INTERNATIONAL MIGRATION
& LOW-SKILL LABOUR
MARKETS: AN AGENT BASED
APPROACH
Annual Review of Work 2012-2013
Huw Vasey & Yaojun Li, Institute for Social Change, University of Manchester
Ruth Meyer, Centre for Policy Modelling, Manchester Metropolitan University
1
Contents:
About the SCID Project
2
Why study migrant labour markets as complex adaptive systems?
3
Model 1 – A highly descriptive approach to modelling the whole UK labour
7
market
Model 2 – A problem-based approach to using ABMs to study low-skilled
9
migrant labour markets
The ‘Other Half’ model: international migration, social networks and
9
the emergence of labour market segmentation in LA
The ‘LaMESt’ model: post-Accession migration and labour market
16
segmentation in Bristol
Future directions
20
Bibliography
22
2
About the SCID project:
This project integrates two very different disciplines – social science and complexity science – in
order to gain new understanding of complex, social issues. It does this by building a series of
computer simulation models of these social processes. One could think of these as serious
versions of the Sims computer games, programmes that track the social interactions between
many individuals. Such simulations allow ‘what if’ experiments to be performed so that a deeper
understanding of the possible outcomes for the society as a whole can be established based on
the interactions of many individuals, as well as examining social phenomena which are difficult
to understand using existing social science methods.
However, a difficulty with the computer simulation of complex systems is that if they are made
realistic (in the sense of how people actually behave) they become very complicated, making the
simulation hard to understand, whilst if they are made simple enough to understand and
rigorously analyse they can be too abstract to provide findings which are useful to our
understanding of ‘real world’ social situations.
This project aims to get around this by making “chains” of related models, starting with a
complicated, ‘descriptive’ model and then simplifying in stages, so that each simulation is a model
of the one “below” it. The simpler models help us understand what is going on in the more
comprehensive ones, whilst the more complicated models reveal the ways in which the less
elaborate ones are accurate as well as how they may over-simplify. In this way this project will
combine the relevance of social science with the rigour of the “hard” sciences, but at the cost of
having to build, check and maintain whole chains of models. Building on an established
collaboration between social and complexity scientists in Manchester, this project will integrate
the two disciplines to produce new insights, techniques and approaches for policy makers and
their advisors.
The social scientists will develop ways of relating these kinds of models to the rich sources of
social data that are available. They will also ensure that the modelling results are interpreted
meaningfully and usefully, in particular guarding against over-interpretation.
For more information and background to the SCID project, please visit our
webpages at:
3
http://scid-project.org
Why study migrant labour markets as complex adaptive systems? An introduction
to ‘employment’ modelling in the SCID project
Introduction – complexity, agent-based modelling and social science:
Over the past decade social scientists have become increasingly interested in what insights the
science of complex physical and biological systems seem to provide for our understanding of
social systems and practices. Authors have explored this in regards to policy making (Geyer and
Rihani 2010; Room 2011), intersectionality (Walby 2009) social theory (Law and Urry 2004;
DeLanda 2006; 2011) and even social research methods (Law 2004).
Whilst there are striking similarities between complex physical and biological systems and social
systems, the connections between the socially complex and complexity science has largely
remained at the level of analogy. So, whilst concepts such as emergence and path dependency
have been useful aids to conceptualising the functioning of society, there has been less success in
routinely incorporating the insights of complexity science into the fabric of how social processes
are investigated, analysed and interrogated. Social complexity approaches are generally only
incorporated post hoc, and as a theoretical-analytical approach rather than as a method in itself.
The SCID project has attempted to integrate the apparent advances of complexity science
approaches through the development of Agent-based models (ABMs) of the complex social
phenomena under considerations. Whilst ABMs have been used in social science research since
at least the 1970’s (see, e.g. Schelling 1971), their use has undergone a resurgence in recent years.
Part of this has been the ability of ABMs to create behaviour which is qualitatively similar to
complex systems – that is, multiple agents interacting with each other, which often produce nonlinear and emergent behaviour. The ‘rules’ by which these agents interact are often extremely
simple, but their collective behaviour may be neither simple, nor directly predicted from their
initial states (Gilbert and Troitzsch 2005).
Migrant labour markets as complex social systems:
Migrant labour markets appear to exhibit many features of a complex adaptive system. To
illustrate this, we can use Geyer & Rihani’s (2010: 29) ‘golden rules’ of complex systems in which
the agents are self-aware (‘conscious’ in the author’s terminology):
4
Partial order: phenomena can exhibit both orderly and chaotic behaviours; e.g. migration flows
often remain stable for many years, before rapidly shifting to a period of alternative order, or
apparent disorder.
