Can CAST be applied to Organisational Cases

1
Emergence in the Irish Housing System
Agent-based modelling for policy analysis
Mary Lee Rhodes School of Business
Studies Trinity College Dublin June 2005
2
What the research is

• Theory-building CAS as a lens for perceiving
  how the behaviour of the agents in public service
  cause systemic outcomes
• Combination of research methods case study /
  systems analysis / process focus
• Empirical data on housing system is driving
  specification of N-K CAS model to generate
• Information source for evidence-based policy that
  directly addresses systems effects
• Insights into strategy of housing agents
• Improvements in the application of CAS theory to
  organisational systems

3
Real World urban housing policy

• Consistent (20th century) policy objective
  housing tenure of choice focus on owner
  occupation
• Planning system introduced in 1960s to foster
  development (following on Lemass/Whitaker report
  1958)
• Urban development stories
• 1970s Brave new world
• 1980s Keeping the wolf from the door
• 1990s Holding on to the tail of the tiger
• 2000s Ich bin ein Socialist

4
Real World housing outcomes

• Good news
• Owner occupation
• Urban development
• Houses / person
• Quality of housing

• Bad news
• Waiting lists
• Homelessness
• Lack of social housing
• Decreasing affordability (renters in particular)
• Cost of housing programmes (capital, SWA-Rent,
  etc.)

5
and its a much more complex system
DoE
LA
LA
LA
Clear mission Rational decisions Hierarchical
control Professional admin
Multiple objectives Mix of public, private,
non-profit Varying forms of control markets hie
rarchy networks Lots of layers of
responsibility
6
Real world - recent observations

• No matter what we did, prices kept rising
  interview with Department of the Environment and
  Local Government 2002
• More than any other since the foundation of the
  state, the last decade (90-00) has been marked by
  significant change in the system of housing
  provision in Ireland - Norris Winston 2004
• Overall it (the Irish housing system) is a
  dynamic but unbalanced system Fahey 2005

7
Is CAS relevant to the problem?

• CAS models represent a genuinely new way of
  simplifying the complex and showing how
  complex outcomes flow from simple schemata
  Anderson 1999
• The (systems) ideas most relevant to public
  policy are those associated with complex adaptive
  systems, soft systems approaches and developing
  learning systems Chapman 2002

8
CAS lots of options to choose from
Anderson 1999, Holland 1998, Kauffman 1993,
Prigogine Stengers 1984, Siggelkow Levinthal
2003

• Key aspects of real world
• Independent agent decision-making
• Interdependence among decisions and agents
• Fitness is exogenously and endogenously defined
• CAS Elements required
• Agents
• Decisions and interdependencies
• Fitness values / dynamics of change
• System state variables
• Initial condition variables (system constraints)

9
What / who are the agents?

• Fundamental agents individuals choosing to
  engage in organising behaviour (Rhodes
  Mackechnie 2003)
• Fundamental agents may also engage in organising
  behaviour forming new agents
• Types of fundamental agents in housing (15)
• Consumption/transfer agents (5)
• Production/supply agents (5)
• Policy influencing agents (5)

10
Agents in the housing system

• Consumption / Transfer Agents
• Buyers, Renters, Investors
• Estate Agents, Finance Providers
• Production / Supply Agents
• Developers / Builders
• Housing Managers
• Product service suppliers
• Policy Influencing Agents
• Civil Service / Planners
• Agencies / Policy Consultants
• Associations

11
What do agents decide?

• Strategy Elements
• Value proposition VFM, Mission, Quality,
  Customer focus
• Arenas geography, product types, customer
  segments
• Organisational form centralised, decentralised,
  outsourced
• Implementation grow/contract, speed, sequence of
  steps
• Compete / Co-operate

• Decision Factors
• External economic, social, political factors
• Internal organisational capacity, quality of
  network linkages
• Adaptive planning system, co-ordination/competit
  ion in sector, innovation dissemination,
  influence of association(s)

12
Decision Factors

• External
• House prices, interest rates, capital
  availability
• Politics / policy
• Labour availability / issues
• Social issues
• Internal
• Organisational capacity
• Staff skills, loyalty
• Network linkages, information access

• Adaptive
• Planning system
• Amount/rate of innovation / learning in sector
• Barriers to entry into sector
• Influence of sector association on policy-makers
• Co-operation / competition level(s)

• Individual
• Trust in counterparts
• Confidence in others skills
• Family influence / life cycle

13
What is meant by interdependence?

• An interaction between two Ns exists if there
  is a statistical correlation between them. This
  may be either a tendency to co-exist or a
  tendency to mutually exclude.

