SIDD Simulator of Individual Dynamic Decisions

About.

The modelling approach.

The Simulator of Individual Dynamic Decisions (SIDD) is a dynamic microsimulation model that projects the evolving histories of a representative population cross-section through time. The model is the product of more than a decade of research and development at the National Institute, and is designed to explore the distributional consequences of discrete changes to the economic environment, including changes to tax and benefits policy. Models of this type are very valuable for distinguishing the effects of policy changes on households with specific characteristics. For example at budget time we are used to statements like “A family with two young children will be better off, but a pensioner household worse off”. In most microsimulation models, economic behaviour is represented by simple statistical relationships. For example, savings functions may be estimated describing savings as a function of age, income and family circumstances. Labour supply may be treated in the same way, or at best treated as the outcome of a static optimisation. Macro-economic modellers have been aware of the Lucas critique for many years. The Lucas critique recognises that many decisions – and particularly those concerning the trade-offs between work/leisure and consumption/savings – are sensibly regarded as intertemporal. It follows that both current saving and labour supply are going to depend on expectations of incomes and relative prices. For example, an increase in state pensions paid to people over 65 should be expected to reduce the saving of people under 65. Or the effects of changes to the tax regime faced by middle-aged savers will depend on the sort of benefit scheme that they expect to find in place when they reach retirement. Statistical estimates of saving or labour supply functions account for these expectations only implicitly, and are therefore ill-suited to adapt to changing expectations in context of policy reform. SIDD adapts to the above observations by projecting family decisions on the assumption that these are the product of dynamic optimisation, given explicit assumptions regarding expectations. The assumption that people engage in some form of optimisation when making their decisions has been a source of criticism for models of the type discussed here.  But a powerful riposte to this argument in the field of policy analysis is that it would be odd to implement policies that work as intended only if they are systematically misunderstood. Understanding the incentives embodied by policy counterfactuals is an essential step in good policy design, even if policy-makers ultimately choose to focus upon other issues of concern when selecting between policy alternatives.  In short, the fundamental premise underlying use of the modelling framework is that it is a useful way of projecting behavioural responses to incentives embodied by policy counterfactuals; and that this is true even if people do not actually make the optimising calculations that are a central feature of the modelling approach. A behavioural model can reveal responses to alternative policy counterfactuals in a way that statistical models cannot. How do unemployment benefits affect individual’s willingness to work? What are the implications for incentives of changes to the tax relief on savings? Who is likely to respond to changes in pensions means testing? These are the kinds of questions that can only be addressed adequately using a dynamic optimisation model. Furthermore, the intertemporal aspect of the model also permits behavioural responses to be considered over the life course. For example, what effect does encouraging employment early in an individual’s life have on their wages when middle aged, and across their entire lifetime?   The analytical approach also makes explicit individual welfare, which facilitates evaluation of policy alternatives. Many policy proposals, for example, imply different effects at different stages of the life course, and for individuals located at different places in the income / wealth distribution. A revenue neutral increase in retirement benefits, for example, may require a parallel increase in tax payments – a policy counterfactual that would benefit retired individuals at the expense of the working population. The model is a useful tool for assessing whether the additional pension benefits that young households will receive in retirement are sufficient to compensate for the additional tax burden that they must bear during their working lifetime. Thus one can say whether, over the life course, a young household is better or worse off.

Dynamic microsimulation made easy.

Current best practice methods of economic analysis of savings and labour supply are notoriously difficult to implement.  This difficulty is arguably the single most important reason why such methods have played a marginal role in practical policy design and reform, despite more than 60 years of intensive research effort. The central purpose of this site is to lower the technical hurdles associated with the practical application of modern methods for analysing behavioural responses to policy reform.  This objective pursued in three ways: adopting a flexible framework for SIDD; public access to programming code; and fostering community interaction.

Pre-progammed flexibility

SIDD has been designed in a way that provides substantial flexibility for modelling alternative policy contexts.  This approach is designed to avoid complex re-programming when reflecting new country specific contexts, so that the analyst can focus upon the important job of identifying appropriate model parameters.  Each alternative adaptation of SIDD is usually given its own name, to distinguish it from others that have been produced. Some examples to date are: NIBAX: The National Institute Benefit and Tax model (UK, 2009) LINDA: The Lifetime INcome Distribution Analysis model (UK, 2016) PENMOD: The PENsions MODel, (IRE, 2010) ITALISIMO: (ITA, 2014)

Open Source coding

It is impossible to anticipate all of the features that are important for adapting a dynamic microsimulation model like SIDD to every conceivable policy context.  This places a premium on code to which public access is freely permitted.  Furthermore, the validity and efficiency of the model framework both derive benefit  from the number of “eyes” that scrutinise its workings.  These observations motivate our decision to adopt an Open Source approach to the model’s code.  See the “downloads” page for further details.

Community

A key factor motivating our decision to make SIDD freely available is the view that the costs of setting up a dynamic microsimulation model are so great that there is very substantial scope for a wide group of individuals to share the load; something economists commonly refer to as economies of scale.  Associated models that current exist are essentially bespoke adaptations, designed to consider special subjects of interest.  Interpretation of results obtained from such models requires significant faith to be placed in those responsible for model development.  Community validation of a common model structure would help to mitigate such concerns. Furthermore, by supporting one another in forums, like the one on this website, a better appreciation might be obtained for what such models can - and just as importantly - cannot say.
SIDD Simulator of Individual Dynamic Decisions

About.