Reductionism and holism: some phenomena are reducible, others are not; e.g. Rational Choice
Theory can explain some elements of the behaviour of economic migrants, but not all. Nor do
predictions based on such behaviours tend to produce accurate predictions of global behaviours;
like many reductionist approaches they do not ‘scale’ well.
Predictability and uncertainty: phenomena can be partially modelled, predicted and controlled;
e.g. as indicated above, many models of migration and the economy work well in highly
prescribed circumstances, but rarely extend well onto a larger scale (temporally, geographically
or socially).
Probabilistic: there are general boundaries to most phenomena, but within these boundaries
exact outcomes are uncertain; e.g. economic migration occurs within certain defined and
undefined limits (i.e. it is not a free-for-all). However, within these parameters the range of
variation of possible forms of organisation is significant and rarely predictable.
Emergence: they exhibit elements of adaptation and emergence; e.g. migration flows and labour
market niches grow, coalesce and fade in ways which are not predictable from initial conditions.
Interpretation: the actors in the system can be aware of themselves, the system and their history
and may strive to interpret and direct themselves and the system; e.g. agents react to known and
assumed properties of the system, learn from the results of past behaviours to guide future
activity, and may seek to alter the system to meet their own needs (e.g. in the case of immigration
policy).
For Geyer and Rihani, such ‘golden rules’ illustrate the ways in which a complexity-based
approach to social phenomena act as a synthesis between orderly rationalist paradigms and
disorderly post-modern approaches. That is, complex systems, such as migrant labour markets,
exhibit both general, predictable, characteristics, and unexpected emergent ones. In essence, we
can understand this in terms of post-Accession migration to the UK in that we may have been able
to predict there would be flows of migrants to the UK from certain countries (e.g. Poland), but we
weren’t able to predict (or, indeed, effectively control) where they went, what work they did, or
the numbers that arrived. Nor have we been able to create adequate post-hoc reconstructions of
all the reasons for what happened. Thus, whilst post-Accession migration to the UK could be
5
understood in a rational, ordered, manner to a limited degree, the forms this order took were
emergent and not predictable from initial conditions.
However, once we have conceptualised migrant labour markets as complex adaptive systems, we
are not immediately confronted with an obvious solution as to how this translates into
researching such phenomena. We may take an extreme variant of what Kwa (2002) terms
‘baroque’ complexity and argue that, because outcomes are not predictable from initial
conditions, then the processes that lead to them must be unknowable; or, rather, that such a
concept as complexity is both fluid and never more than analogous to ‘reality’. This is very close
to the position Deleuze often finds himself in (see, e.g. 1988, 1993) and one which many social
theorists have tended to abide by when they utilise the notion of social complexity (see, e.g. Law,
2004; Urry, 2007). Conversely, many agent-based modellers with an interest in complexity have
illustrated how complex phenomena can arise from very simple rules of interaction (see, e.g.
Simon, 1996; Waldrop, 1992). A ‘classic’ example of such a process is the ‘Schelling’ model, which
illustrates how residential segregation can develop from very simple rules regarding the number
of non-similar neighbours an agent will tolerate before looking to move (Schelling, 1971).
You can find a version of the Schelling model here:
http://ccl.northwestern.edu/netlogo/models/Segregation
This would suggest that it is not the social processes themselves which are unknowable, rather it
is the results of repeatedly activating such inter-related processes which is unpredictable. Thus
small differences in initial conditions can produce vastly different outcomes (whilst even the
same initial conditions will often produce differing results). This insight is what makes using
agent-based models of migrant labour markets so appealing – if we can get the basic rules of agent
behaviour correct, we should be able to provide real insights into the often confusing and hidden
processes which produce and sustain migrant labour markets.
Whilst social scientists with an interest in migrant labour markets have been highly successful in
illustrating the inequities in the labour market, the tightly intertwined webs of actors, networks
and institutions which make up the landscape of work and employment have made definitive
statements about why such differences emerge and persist in such an environment all but
impossible. Much work in quantitative sociology has helped to provide convincing evidence as to
both the existence of inequities in the labour market and their impact. Meanwhile, qualitative
6
researchers have provided rich and nuanced accounts of how such inequity is played out in
everyday settings. However, both approaches have struggled to provide us with satisfying
accounts of how processes of differentiation and advantage emerge, become ‘successful’ and are
sustained. Social theories too have struggled to bridge this gap, with both agency- and structurebased approaches proving unsatisfactory to a growing number of researchers. Indeed, the growth
in recent years of social science approaches inspired by complexity theory (Law and Mol 2002;
Walby 2009; Geyer and Rihani 2010; Room 2011) seems indicative that now may be the time to
explore this in more sustained depth.
