Swan's missing middle misses analysis

Wayne Swan presents an analysis of Labor’s 2001 federal election defeat in terms of Theda Skocpol’s ‘missing middle’ thesis of the decline of America politics.

Swan’s line of reasoning is that while Labor successfully garnered the support of the ‘battlers’ on issues like the GST (only to have some peel away in light of the ‘Tampa episode’), Labor failed to "convince enough of what I would term the outer suburban middle class".

This latter group are in his words the "‘Westfield Mallers’ because the shopping centre is the hub of their social interaction in the new and emerging suburbs". The ‘battler’ and the ‘Westfield Mallers’ are, in Swan’s eyes, Australia’s version of Skocpol’s ‘missing middle’.

A blunt instrument

The first flaw in Swan’s analysis is that the descriptor ‘outer suburban middle class’ does not quite live up to expectations. On the surface of things, the label is appealing and pretends to both describe and explain. It doesn’t really do either.

On the descriptive front, other than a ‘spatial’ dimension (outer suburban) the notion of ‘middle class’ in Swan’s terms doesn’t tell us much at all.

According to Swan, the ‘outer suburban middle class’ "is a very diverse group – some are teachers and nurses, some are sub-contractors and small business owners who have done well in life and want their children to do better." He goes on the say that for these people the "value of their home is important and they will often have a second investment property. There is a diversity of incomes within this grouping".

On these terms, Swan’s ‘outer suburban middle class’ is a bit of a catch-all. It’s loose enough to describe a very broad range of people on the basis of where they tend to live (which is, of course, superficially appealing), but not really precise enough to tell us much about how these diverse demographic factors played out in terms of political behaviour.

A similar criticism can be applied to ‘the battler’ as an analytical descriptor.

It isn’t clear in Swan’s thesis whether or not the common factors behind what he admits are disparate groups of people are spatial (outer suburbs), demographic or cultural.

For example, Swan invokes Skocpol’s ‘missing middle’, which he believes are represented in Australia by the ‘battler’ and the ‘Westfield Mallers’. What unifies these two groups is that they are "families who live on modest wages or wages made modest by the cost of their responsibilities to their children". So, it would appear the common denominator is ‘disposable income’.

I’m sorry, but I find it hard to think that there is much analytical purchase to be gained from a framework that lumps together families whose incomes range from $80 a week to $800 on the basis that after taking out private insurance and paying private school fees the level of disposable income among families at each end of this spectrum is a basis for similar political behaviour.

On the contrary, the statistical analysis discussed below shows that, in fact, the political behaviour of people within this income range varied quite significantly at the 2001 Federal Election.

On top of disposable income, Swan believes that these disparate groups of people are united in their cultural outlook. Both the ‘battler’ and the ‘Mallers’ have a dislike for welfare, want governments to stand up to vocal minorities and vested interests, and champion the view that rewards hard work.

How the connection is made between disposable income and cultural outlook on these matters isn’t explained. It is, of course, a classic reductionist move that’s common in a range of social science. But if it’s to be useful politically, it would be worth knowing (a) whether or not such views are prevalent across these diverse demographic groups and (b) if so, what is the relationship between cultural outlook and demographic characteristics.

Without these sorts of answers, one is left wondering whether or not ascribing these attitudes to this spatially and demographically defined group – ‘the missing middle’ – is a function of transference rather than sober analysis.

An alternative analysis

I’ve already published a demographic analysis of the 2001 election elsewhere, so in what follows I will concentrate on three demographic characteristics and the voting behaviour associated with each.

These are measures of personal and household income, monthly mortgage levels and type of schools that children attend. These have been chosen because they are central to Swan’s analysis, in the way he describes the ‘missing middle’.

The demographic correlation analysis presented here is based in multi-factor regression modeling undertaken using the demographic model developed by Australian Development Strategies. The model – called Elaborate – includes over 450 ABS demographic variables derived from the Census and privately derived data. (Details of the analysis can be found at www.australiandevelopmentstrategies.com.au).

