Sequentially calibrating a Bayesian microsimulation model to incorporate new information and assumptions

Background Microsimulation models are mathematical models that simulate event histories for individual members of a population. They are useful for policy decisions because they simulate a large number of individuals from an idealized population, with features that change over time, and the resulting event histories can be summarized to describe key population-level outcomes. Model calibration is the process of incorporating evidence into the model. Calibrated models can be used to make predictions about population trends in disease outcomes and effectiveness of interventions, but calibration can be challenging and computationally expensive. Methods This paper develops a technique for sequentially updating models to take full advantage of earlier calibration results, to ultimately speed up the calibration process. A Bayesian approach to calibration is used because it combines different sources of evidence and enables uncertainty quantification which is appealing for decision-making. We develop this method in order to re-calibrate a microsimulation model for the natural history of colorectal cancer to include new targets that better inform the time from initiation of preclinical cancer to presentation with clinical cancer (sojourn time), because model exploration and validation revealed that more information was needed on sojourn time, and that the predicted percentage of patients with cancers detected via colonoscopy screening was too low. Results The sequential approach to calibration was more efficient than recalibrating the model from scratch. Incorporating new information on the percentage of patients with cancers detected upon screening changed the estimated sojourn time parameters significantly, increasing the estimated mean sojourn time for cancers in the colon and rectum, providing results with more validity. Conclusions A sequential approach to recalibration can be used to efficiently recalibrate a microsimulation model when new information becomes available that requires the original targets to be supplemented with additional targets.

The Distributional Effects of Value-Added Taxes in OECD Countries

<p>Most OECD countries’ value-added tax (VAT) systems apply reduced VAT rates to a selection of expenditure items in order to achieve distributional goals, and – to a lesser extent – social, cultural and employment-related goals. This thesis investigates the distributional effects of the VAT in OECD countries, and the merits of using reduced VAT rates to achieve distributional goals. The research adopts a microsimulation modelling approach that draws on household expenditure microdata from household budget surveys for an unprecedented 27 OECD countries. A consistent microsimulation methodology is adopted to ensure cross-country comparability of results. Non-behavioural VAT microsimulation models are first built to examine the overall distributional impact of the current VAT systems in each country. The research assesses the competing methodological approaches used in previous studies, highlighting the misleading effect of savings patterns on cross-sectional analysis when VAT burdens are measured relative to income. Measuring VAT burdens relative to expenditure – thereby removing the influence of savings – is found to provide a more reliable picture of the distributional impact of the VAT. On this basis, the VAT is found to be either roughly proportional or slightly progressive in most of the 27 OECD countries examined. Nevertheless, results for a small number of countries (Chile, Hungary, Latvia and New Zealand) highlight that broad-based VAT systems that have few reduced VAT rates or exemptions can produce a small degree of regressivity. Results also show that even a roughly proportional VAT can still have significant equity implications for the poor – potentially pushing some households into poverty. Behavioural VAT microsimulation models are then built for 23 OECD countries to investigate whether reduced VAT rates are an effective way to support poorer households, and whether the use of targeted cash transfers would be more effective. The behavioural microsimulation methodology follows the Linear Expenditure System based approach of Creedy and Sleeman (2006). Complementing this approach, a Quadratic Almost Ideal Demand System (QUAIDS) is estimated specifically for New Zealand, thereby providing the first estimates of a QUAIDS model based on New Zealand data. Simulation results show that, as a whole, the reduced VAT rates present in most OECD countries tend to have a small progressive impact. However, despite this progressivity, reduced VAT rates are shown to be a highly ineffective mechanism for targeting support to poorer households: not only do rich households benefit from reduced rates, but they benefit more in aggregate terms than poor households do. When looking at reduced VAT rates applied to specific products, results are found to vary considerably. Reduced VAT rates specifically introduced to support the poor (such as reduced rates on food consumed at home and domestic utilities) are generally found to have a progressive impact, though rich households still receive a larger aggregate benefit than poor households. In contrast, reduced VAT rates introduced to address non-distributional goals (such as reduced rates on restaurants, hotels, and cultural and social expenditure) often have a regressive impact. Additional simulation results show that an income-tested cash transfer will better target support to poorer households than reduced VAT rates in all countries. Furthermore, even a universal cash transfer is found to better target poorer households than reduced VAT rates. However, results also show that it is very difficult for an income-tested cash transfer to fully compensate all poor households for the removal of reduced VAT rates. This is due to the significant variation in the underlying consumption patterns across households. While a small number of poor households lose out from replacing reduced VAT rates with targeted cash transfers, those that receive support are instead determined by income and family characteristics as opposed to consumption tastes – thereby increasing horizontal equity. Furthermore, many households are lifted out of poverty as revenue previously transferred to richer households is now transferred to poorer households. These results empirically confirm the theoretical expectation that, where available, direct mechanisms (whether via the income tax or benefit system) will better achieve distributional goals than reduced VAT rates. Countries that currently employ reduced VAT rates to achieve distributional goals should therefore consider removing these reduced rates and adjusting their income tax or benefit systems to achieve these distributional goals instead. Countries should also consider removing reduced VAT rates aimed at non-distributional goals where a more effective instrument is available to achieve the particular policy goal. At a minimum, the merits of these reduced VAT rates should be reassessed in light of their negative distributional impact.</p>