Indeed, the potential of complex, descriptive ABMs to illustrate the hidden processes of social life
is one of their most-attractive features. By ensuring that a deep representation of the multiple
social interactions which make up our everyday lives is represented in the models, we are
producing a simulation of the labour market which allows us to uncover the interactive processes
of inequality which are normally hidden to us. Furthermore, by manipulating the parameters of
the model, we can envision what small changes in one area may produce elsewhere. Therefore,
this approach allows us both to ask why certain outcomes are more commonplace than others,
but also what if circumstances were different. The advantages of being able to ask such questions
is clear, both for a greater understanding of the work and employment, and for providing more
satisfying answers as to how we can change the labour market for the better.
7
Model 1 – A highly descriptive approach to modelling the whole UK labour
market: Lessons learnt
Our first attempt at using ABM approaches in the ‘employment’ strand of the SCID project was to
explore ethnic occupational attainment in the UK labour market. The idea here was to develop a
highly descriptive model of the UK labour market which we could use to explore issues around
differences in occupational attainment. Whilst this fitted well with both SCID’s stated aims and
current research on work, ethnicity and immigration, it proved an extremely tough challenge.
The original conceptualisation of this approach was, “to i) explore the potential of adapting
techniques from complexity science towards the purpose of expanding the tools at the disposal
of the social researcher as outlined above, and ii) to use such innovative techniques to investigate
occupational attainment in the labour market with particular reference to ethnic diversity and
immigration.” Furthermore, we were also keen to attempt an alternative approach to building
ABMs of complex social phenomena, beginning with a highly descriptive ‘close to reality’ model,
before simplifying it in stages, so that each simulation is a ‘model’ of the one below it.
Our initial approach to resolving this dilemma was to model the demographics, behavioural rules
and interactive behaviour of the whole UK labour market, with the proviso that this ultracomplicated model could then be iteratively simplified. However, we soon ran into two related
problems: i) too many processes and ii) too little information about their functioning.
If we think of all the possible processes which may impact on the occupational attainment of
different ethnic groups in a population, we will rapidly uncover an ever growing list. Some of
these will be demographic – age, gender, education etc… – of which we often have very reliable
sources of information. Others will be life-course events with their own complicated causes and
effects, such as marriage, having children and ill-health. Others still will be hugely important, but
little understood, processes such as direct and indirect discrimination. Furthermore, for each
complication added to the model, several highly interdependent processes were revealed,
complicating the model further in turn.
This provided us with a very practical problem – to study the issue of occupational attainment in
the manner we had original envisaged required a complete model of the labour market. However,
the more we progressed, the further away completion appeared to be and the harder it was to
keep track of all the inter-related pieces of the model. This issue revealed a truth about
contemporary social science academia – spending four or five years on a project which may
8
eventually fail to produce the desired results was simply impossible. Funders and employers
(current and potential) are rarely impressed by such an endeavour, such is the pressure to
produce papers and results for academics at all levels. Furthermore, gaining a coherent
understanding of a system as complicated as a labour market is something academics spend their
lives working towards and rarely achieve – indeed, those who believe they have achieved it, are
generally proved wildly wrong by their colleagues. To attempt such a feat within the timescale of
a funded project is, at best, highly taxing.
So whilst a highly descriptive and complicated model would appear theoretically to be the right
approach, it currently appears an impractical way to use ABMs to advance our sociological
understanding of contemporary labour markets.
9
Model 2 – A problem-based approach to using ABMs to study low-skilled migrant
labour markets
The problems outlined above raise the question as to whether we can utilise the promise of ABMs
in the study of complex social systems in a manner which is both practical for contemporary social
science academics and provides genuine insights into our subjects of study, without reducing
them to abstract ‘toy’ models?
In light of this, we are currently exploring what we term a ‘problem-based’ approach to utilising
ABMs in social science – an approach with parallels to Edmund Chattoe-Brown’s recent overview
of the potential of ABMs for theory building and testing in sociology (Chattoe-Brown, 2013). In
this approach, the initial ‘target system’ to be modelled is determined by the problem under
exploration, which itself is drawn from a close reading of contemporary sociological research in
the area of concern. Thus the aim moves from building a ‘whole world’ model which can be used
to test and build sociological understandings of the world, to creating a scenario in which specific
scenarios and interdependent relationships can be explored and tested. This allows model
builders to pare down the breadth of processes to be modelled, with the downside of moving
further away from ‘reality’. However, by basing such abstractions on pre-existing theoretical and
empirical findings it is possible to retain relevance to contemporary sociological debates.
