When the insecure are not the disadvantaged

Amidst all the other excitements of the summertime, you might have missed a couple of significant papers by top scholars in top American journals. So just to mentally prepare you for the autumn – don’t worry, it’s not here yet! – I thought that over two posts, I’d cover two papers that seemed likely to be important.

The first argues that we have failed to see a critical dimension of support for the welfare state: the extent to which at-risk groups overlap with the disadvantaged. 

The argument

The paper is by Philip Rehm and colleagues in the American Political Science Review, and was done with The Great Risk Shift and Winner-Take-All Politics author Jacob Hacker, and their argument goes something like this. There are two conventional ways of looking at the link between socioeconomic status and support for the welfare state.

• Firstly, there is the power resources school which focuses on class – the more disadvantaged you are, the more it’s in your interests to demand redistribution.

• Secondly, there is the ‘revisionist’ school (in the words of Rehm et al), which focuses on risk – the welfare state functions as a social insurance mechanism (as I’ve previously mentioned), and the people who are most insecure are most likely to support the reassurance of insurance.

Rehm et al argue that previous scholars have missed the combination of these two ideas – that the overlap between class and risk is important. As they put it, “we argue that the breadth of popular support for social programs depends crucially on whether those programs unite lower income citizens (who support them primarily because of their redistributive impact) and more affluent citizens (who support them primarily because of their insurance function).” They predict that a low overlap between income and risk means that (i) there’s less opposition; (ii) less polarization; and (iii) usually more support for social programmes.

The evidence

So, are their predictions met? Well, you’ll be entirely unsurprised to hear that papers that make into a journal as good as APSR tend to find some support for their theories! But the evidence itself is interesting, and comes in three parts.

As a prelude to the main analysis, they show that risk is associated with income, and that both risk and income are independently associated with support for social programmes. They do this focusing on unemployment benefit across 13 countries, looking at (relative) household income and the level of unemployment risk for people employed in a specific occupation.   The results are shown in Table 1 below – and show everything that you would expect.

They then provide evidence for their hypothesis, that it matters if the at-risk and the disadvantaged are the same people. The first analysis looks at support for unemployment benefits across countries, using the question from the International Social Survey Programme 2006:

“On the whole, do you think it should be or should not be the government’s responsibility to provide a decent standard of living for the unemployed?”  

They then look at whether there is more support for unemployment benefits in countries with less overlap between low income and being at-risk of unemployment. As the figure shows below, their empirical findings support their hypothesis – and moreover in later tables, they show this relationship holds even when they control for other country-level factors (economy-wide unemployment, inequality, social expenditure, and type of unemployment system).

Lower overlap between unemployment risk and relative income (to the right) is associated with higher support for unemployment benefits (to the top) across countries

Their second piece of evidence looks within a single country (the US) but across multiple policy domains. They did this by asking a set of questions in the American National Election Study, about (i) how much people worry about a series of risks, and (ii) how common some risks will be in the next year ‘for people like them’.  Again, they find a high correlation between (i) the overlap of particular risks (this time subjective rather than objective risk) and income, and (ii) lower support for policies in that particular area.

Lower overlap between domain-specific risks and relative income (to the right) are associated with greater support for that real/imagined policy (to the top) across different domains in the US

[This graph is taken from the paper and is truly atrociously labelled – there is no scale on the horizontal axis, but these are all negative correlations, so dots further to the right have weaker overlap between income and risk]

So what does this all mean?

The full implications of this idea need further fleshing out and testing, but it already prompts a cascading series of thoughts about what drives support for the welfare state across countries.

Firstly, it makes us think even harder about the links between economic changes and support for the welfare state. Policy wonks on both sides of the Atlantic are thinking hard about the economic system they want – and how they could achieve it – and this reminds us that any changes could  have far-reaching impacts on public opinion.

Secondly,  Rehm et al point out that it may help explain when the recession will lead to greater support for social policies, and when it won’t – “economic shocks whose effects are felt mostly by the less advantaged are unlikely to shift the structure of public opinion in a direction conducive to greater welfare state generosity.”

Finally, and reflecting my longstanding interest in universal vs. selective benefits, Rehm et al argue that this is one component of the widely-found link between universality and public support.  In their words, “what makes programs more universal is not just the scope of the risks that they cover but also the degree to which those risks affect a broad cross-section of citizens, not just the economically disadvantaged.”

As my comment below notes, I have a couple of questions about the methodology. But it’s a really interesting paper all the same.