The Distributional Effects of Value-Added Taxes in OECD Countries

<p>Most OECD countries’ value-added tax (VAT) systems apply reduced VAT rates to a selection of expenditure items in order to achieve distributional goals, and – to a lesser extent – social, cultural and employment-related goals. This thesis investigates the distributional effects of the VAT in OECD countries, and the merits of using reduced VAT rates to achieve distributional goals. The research adopts a microsimulation modelling approach that draws on household expenditure microdata from household budget surveys for an unprecedented 27 OECD countries. A consistent microsimulation methodology is adopted to ensure cross-country comparability of results. Non-behavioural VAT microsimulation models are first built to examine the overall distributional impact of the current VAT systems in each country. The research assesses the competing methodological approaches used in previous studies, highlighting the misleading effect of savings patterns on cross-sectional analysis when VAT burdens are measured relative to income. Measuring VAT burdens relative to expenditure – thereby removing the influence of savings – is found to provide a more reliable picture of the distributional impact of the VAT. On this basis, the VAT is found to be either roughly proportional or slightly progressive in most of the 27 OECD countries examined. Nevertheless, results for a small number of countries (Chile, Hungary, Latvia and New Zealand) highlight that broad-based VAT systems that have few reduced VAT rates or exemptions can produce a small degree of regressivity. Results also show that even a roughly proportional VAT can still have significant equity implications for the poor – potentially pushing some households into poverty. Behavioural VAT microsimulation models are then built for 23 OECD countries to investigate whether reduced VAT rates are an effective way to support poorer households, and whether the use of targeted cash transfers would be more effective. The behavioural microsimulation methodology follows the Linear Expenditure System based approach of Creedy and Sleeman (2006). Complementing this approach, a Quadratic Almost Ideal Demand System (QUAIDS) is estimated specifically for New Zealand, thereby providing the first estimates of a QUAIDS model based on New Zealand data. Simulation results show that, as a whole, the reduced VAT rates present in most OECD countries tend to have a small progressive impact. However, despite this progressivity, reduced VAT rates are shown to be a highly ineffective mechanism for targeting support to poorer households: not only do rich households benefit from reduced rates, but they benefit more in aggregate terms than poor households do. When looking at reduced VAT rates applied to specific products, results are found to vary considerably. Reduced VAT rates specifically introduced to support the poor (such as reduced rates on food consumed at home and domestic utilities) are generally found to have a progressive impact, though rich households still receive a larger aggregate benefit than poor households. In contrast, reduced VAT rates introduced to address non-distributional goals (such as reduced rates on restaurants, hotels, and cultural and social expenditure) often have a regressive impact. Additional simulation results show that an income-tested cash transfer will better target support to poorer households than reduced VAT rates in all countries. Furthermore, even a universal cash transfer is found to better target poorer households than reduced VAT rates. However, results also show that it is very difficult for an income-tested cash transfer to fully compensate all poor households for the removal of reduced VAT rates. This is due to the significant variation in the underlying consumption patterns across households. While a small number of poor households lose out from replacing reduced VAT rates with targeted cash transfers, those that receive support are instead determined by income and family characteristics as opposed to consumption tastes – thereby increasing horizontal equity. Furthermore, many households are lifted out of poverty as revenue previously transferred to richer households is now transferred to poorer households. These results empirically confirm the theoretical expectation that, where available, direct mechanisms (whether via the income tax or benefit system) will better achieve distributional goals than reduced VAT rates. Countries that currently employ reduced VAT rates to achieve distributional goals should therefore consider removing these reduced rates and adjusting their income tax or benefit systems to achieve these distributional goals instead. Countries should also consider removing reduced VAT rates aimed at non-distributional goals where a more effective instrument is available to achieve the particular policy goal. At a minimum, the merits of these reduced VAT rates should be reassessed in light of their negative distributional impact.</p>