Furthermore, such foundational models can be made increasingly intricate through successive
iterations, allowing researchers to develop more nuanced and sophisticated understandings of
sociological processes. It is hoped that such an approach will allow us to create models of complex
social systems which will be practical to produce, yet will retain a relevance to our understanding
of contemporary sociological problems.
The ‘Other Half’ model – international migration, social networks and the emergence of
ethnic labour market segmentation in Los Angeles
Given our stated interest in the occupational attainment of ethnic minorities, what would be a
sensible sociological problem to initially focus on? What, in essence, could we conceptualise as
the basis of current work on the sociology of ethnic labour markets? Of course, every researcher
in the field would able to supply a different answer, but we decided to concentrate on a notion
that, implicitly or explicitly, is widely supported by social scientists working in this area – that
social networks matter. And, if this was to be our starting point, Waldinger & Lichter’s (2003) How
the Other Half Works seemed an eminently sensible place to start.
10
In this seminal study of immigration and the social organisation of labour in Los Angeles,
Waldinger & Lichter not only provide a neatly defined area of study – LA in the 1990’s – but also
supply some highly testable scenarios which are interesting, not only from a sociological
perspective, but also from a complexity one. For the former, the question could be stated as, ‘can
social networks alone drive ethnic segmentation in a low-skill labour market?’ For the latter, we
can extend this to include whether such processes are emergent and whether they lock-in over
time. These are important questions in the sociology of immigration and work, because whilst we
are well aware such segmentation happens (often rapidly), there is less clarity about the
processes that drive it. Is it simply an effect of the functioning of social networks, or do we need
to take into account other socially embedded processes, such as habitus (Bauder, 2006), dual
interpellation (McDowell et al, 2007), discrimination and job queuing (Wills at al., 2010), or the
structural forms of the labour markets in global cities and their tributaries (Sassen, 1998)?
Indeed, even if we agree that many of the factors do have an impact, the extent of the effect social
networks have is difficult to gauge by other sociological methods; perhaps an ABM could cast light
on this?
Waldinger & Lichter’s conceptualisation of how social networks impact on labour market
segmentation is clearly expressed in the chapter entitled, ‘Networks, bureaucracy and exclusion’
(2003: 83-99) and is summarised below:
Q) ‘Why do social networks so heavily influence the way workers find jobs and bosses find
help?’ (ibid: 83)
Assumption 1: ‘Most job-seekers activate their social connections to find jobs’ (ibid)
Assumption 2: ‘Employers use ties linking the workers whom they know to the new people they
may like to hire’ (ibid)
Note: The problem here is that the question stated above appears to have been
answered by the two assumptions presented in the same initial paragraph of
the chapter – if assumptions 1 & 2 are correct then the answer to the question
11
posed is implicit. However, the mechanics of this are not immediately clear.
Therefore, I suggest reframing the question as:
How do social networks so heavily influence the way workers find jobs
and bosses find help?
Waldinger & Lichter suggest four elements to consider (according to their interpretation of the
current sociological literature):
1) Networks provide information – ‘telling job seekers about opportunities and informing
employers about the characteristics of applicants’
2) Networks are instruments of influence –‘allowing job seekers to put themselves on the
inside track by proxy’
3) Networks can be used to enforce obligations – ‘so that the employer is assured that the
favors he or she does for the job-seeker and his or her accomplices will be repaid’
4) Networks can cement implicit contracts – because networks are carriers of information
and obligation, they can be used to impose the ‘rights and responsibilities of each party of
the employment exchange.’ Furthermore, ‘[t]o the extent that a group of workers feels
bound by these understanding, the employer can count of on its exercise of social control
to keep recalcitrant fellows in line’ (ibid).
This suggests two types of agent to be modelled:
1) Worker agents – they look for jobs and are enmeshed in social networks. They can pass
informationabout jobs through these networks, and also appear to get a boost to their
likelihood of getting a job they have heard about through such a connection (influence).
However, getting a job through such a network implies responsibilities to both the
employer (implicit contracts) and their social network (obligation), suggesting that social
12
networks increase your chance of getting a job, but decrease your scope for resistance to
negative actions by your employer (such as decreased pay, poor conditions etc…).
2) Employer agents – they give out jobs and use the social networks of their existing
workforce for recruitment. They have a preference for recruiting workers in the same
social networks as their existing workforce (influence), as this ensures new recruits are
trustworthy and effective (obligations). However, it is also implied that they can use these
same networks to constrain workers’ rights (such as poor pay and conditions, breaking
off of privileged access to the social networks of recalcitrant workers).