Income and voting behaviour

According to the correlation analysis, Labor gained the support of very low-income males where positive swings were also achieved, but tended to lose support among very low income females (individuals earning less than $200 a week – 1996 Census figures).

Labor continued to receive support from men earning between $79-199 a week, and gained swings to it from men earning between $39 and 119 a week. Strong two-party preferred (2PP) support was also evident among middle-income males (those earning $600-699 a week), though swings against the party were evident in these two groups.

Statistically significant correlations for support are found between Labor’s 2PP vote and men earning $79-159 a week. As for swings gained by Labor, a statistically significant correlation was evident for men earning up to $39 a week.

As for women, Labor received support from women earning between $159-199 a week, with women earning $120-159 a week being the strongest predictor of 2PP support for Labor. However, despite continued support among these lower income groups, Labor suffered swings against it among women in lower income groups. Interestingly, Labor gained swings from women earning over $600 a week – whereas it suffered swings against it from men in the $600-plus income groups.

The data confirms Swan’s claims that Labor made gains among the lower income groups (the so-called ‘battlers’), though this is really only the case with regards to men. Middle income men also supported Labor (weekly earnings $500-699), though these groups did swing against the ALP.

However, apart from this, the analysis shows that Swan’s analysis is insufficiently subtle. Importantly, it shows that contrary to Swan’s claims Labor actually gained swings among middle income females (weekly earnings in excess of $600 a week).

This is confirmed by correlations between the 2PP vote and swings, and measures of family income. According to the statistical analysis, Labor suffered swings against it among families earning between $500-1,499 a week, but gained from families earning up to $499 a week. However, on both these fronts the correlations were not statistically significant.

The only significant correlations between 2PP and family income was found in Labor’s support from families earning between $500-699 a week, and the Coalition’s support from families earning over $2,000 a week.

Mortgage payments

Statistically speaking, the correlations between voting behaviour and monthly mortgage payments are far more significant than are the correlations for income.

The statistical analysis shows that Labor gained swings to it from people paying up to $599 a month (remember, these are 1996 figures), but suffered swings against it from mortgage payers with monthly repayments of between $600-1,499.

Statistically significant correlations are evident between the swing to Labor from mortgage payers paying between $200-399 a month, and the swing against Labor from those paying between $600-1,199 a month.

What this shows is that the political behaviour of lower level and medium level mortgage payers was quite different. Contrary to Swan’s analysis, the voting behaviour of these two groups is not the same.

Children attending school

Lastly, an analysis of how voting behaviour correlated with the type of schools children attend confirms that Swan’s analysis is too blunt to be of much value.

In terms of Labor’s 2PP vote, an extremely strong correlation was evident from parents whose children attend Catholic primary and secondary schools. These parents also swung towards Labor.

However, parents with children attending non-Catholic private schools swung strongly to the Coalition, though a positive correlation was evident among this group and Labor’s 2PP vote.

For Swan, many of the ‘missing middle’ "have private school fees to pay and don’t readily see the benefit to them of greater public investment in these areas". The statistical analysis shows that in fact, private schools per se are not a unifying predictor of voting behaviour.

On the contrary, the correlations show Labor gaining support among parents with children attending private Catholic schools, but losing support among parents with kids going to non-Catholic private schools. Speculatively, this probably has as much to do with the extent to which these parents support or oppose private investment in education as it has to do with Labor’s concerted attack on government funding of the private school system during the campaign.

Concluding observations

Wayne Swan has co-opted an academic framework developed in the context of American politics to explain political behaviour in Australia and to define the challenges for the ALP.

The analysis presented here shows that this move is fraught with real danger. The adoption of conceptual terms divorced from an examination of the empirics can lead to inaccurate political diagnosis (and by extension prognosis).

The bottom line is that Swan’s analysis is not sufficiently subtle, and is in important respects demonstrably inaccurate. It does, however, raise interesting questions about the role of transference (of an analyst’s prejudices) in ascribing political and social attitudes to pre-defined demographic groups.