What Can Simulation Contribute to the Study of Labour Markets?

1
What Can Simulation Contribute to the Study of
Labour Markets?
Edmund Chattoe-Brown (ecb18_at_le.ac.uk)
Department of Sociology, University of Leicester
http//www.simian.ac.uk
2
Thanks

• This research funded by the Economic and Social
  Research Council as part of the National Centre
  for Research Methods (http//www.ncrm.ac.uk).
• The usual disclaimer applies particularly
  regarding Nigel Gilbert (co-PI SIMIAN, Sociology,
  Surrey).

http//www.simian.ac.uk
2
3
The implications of innovation

• Despite the quest for innovation, there doesnt
  seem to be much reflection on its implications in
  social science (particularly as regards research
  methods). Equal timing example Trivial but
  interesting.
• Most innovations (as anywhere) are more of the
  same (focus groups and semi-structured
  interviews).
• In the 1920s-1930s both ethnography and
  statistical models were innovations in sociology.
• There is very uneven coverage of methods across
  the social sciences and the number of major
  methods in each discipline is small. It is thus
  very easy to mistake local conditions for
  global reality Almost no qualitative research
  in economics, very few sociological (as opposed
  to economic or psychological) experiments in
  sociology.

http//www.simian.ac.uk
3
4
A thought experiment

• Suppose there could be radical innovation in
  research methods (rather than import of existing
  methods), what impact would that have?
• A whole new place to stand with respect to all
  research conducted until that point.
• A truism that each research method shapes the
  questions raised and the kind of solutions put
  forward?
• A radically innovative method could thus
  potentially cast light on the biases, lacunae and
  theoretical preconceptions of whole substantive
  research fields.
• This opportunity seems hugely important and yet
  we give little thought to blue sky questions
  like How could such radical innovations in
  methods be found?

http//www.simian.ac.uk
4
5
Part of a wider problem

• Ethnographers criticise statisticians for over
  simplifying social processes. Statisticians
  respond that their models fit and ethnographers
  cannot show that the additional complexity
  makes a difference to statistical analysis.
• We have an impasse because neither side can
  express insights within the framework of the
  other. It becomes a matter of religious faith
  how complex a particular social phenomenon
  actually is. This is not relativism. There is a
  fact of the matter but these methods cant
  access it.
• Methods show how well they apply by their own
  criteria but not which method should be applied
  in the first place No wrong method error.
  Statisticians never conclude from a low R
  squared that they should become ethnographers!

http//www.simian.ac.uk
6
A large complex system

• Workers are recruited into organisations and may
  be promoted or fired. Ultimately they retire,
  leave or die.
• Workers acquire skills through education and in
  work. It is these skills that fit them for
  certain jobs.
• Organisations must combine skills into jobs to
  create output with value exceeding the cost
  of input. (This includes charities and the
  probation service too.) If they cant, they fail.
  The population of organisations is relatively
  stable but, as well as failure, founding and
  self-employment occur.
• Organisations have internal structure permitting
  promotion, control, technical innovation and so
  on.
• Producers are also consumers.

http//www.simian.ac.uk
6
7
Now what?

• How do we engage with this large complex system?
• This isnt really a theory. Each broad
  description can be progressively subdivided into
  observable social processes (recruitment and
  thence short listing, interviewing and final
  negotiations to appoint).
• We see the competing methods clichés operating
  here. Ethnography can tell us in exhaustive
  detail what happens in a job interview but not
  what the resulting aggregate picture would be.
  Statistical analysis can tell us how attributes
  shape outcomes but not what dynamic
  micro-foundations are giving the patterns.
• The same style of argument might be applied to
  sub-components of the system.

http//www.simian.ac.uk
7
8
Sleight of hand?

• Realistically, I cant talk through a detailed
  simulation of the labour market. It would take
  too long and almost certainly bog down in
  discussions about assumptions.
• What I am going to do instead is to explain a
  very simple simulation that has nothing to do
  with labour markets to show how the method works
  and then show you a simulation of labour markets
  very briefly to prove that I am not selling a
  Reliant Robin by advertising a Rolls Royce.
• This is consistent with what anyone would do when
  explaining an existing method to novices. (You
  dont start teaching regression with 25 variables
  and a whole bunch of interaction terms.) Oddly,
  however, for a novel method, it makes people
  incredibly sceptical.
• Dont need to know what goes on under the
  bonnet regarding programming either?

http//www.simian.ac.uk
8
9
The Schelling segregation model

• Agents live on a square grid. Each has maximum
  8 neighbours.
• There are two types of agents (red and green).
  Some grid spaces are vacant. Initially everything
  distributed randomly.
• All agents decide what to do in the same very
  simple way.
• Each agent has a preferred proportion (PP) of
  neighbours of its own kind (0.5 PP means you want
  at least half your neighbours to be your own kind
  - but you would accept all of them i. e. PP is
  minimum.) Vacant grid spaces dont count which
  is why the PP is a fraction not a number.
• If an agent is in a position that satisfies its
  PP then it does nothing otherwise it moves to a
  vacant site chosen at random.

http//www.simian.ac.uk
9
10
Initial random state
http//www.simian.ac.uk
10
11
Clustering
Schelling (early seventies) was interested in
urban residential segregation of ethnic groups.
http//www.simian.ac.uk
11
12
Two questions

• What is the smallest PP that will produce
  clusters?
• What happens when the PP is 1?

