They also

belong, however, to what Brady calls the “latent coalition for egalitarianism” (2009: 102),

and contribute to decommodify labour markets, supporting various laws and measures

that reduce market pressures on workers. Trade union density may also be a correlate

of the mobilization capacity of collective actors in a given society. It is plausible, for

instance, but difficult to demonstrate, that the women movement or associations

defending the rights of the poor are more powerful in countries where the labour

movement is stronger. Whatever the case, the poor benefit from the presence of trade

unions, even though they are not members. The legacy of parties of the left, however, is

not so positive. In pooled correlations, left power is associated with benefits adequacy,

but this result is not observed in cross-sectional or time-series cross-sectional analysis.

There is, in fact, a significant negative relationship between left power and adequacy in

the 2010 cross-sectional analysis. More importantly, in time-series cross-sectional

models, one standard deviation in the cumulative power of the left predicts a decrease of

0.14 standard deviations in adequacy. In later years, at least, the left has not been

particularly helpful to the poor. Some may think the Third Way or social investment left

favoured labour market activation at the expense of redistribution. A quick look at the

data contradicts, however, the idea of a simple trade-off between active labour market

policies and transfers toward the poor. On this question, more research needs to be

done.

26

Every advanced democracy provides a minimum income to able-to-work single

adults without market incomes, family support, or assets. This income can be extremely

low. In the United States, for instance, it comes mostly as food stamps and leaves many

persons far below the poverty line. Minimum incomes nevertheless define, in the eyes of

many, a country’s welfare state. Almost everywhere, in the last twenty years, these

minimum incomes have gone down, but the starting points and national evolutions

varied significantly from one country to the other, and the different national evolutions

were undoubtedly political. A high domestic income provided no guarantee of

generosity, although it seemed better not to have an important public debt. In line with

the paradox of redistribution argument, the most important determinant of benefits

adequacy was the overall character of the welfare state. When social protection was

good for all, it was also better for the poorest. Strong trade unions also helped, but the

effect of parties of the left was more uncertain, which says something perhaps about the

current difficulties of social democracy.

One should keep in mind that the workless poor remain perennial outsiders. For

all its variations, minimum income protection always remain minimum, usually quite

below the 50 per cent of median income line often taken as a poverty line. Governments

worry more about work incentives than about basic needs, even though the gap

between low wages and social assistance benefits remains “quite substantial: “it is hard

to argue,” note Marx, Nolan and Olivera, “that long-term dependence on social

assistance benefits is an attractive financial proposition in most of Europe” (2014: 29).

As governments and groups around the OECD begin to evoke a basic income or a

guaranteed annual income, we should remember that no country stands even near the

possibility or the levels of a basic income. The road to get there, if it exists, appears very

narrow.

27

References Achen, Christopher (2000). “Why Lagged Dependent Variables Can Suppress the

Explanatory Power of Other Independent Variables,” Paper presented at the annual meeting of the Society for Political Methodology, UCLA.

Bahle, Thomas, Vanessa Hubl, and Michaela Pfeifer (2011). The Last Safety Net: A

Handbook of Minimum Income Protection in Europe, Bristol, Policy Press. Beck, Nathaniel and Jonathan Katz (1995). “What To Do (And Not To Do) with Time-

Series Cross-Section Data,” American Political Science Review, 42, 4: 634-47. Beck, Nathaniel and Jonathan Katz (2001). “Throwing Out the Baby with the Bath Water:

A Comment on Green, Kim, and Yoon,” International Organization, 55, 2, March: 487-95.

Bradshaw, Jonathan and Fran Bennett (2009). Minimum Income Schemes in the United

Kingdom: A Study of National Policies, Peer Review in Social Protection and Social Inclusion and Assessment in Social Inclusion, Brussels, European Commission, April.

Brady, David (2009). Rich Democracies, Poor People: How Politics Explain Poverty,

Oxford, Oxford University Press. Brady, David, Evelyne Huber and John D. Stephens (2014). Comparative Welfare

States Data Set, University of North Carolina and WZB Berlin Social Science Center.

Cantillon, Bea (2014). “Beyond Social Investment. Which Concepts and Values for

Social Policy-Making in Europe?,” in Bea Cantillon and Frank Vandenbroucke (eds.), Reconciling Work and Poverty Reduction: How Successful are European Welfare States?, Oxford, Oxford University Press, pp. 286-318.

Duvoux, Nicolas (2009). L’autonomie des assistés: sociologie des politiques d’insertion,

Paris: PUF. Eardley, Tony, Jonathan Bradshaw, John Ditch, Ian Gough, and Peter Whiteford (1996).

Social Assistance in OECD Countries. Volume I: Synthesis Report, A study carried out on behalf of the Department of Social Security and the OECD by the Social Policy Research Unit, London: HMSO.

Esping-Andersen, Gøsta (1990). The Three Worlds of Welfare Capitalism, Princeton,

Princeton University Press. Hemerijck, Anton (2013). Changing Welfare States, Oxford, Oxford University Press.

28

Huber, Evelyne and John D. Stephens (2012). Democracy and the Left: Social Policy and Inequality in Latin America, Chicago: University of Chicago Press.

