From Data to Policy Analysis: Tax-Benefit Modelling Using SILC 2008
Callan, Tim, Keane, Claire, Walsh, John R., Lane, Marguerita, Journal of the Statistical and Social Inquiry Society of Ireland
Policy makers and the body politic have a strong interest in ensuring that the tax transfer system functions well. This common interest in policies that are efficient in achieving their objectives including economic efficiency and fairness--is heightened by the scarcity of resources during the current crisis. The effects of tax and welfare policy changes are wide-ranging and complex, varying with the characteristics of individuals and their family and household situation. Changes in income tax and social welfare can alter the distribution of income and the incidence of poverty (whether measured in terms of income alone ("risk of poverty") or in terms of income and material deprivation (consistent poverty)). Tax and welfare changes can also have significant impacts on financial incentives to work, potentially affecting decisions about labour market participation and hours of work.
Tax-benefit models have been widely used as tools for the analysis of potential policy reforms--(for a recent review see Bourguignon and Spadaro, 2009). Microsimulation modelling has been widely used in the UK and the US for many years in order to explore policy choices and the impact of potential or actual policy changes. In Ireland, a tax-benefit model was developed at the ESRI to undertake a similar role. The SWITCH model (Simulating Welfare and Income Tax CHanges) was initially based on the ESRI's 1987 Survey of Income Distribution, Poverty and Usage of State Services, and later was re-based to use data from the Living in Ireland Surveys. In more recent years, the model has been redeveloped to use data from the CSO's Survey on Income and Living Conditions (SILC)--first based on data for the year 2005, and now based on the most recent year currently available, 2008.
This paper looks more closely at the technical issues which arise in using SILC data as the basis for the SWITCH tax-benefit model. Some specific applications are used to illustrate particular points, but the main focus of the paper is a more technical one. However, Section 2 begins by discussing the benefits and capabilities of a tax benefit model and the contribution it makes to policy evaluation. Section 3 discusses the broad data requirements of a tax benefit model. Section 4 documents the procedures and issues involved in creating a database for the model based on data from the CSO's Survey on Income and Living Conditions (SILC). Section 5 considers issues regarding the degree to which the database represents the income tax base and social welfare client population. The potential role of weights designed specifically to address this issue is considered. Adjustment of the survey database to represent the next budgetary year is also discussed. Some key issues are then drawn together in the concluding section.
2. MODELLING TAX AND WELFARE POLICY OPTIONS
Very often policy changes are considered in terms of their effects on a number of "hypothetical families". This approach has severe limitations. For example, less than one family in 20 falls into the category of "one-earner couple with 2 children" which attracts so much attention at budget time. Furthermore families within this category differ in terms of income, housing tenure, and other characteristics that affect their tax-benefit position. More fundamentally, analysis of hypothetical families--no matter how well chosen--simply cannot give an overall picture of the impact of a policy change on incomes and work incentives.
Up until 2007 SWITCH was based on data from the Living in Ireland (LII) Surveys. In 2007 the model was rebased using data from the 2005 wave of the CSO's Survey on Income and Living Conditions (SILC). One advantage of the SILC data over that of the LII is the size of the sample interviewed, with the SILC sample size being about 50% greater than that for the earlier surveys. By the final year of the LII Survey (2001) the sample size stood at 2,865 households with 6,518 individuals. …