Do neighbourhoods matter?

My AFR oped today looks at social inclusion and neighbourhood disadvantage. Full text over the fold.

Moving away from trouble, Australian Financial Review, 12 February 2008 

As the machinery of government passes to Labor, a bevy of new buzzwords has hit Canberra. Less talk of free nations, markets, and efficiencies. More discussion of working families, apologies, and challenges. And among the new terms being bandied about is “social inclusion”. A new Social Inclusion Unit has been established in the Department of the Prime Minister and Cabinet, and the government is presently canvassing for a Social Inclusion Board.

If the example of British Labour’s Social Inclusion Unit is anything to go by, a core focus of the Australian social inclusion initiative will be on neighbourhood disadvantage, and on answering the question that has puzzled social scientists for decades: Would a poor family do better if they lived in a middle-class neighbourhood than if the same family lived in a low-income community?

From a theoretical standpoint, there are good reasons to think that neighbourhoods might matter. If getting a good job depends on informal ties, then it will be easier to find work if most people in your street are employed. If community norms count for a lot, then poor children who grow up in poor neighbourhoods may find it harder to break out of the poverty cycle. And because low-income communities also tend to have worse public amenities and higher crime rates, living in these places may be bad for your physical and mental health. If these theories hold, then they have major implications for housing policies; suggesting that mixed-income neighbourhood should be the name of the game. 

But separating the effects of being poor from living in a poor place turns out to be a tricky research problem. To find good answers, we have to cross the Atlantic to the United States, where an ambitious five-city randomised trial has provided some of the best evidence to date on how neighbourhoods affect individuals’ life chances.

Implemented in 1994, Moving to Opportunity offered families in public housing projects in Baltimore, Boston, Chicago, Los Angeles and New York a voucher that would enable them to rent in a lower-poverty neighbourhood. Because demand outstripped the number of available vouchers, the vouchers were randomly assigned through a lottery. As with a randomised medical trial, this ensured that at the outset, those who moved to lower-poverty neighbourhoods (the treatment group) were identical in all respects to those who stayed (the control group).

In the early-2000s, economists Jeffrey Kling, Jeffrey Liebman, and Lawrence Katz followed up the experiment, to judge how a change of neighbourhood affected those in the treatment group. Contrary to some initial expectations, they found no significant impacts on adults’ employment outcomes. Movers were no more likely to have a job than stayers, nor did movers tend to earn higher wages. Summing up the evidence on earnings, the researchers concluded that “housing mobility by itself does not appear to be an effective anti-poverty strategy”.

But money isn’t all that matters. Asked why they wanted to leave the housing projects, many participants said “to get away from drugs and gangs”. Consistent with this, the follow-up study found that those who moved to a lower-poverty neighbourhood had better mental health. Indeed, the psychological benefits of moving were so large and consistent that they alone could have justified the cost of the program.

Might Moving to Opportunity have implications for Australia? According to a paper by Australian National University researchers Bob Gregory and Boyd Hunter (in my view, the best piece of unpublished economic research in Australia), the economic indicators in Australian neighbourhoods have diverged markedly since the 1970s. If you walked across Australia in the mid-1970s, you would have seen much more similarity in employment and earnings than if you trod the same path today.

This growth in neighbourhood inequality led Gregory and Hunter to warn of the growth of Australian ‘ghettos’, and to suggest that better understanding patterns of poverty should be placed high on the national agenda. One way of building on this research would be for Australia to conduct its own Moving to Opportunity experiment – offering randomly selected families in large public housing projects the chance to move to a middle-income suburb.

The best way to redress disadvantage is to put our ideas to the test. As 19th century British economist Alfred Marshall once said, we should combine “cool heads and warm hearts”.  If the federal government’s social inclusion agenda prioritises evidence and results over ideology and rhetoric, it will be off to a fine start.

Dr Andrew Leigh is an economist in the Research School of Social Sciences at the Australian National University.

This entry was posted in Inequality, Low Wage Work. Bookmark the permalink.

14 Responses to Do neighbourhoods matter?

  1. Kevin Cox says:

    Surely a simpler way is to try to remove the “drugs and gangs” or whatever it is that causes the mental health problems from the poorer neighbourhoods – if there is a problem. That way you help everyone and not just the ones who stay.

    It may well be in Australia that some poorer neighbourhoods have a more supportive social environment – some small vibrant towns come to mind.

  2. derrida derider says:

    Bob and Boyd’s paper was not unpublished – it was in the Economic Record, along with a fairly vigorous Comment on it by Peter Whiteford. Peter’s critique was basically along the lines of “correlation is not causality”, which is actually grist to your mill of calling for randomised experiments.

    But while I agree randomised expereiments are the way to go for lots of policy question, I’m not sure how practical this one would be in Oz. We just don’t have an equivalent to the south side of Chicago – not even Redfern approaches it in scale and intensity.

