Segregation could be bad,and it could be good – but what is it?

What is segregation?

The first sentence of his section headed ‘Segregation’ (p. 2) reads:

In its broadest sense segregation refers to separation – or spatial concentration – as defined by some characteristic, such as race, ethnicity, socioeconomic status, political affiliation, gender, religion, employment status or language. (His emphasis)

This statement – segregation matters: measurement matters – has been at the core of substantial debate about how segregation should be measured. As all involved in that debate would agree, no measure is either perfect or all-embracing but, as some have argued in their promotion of particular metrics, some are more imperfect than others. This is not the place to review the various points of view expressed, nor some of the many measures offered, but one element of those debates is key to my major, positive, critical comment on Merry’s ‘defense of separation’.

Milieu composition

Very many of the indices of segregation generally deployed by analysts (largely based on the ‘classic five’ identified by Massey and Denton 1988) provide single-number descriptors only; they are average values which say little, if anything, about either variation around that situation or the nature of individual milieux – such as particular neighbourhoods. They can indicate, for example, what proportion of an ethnic group would have to be moved to different areas of a city so that its overall distribution across its component areas was the same as that of the remainder of the population but say nothing about the population composition of any of those component parts. A group may be very unevenly distributed across the city, but not strongly concentrated (i.e., comprising a substantial component of the local population) in any one neighbourhood. For that reason an alternative approach was proposed that explicitly focused on the (ethnic) composition of each of a city’s component parts. If – as many studies seek to show (as, e.g. in almost every issue of journals such as Social Science and Medicine and Health and Place) – segregating a group’s members in particular areas can have either a positive or a negative impact on some aspect of their life chances and/or behaviour, then knowing how segregated an area is crucial; knowledge of the citywide average situation is of no value at all. (The original paper – Poulsen et al. (2001) – was followed by a number of applications, in some of which the technique was slightly modified: for a full review see Johnston et al. (2014).)

Why is this important with regard to Merry’s book? He argues that (voluntary) separation can be positive for the segregated individuals/groups/population in certain circumstances. ¹ But he eschews the measurement issue and instead adopts a simple (one might almost say simplistic) binary – either a school is segregated or it is not. But where is the dividing line between the two categories? What percentage of the school’s population has to be drawn from the relevant ethnic group before it is considered segregated – 50%, 67%, 75% and 100%?And does it matter? In discussing social class segregation, Merry argues that ‘a critical mass of middle-class children bring the social capital of their parents with them’ (p. 151), but what is a critical mass in this context? The implication is that if a school has less than a certain percentage of its students from the middle class, then they will be insufficient in number to provide the positive role models from which the (majority) working-class students will benefit. (And the same presumably occurs with the ethnic composition of a school: if one group is too small, it will not have a positive influence on the other?) But Merry does not review any literature which explores what the critical mass might be. He refers to what he terms the ‘conceptual vagueness of an adequate education’ (p. 151); he in turn could be criticized for the conceptual vagueness of the key concept in his argument – segregation!

In most empirical studies of segregation it is treated as a continuous not a binary variable: groups are or more less segregated on a ratio scale (usually, but not invariably, from 0 to 100) and the greater the segregation the greater the consequence. (For example, the larger the percentage of an area’s population who are left wing in their politics, the greater the probability that those who – if only their individual characteristics were taken into account – are not predisposed towards a left-wing position nevertheless do so because of the influence of contacts with neighbours in their local social networks.) Merry does not incorporate this into his arguments: segregation to him is a binary not a continuous phenomenon. It is also treated – again implicitly – as temporally stable, a sort of climax situation that is attained and then is more or less fixed. But empirical studies show that this is far from the case in many situations; neighbourhood and school change is not rare – save in certain cases such as US Black ghettos for much of the 20th century. Details of how segregation and segregation change are measured may be irrelevant to Merry’s argument, but recognition that change occurs in many places – sometimes fast, sometimes slow, although sometimes not at all – is relevant to the social engineering he aspires to. The binary ideal–typical situation at the core of his theory needs modification to reflect the messier ‘real world’.

