Multiple households

Rising inequality, deprivation, and poverty are direct concern of the economic policy. In addition, policy goals are influenced by indirect effects of differential development of the individual social groups. Since the savings rate and consumption pattern of the strata are quite different, demand for the products and hence employment of the individual industries depend very much on the within-household sector income distribution. Similarly, shifts within the household sector influence other macroeconomic aggregates too, like import or investment.

Representing multiple households and their relationships with the labour market and income distribution is a common practice in CGE models. Several applied modelling experiences can be found in the literature, in particular within the stream of CGE models used for policy analysis in the World Bank (see, for example, Harrison et al [2002], Kalb [2000], Shoven, J. B. and J. Whalley [1998]). Among the project participants, P. Capros, D. Van Regemorter and Zalai and Révész have also conducted, separately and collectively, several successful CGE modelling exercises with multiple households in the past.

However, including more than one representative household into a CGE-model arises various conceptual and technical problems.

In the models of multiple households each class of them receive labour and capital income from each sector according to specific distribution schemes. Household classes also receive government transfers, pay income taxes. Their savings can be represented either as a fixed proportion of their after-tax income, or determined together with their consumption of goods and services based on the usual assumption of utility maximization subject to their budget constraints, each of the household classes having its own preferences.

The construction of the household accounts usually proceeds as follows. First a wage and salary distribution matrix is created by combining industry-occupation data with average salaries, and then mapping the earnings by individuals according to their occupations to households. Capital-related income by industry is aggregated into an economy-wide enterprise account and then mapped to household classes according to data derived from individual income tax returns and/or household budget surveys (HBS)¹¹. The statistical data on individual income tax returns forms the basis also for calibrating household related income taxes and government transfers. Personal consumption is disaggregated according to various household classes using data from the Consumer Expenditure Survey. Household savings can be determined as the residual income (income-expenditure) or its estimate can be based on the data of specific sociological surveys on savings behaviour and its amount.¹²

The consumption behaviour of each household class can be modelled by using the same type of expenditure system as in the standard single household GEM-E3 models, but their coefficients will of course differ from one household class to another, reflecting differences in tastes and habits. As a matter of fact, empirical research shows that preferences differ across

11 In some countries this aggregation method can be rather problematic when certain household types (e.g. farmers) capital endowments concentrate on specific sectors (agriculture)

12 The training document MultHh-Hu05.doc provides an example how HBS data can be used to disgregate the household related data of the GEM-E3 model.

household classes according to differences in income levels, occupational, educational positions and urban/rural residence of the classes. The time series of the EU Consumer Expenditure Surveys provide adequate information basis for the disaggregated representation of consumer expenditure systems in the model.

The choice of the household classes to represent in the model is a practical issue that depends first of all on available statistical data. The choice can be based on the cross section between three sets of statistical information: the consumer expenditure surveys, the industry-occupation data and the income tax return statistics. Pure income classes are not desirable, because empirical research shows that income is not the only, not even the dominant factor in explaining the differences in the consumer expenditure systems of various household groups. A more common factor that explains these differences is the occupational-educational position of the family head. On that basis, cross mappings with labour skills and tax income categories is also easier to define. The occupational-educational distribution of households is strongly correlated with income distribution. Unfortunately, in some countries - because of the poor availability of statistical data - it is not possible to differentiate classes according to urban-rural criteria.

An advantage of using the occupational-educational dimension for definition of various household classes is that it makes it possible to represent societal and demographic evolutions in the long term scenarios prepared for the model. As a matter of fact, a multitude of factors that are not necessarily resulting from changes in the economic indicators simulated with the model may explain the evolution in the future of the number of households per class as defined according to the occupational-educational dimension. There have been attempts in the literature to link such an evolution with the projections of economic growth, income and the labour market, however, there is poor evidence about their direct causal (functional) relationships that would allow the modellers to include such a mechanism in the model.

Therefore, it is advisable that in the forecast of the number of households per class be exogenous in the model, and be quantified as part of the construction of the exogenous assumptions of scenarios. Evidently, to build such a scenario, one should implicitly relate the evolution of classes with other exogenous assumptions, such as population growth and regional welfare convergence trends. Of course, one may vary these assumptions across scenarios.

While the number of households per class is exogenously determined in the model, all other changes related to their behaviour shall be endogenously determined. For example, linkages between consumer choices per class, income per class in relation to the labour markets, tax-income policy and industry-occupational levels per class will all be endogenous variables in the model.

Expenditures and labour supply in each class of households should be modelled separately, using the type of nested linear expenditure system typically applied in the GEM-E3 models.

Labour per class and capital can be assumed to be perfectly mobile across sectors, but imperfectly mobile across countries. Within each country the labour market must clear according to an imperfect competition mechanism (see for example the labour market clearing in the Worldscan model which is inspired from the research work of G. Pissarides), through a

nested scheme involving clearing for labour demand, supply and salaries at each class level and globally at each country's economy.

4.3. Illustrative programs

The special programs accompanying this training material contain illustrations for household disaggregation possibilities and socio-economic group-formation criteria. The MultHhMod-MENG.doc file describes the MULTHH.GMS program, a CGE model calibrated for Hungarian data for 1998 distinguishing 3 sectors and 10 household groups. The accumulation capacity of the various household groups differ much are more than their consumption levels and patterns, therefore only dynamic models can illustrate properly the different impacts of economic policy measures on the prospects of the individual social groups.

To render this possible we introduced into our model group specific human capital (which also can be accumulated by the “productive” part of the personal consumption), capital incomes (imputed rent income and net interest) and group specific financial wealth. Consumption of foreign tourists is also treated separately in our model, since it does not belong to any resident household groups. To make the model more flexible and realistic we also introduced a CET-labour supply function, and alternative closures rules.