Introduction

This chapter serves as an introduction to the book Practical Microsimulation Modelling. It provides as context a description of microsimulation modelling, a simulation-based tool with a micro-unit of analysis that can be used for ex-ante analysis. The methodology is motivated as a mechanism of abstracting from reality to help us understand complexity better. It describes the main analytical objectives of users of microsimulation models in the field of income distribution analysis. The chapter then describes in turn the main methods of microsimulation considered in the book: hypothetical models, static models, behavioural models (labour supply and consumption), environmental models, decomposing inequality, dynamic microsimulation models, and spatial microsimulation models. The chapter concludes by providing an outline of the book.

Labour-Supply Behaviour

In the preceding chapters, the focus was on simulating policies that aim to reduce poverty, generate revenue, or redistribute resources. However, many public policies also try to incentivize behaviour, such as those to improve labour participation or supply, or to change behaviours in relation to savings or pollution. Social- and fiscal-policy instruments face a fundamental trade-off. An instrument that performs well from an income-maintenance perspective may have unintended behavioural consequences. This chapter considers the structure of instruments that have an explicit goal to improve behavioural response, particularly focusing on in-work benefits. The chapter also describes how to use a microsimulation mode to simulate the inputs required for the estimation of a behavioural-econometric model, and then estimates a revealed-preference-choice model. The chapter then describes a method often used in microsimulation models to calibrate choice models for simulation purposes. In terms of measurement issues related to the behavioural analysis, we describe the design and use of replacement rates. The chapter concludes by undertaking a simulation of the introduction of a change in in-work benefits.

Anti-Poverty Policy

This chapter discusses the development of a static microsimulation model for the purpose of undertaking an anti-poverty policy reform. Microsimulation models, which simulate the legislative detail of poverty-reduction instruments, can be used to make social-protection instruments more effective in this objective by helping to improve the targeting of these instruments. This chapter describes firstly the structure of the dataset required for microsimulation modelling. It then creates a theoretical understanding of the structure of social transfers, and of the concept of a hypothetical microsimulation model. Although the model developed in this chapter abstracts from the population complexity described in Chapter 1, it allows us in a simpler way to understand the targeting and structure of anti-poverty policies. Some of the issues that arise in creating a base dataset for a microsimulation model are discussed. As validation, debugging, and error checking are paramount in model development, the use of a hypothetical family model to use for validation purposes is introduced. We define some concepts used to calculate the poverty efficiency of a social-protection instrument. Finally, the chapter undertakes a simulation of the development of a means-tested benefit.

Spatial Inequality

There has been a growing emphasis on the spatial targeting of policy options in the areas of poverty and social exclusion. In this chapter, the focus will be on using spatial microsimulation models to look at the local impact of policies related to inequality and poverty. Spatial data typically exist in national census datasets, but very frequently these data do not contain information on incomes. The challenge, therefore, is to generate datasets that are spatially consistent, in order to facilitate the linkage of spatially defined data, such as local-area census data, with nationally representative surveys that contain labour, demographic, and income information. Spatial microsimulation modelling helps with this. The purpose of this chapter is to provide an insight into the rationale, development, and application of the spatial microsimulation method for analysing the spatial distribution of inequality. The policy context for spatial-inequality analysis is discussed initially, before considering the statistical method for synthetically generating spatially consistent, household-income-distribution data. Approaches to validating these methods are then discussed, before applying quantitative methods to measuring spatial inequality in a national setting.