Huo, Jingjing (2009). Third Way Reforms: Social Democracy after the Golden Age,

Cambridge: Cambridge University Press. Immervoll, Herwig (2009). “Minimum-Income Benefits in OECD Countries: Policy

Design, Effectiveness and Challenges,” IZA Discussion Paper, No. 4627, Bonn, Institute for the Study of Labor, December.

Kenworthy, Lane (2011). Progress for the Poor, Oxford, Oxford University Press. Korpi, Walter and Joakim Palme (1998). “The Paradox of Redistribution and Strategies

of Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries,” American Sociological Review, 63, 5: 661-87.

Larocque, Florence and Alain Noël (2014). “The Politics of Poverty in the European

Union: How States Respond to the Open Method of Coordination on Social Inclusion,” Policy & Politics, 42, 3, July: 333-50.

Lødemel, Ivar (1997). The Welfare Paradox: Income Maintenance and Personal Social

Services in Norway and Britain, 1946-1966, Oslo, Scandinavian University Press.

Marchal, Sarah, Ive Marx, and Natascha Van Mechelen (2014). “The Great Wake-Up

Call? Social Citizenship and Minimum Income Provisions in Europe in Times of Crisis,” Journal of Social Policy, 43, 2, April: 247-67.

Marx, Ive and Kenneth Nelson (2013). “A New Dawn for Minimum Income Protection?,”

in I. Marx and K. Nelson (eds.), Minimum Income Protection in Flux, Houndmills, Basingstoke, Palgrave Macmillan, pp. 1-27.

Marx, Ive, Brian Nolan, and Javier Olivera (2014). “The Welfare State and Anti-Poverty

Policy in Rich Countries,” IZA Discussion Papers, No. 8154, Bonn, Institute for the Study of Labor, April.

Montanari, Ingalill, Kenneth Nelson, and Joakim Palme (2008). “Towards a European

Social Model? Trends in Social Insurance Among EU countries, 1980-2000,” European Societies, 10, 5: 787-810.

Nelson, Kenneth (2003). Fighting Poverty: Comparative Studies on Social Insurance,

Means-Tested Benefits and Income Redistribution, Stockholm, Swedish Institute for Social Research.

Nelson, Kenneth (2007). “Universalism versus Targeting: The Vulnerability of Social

Insurance and Means-Tested Minimum Income Protection in 18 Countries, 1990-2002,” International Social Security Review, 60, 1: 33-58.

29

Nelson, Kenneth (2008). “Minimum Income Protection and European Integration: Trends

and Levels of Minimum Benefits in Comparative Perspective, 1990-2005,” International Journal of Health Services, 38, 1: 103-24.

Nelson, Kenneth (2010). “Social Assistance and Minimum Income Benefits in Old and

New European Democracies,” International Journal of Social Welfare, 19: 367-78.

Nelson, Kenneth (2012). “Counteracting Material Deprivation: The Role of Social

Assistance in Europe,” Journal of European Social Policy, 22, 2: 148-63. Nelson, Kenneth (2013). “Social Assistance and EU Poverty Thresholds 1990–2008. Are

European Welfare Systems Providing Just and Fair Protection Against Low Income?,” European Sociological Review, 29, 2: 386-401.

Noël, Alain and David Deault Picard (2015). “The Politics of Minimum Income Protection

in the Canadian Provinces,” Paper presented at the Annual Conference of the Canadian Political Science Association, University of Ottawa, June 3rd.

OECD (2016). Income Distribution and Poverty Database, Paris: OECD. Pierson, Paul (1994). Dismantling the Welfare State? Reagan, Thatcher, and the Politics

of Retrenchment, Cambridge, Cambridge University Press. Plümper, Thomas, Vera E. Troeger, and Philip Manow (2005). “Panel Data Analysis in

Comparative Politics: Linking Method to Theory,” European Journal of Political Research, 44: 327-54.

Rueda, David (2005). “Insider-Outsider Politics in Industrialized Democracies: The

Challenge to Social Democratic Parties,” American Political Science Review, 99, 1, February: 61-74.

Scruggs, Lyle, Detlef Jahn, and Kati Kuitto (2014). Comparative Welfare Entitlements

Dataset 2. Version 2014-03. University of Connecticut and University of Greifswald.

Swank, Duane (2013). Comparative Political Parties Dataset: Electoral, Legislative, and

Government Strength of Political Parties by Ideological Group in 21 Capitalist Democracies, 1950-2011, Department of Political Science, Marquette University.

Van Kersbergen, Kees and Barbara Vis (2014). Comparative Welfare State Politics.

Cambridge: Cambridge University Press. Van Mechelen, Natascha (2009). Barriers to Adequate Social Safety Nets, Ph.D. Thesis,

Faculty of Political and Social Sciences, University of Antwerp.

30

Van Mechelen, Natascha and Sarah Marchal (2013). “Struggle for Life: Social Assistance Benefits, 1992-2009,” in Marx and Nelson (eds.), Minimum Income Protection in Flux, pp. 28-53.

Walker, Carol (1993). Managing Poverty: The Limits of Social Assistance, London:

Routledge.