My research interests all belong to what can be broadly defined as “applied welfare analysis”: assessment of the impact of tax reforms, analyzing changes in inequality and/or poverty (also at the global level), investigating the sensitivity of the assessment to the concepts used. Since the focus is on distributional issues, most research is based on the empirical analysis of microdata: household budget surveys, household income surveys, and administrative data. We have built and maintained several microsimulation models (indirect taxes, tax‐benefit), often combined with behavioural models of labour supply or full demand systems.
We applied the methodology of “marginal indirect tax reform” to data of the Belgian budget survey (with E. Schokkaert). In this context, and inspired by the work of A. Barten, I also estimated detailed demand systems for Belgium (both Rotterdam, AID and QUAID systems, under weakly separable preferences to disaggregate as far as possible). This research resulted in the construction of a microsimulation model for indirect taxes (ASTER) (started in september 1992, joint work with mainly G. Van Camp and F. Vermeulen). This is a user‐friendly program, that was disseminated in 1994. Since then, it has been used by academics, civil service and non‐profit organisations. One of its distinguishing features was the elaborate evaluation module, which allows to assess the impact of the reforms on different groups in the population by means of the most up‐to‐date analytical tools of distributional analysis. A second version (ASTER2.0) was released in May1995 and the program was thoroughly actualised in 2000 (ASTER 3.0) through the actualisation of the database and the improvement and fine‐tuning of the simulation routines.
In 2003 we started the construction of a much broader tax benefit microsimulation model for Belgium (MIMOSIS), covering personal income taxes, social transfers, and social security contributions. This was joint work with the university of Antwerp (Bea Cantillon of the Centre of Social Policy) and the university of Liège (Sergio Perelman) with whom we formed a research team of which we in Leuven take the coordination (joint work with Guy Van Camp). As a follow‐up to this research we obtained a grant from the Flemish Research Foundation to build a microsimulation for Flanders (one of the three Belgian regions). We extended our project team with demographers of the university of Brussel (Patrick Deboosere), and the EUROMOD‐team of ISER‐Essex under the supervision of Holly Sutherland. The latter made it possible to use EUROMOD as the core engine of our new microsimulation model for Flanders (christened MEFISTO). We also made the model available to a wide public and policy makers by means of a web interface (see www.flemosi.be). To incorporate behavioural reactions in microsimulation models, we estimated labour supply models on Belgian data (PhD’s of Orsini and Vanleenhove), and used this labour supply model to evaluate tax reform proposals, such as flat tax proposals. After the estimation of the standard multinomial logit‐model, we move to the estimation of a model developed in Statistics Norway, which incorporates some aspects of labour demand in the labour supply model. The microsimulation models played a crucial role in the comparative assessment of the electoral programs which we carried out during the federal elections of june 2014 (Rekening14). The estimated labour supply models were used to implement recently proposed new individual welfare metrics, in which preferences play a crucial role (work with P. Haan).
During a stay in UNU‐WIDER, Helsinki, I constructed, in collaboration with T. Shorrocks, a first version of a genuine microsimulation model for the Russian tax and benefit system. Also this model has been available on line for interested users: http://www.wider.unu.edu/darts_web/splash.php. The experience on the Russian microdata was exploited in a study of the distributional consequences of the transition to full cost coverage in the housing and utilities sector in Russia (work commissioned by the World Bank).
A lot of research effort has been devoted to link different sources of data by means of statistical matching techniques (work with Guy Van Camp and Kris De Swerdt). In order to simulate simultaneous reforms in personal income taxes and indirect taxes, one requires a single data set with expenditure and income figures. Such data sets are only rarely available. On the one hand there is a Household Budget Survey, with very detailed expenditure figures, and hence well suited to simulate indirect taxes. On the other hand, there is an administrative fiscal data set with detailed income figures from the annual tax forms, entered by taxpayers. This, evidently, is the obvious candidate to simulate personal income taxes. These data sets cannot be linked in an exact way since there is no exact overlapping information such as a unique identification number. But, both data sets contain overlapping variables that we have exploited in statistical matching techniques to construct an integrated data set for Belgium with both expenditure figures and gross income figures for 1998. The datasets available have been used to assess the distributional impact of a shift of personal income taxes to indirect taxes, and to measure the contribution to the progressivity in the Belgian personal income taxes, of different components (deductions, allowances, tax credits, rate structure). In 2001 we repeated the analysis by matching the budget survey of 2000 with income data in the fiscal forms of 2001. We improved the methodology by using semi parametric Engel curves to impute expenditures (work with Kris De Swerdt). Under the EC‐ funded project AIM‐AP (directed by the EUROMOD‐team), we also applied the methodology for an extension of EUROMOD to include indirect taxes, and continue to work on this EUROMOD extension with funding of the Joint Research Centre of the European Commission in Sevilla.
In 2003 we made a descriptive analysis of the Belgian housing market on the basis of the 1997‐98 Belgian budget survey (work with B. Capéau). A formula was designed to calculate the user cost of owner‐ occupied houses (about 70% of Belgian households are owners), which embraces the main aspects of the Belgian tax treatment of housing. This formula was used to calculate the impact‐effect of a change in the transaction tax tariffs on houses bought, keeping the housing decisions (ownership, timing and size of the investment) constant. This methodology was used to make a first welfare analysis of the recent reduction of registration fees in Flanders. We continued this work in 2004 (work with Bart Capéau and Frederic Vermeulen) by estimating an ordered logit model to explain the timing in the life cycle of the first move to homeownership. The model has been used to simulate the impact of the lowering of the registration fees in Flanders. In 2005 we investigated how to improve upon the existing practice of house price indices by estimating a hedonic price index for housing sales sold on the secondary market in Belgium (work with Kris De Swerdt).
My research includes an investigation of the sensitivity of distributional measurement to different choices of the income unit: household, person, or equivalent persons, application of different methodologies to take into account differences in needs (work with E. Ooghe), an assessment of changes in the world income distribution, and the different views about it (work with Kristof Bosmans, Bart Capéau, Koen Decancq and Erik Schokkaert), and an analysis of the distributional consequences of some flat tax proposals for Belgium.
I actively participate in building a bridge between theoretical insights from the literature on Fiscal Federalism literature and the practice of constructing a new ‘Special Finance Act’ in Belgium. This Act determines the extent and form of the fiscal autonomy of the Belgian regions and communities, the grants which filled the vertical fiscal gap, and the fiscal equalization system. In 2017 we updated estimates of interregional financial flows in a study for the Flemish government (with W. Sas).