Part of Merry’s argument for ‘voluntary separation’ (or segregation) is that there are circumstances in which a stigmatized minority can better achieve equality through ‘the cultivation of self-respect’ (p. 89) if separated from other groups than if they were in some – presumably forced – ‘integrated’ situation. He – undoubtedly correctly – argues that integration cannot be conflated with equality; just because a city’s schools are integrated – or not segregated – does not mean that members of the various ethnic groups comprising their student populations get an equal quality of education. But what do segregation and integration mean in this context? Does the promotion of self-respect and equality of citizenship for a group only succeed when it is totally segregated – that is, isolated – from the rest of society? Save in very special circumstances – such as the Dutch faith schools that Merry discusses in one of his case studies – such extremes are unlikely; there is bound to be some ‘ethnic mix’. And if there is, what is the situation of the ‘majority–minority’ group in such a context, such as those few in a school dominated by members of an ethnic minority who come from the city’s ethnic majority? Do they suffer negative consequences of being in a place where others dominate? On occasion Merry does refer to the ethnic mix in a school but – after a reference to ‘segregation index levels across Europe and North America … either holding steady or worsening’, which is not in line with most recent empirical studies – says that it is ‘simply unreasonable to expect academic success to hinge on the elusive benefits that mixed settings ostensibly provide’ (p. 83), implying that complete segregation/voluntary separation is the only way to obtain the gains that integration (of whatever form) cannot. Where there is such integration, he claims, it is not ‘occurring on a significant level … [whatever significant means in this context] … in every case where the phenomenon has been studied’. (There are no references to sustain this claim; elsewhere he states that ‘Various initiatives have been adopted to engender more integrated communities, yet even when modest integrated patterns emerge, these typically do not yield the expected outcomes’ (p. 152); again, there are no supporting references, though see Goodman and Wright (2015).)

Which introduces a related point – spatial scale. Most empirical studies of segregation have been conducted at a single spatial scale only, usually reflecting the nature of the available data. Fowler (2015) has recently argued that segregation is a multi-scalar phenomenon (as have previous studies: Fischer et al., 2004; Johnston et al., 2003; Voas and Williamson, 2001) and others have even more recently proposed its evaluation in a multilevel modelling framework (Jones et al., 2015; Manley et al., 2015). The intra-city locational decisions made by households – in many cases ‘guided’ by the gatekeepers who oversee housing market operations – occur at a number of scales (different segments of a city; different neighbourhoods within those segments, etc.) and the aggregate of all those decisions produces the, ever-changing, maps of urban segregation. Merry recognizes this: Writing about religious separation in the Netherlands he notes that ‘very few segregated neighbourhoods are isolated. Mainstream culture is pervasive’ (p. 108). Few neighbourhoods in most cities (the situation of African Americans and Hispanics in many contemporary US cities providing the main exceptions: Johnston et al., 2007) are exclusive to a single ethnic group and most neighbourhoods where one minority group forms the largest component of the local population, if not a majority, are fairly close to others where the ethnic mix of the population is quite different. How do these scale differences impact upon the negative and positive consequences associated with segregation? Is a segregated neighbourhood ‘less bad’ for the members of the group concentrated there, if an adjacent neighbourhood is dominated by another group (perhaps the majority) than if all the adjacent neighbourhoods have the same ethnic composition as it? (Sampson, 2011, offers some pointers to answers to that question.) In a school, for example, a class may be segregated but the entire school integrated?

This issue is important to Merry’s main concern with education. Most schools in most cities to a greater or lesser extent draw their students from their immediate and adjacent neighbourhoods so that unless one group predominates across large tracts of a city (such as ‘white suburbia’), a school’s population (especially that of a secondary school, most of which are much larger than their primary school feeders) is likely to be more mixed than any one of the neighbourhoods that it serves. Parental choice and other mechanisms might alter this somewhat, but the probability is that in most cities many of their schools – the exact proportion depending on the relative size of the majority and minority populations – will be more integrated than segregated, relative to the situation in the neighbourhoods that they serve. For this reason, it is impossible – as Merry implicitly does – to study schools in isolation from the local milieux in which they are located. Their students may encounter quite different role models in those two separate but interacting environments. Which is the more influential?

Conclusions

Merry’s book is a very important original contribution to the study of segregation, especially ethnic school segregation. Built on very firm foundations in philosophy, he has explored the potential positive benefits that might accrue from (voluntary) segregation in certain circumstances. But by entirely eschewing the issue of measurement he has not been able to explore his argument’s full implications. Segregation may in certain theoretical contexts be a binary – either a school or a neighbourhood is segregated or it is integrated – but the empirical world is much messier. Only by exploring that messy world, of how the extent and degree of segregation varies from (multi-scalar) place to place and how that variation impacts upon the individuals, families, households and groups involved, can Merry’s arguments be fully assessed and evaluated. He has laid the foundation for an important research enterprise, extending the study of segregation in important new directions, indicating that it should be much more than the numerical measurement of who lives where and among whom. Nevertheless placing measurement at the heart of the analysis is necessary for a full evaluation of his arguments.