EUROPEAN COMMISSION EUROSTAT Directorate F: Social Statistics and Information Society Unit F-3: Living conditions and social protection statistics Doc Net/2009/02 ESSPROS TASK FORCE MEETING ON NET BENEFITS 26 TH November2009 At 9.30 Eurostat-Luxembourg, BECH Building, room B2/464 EU SYNTHESIS REPORT ESSPROS MODULE ON NET SOCIAL PROTECTION BENEFITS (RESTRICTED APPROACH)
ESSPROS MODULE ON NET SOCIAL PROTECTION BENEFITS (RESTRICTED APPROACH) EU SYNTHESIS REPORT INTRODUCTION The pilot data collection of ESSPROS net social protection benefits was established through regulation (EC) 458/2007 of the European Parliament and Council and represents the first time that Eurostat has systematically collected data on net benefits from all countries on the basis of agreed methodological guidelines 1 (following the so-called restricted approach 2 ). The regulation specifies that the future continuation of the data collection as a routine exercise depends on the outcome of the pilot being positive in the majority of countries, which requires that the exercise is feasible for most countries and that the data satisfy quality standards. All twenty nine participating countries submitted a completed questionnaire for the pilot data collection. These questionnaires and the accompanying documentation were subjected to a series of validation steps to assess compliance with the methodology and to identify possible errors and improvements that could be made. These steps included a qualitative assessment of benefits shown to be liable to taxation and comparison with other sources of information (primarily MISSOC), a review of the methods used to derive tax rates and the underlying sources of data, and a quantitative assessment of the tax and social contribution rates and fiscal benefits reported, including a comparison with data provided to the OECD-SOCX database 3. The resulting validation reports were then circulated to national delegates in order to have an exchange about issues arising and to improve the quality of data where possible. This document gives an overview of the data sources and methods used to calculate tax rates and the coverage of the fiscal system in order to assess the overall quality of the data collected and the feasibility of continuing with the data collection as a routine exercise. A first analysis of data collected i.e. the actual value of net benefits, effective tax rates, proportion of benefits liable to taxation, etc. is presented in a draft Statistics in Focus: Net expenditure on social protection benefits, 2005. Results of a pilot data collection. Overall, data constraints in many countries prevent a 100% complete and precise coverage of the effects of the fiscal system on social protection benefits. However, in the majority of cases it can be assumed that these gaps will not result in a significant difference between the recorded value of net social protection benefits and the actual value of net social protection benefits. Some aspects of the methods used by some countries to provide AITRs and AISCRs reduce the precision of their estimates but in general these can be considered to result in only minor inaccuracies in the final estimates of net expenditure. The net data provides a much fairer platform for comparing social protection benefits between countries than is possible using the gross data. 1 Doc. NET/2008/05 ESSPROS Net Social protection benefits: Coverage and definitions 2 The restricted approach means that the scope of the data collected should not change from that covered by the ESSPROS Core system it attempts to measure the net value of social protection benefits reported to the Core system. In other words it covers taxes and social contributions paid on gross benefits, and reliefs on those levies, but not reliefs that apply to taxes paid on other sources of income. See Doc SP-NET/2009/03/EN Review of fiscal benefits. 3 A detailed description of the validation steps is given in Doc SP-NET/2009/02/EN Summary of validation and quality assessment. DOC 2 _ EU Synthesis report.doc\alphametrics Ltd.\20 November 2009 2/8