http//www.simian.ac.uk
12
13
Answers

• About 0.3.
• No clusters form.
• Challenge Had you seen the cluster data
  generated by PP0.3, might you (if you were of a
  particular political or social scientific
  persuasion) have attributed xenophobia to the
  system? (Or individual preference?)

http//www.simian.ac.uk
13
14
Simple individuals/complex system
Counter-intuitive macro (social) results from
simple micro interactions. A non-linear (and
complex) system.
http//www.simian.ac.uk
14
15
The punch line

• Simulation is a macroscope because it allows us
  to see complexity somewhat like a microscope
  allows us to see very small things.
• Even in this trivially simple (and behaviourally
  implausible) system, the link between micro and
  macro is not intuitive.
• We can now see why grossing up practices or
  drilling down from aggregates is problematic.
• The solution offered by simulation is an explicit
  process specification represented as a computer
  programme (compare narratives/equations).

http//www.simian.ac.uk
15
16
A new role for a new method?

• In this framework, research methods can do almost
  (but not quite) what they always did. Simulation
  wont take over and the worst work in
  simulation comes from avoiding data.
• Ethnography becomes more purposive because it
  really builds process theory (rather than
  typology or journalism).
• Statistics does not compare models with data but
  real data with simulated data How often people
  move house, how big clusters are, what shape and
  so on.
• What we want to know How people see
  neighbourhoods, how (and where) they decide to
  move, what opportunities and constraints exist
  back to properties of a large structured system,
  compare How do people look for jobs?
• To hint at an answer to a previous question
  Innovative research methods might exist in parts
  of the research process neglected by existing
  ones (here micro-macro theory building).

http//www.simian.ac.uk
16
17
The labour market simulation

• Work in progress with Nacho Gª-Valdecasas Medina
  using some ideas suggested by Sean Moley.
• An attempt to capture the large scale structure
  of the labour market already described.
• Very simple at this point Better to have
  something for every process than detail in a
  few and some processes missing altogether. (A
  hypothesis about abstraction of process.)
• Needs to reach steady state after
  initialisation.
• Start by setting parameters of various kinds then
  access dynamic outputs of a range of types
  (potentially to be compared with real data).
• Firms create jobs (sets of minimum skill
  requirements and associated wages) and then find
  workers to fill them but the set of jobs must be
  profitable to the firm as a whole.

http//www.simian.ac.uk
17
18
Example output 1
Steady state visible above. Firms arent adapting
to use all worker skills.
http//www.simian.ac.uk
18
19
Example output 2
Comforting that without being told the system
fails to employ the least skilled workers.
http//www.simian.ac.uk
19
20
Example output 3
Again, wages (arbitrary units) have sensible
values and so do unemployment rates. Distribution?
http//www.simian.ac.uk
20
21
Example output 4
To show Im not selling the method Self
employment plainly not sensible. Problem with
production function for single person firms.
http//www.simian.ac.uk
21
22
Conclusions

• Could simulation be a radical innovation? If
  so, its main contributions could be to revisit
  what we already know and represent a large
  complex system effectively.
• Examples of new questions raised by the
  approach
• What, if anything, stops all non transitional
  unemployment occurring in the lowest educated
  group?
• How are the skill requirements of job sets
  created?
• What are the market implications of rising skill?
  Firms need to create hierarchies to retain more
  skilled staff.
• How, mediated by the labour market, is a match
  between skill supply (education) and skill demand
  (employment) achieved?
• What might happen in this system if meritocracy
  was complicated by issues of gender and
  ethnicity?
• Not an attempt to favour simulation. Just
  another research method that advantageously
  arrived late. There may be more. Equally clearly,
  however, not just a fix to the limitations of
  an existing method and thus not more of the
  same.

http//www.simian.ac.uk
22
23
Further resources

• Simulation Innovation, A Node (Part of NCRM
  research, training and advice)
  lthttp//www.simian.ac.ukgt.
• NetLogo (software used here, free, works on
  Mac/PC/Unix, with a nice library of examples)
  lthttp//ccl.northwestern.edu/netlogo/gt.
• Simulation for the Social Scientist, 2nd edition,
  2005, Gilbert/Troitzsch. Dont get first
  edition, not in NL!
• Agent-Based Models, 2007, Gilbert.
• Journal of Artificial Societies and Social
  Simulation (JASSS) lthttp//jasss.soc.surrey.ac.uk
  /JASSS.htmlgt. Free online and peer reviewed.
• simsoc (email discussion group for the social
  simulation community) lthttps//www.jiscmail.ac.uk
  /cgi-bin/webadmin?A0SIMSOCgt.

http//www.simian.ac.uk
23
24
Extra slides

• If needed for questions.

http//www.simian.ac.uk
24
25
Why? So what?

• Because PP is a minimum, people are always happy
  inside a cluster of their own kind.
• If a cluster is full (no internal vacancies)
  then it cannot be disrupted except at the
  edges.
• Whether clusters form thus depends on whether
  their shape is compatible with the PP for edge
  agents. (No sharp corners possible Minimum
  cluster size?)
• When PP is 1, no shape of the cluster edge is
  compatible with the satisfaction of edge agents
  so the cluster cannot form.
• An aggregate entity (the cluster) thus becomes a
  structuring principle for individual behaviour. A
  toy version of structuration?

http//www.simian.ac.uk
25