A New Approach to Social Mobility Models: Simulation as “Reverse Engineering”
Download as PPT, PDF2 likes356 views
The document discusses a new simulation approach to social mobility, emphasizing the challenges sociologists face in accepting simulation as a tool for understanding social phenomena. It explores the concepts of inter and intra-generational mobility, the utility of microsimulation in modeling educational and labor market dynamics, and the potential for experiments that can validate social theories. The authors aim to enhance current models by incorporating structured populations and various social dynamics to create more accurate simulations of social mobility.
A New Approach to Social Mobility Models: Simulation as “Reverse Engineering”
- 1. A New Approach to Social Mobility Models: Simulation as “Reverse Engineering” Edmund Chattoe and Anthony Heath Department of Sociology University of Oxford 3 George Street Mews Oxford OX1 2AA http://www.sociology.ox.ac.uk edmund.chattoe@sociology.ox.ac.uk
- 2. Plan of the Presentation • Why are sociologists so hard to convince about simulation? • What is social mobility? • A description of MOBSIM • Initialising populations with structure: a simulation puzzle
- 3. Sociology and Simulation • Fundamental split: “qualitative” versus “quantitative” approaches • Qualitative sociologists don’t like simulation because they don’t like any modelling • What about quantitative sociology? • That shares a view of regularities as social phenomena not “laws”.
- 4. “Real Problems” • Quantitative sociologists want to study “real” problems: – Lots of data, direct and surrounding – Social institutions (schools, factories) and social structure (family relations) – Problems set by academic consensus (tradition) rather than choice
- 5. Two Kinds of “Theory” MACRO MACRO MACRO Statistical Model World Simulation MICRO MICRO MICRO ? ?
- 6. Problems with Statistical Models • Curve fitting (rather than cross checking at different levels) is not real testing • “Theories” of micro behaviour compatible with statistical approaches must either be very “tidy” (RCT) or run the risk that the macro behaviour emerging from it not what was anticipated
- 7. Reverse Engineering • “Pick apart” dependent variables and set in order: the problem of “scope” • Do “normal simulation stuff”: experiments, validation, exploring theory lacunae • Generate original patterns (and ideally others) thus producing a simulation that is both “relevant” and falsifiable
- 8. The Scope of Models Labour Markets Demography Education ? ?
- 9. What is Social Mobility? – Movement between classes over time: • Inter and intra generational • Relative and absolute mobility – Motivations for study: • Efficiency • Social Stability • Morality – Findings • Meritocracy and myth • Comprehensive schools • Considerable constancy of ASM
- 10. MOBSIM • A microsimulation with behaviour • Initialised at “1901” • Key Features – Identities – Classes (equated with jobs) – Education (“epoints”) – Status – “Institutions” and “Biology” – Labour market
- 11. Constructive Surprise Example • Models of more than minimal complexity are falsifiable • People with less education can hold better jobs • This occurs because even though hiring is meritocratic, firing is probabilistic • Observers might attribute this to discrimination • It also suggests new research
- 12. Distribution of Education Points By Class 0 0.5 1 1.5 2 2.5 0 200 400 600 800 1000 1200 Education Points Series1
- 13. Developing the Model • Endogenising norms • Disaggregating the school sector • Dealing with anachronisms • Disaggregating the labour market • Removing “arbitrary” parameters • Adding material resources • Adding spatial location • Adding social networks
- 14. Possible Experiments I: Cultural Shifts • Homogamy • Women’s labour market participation and higher education • Divorce • Birth control, demography and genetics • Deliberate policy • Ascription and meritocracy
- 15. Possible Experiments II • Validating Statistical Assumptions – Class Aggregates – Time at Measurement – Transmission through the father • “Exogenous” Factors – War – Sectoral Change and International Adjustment
- 16. The Problem of Structured Populations I • Three kinds of structure – No structure: agents with separable attributes initialised randomly – Some structure: microsimulations in which attributes like “number of children” matter – Detailed structure: social mobility, attributes like “my father” matter • Again an unremarked feature of these models
- 17. The Problem of Structured Populations II Pseudo Histories Pseudo Parents TWO POSSIBLE SOLUTIONS
- 18. The Problem of Structured Populations III • Pseudo Histories – Consistency of pseudo history distributions with actual run distributions – Problem of “linking” pseudo histories to each other • Pseudo Parents – Hard to generate stable populations – Large “overhead” of lead time and boom and slump in initial pseudo population
- 19. The Problem of Structured Populations IV • Should we worry that this problem has never come up before? • Just how much structure is plausible for real social systems?
- 20. What is Microsimulation? • Agents and attributes • Environment and different updating rules • Initialisation