There are many answers to that question. I could, for instance point to the lack of appreciation for the temporal scale – the theory that has been laid out for the neighbourhood to impact the individual relies on a far longer temporal scale than the time frame most analysts (if not all) allow for the analysis. Or, I could have lamented the lack of thinking about causal pathways: open any health related journal and I guarantee you’ll find a paper analysing the relationship between public open space and improved health outcomes (quite how the mere presence of local open space is sufficient is never disclosed in these papers. My grandmother lived very near a number of parks but she rarely, if ever, ventured into them). Or, I could have pointed to near silence in the literature that recognises that where people are located in residential space owes a lot to the spatial expression of inequalities and class struggles so that the urban differences between places are not accidental outcomes (but others blogging here are far better placed to talk about that than I am!). No, when I was asked to respond to Joe’s blog on neighbourhood effects I thought long and hard about Bourdieu and what his work could mean for the field. Then I thought harder and longer about the dominance of the disciplines represented in the near 20,000 papers reporting neighbourhood effects. My gaze landed on the econometricians and I wondered how they would engage with the ideas of the philosopher: whilst I may be doing a disservice to some econometricians, many within the field are likely not to be listening. So, the thing that keeps me awake at night and where I think we (by we in this instance I mean people interested in neighbourhood effects) need a step change is to wrestle the term and analyses back from the economists and econometricians.
What is wrong with the econometricians you ask? I’ll tell you: in the world of the econometricians complex modelling approaches can be used to overcome some pretty serious problems. What’s wrong with that, after all, that’s what all quantitative modellers are doing isn’t it? Well, no: for instance ‘things’ that we have not been able to measure (this could be personality, how risky an individual is willing to be in their financial behaviour to name two) can be ignored through the use of complex modelling techniques (the ‘fixed effect’ approaches). The econometricians are happy with these models because they allow them (they believe) the chance to estimate values to attached to neighbourhoods that are ‘unbiased’ and not altered by those inconvenient things that we do not know or cannot measure. The problem is, in reaching those ‘unbiased’ estimations a lot of other important information that we do know and that is important has been thrown out: for instance, in their simplified representation of reality (for that is all a model really is) other variables have also been discarded because the model cannot cope with information that does not change. In short the baby is thrown out with the bath water and variables such as ethnicity and gender, to name two, are omitted from the model and any effect that these variables may have (and there is reason to suggest that they may be important!) is lumped together as ‘error’ with those other ‘things’ we don’t know. So far, so mechanical. But this blog is about more than just the specification of the variables in a model. Because, the same assumptions that apply to the econometrician’s variables also apply to their neighbourhoods!
Reading through the literature (and I am a part of this literature so I must shoulder some of the guilt) the most important component of the investigation – the neighbourhood to which we are ascribing these effects – is the part that receives the least attention. Indeed, in many cases the neighbourhood is used as a non-spatial entity. We (and this time I am using the ‘we’ for geographers!) should be the front of this literature using our considerable spatial arsenal to explain, examine, critique and explore how space matters. Neighbourhoods (whatever they may be – that is another blog to be written at another time) are fundamentally about the organisation of individuals into spatial entities. They may be spatially contiguous – that is next to each other like residential neighbourhoods – they may be disjointed – like work, leisure or cultural neighbourhoods – or they may not exist in a physical sense but all are important. The neighbourhood must be the most important part of any study trying to determine if there are linkages between places and individual outcomes. And of course, neighbourhood is a highly contested and debated object at an atomistic level neighbourhood has a unique meaning to each individual in the data. Yet, it is also the piece of information that received the least attention in much of the literature: neighbourhood is frequently used to mean purely the residential context and is derived from standard administrative units created to satisfy the delivery of state statistical data. They have no meaning for the activity space of individuals, of the spaces through which people travel or interact, and have no meaning for the spaces in which people inhabit. Moreover, the kind of neighbourhood that you would use for, say, trying to understand peer group effects on children are very different to those that you would employ for understanding Similarly, the things that we measure in the neighbourhoods are equally important. Much of the neighbourhood effects work uses the percentage of X, or Y and then attempts to make an assertion that the more (less) of X or Y the worse (better) things will be for individuals.
So, until we engage spatially then we are going to continue to look for effects without getting a handle on where they may (or may not) exist. In doing so we may not find the needle in the haystack, but at least we’d be looking in the right place!