Introduction Creating simulations of future outcomes on the basis of model assumptions is essential for resource planning and policy formation across a wide range of disciplines. For example, population projections inform infrastructure planning across a wide range of sectors (e.g. housing, transport, utilities) and take into account past trends and potential future demographic rates; ensuring adequate healthcare provision requires assumptions to be made about the health and lifestyle of the population; and pension planning requires assumptions about life expectancy and mortality rates. Inherent in these models are endogenous factors which relate to the behaviour of the population being modelled (do they have a healthy lifestyle, when do they plan to retire) and exogenous factors (what policies are in place to curb exposure to unhealthy activities, what is the state pension age). Changing both the endogenous and exogenous factors in a model allows us to undertake simulations and experiment with different policy or behavioural interventions.
This project brings together two well established methods, choice modelling and dynamic microsimulation, to create a new modelling framework for assessing a range of policy outcomes based on decisions made by individuals as well as assumptions about external factors such as policy change. The framework will be applied to a range of application areas, but with a specific focus on health (where both techniques are well established) and crime (a field where these types of simulations are rare).
Choice Modelling Choice models are mathematical structures used to describe, explain and predict why people make certain choices when faced with specific circumstances. Applications cover numerous areas, from transport to health and education. Factors explaining the choices can include the socio-demographic characteristics of the individual, the environmental context, the choices made in the past that might have affected that outcome or the characteristics of the city/neighbourhood the individual lives in. For each one of these factors, parameters are estimated that reflect the effect, including directionality and size/weight. Choice models are particularly relevant in the context of intervention planning and policy making, since they allow researchers to explore why people in particular circumstances behave the way they do, and what is likely to happen when context changes. In addition, the mathematical structures themselves are also well suited for explaining outcomes rather than just “choices”, for example health outcomes or educational outcomes.
Microsimulation Dynamic microsimulation is used to model individual units (people, households, etc.) over time. Transition probabilities that represent the likelihood an individual will go from one state to another are estimated and applied to individuals during the simulation, for example the probability of dying (given age, sex, health status, etc.), developing a chronic disease, or retiring from the workforce. These probabilities are typically estimated using longitudinal survey datasets such as Understanding Society or the English Longitudinal Study of Ageing. Microsimulations have been used to answer a range of ‘what if?’ questions including estimating health outcomes (Goldman et al., 2014), pensioner incomes (Emmerson et al., 2004) and education and labour force outcomes (Fredriksen and Stolen 2007). Undertaking policy experiments using microsimulation involves the researcher perturbing transition probabilities based on external evidence, and examining the impacts of these changes on the simulated population.
The project is in competition for a 3.5 years EPSRC DTP 2020 Environment scholarship which will include tuition fees (£4,500 for 2019/20), tax-free stipend (£15,009 for 2019/20), and a research training and support grant.