Based on a contingency table analysis
14
Where does fitness fit?

• Fitness is the relative value of different
  decision combinations. Value is both exogenous
  and endogenous to the system with certain agents
  having more influence than others over the value
  of various combinations and the possibility of
  particular Ns.

15
Landscapes are agent-specific, but

• Overlapping
• Subject to change through agent mutation or
  exogenous influence

K
K
K
K
K
K
K
K
K
K
16
How does the system change?

• Entry/exit, variation, selection, feedback
• Agents form in order to fulfil a purpose. Their
  fundamental purpose is to create value that, in
  the absence of organisation, would not be
  possible. But the definition of value is agent
  specific and is represented in the system by the
  assignment of fitness value to each of the N
  choices that are possible.
• The initial conditions and/or structure of the
  system at the time of agent formation will
  influence the assignment of fitness values.
• An agents initial configuration of choices is
  random (though some constraints may apply).
• Agents search for greater fitness through
  search processes that are constrained based on
  the organisational structure of the agent and the
  interdependencies amongst choices (K).
• Outcomes of the system result from the number of
  agents choosing combinations of choices
  (landscape population).

17
Pause for breath

• CAS Elements required
• Agents
• Decisions and interdependencies
• Fitness values / dynamics of change
• System state variables
• Initial condition variables (system constraints)

18
What system states are of interest?

• House prices defined as mortgage/rent/
  service/maintenance cost as of disposable
  income
• Capital appreciation (wealth accumulation) rate
• of people in unsuitable / no dwellings
• Time capital required to produce new dwelling
• Quality of housing stock
• of GDP in provision of housing ratio of
  private to public investment
• Housing policy stability

Indicates that a particular subset of the group
is interested in this state variable.
19
Whats left?

• Complete analysis of decision interdependencies
• Figure out how to model fitness based on
  empirical data
• Figure out how to link fitness with systems
  outcomes (aggregate location of agents on
  performance landscape corresponds to system
  outcomes?)
• What is the nature / effect of initial
  conditions?
• Find appropriate software to build model
• Collect/define measures of actual systems
  outcomes over a specified time period for use in
  validating model
• Develop / apply validation approach for model.

Just a few minor details
20
Real World policy implications so far

• Economic models with simplifying assumptions
  miss a significant part of the system and we
  can list them!
• Policy based only on assumptions of competitive
  dynamics in the system is likely to be flawed
• Measures of systems outcomes could be improved to
  support evidence-based policy making
• Policy-makers need to consider the type of
  decision, type of agent and relevant decision
  factors that influence desired outcomes in
  developing policy instruments
• Irish policy bias towards OO advantage may no
  longer be appropriate

21
Research method(s) a bit of a hybrid
investigating complex phenomena to develop
hypotheses out of a rich contextual framework
(Rhodes 1997)
Case Study
Total Systems Intervention uses a range of
systems metaphors to encourage creative thinking
about organisations and the difficult issues
their managers have to confront Flood Jackson
1991
Systems Analysis
Process Approach
tracking the people, ideas, transactions,
contexts and outcomes of events relevant to the
particular outcome of interest (Poole et al 2000)
22
What sort of empirical research is required?

• What agents exist and what combination of
  characteristics do they display?
• How do (individuals in) initiatives perceive the
  environment? What factors are important to them?
  
• How do (individuals in) initiatives make
  decisions? What information do they use? How
  does the process unfold?
• What is the nature of the connections among
  initiatives?
• What outcomes of the system, if any, are
  important to the participants? How have these
  changed over the period of study?

Existing studies
Surveys
Interviews
Longitudinal outcome data