Incorporating the value of government services and cost of taxes into household incomes raises a range of methodological and conceptual questions. Household surveys generally do not contain information on taxes or benefits or, at least, not with the required level of disaggregation, and little consensus exists on the best way of valuing these services and distributing the result across individuals, matters which can importantly affect the results.

The use of incidence analysis techniques is widely exemplified by Euromod (2009) and the OECD (2008a). The work carried on by ECLAC (2007) and the World Bank (Breceda et al., 2008; and Goñi et al., 2008) are regional examples of this technique. Finally, national studies such as the Chilean Planning Ministry (Mideplan, 2007) and the Mexican Ministry of Public Finance and Credit (2008) use this approach to evaluate the outcomes of policies captured by household surveys.

The methodology we have adopted is similar to these examples. The main data sources and methods are described below.

Data sources

Tax-benefit incidence analysis relies on diverse sources of information and uses imputation techniques to splice them together. In order to estimate the impact of taxes and benefits the following information was used:

  • Household surveys: Individual records from the 2006 National Characterisation Socio-economic Survey (CASEN) for Chile and the 2006 Household Income Survey (ENIGH) for Mexico. Both surveys provide data on income of households as well as information on their economic characteristics that can be used to impute public services and taxes to individuals. In Chile, estimates of the effects of value-added taxes and excise duties drew also on the 2006-07 Family Budget Survey (EPF).
  • Government statements and institutional records: The analysis covers health and education services, using data on public expenditures at institutional level from the Chilean National Budget Office (DIPRES) and the Mexican Ministry of Public Finance and Credit (SHCP). In addition, the distributive impact of health in Chile relies also on the Satellite Account for Health.
  • Tax records: Statistics drawn from personal income-tax returns provide another source of information about the tax base. In the case of Chile, specially commissioned data was obtained from the SII, analysing the number of taxpayers, their assessed income, its composition and the taxes paid by income bracket.

In terms of coverage, the analysis covers 72% and 66% of total social expenditures for Chile and Mexico, respectively; while on the other side it includes 69% and 71% of total tax revenues.

Determination of tax burdens and benefits

The boundaries of what items can be imputed to households are not always obvious. Certainly items such as health care and education are good candidates. However, any public expenditure or tax is in theory a candidate, having at least some direct or indirect impact on households’ consumption possibilities. For the purposes of this analysis, the approach must be a pragmatic one, with the inclusion of questions on specific programmes in household surveys driving the extent to which we can include such items in the analysis. Though in practice the impact is typically at the level of the individual, we treat it as evenly distributed across household members.

  • Cash transfers: Since they are generally targeted at people in the lower income strata, in developing countries these programmes are usually among the most visible types of social spending. Household surveys treat them directly, and our calculations take the value that families surveyed declared as received.
  • In-kind transfers: Following OECD (2008a), the incidence of education is obtained applying the actual-use approach (beneficiaries are those students using the educational services) and for health care the insurance-value approach (imputing the insurance value of coverage to each person based on specific characteristics, such as age and sex). Because of the lack of market prices, the value of the transfer is assumed equal to its production cost. Even when this approach neglects differences across countries in terms of quality and efficiency in the provision of the service and in the value individuals assign to these services, similar assumptions are a regular feature in the specialised literature (including OECD, 2008a; and Euromod, 2009).
  • Direct taxes: Personal income taxes are estimated for each individual according to their reported income in the household survey, the tax law in force in the survey year and information on effective income tax collection. Some income reported in household surveys is collected on an after-tax basis. Therefore, a first step was calculating the incidence of taxes paid in 2006 to construct pre-tax estimates for these items. “Income taxes” in Chile include the second category (tax on income from dependent employment) and the withholding income tax, and in Mexico they are the taxes on personal labour income, income derived from interest, rents and self-employment activities. Statutory tax rates are then applied in order to obtain the income tax that individuals should pay. These figures are then compared with the effective tax collection. In the case of Chile, tax-return information was available and the amount of income tax that individuals chose to pay was computed as follows. The number of non-filers in each decile was estimated as the difference between the number of individuals in the household survey with incomes high enough to be subject to the income tax, and those who actually filed a tax return, and then imputing these randomly within the survey. Then, for the tax filers the proportion of income tax due that individuals actually paid was estimated from the tax-return information and then distributed in the survey proportionately to the estimations of income tax due.
  • Indirect taxes: The total tax take for indirect taxes is estimated from the effects that both value-added taxes and excise duties have on the price of final goods. Following Euromod (2009), the total tax liability Tifor commodity i is calculated on the basis of observed expenditures

The effect of each tax is then constructed by applying the statutory tax rates and deductions in force for each type of product in the survey and then aggregating these into 17 categories of goods and services. Then, the proportion of indirect taxes that households actually pay is adjusted to the effective tax collection on these items that is transferred to private consumption and then distributed in the survey proportionately to the total tax liability. The amount of indirect taxes that is transferred to private consumption is estimated from the Tax Matrix information in National Accounts.

In the case of Chile, a matching procedure was used to impute household expenditure from the input data (EPF) into the survey on the basis of budget shares for different population groups identified by disposable income and the largest set of demographic variables – age, gender, educational level, professional status, and number of adults and children – common to both datasets.

  • Health-care social-security contributions: In Mexico contributions include those made in respect of the sickness and maternity insurance within the compulsory scheme (seguro de enfermedades y maternidad del régimen obligatorio). In Chile contributions were calculated according to the scale applicable to the different FONASA health groups. These groups are defined by household characteristics such as income level and number of beneficiaries.

Measurement errors and under-reporting

Household and expenditure surveys are an important source of information on the allocation of tax benefits within households. Nevertheless, systematic misreporting of some income sources, such as capital income, income from self employment or income from social transfers, can provide a misleading view of the income distribution and redistribution profiles.

Reconciling household-survey data and national-accounts data is a well-known problem. Macro aggregates from household survey data normally present discrepancies with published national accounts, even though the sample weights are designed to represent the national population. Table 4.A1 illustrates the extent of such discrepancies in recent household budget surveys in Chile and Mexico.

Table 4.A1. Comparison of national accounts and household survey estimates

Table 4.A1. Comparison of national accounts and household survey estimates

The differences between the surveys and estimates from national accounts highlight potential biases in the totals. In particular, household surveys tend to under-report household incomes. A common approach in the literature has been to adjust aggregate reported household incomes so as to match the corresponding items in national accounts, though no agreement exists on the best way to do this – even assuming that national-accounts aggregates are correct. Assumptions are needed, for example, in order to assign under-reported income across the population and such assumptions can be material to the results, particularly when discrepancies are high. Allocation of income from capital is a good example, since such income in practice tends to be found only among upper-income households.

Following OECD (2008a), we have made no adjustments to household-survey income aggregates and all calculations were based on data gathered directly from published records. In the case of Chile, official data are already imputed using estimates from the national accounts (more details about this procedure can be found in Mideplan, 2006); while for Mexico income is not adjusted in the survey. For the interested reader, this effect is examined in Mexican Ministry of Public Finance and Credit (2008).