  3. conrad says:

    “Surely a simpler way is to try to remove the drugs and gangs or whatever it is that causes the mental health problems from the poorer neighbourhoods”

    Please tell me how Kevin (and, for that matter, what to do with those with mental health problems in the neighborhoods to start with).

  4. Andrew Leigh says:

    DD, I’m pretty sure that was a different paper (I did run my line past Boyd last week).

    As to poverty intensity, you’re quite right – but this doesn’t necessarily mean that neighbourhood effects are less interesting here (or conversely that they’re more interesting in Brazil).

  5. Eric says:

    A major predictor of poverty for any area across Adelaide is public housing. SA used to have the highest % of public housing of all the states (not the terrirtories) with much of this in the former Enfield LGA, particularly the western half (the Parks area), and this is one of the two major patches of disadvantage in Adelaide. The strategy has been to sell much of it off and have the area developed with upmarket housing, which is in high demand, being close to the city (unlike Elizabeth, which is the other main area of disadvantage).

    But Adelaide’s public housing (as well as rapidly falling toward the levels of the other states) is being spread around more than before, such that you see African refugees spread all over Adelaide, while last generation’s Vietnamese refugees were mostly around Parks.

  6. Peter Whiteford says:

    DD and Andrew

    It wasn’t in the Economic Record, but the original paper was CEPR Discussion Paper No. 325 – which appears not to be downloadable. (Andrew – I presume this means that an ANU discussion paper doesn’t count as a publication?)

    My critique was in the June 1995 issue of the Social Security Journal and Boyd and Bob responded in the same issue (which is also not downloadable but proof of existence is at

    My main critique related to the implication – which was the main point made in media reporting at the time – that poor people had got poorer in Australia over the period. What in fact happened – in my view at least – is that by increased targeting of public housing and introducing income-related rents, poor areas got a higher concentration of poor people because not so poor people moved out or no longer got in to public housing in the first place. In fact, the poorest people actully got better-off as they received a higher share of a valuable subsidy.

    This doesn’t necessarily mean that increased targeting of public housing was necessarily a good idea, if the neighbourhood effects of living in public housing are more strongly negative as a consequence of increased targeting.

    Of course, Australia does effectively have a housing voucher scheme already. It’s called Rent Assistance and goes to benefit recipients and low income working families with children. It’s a payment to individuals (a demand subsidy) that is related both to family income and the amount of rent paid (you don’t get it if your rent is below a specified level), and you get it whomever is your private landlord (i.e. its fully transferable).

    However, overall I agree that some areas appear to have a high concentration of disadvantage, but I suspect not so much as in the UK or the US, and I’m certainly in favour of randomised trials, but I’m not sure what we should be trialling!

  7. Anthony says:

    Yes, I have the Boyd and Gregory CEPR Discussion paper sitting on my shelf, so it seems it was ‘published’. And I think Peter’s response at the time was important.

    The way the B&G paper was reported in the media was to take it as evidence of growing inequality, whereas it might just have been evidence that people were moving around. (Bob Gregory’s other work on inequality at around this time and after was, I think, more convincing). And as Peter pointed out at the time, both real estate markets and targeted housing assistance (including State housing commissions) tended to ‘sort’ people out geographically by income. B&G used the word ‘ghetto’ in the paper which was an emotive term, and no doubt helped to get press coverage. But a lot of the discourse around ‘ghettoes’ in the US is just a step up from the old ‘culture of poverty’ discourse: too many poor people hanging out together and reinforcing each other’s bad habits. Do we have ghettoes in Australia? Broadmeadows ranks high on disadvantage, but most poor people don’t live in Broadmeadows and most people who live in Broadmeadows aren’t poor. So what should be our response to Broadmeadows’ high indicators of disadvantage? I don’t know, just sayin’, that’s all.

  8. Brendan says:

    So let me get this straight: poor people create negative externalities?

  9. derrida derider says:

    Brendan, that’s precisely the point in dispute – whether socalled “neighbourhood effects” exist. Conventional wisdom says they do if the ghettoisation is strong, as in lots of the US. But the Moving to Opportunity experiment casts some doubt on this even in the US (which is the point of Andrew’s post) and it’s certainly not clear in Australia.

    BTW people I’m sorry for my misremembering about the publication.

  10. Bruce Bradbury says:

    Another (potential) problem with the Hunter and Gregory paper was a possible regression to the mean artefact. That is, regions were sorted according to their income level as it applied towards the end of the period. Random changes would thus imply that inequality would be automatically appear lower for the earlier periods.

    This effect might or might not be empirically important. It is a real shame that more research on this has not been done. Unfortunately, this sort of basic descriptive research tends to fall between the main funding sources for Australian social research. It is not theoretically interesting enough for ARC funding, and insufficiently policy relevant for govt funding.