Additionally, the following are also required:
3) Social networks – these connect groups of workers. They provide information about
available jobs at the employers of social network group members (information). They
provide an unspecified advantage to getting a job where other network members are
already employed in an organisation (influence). They also act as conduits for obligations
and implicit contracts (though it is not immediately clear how this occurs).
The model is initiated in a scenario similar to the early 1950’s in LA – low-skilled jobs are
arranged in a variety of small, medium and large organisations (based on Table A9, pg. 251). Most
of these jobs are initially filled by a pre-existing majority population (invisible in the model) who
slowly vacate the jobs as they retire and do not compete for newly vacated jobs – equivalent to a
‘white’ working class moving up the labour market hierarchy in the post-war era (ibid: 9). The
rest of the filled jobs are initially taken by a ‘native’ minority group (eth0 – coloured orange in the
model), equivalent to US-born African-Americans. A small number of seed-corn agents are also
present – equivalent to ‘pioneer’ Mexican and ‘Asian’ immigrants – who visit organisations
looking for work. All agents are initiated with, on average, 3 network ties with agents in
geographic proximity who share their ethnicity. Agents are then provided information regarding
job vacancies at the organisations of linked agents.
Employers collect ‘job applications’ every tick (2 weeks) and employ their chosen agent (they
have a preference for those with social links to pre-existing employees). Employed agents may
then form new links with co-ethnics with whom they work – though agents do not form crossethnic links in this model (mimicking initial linguistic barriers).
13
The immigration rate of the non-native ethnic groups (eth1 and eth2) depends on the labour
market success of co-ethnics already in the model – the lower the unemployment rate, the higher
the immigration rate. However, once employment falls below a certain level, immigration will
cease until that level is again exceeded (this is subject to a 2-tick information delay) (see, e.g.
Stalker, 2000).
Model outcomes:
It is clear that segmentation within the model can occur – some organisations become dominated
by one of the three ‘ethnic’ groups. Furthermore, it is clear that this process is emergent, i.e. it
cannot be predicted from the initial conditions.
The still of the ‘mature phase’ model (Figure 2) shows an organisation (circled in yellow),
dominated by eth1; the numbers under the blue organisation symbol show numbers of eth0, eth1
and eth2 employed there in that order. The graph overleaf (Figure 1) shows the process of
segmentation in a similar scenario in a large organisation.
Such scenarios are consistent with Waldinger & Lichter’s hypothesis about the development and
persistence of labour market segmentation in low-skill work – once a successful labour market
niche is formed, social network processes will tend to reinforce that nascent advantage, leading
to the employment of more co-ethnics.
Therefore, it is plausible that social networks lend themselves to the emergence of locked-in
processes of labour market segmentation in low-skill work.
Figure 1 – An example of a segmenting organisation in the ‘Other Half’ model:
14
However, as the organisations circled in purple in Figure 2 (overleaf) illustrate, it is also possible
for organisations dominated by a pre-existing ethnic group to ‘lock-out’ immigrant groups. This
is also a feature described in Waldinger & Lichter’s work, whereby certain organisations
(particularly those with more bureaucratic hiring regulations) tend to remain dominated by
African-American workers. Indeed, it is this process we are looking to develop in the next
iteration of the model – introducing linguistic skills and bureaucratic hiring processes to replicate
a more nuanced labour market simulation.
Figure 2 – A still of the ‘Other Half’ model after 25 years:
15
Whilst we are developing a number of ways in which to validate the model both qualitatively and
quantitatively, we will focus here on unemployment rates. Figure 3 illustrates US Census data for
working-age populations in LA county equivalent to our model agents. Figure 4 provides a 5-run
mean of unemployment rates in the model.
Considering the current model has no exogenous shocks (such as recessions) and can therefore
be expected to be more stable, the range of unemployment rates seem a reasonable fit to ‘real
world’ data.
Figure 3: Unemployment rates by ethnicity 1950 - 2000 for ages 20-65 in LA County without
High School-level qualifications
Black US born
Hisp f born
Asian f born
UNEMPLOYMENT RATE
25%
20%
15%
10%
5%
0%
1950
1960
1970
1980
YEAR
Source: US Bureau of Labor Statistics
1990
2000
16
Figure 4 – Unemployment rates in the ‘Other Half’ Model (a 10-run average):
Thoughts and comments:
Though the ‘Other Half’ model is clearly a highly abstracted version of the low-skill labour market
in Los Angeles, it does illustrate how ABMs can be used to discuss issues of fundamental
importance to our understanding of the sociology of immigration and work. Furthermore, it
demonstrates that social networks alone can produce emergent labour market segmentation
based on ethnically homogenous social networks. This segmentation may not be as extreme as
that described in other studies (see, e.g. Wills et al., 2010), but it does indicate that labour market
segmentation (and its persistence) is not only a matter of negative discrimination against
immigrants groups, but may also be partially explained by positive discrimination in favour of
known and trusted immigrant networks, reinforced by the use of social ties as the pre-eminent
source of information about job vacancies in the low-skilled sectors of the labour market. Thus
some degree of labour market segmentation persists even when socially discriminatory
processes, such as habitus, dual interpellation or job queuing, are discounted.