The modelling approach.

The Simulator of Individual Dynamic Decisions (SIDD) is a dynamic microsimulation model that projects the evolving histories of a representative population cross-section through time. The model is the product of more than a decade of research and development at the National Institute, and is designed to explore the distributional consequences of discrete changes to the economic environment, including changes to tax and benefits policy. Models of this type are very valuable for distinguishing the effects of policy changes on households with specific characteristics. For example at budget time we are used to statements like “A family with two young children will be better off, but a pensioner household worse off”. In most microsimulation models, economic behaviour is represented by simple statistical relationships. For example, savings functions may be estimated describing savings as a function of age, income and family circumstances. Labour supply may be treated in the same way, or at best treated as the outcome of a static optimisation. Macro-economic modellers have been aware of the Lucas critique for many years. The Lucas critique recognises that many decisions – and particularly those concerning the trade-offs between work/leisure and consumption/savings – are sensibly regarded as intertemporal. It follows that both current saving and labour supply are going to depend on expectations of incomes and relative prices. For example, an increase in state pensions paid to people over 65 should be expected to reduce the saving of people under 65. Or the effects of changes to the tax regime faced by middle-aged savers will depend on the sort of benefit scheme that they expect to find in place when they reach retirement. Statistical estimates of saving or labour supply functions account for these expectations only implicitly, and are therefore ill-suited to adapt to changing expectations in context of policy reform. SIDD adapts to the above observations by projecting family decisions on the assumption that these are the product of dynamic optimisation, given explicit assumptions regarding expectations. The assumption that people engage in some form of optimisation when making their decisions has been a source of criticism for models of the type discussed here.  But a powerful riposte to this argument in the field of policy analysis is that it would be odd to implement policies that work as intended only if they are systematically misunderstood. Understanding the incentives embodied by policy counterfactuals is an essential step in good policy design, even if policy-makers ultimately choose to focus upon other issues of concern when selecting between policy alternatives.  In short, the fundamental premise underlying use of the modelling framework is that it is a useful way of projecting behavioural responses to incentives embodied by policy counterfactuals; and that this is true even if people do not actually make the optimising calculations that are a central feature of the modelling approach. A behavioural model can reveal responses to alternative policy counterfactuals in a way that statistical models cannot. How do unemployment benefits affect individual’s willingness to work? What are the implications for incentives of changes to the tax relief on savings? Who is likely to respond to changes in pensions means testing? These are the kinds of questions that can only be addressed adequately using a dynamic optimisation model. Furthermore, the intertemporal aspect of the model also permits behavioural responses to be considered over the life course. For example, what effect does encouraging employment early in an individual’s life have on their wages when middle aged, and across their entire lifetime?   The analytical approach also makes explicit individual welfare, which facilitates evaluation of policy alternatives. Many policy proposals, for example, imply different effects at different stages of the life course, and for individuals located at different places in the income / wealth distribution. A revenue neutral increase in retirement benefits, for example, may require a parallel increase in tax payments – a policy counterfactual that would benefit retired individuals at the expense of the working population. The model is a useful tool for assessing whether the additional pension benefits that young households will receive in retirement are sufficient to compensate for the additional tax burden that they must bear during their working lifetime. Thus one can say whether, over the life course, a young household is better or worse off.

Dynamic microsimulation

made easy.

Current best practice methods of economic analysis of savings and labour supply are notoriously difficult to implement.  This difficulty is arguably the single most important reason why such methods have played a marginal role in practical policy design and reform, despite more than 60 years of intensive research effort. The central purpose of this site is to lower the technical hurdles associated with the practical application of modern methods for analysing behavioural responses to policy reform.  This objective pursued in three ways: adopting a flexible framework for SIDD; public access to programming code; and fostering community interaction.

Pre-progammed flexibility

SIDD has been designed in a way that provides substantial flexibility for modelling alternative policy contexts.  This approach is designed to avoid complex re-programming when reflecting new country specific contexts, so that the analyst can focus upon the important job of identifying appropriate model parameters.  Each alternative adaptation of SIDD is usually given its own name, to distinguish it from others that have been produced. Some examples to date are: NIBAX: The National Institute Benefit and Tax model (UK, 2009) LINDA: The Lifetime INcome Distribution Analysis model (UK, 2016) PENMOD: The PENsions MODel, (IRE, 2010) ITALISIMO: (ITA, 2014)

Open Source coding

It is impossible to anticipate all of the features that are important for adapting a dynamic microsimulation model like SIDD to every conceivable policy context.  This places a premium on code to which public access is freely permitted.  Furthermore, the validity and efficiency of the model framework both derive benefit  from the number of “eyes” that scrutinise its workings.  These observations motivate our decision to adopt an Open Source approach to the model’s code.  See the “downloads” page for further details.

Community

A key factor motivating our decision to make SIDD freely available is the view that the costs of setting up a dynamic microsimulation model are so great that there is very substantial scope for a wide group of individuals to share the load; something economists commonly refer to as economies of scale.  Associated models that current exist are essentially bespoke adaptations, designed to consider special subjects of interest.  Interpretation of results obtained from such models requires significant faith to be placed in those responsible for model development.  Community validation of a common model structure would help to mitigate such concerns. Furthermore, by supporting one another in forums, like the one on this website, a better appreciation might be obtained for what such models can - and just as importantly - cannot say.