    As for causal neighbourhood effects, we should not forget that for many policy purposes the causal direction is not important. If we want to help the most disadvantaged, then targeting services to disadvantaged regions is a sensible strategy – irrespective of whether living in the location causes disadvantage or whether this simply reflects the characteristics of the people living there.

  11. christine says:

    Re Peter W: it’s indeed an interesting question whether an ANU working paper does count as a publication.

    Re Brendan: it’s not really the poorness of your neighbours/peers that would create a negative externality. However, there might be negative externalities (eg on propensity to go to university) of living in a place full of people with little education. And people with little education might tend to be poor. Or being in a neighbourhood with lots of teenage mums might influence other girls’ decisions regarding early childbearing. Again, neighbourhoods with lots of teenage mums might be relatively poor. (For some work on this, which finds not much effect, see: Evans, Oates, Schwab, JPoliticalEconomy, 1992).

    Parents say this sort of thing matters in their choice of housing location/school. They don’t tend to say family income matters.

    And as Andrew points out, actually doing really good empirical work on this is quite difficult, unless you can move people around ‘experimentally’.

  12. Boyd Hunter says:

    Should CEPR Discussion Paper No. 325 be considered to be published? The Government (formerly DEST and presumably now DEEWR) do not recognize it as published in their funding formula. Certainly none of my colleagues believe it to be published.

    Some of the confusion seems to arise from the fact that there are two main aspects to the Gregory and Hunter paper: The identification of the dramatic increase in neighbourhood inequality in Australian cities and the interpretation of this stylized fact. I certainly published a few things from my doctoral thesis that attempted to have a go at analyzing the stylized fact but none of these articles achieved the synthesis or breadth of interpretation attempted in Gregory and Hunter. I was asked to contribute to the Bobfest conference, which eventually was published in Economic Record. For that conference, I was keen to point out Bob’s unique contribution to the CEPR paper that integrated macroeconomic, sociological, and geographic perspectives to interpret the data. One could argue that my rehearsing some of the arguments in the Bobfest article means that it has now been published. However, I am reticent to endorse this position since that would mean that I got a publication at Bob’s expense.

    Why was the CEPR paper not published in its own right? The geographers did not like the way it treated spatial data as an economist would. Sociologist seemed to think that it was too naïve in its treatment of social processes. To my mind it didn’t matter because it has certainly stimulated enough debate, which after all is the point of intellectual endeavour.

    Speaking of which I think that the issue of public housing raised by Whiteford was a good point. However, Bob and I wrote an unpublished report that showed that increased targeting of public housing was not the whole issue as neighbourhood inequality increased substantially in non-public housing areas. The issue of regression to the mean raised by Bradbury is just a complicating factor for the interpretation of the inequality in a few inter-censal periods examined in the CEPR paper. However, I calculated a socioeconomic index for 1976 in a subsequent SPRC working paper to control for this possibility and there was little change in the underlying stylized fact.

    I always felt that the weakness of the original paper was that I only had access to neighbourhood averages and hence it was virtually impossible to tease out the extent to which it was neighbourhood were important rather than individual, families or even the macroeconomy. Cross-level inference is impractical in the presence of social externalities such as neighbourhood effects as one needs simultaneous collection of data from all the different levels of analysis (with sufficient variation in each level). Andrew’s suggestion of randomised trials is definitely worth considering especially if it assist to tease out the role of these various levels of data.

  13. Andrew Leigh says:

    Thanks for the many interesting responses, esp from key players. FWIW, I don’t think most social scientists regard a discussion paper as a prior publication – the point of DP or WP series in the social sciences is to put out a draft paper, and get comments on it before sending it off to a journal.

  14. Nana Levu says:

    Boyd, why not download the Bob and Boyd ghetto paper here and help this thread become this year’s “Best of the Blogs”.

    I am interested in the broadly different cultural aspects of disadvantage and exclusion you see in suburbs with high state housing like Airds, Bidwill, Flemington, and no or very low state housing like Lakemba, and parts of Cabramatta. In the first there is a high proportion of sole parents, high disability support pensioners. Compare this to hig privatge renting, high unemployment or working poor with high non-English speaking at home – either Arabic or Cantonese or Manderin, but lower proportion of sole parent families.

    I note that the Department of Immigration lease sections of state housing compexes, such as the old Mia Mia complex in Mirrabooka WA. These are filled with wave after wave of refugees who are then dispersed when their visas change and they are no longer able to be housed in these DIAC leases.

    Could it be that after the experience of living side by side with the particular cultural expression of the old Aussie disadvantage, sole parents, alcoholism, and still suffering from the trauma that displaced them from their home countries, refugees retreat to form ghettos? If they had been settled by the Department of Immigration in a more dispersed fashion thoughout areas without concentrations of the old Aussie culture of disadvantage, they may not have so retreated into ghettos.

Comments are closed.