The ‘LaMESt’ Model – post-Accession migration and labour market segmentation in Bristol:
Keen to test the ‘Other Half’ model in a UK setting, we decided to shift our focus to the UK and
specifically Bristol. Bristol was chosen because we wanted to develop the model in an area for
which we had plenty of qualitative as well as quantitative data and which had seen a recent influx
of migration after the accession of countries such as Poland to the EU in 2004. Whilst London was
tempting, its position as a super-diverse global city arguably made it both atypical and difficult to
17
model. Bristol was a more manageable target and we had the bonus of Huw Vasey’s PhD research
in the area to use as supporting qualitative data. Thus the Labour Market and Ethnic
Segmentation (LaMESt) model was born.
However, transferring a model to a new temporal and geographic setting is never easy, even when
the model is as abstracted as the ‘Other Half’. Bristol in 2001 (when the model was initialised)
was very different from LA in either the 1950s (the start point of the ‘Other Half’ model) or the
1990’s (its endpoint). Whilst many of the demographic changes were relatively straightforward
to adapt (such as birth rates, age ranges and ethnic composition), other factors were surprisingly
complicated – for example, getting the right balance and range of large, medium and small
organisations was far more of a modelling challenge than anticipated.
Therefore, whilst we have tried to keep this initial LaMESt model as close as possible to the LA
model, it has been necessary to incorporate a number of changes:
Worker agents – The two immigrant groups, plus one ‘native’ ethnic minority setup of the LA
model has been replaced by single ‘native’ and ‘immigrant’ categories of agent. Secondly, there is
no longer an invisible retreating majority population leaving the model as they retire. The
rationale for these changes is that the conditions which produce a labour market expansion, and
the resultant upward mobility of the majority ethnic population in 1950s LA were not relevant to
21st century Bristol. Secondly, though Bristol has a number of settled ethnic minority populations
(most notably those from Caribbean or Pakistani backgrounds), including these groups would not
reveal much about the initial development of labour market niches, because they had already
gone through this process. Replicating any pre-existing niches would make it difficult to provide
any useful insights into the growth of new niches in the labour market after the EU accessions in
2004.
Demographic data for these groups were updated from 2001 Census data (with the exception of
fertility rates, which were drawn from 2001 ONS data). The ‘Other: White’ category was used to
represent the small number of A8 migrants in Bristol in 2001 – replicating the role of ‘seed corn’
agents in the LA model.
Agents used the same processes to form social ties and search for jobs as in the previous model.
Employer agents – Unlike the ‘Other Half’ model we had no survey data regarding the types and
sizes of potential migrant-heavy organisations to draw on in Bristol. However, we did have plenty
of data about organisation size and industrial classification (SIC code), and employee socio-
18
economic classifications (NS-SEC codes) at the level of the Unitary Authority. Scaling and
replicating the proportions of low-skilled jobs in organisations was a challenge which involved
incorporating numerous data sets, and producing a pool of ‘realistic’ organisations which would
provide a reasonable facsimile of the low-skilled labour market in Bristol in 2001. When the
model is initiated approximate proportions of large, medium and small organisations are selected
from the pool until there are sufficient jobs.
Employers then use the same ‘rules’ to select employees as in the previous model, but the ‘churn’
rate (i.e. how many employees leave by choice or otherwise) is higher (4.2% - taken from LFS
data) than in the previous model to ensure the model has a realistic turnover of staff.
Social network formation and development remains unchanged.
Initial findings:
It may initially appear logical that we would see very similar outcomes in our simulated Bristol,
as we did in our model of Los Angeles; after all, we are using the same rules of behaviour to govern
how our agents act. However, the changes outlined above produce an intriguing scenario where
labour market segmentation and niche formation happens, but with a much delayed onset to the
previous model. For example, the organisation in Figure 5 (below) is typical in that the migrating
group does not become dominant in the workplace until a full two decades into the model run,
whereas this occurred much faster (between 3-5 years) in the previous model.
