Transcript Document

Migration in an agent-based
labour market
Can ABM be used to influence
policy decisions?
Jeremy Franklin and Kate Mieske
March 2011
Contents
• Motivation and policy context
• Developing an agent-based labour market
▫
▫
▫
▫
Empirical regularities
Model structure
Model validation
Policy experiment
• Conclusions and limitations
The views expressed are of the authors only, and do not represent the views or
policy of the Home Office and Migration Advisory Committee!
Motivation
• Trained economists, employed in the Government Economics
Service (GES)
• Interested in new modelling techniques
• Heard a lot about ABM
▫ Origin of Wealth (Beinhocker)
▫ Work of Sante Fe Institute
▫ Presentations to the GES by Nigel Gilbert and Paul Ormerod
• Wanted to find out what it was all about and how it could be applied
to public policy
• So applied for User Fellowships with SIMIAN
Policy context
• Formerly employed by the Migration Advisory Committee (MAC)
• MAC are a NDPB that provide advice and recommendations to
Government on key migration issues (most have previously been
accepted)
• The MAC is made up of 5 leading academic labour market
economists (independent of Government)
• The secretariat is staffed by Civil Servants (both policy officials and
economists)
• Key questions that MAC deal with are:
▫ How to identify shortages in the labour market?
▫ What should be the level of limits for economic migration?
▫ What is the economic/social/public service impact of migration?
• Further information and past reports can be found here:
www.ukba.homeoffice.gov.uk/mac
Why could ABM be useful?
• Benefits
1.
2.
3.
4.
Useful way of incorporating implications of micro behaviours and
aggregating up
Can vary underlying assumptions very easily
Include more realistic behavioural theories (people aren’t always
‘perfectly rational’)
Can introduce geography, look at regional effects
• Key challenges
1.
2.
3.
Likely to be limited to theory (very difficult to generate calibrated
model to aspects of UK economy – too many factors to consider)
What are the key macro patterns – are these sufficient?
How can conclusions be interpreted both in economics but also agentbased literature
……. plus many more
Developing a simple agent-based
labour market
• Aim: Scope the potential for using ABM to influence policy, by:
1.
2.
3.
Designing a model based on both existing economic theory/evidence
and latest agent-based studies
Validating to key empirical regularities
Testing the model implications under different policy situations
Past literature
• Economics:
▫ Big move towards ‘micro-foundations’ in macro economic models
▫ ‘Search-matching’ (Pissarides and Mortensen) – frictions that affect job
search and vacancy creation have real effects on macro-economy
▫ These frictions lead to an ‘aggregate matching function’ – the rate at
which unemployed workers are matched to new vacancies
▫ The Beveridge curve (a plot of vacancies against unemployment) can
show empirically the scale of mismatch in the economy
• Agent-based:
▫ Several studies unpick the ‘aggregate matching function’ and examine its
micro-properties (Richiardi (2006), Neugart (2004), Dosi et al (2004)
▫ Undertake basic validation of their models using three empirical
regularities…
Empirical regularities
370
2.9%
360
2.7%
real earnings
2.3%
2.1%
340
330
1.9%
320
1.7%
310
1.5%
4.0%
3) Wage
curve
350
2.5%
Vacancy rate
1) Beveridge
curve
300
5.0%
6.0%
7.0%
8.0%
0%
9.0%
2%
10%
% change in earnings
% change in GDP
8%
10%
4%
8%
6%
4%
2%
0%
-2%
-4%
2%
(Wage
curve in
% changes)
0%
-2%
-4%
-6%
-6%
-8%
-40%
6%
6%
12%
2) Okun’s
curve
4%
unemployment rate
Unemployment rate
-20%
0%
20%
40%
% change in the unemployment rate
60%
-8%
-20%
-10%
0%
10%
20%
30%
40%
50%
% change in the unemployment rate
… but there are also many others!
Model structure
• 2 types of agent: firms and workers, who are randomly distributed
in an environment
Firms:
▫
▫
▫
▫
produce one type of good using only labour (L) and technology (A)
labour exhibits diminishing returns, Y = AitLit1/2
maximise profit according to, ∏ = pAitLit1/2 – wL
simplify decision making by assuming good traded on global markets
(so sell everything produced, at a fixed price)
▫ therefore firms only compete for workers
Workers:
▫ want to work and search for jobs
▫ wish to maximise their earnings, whether in employment already or
unemployed.
▫ there is a universal reservation wage below which no worker will work
Model processes
(1) Decide optimal size of workforce.
 If vacancies, then advertise.
 If excess staff, then fire workers
randomly.
Firms
(5) Produce goods
(6) Count profits and update wage offer.
 If still have vacancies , then
increase the wage (only if
profitable to do so)
(7) Each firm receives a random
productivity shock.
(4) Count job offers and accept the one
offering the highest effective wage.
Workers
(3) Count number of applications and
issue job offers
(2) Search for vacancies.
42%
70
37%
60
32%
50
average wage
1) Beveridge
curve?
vacancy rate
Basic model validation
27%
22%
17%
3) Wage
curve?
40
30
20
12%
10
7%
0
2%
4%
14%
24%
34%
44%
0%
54%
10%
20%
8%
2.0%
6%
1.8%
% change in average wage
% change in total output
2) Okun’s
curve?
4%
2%
0%
-2%
-4%
-6%
-8%
-8%
30%
40%
50%
unemployment rate
unemployment rate
(Wage
curve in
% changes?)
1.6%
1.4%
1.2%
1.0%
0.8%
0.6%
0.4%
0.2%
-6%
-4%
-2%
0%
2%
4%
change in the unemployment rate
6%
8%
0.0%
-8%
-6%
-4%
-2%
0%
2%
4%
change in the unemployment rate
6%
8%
Policy experiment: costs to moving
• Introduce costs to moving that are proportional to distance
between firms and workers (therefore, workers consider cost of
moving in decision to accept new job)
• Examine prevalence of labour shortages and wage volatility
• Firms find it harder to recruit in period.
• Wage volatility (relative to productivity) increases (firms need to
increase wages by more to expand)
Cost = 0
Cost = 5
Conclusions and limitations
• No ground-breaking results, but…
▫ Provided Committee with useful introduction to ABM for considering
future research possibilities
▫ Highlighted the key benefits and challenges with applying ABM to
economic issues
▫ Constructed a basic ABM framework for thinking about labour market
issues, that the Home Office/ MAC can develop further
▫ Identified a number of avenues to take research further
• Limitations:
▫ Theoretical model only – Government policy places much greater weight
on empirical studies
▫ Very simplified – although for ABM to add any value to economic theory,
must be able to identify causal links (Haldane and May (2011) and
Farmer et al (2009) are great examples!) and not become a ‘black box’
Scope for further developments
• Too many to list!
▫ Introduce migration into the economy – need to use evidence on
assimilation, but the story is mixed
▫ Introduce capital into the system – allowing firms to choose their
production methods
▫ Introduce heterogeneity across agents – such as skill level
▫ Endogenise entrepreneurial activity within the system (unsuccessful
firms die and workers can save and become new firms)
▫ Allow firms to compete in prices and quantities produced
▫ Compare traded and non-traded goods
• In summary, ABM presents a very exciting methodology for
examining economic issues and we look forward to seeing new
developments!