Mosi: An Account System For The Dutch Social Security

The Dutch social security system is quite elaborate. From a budget point of view, it is divided in two parts. The larger part is financed by insurance premiums, the remainder by taxes. The insurance part takes care for allowances for people with little or no income: the elder, the disabled, the unemployed; moreover, a large part of the health care system is financed in this way. The social security system is organised in about 25 different funds. Our aim was to develop a new system to predict incomes and expenditures per fund, taking into account all legal interdependencies between funds. The system should give an accurate estimation of the total amount of income tax, gross allowances and other variables of interest. Moreover, effects of policy changes should easily become clear.Broadly, the social security funds can be divided into three groups. The first one is aimed at all inhabitants, the second one only for employees and the last one only or specific groups. The financing is rather complicated. Partly the system is paid by employers, partly by individuals, the rest by taxes. Every insurance has a specific premium percentage, which is applied to the individual income below a fixed maximum. Premiums paid for an employee's insurance are deductible from gross income before general insurances and taxes are levied.Given an estimation of allowances, premium-percentages have to be computed for the different funds. This is complicated, as premiums paid for one insurance scheme can be deductible for another one. Furthermore, there are mutual financial streams between funds. To sum up, the complete system is complicated and changes continuouslyTo compute premium-percentages and other key-variables we built a macro-economic model of approximately 2000 equations. These equations describe the income and expenditure part of all relevant funds, the relationships between funds, the complicated way in which allowances are linked to the legal minimum wages and the amount of income liable to taxes and specific premiums. This model is basically in the well-known economic model form: each endogenous variable is written as a function of endogenous or exogenous variables, not necessary linear. Due to many legal and administrative changes the macro-models differ from year to year. Because of the many interdependencies it is simultaneous. Of course, such a model is easily solved by standard techniques.For this model the total sum of individual incomes liable to premium payment is given. However, as stated before, the problem is, that the premiums influence individual incomes, while the premium proceeds in turn depend on individual incomes. The way in which gross to net incomes are computed is contained in a so called micro-model, consisting of some 500 equations. To find the total of income liable to premiums and taxes, this micro-model is fed with a adapted sample of about 13500 individual incomes. This means our system is double simultaneous. Firstly the macro-economic model is simultaneous, secondly changes in premium-percentages influence incomes and vice versa.Finally, the last complication is that we have to generate future income distributions to make predictions. This is accomplished by a technique called calibration. The income distribution is corrected to obey restrictions like the total amount of income, the total number of tax-payers, the total number of employees or other restrictions, which can be freely formulated in the macro-model and corresponding datafile.Additional requirements for the new system were inter alia : - the combination of a macro-economic part - both model and data - and a micro-part - both database and computational rules - to describe the interactions between policy measures and the effects on individual incomes - accessibility: all models and data files should be easily accessible by nowadays common tools like editors, word-processors and spreadsheets; - model-flexibility: different models of every size and contents should run without changes in software; even only a macro-model without the micro-model should run; for every year within a simulation period a different model should be permitted ; - pin-pointing endogenous variables on specific values and automatic generation of model-consistent residuals; - system-flexibility: all kinds of system parameters, like convergence criterion, trace-parameters, etc. should be changed easily; - platform independent: the system should run on at least a UNIX-machine and in a Windows environment; - portability: a strict distinction between software on the one hand and model and datafiles on the other.The software is written in a widely available fourth generation language, without any need for compilations. The same program runs both under UNIX and under Windows. In effect we wrote our own compiler/ interpretor to analyse the job to be done: what kinds of actions have are requested, which models and datafiles have to be used, what kind and form of output is required and so on.The system is nowadays taken into production and will play an important role in the development and implementation of policies in the field of social security and income distribution. The project was done by the CPB, the National Bureau for Economic Policy, with support of the Ministry of Social Affairs and Employment.