Figure 5 – An example of segmentation in an organisation in the LaMESt model
19
Why the change? The most obvious candidate is the removal of the openings caused by our
upwardly mobile majority ethnic group leaving the LA model (to pursue unmodelled higher level
jobs). This means the number of available slots are reduced substantially, making it much harder
for new migrants to gain a foothold in any one organisation and thus reducing the level of
immigration (because current migrants are largely unsuccessful) and delaying any emergent
segmentation between organisations for a considerable time.
It is also striking that this is nothing like what we observed in Bristol (and the UK in general),
after the accession of Poland to the EU in 2004, where we saw a rapid concentration of Accessionstate nationals in certain occupations in the low-skilled labour market (Bryant et al, 2006). These
occupations were poorly paid, of low status and generally suffered from very high staff turnover
(ibid). Furthermore, they were heavily reliant on the use of employment agencies to service their
staffing needs – the same agents who quickly became central to the employment of newly arrived
economic migrants to the UK (see, e.g. Garapich, 2008; McDowell et al., 2008). Add to this that
many new migrants were keen to get into work as soon as possible, rather than waiting for a
‘good’ job to come along (see, e.g. Ellis et al., 2007), and we would appear to have the conditions
for the rapid emergence of ethnic niches in the low-skilled labour market.
20
Future directions – extending the modular problem-based approach
So whilst ethnic social networks go a long way to explaining the emergence of labour market
niches in the conditions described by Waldinger & Lichter in Los Angeles, they are far more
constrained in the more restricted labour market environment of a simulated Bristol. This
provides us with the interesting – but hardly shocking – observation that labour market niches
are much more likely to occur where there is a pre-existing labour shortage. However, in order to
gain a more nuanced insight into the processes of labour market segmentation in early 21st
century Bristol, we need to add to our model.
Our initial findings from the LaMESt model, when taken alongside the existing literature on postAccession labour migration, imply that we first need to concentrate on two areas – poorly-paid,
low status jobs with a high turnover of staff (such as food production and processing), and labour
market intermediaries. In the second part of the workshop, we would like you to help us develop
these elements, as well as discussing and suggesting other future developments. However, I will
briefly outline our initial thoughts on developing the model below:
‘Undesirable’ jobs – we know from Worker’s Registration Scheme (WRS) data that most A8
migrants worked in poorly paid, manual occupations, often in the food sector. In essence, they
were working in jobs the ‘native’ population were unwilling to do (Wills et al., 2010).
Furthermore, such jobs often seem to be concentrated in the same organisations in the same
sectors (such as food processing). This would suggest we need to develop some way of
distinguishing between desirable and undesirable jobs. However, this isn’t simply a matter of
money – there are many poorly paid roles in the labour market, which this generation of migrants
did not initially colonise (such as shop work) – so adding wage levels would not appear to be a
satisfactory solution. In lieu of having data on the perceived desirability of different jobs in the
low-skill labour market, we can either try to find and use data on ‘hard-to-fill’ vacancies in lowskill work, or we can design another way of denoting the relative desirability of different
workplaces in the model (and the relative tolerance of ‘native’ and ‘migrant’ agents to taking less
desirable roles).
Labour market intermediaries – it has been widely argued (not least at the previous review of
work for this model) that labour market intermediaries are central to the functioning of postAccession migration to the EU, as they acted as a conduit controlling and directing the flow of new
21
migrants into local low-skill labour markets in the UK. These intermediaries may have been
resisted, disliked and distrusted, but their role is undeniable. However, the exact processes by
which this role functioned is less clear – we have found little data about the way in which these
intermediaries controlled access to the labour market. In one recent study of post-Accession
migration to the south-west of England, most interviewees had found their first working role in
the UK through an employment agency, regardless of their linguistic ability, or their use of other
forms of job searching behaviour (Vasey, 2011). Additionally, the importance of such
intermediaries seems to cut across Waldinger & Lichter’s notion of the importance of ethnic social
networks in the process of approaching and vetting potential new recruits. Should we assume
this process is simply outsourced to intermediaries? Does this mean that employers no longer
accept ‘walk-up’ applications?
Language and skills – many theorists, particularly those from a human capital background,
stress the importance of linguistic ability to the job prospects of recent migrants (Dustmann &
Fabbri, 2003; Duvander, 2001). At the very least we can assume a reasonable proficiency in the
language of the host country allows for a widening of social networks and job prospects. Indeed,
many migrants stress that the inability to improve their language skills (because of the lack of
opportunities for practice in a workplace dominated by co-ethnics) is a major barrier to labour
market advancement (Vasey, 2011). Thus, a future iteration of the project will seek to model how
language skills develop amongst migrants and what effects this has on job advancement and
labour market niches. However, in order to do this, we will need to provide more nuanced
differentiation between jobs – there is no point in introducing a skill (such as language ability) if
employers do not value it. This would imply a substantial expansion of the model to incorporate
higher level jobs and the modelling of both pre-existing language skills of migrants and the
process of language learning once in the country (for which there is relatively little source
material).
22
References:
Bauder, Harald (2006) Labor Movement: How Migration Regulates Labor Markets. Oxford:
Oxford University Press.
Bryant, Lynn, et al. (2006) A Study of Migrant Workers Employed in the Food and Drink Sector
of the South West of England. Plymouth: SRRU
Chattoe-Brown, Edmund (2013) ‘Why Sociology should use agent-based modelling’ in
Sociological Research Online 18(3) 3
DeLanda, M. (2006) A New Philosophy of Society: Assemblage Theory and Social Complexity. New
York, Continuum.
_______ . (2011) Philosophy and Simulation: The Emergence of Synthetic Reason. London,
Continuum.
Deleuze, Gilles (1988) Foucault. Trans. S. Hand. London, Continuum.
_______ . (1993) The Fold: Leibniz and the Baroque. Trans. T. Conley. Minnesota, University of
Minnesota Press.
Dustmann, Christian, and Francesca Fabbri. 2003. Language proficiency and labor market
performance of immigrants in the UK. The Economic Journal 113 (489): 695-717.
Duvander, Ann-Zofie E. 2001. Do country-specific skills lead to improved Labor Market
positions? An analysis of unemployment and Labor Market returns to education among
immigrants in Sweden. Work and Occupations 28 (2): 210-233
Ellis, Mark, Richard Wright, and Virginia Parks. 2007. Geography and the immigration division
of labour. Economic Geography 83 (5):255 - 281.
Garapich, Michal P. 2008. The migration industry and civil society: Polish immigrants in the
United Kingdom before and after EU enlargement. Journal of Ethnic and Migration
Studies 34 (5): 735-752.
Geyer, Robert and Samir Rihani (2010) Complexity and Public Policy: A New Approach to 21st
Century Politics. London, Routledge.
Gilbert, Nigel and Klaus G. Troitzsch (2005) Simulation for the Social Scientist. Second Edition.
Maidenhead, Open University Press.
Kwa, Chunglin (2002) ‘Romantic and baroque conceptions of complex wholes in sciences’ in J.
Law & A. Mol, eds. Complexities. London: Duke University Press; pp. 23-52.
Law, John (2004) After Method: Mess in Social Science Research. London, Routledge.
_______ and Annemarie Mol, Eds. (2002) Complexities: Social Studies of Knowledge Practices.
London, Duke University Press.
_______ and John Urry (2004) ‘Enacting the social’ in Economy and Society 33(3): 390-410.
McDowell, Linda, Adnina Batnitzky & Sarah Dyer (2007) ‘Division, segmentation, and
interpellation: the embodied labours of migrant workers in a Greater London hotel’ in
Economic Geography 83(1): 1-25
_______ . (2008) Internationalization and the spaces of temporary labour: the global assembly of a
local workforce. British Journal of Industrial Relations 46 (4): 750-770.
23
Room, Graham (2011) Complexity, Institutions and Public Policy: Agile Decision-Making in a
Turbulent World. London, Edward Elgar.
Sassen, Sakia (1998) Globalization and its Discontents: Essays on the New Mobility of People and
Money. New York, The New Press.
Sawyer, R. Keith (2005) Social Emergence: Societies as Complex Systems. Cambridge, Cambridge
University Press.
Schelling, Thomas C. (1971) ‘Dynamic models of segregation’ in Journal of Mathematical
Sociology 1: 143-186.
Simon, Herbert A. (1996) The Sciences of the Artifical. London, MIT Press.
Stalker, Peter (2000) Workers Without Frontiers: The Impact of Globalisation on International
Migration. London: Lynne Rienner & ILO.
Vasey, D. Huw (2011) A Complex Work of Migration: Work, Knowledge and Migration in the
South West of England. Unpublished PhD Thesis. Exeter: University of Exeter.
Walby, Sylvia (2009) Globalization and Inequalities: Complexity and Contested Modernities.
London, SAGE.
Waldinger, Roger and Michael I. Lichter (2003) How the Other Half Work: Immigration and the
Social Division of Labor. London, University of California Press.
Waldrop, M. Mitchell (1992) Complexity: The Emerging Science at the Edge of Order and Chaos.
New York, Simon & Schuster.
Wills, Jane et al. (2010) Global Cities at Work: New Migrant Divisions of Labour. London: Pluto
Press.