METHODS AND DATA USED The methods employed by each country to evaluate the effects of the fiscal system on social protection benefits are directly linked to the types of data that are available and are practical to use. Inevitably, different types of data allow different levels of detail to be exploited and may, therefore, result in different levels of precision in the estimates of net social protection benefits. This section considers different aspects of the data and methods used and the extent to which these impact on final results. In most cases, the analysis of data sources and methods used by countries to compile data for the ESSPROS net social benefits pilot data collection relies on accompanying qualitative information which was delivered with the questionnaires or in response to queries raised in validation. Only one country failed to provide any accompanying documentation (Ireland), though the methods used can to some extent be deduced from the quantitative data. In a few further cases the level of detail in the accompanying information was limited so that some issues may remain hidden. However, the majority provided enough detail give a clear picture of how levy rates were estimated and the issues that they were faced with. This allows an overview of the methods used across the data collection to be constructed. When evaluating the effect of different methods it is important to bear in mind the proportion of benefits to which it has been applied. For instance, in some countries the amount of gross benefit that is subject to levies is less than 10% of total (Table 1). Therefore issues that arise in the calculation of AITRs and AISCRs as a result of a specific method may not have a significant effect on the data as a whole. Concerns arise only when a method which can cause inaccuracies is applied to significant proportion of the AITRs or AISCRs and a significant proportion of the gross data is subject to these levies. Table 1: Share of social benefits subject to levies Taxes on income (AITRs) Count Social contributions (AISCRs) Count Share of TOTAL benefits subject to levies: 0% BG, SK 2 BG, SK, CZ, LT, MT, RO, SE, SI 8 <10% SI, HU, LT 3 EE, IT, HU, LV, PT 5 10-50% CZ, IE, CY, RO, UK, IS 6 BE, DE, DK, IE, ES, CY, UK, IS 8 >50% BE, DE, DK, EE, EL, ES, FR, IT, LV, LU, MT, NL, PL, PT, FI, SE, AT, NO 18 EL, FR, LU, NL, PL, FI 6 29 27 Share of CASH benefits subject to levies: 0% BG, SK 2 BG, SK, CZ, LT, MT, RO, SE, SI 8 <10% LT 1 EE, IT, LV, PT 4 10-50% CY, SI, HU 3 IE, ES, CY, UK, IS, HU 6 >50% BE, CZ, DE, DK, EE, IE, EL, ES, FR, IT, LV, LU, MT, NL, PL, PT, RO, FI, SE, UK, IS, AT, NO 23 BE, DE, DK, FR, LU, NL, PL, FI, EL 9 29 27 Identification / levels of breakdown The net benefits questionnaire allows for AITRs and AISCRs to be applied separately by detailed type of benefit (case, lump-sum, means-tested, non-means-tested) and by scheme. However, the available source data often do not facilitate the breakdown of data to the this level of detail and the resulting AITRs and AISCRs will reflect this. DOC 2 _ EU Synthesis report.doc\alphametrics Ltd.\20 November 2009 3/8
Distinction between types of levies (taxes and social contributions) Of the twenty-one countries that have social protection benefits that are liable to both taxes and social contributions, nineteen were able to provide separate estimates for AITRs and AISCRs (Table 1). The remaining two countries provided only a combined tax and social contribution rate because the underlying data did not allow the relative proportions of the two types of levy to be distinguished. For example, this includes the use of data reporting on the value of net benefits which is already net of both income taxes and social contributions, or tax return data which does not separate social contributions from taxes on income. Being unable to distinguish between taxes and social contributions does not present an issue for the final estimates of net social benefits but does prevent separate analysis of the impact of taxes on income and of social contributions for the affected countries. Table 2: Distinction between taxation and social contributions Liability of social benefits: Countries Total Liable to taxes on income and social BE, EE, CY, DK, DE, LV, PL, IS, IT, FR, NL, LU, ES, IE, UK, 21 contributions HU, PT, AT, FI, NO, EL Separate rates for taxes and social BE, EE, CY, DK, DE, LV, PL, IS, IT, FR, NL, LU, ES, IE, UK, 19 contributions HU, PT, FI, EL Combined tax and social contribution AT, NO 2 rate Liable to taxes on income only CZ, LT, MT, RO, SI, SE 6 Liable to social contributions only (None) 0 Not liable to any levies BG, SK 2 Level of detail in the breakdown of levy rates; grouping of benefits In several cases limitations in the available source data prevent tax and/or social contribution rates being calculated for each different (liable) benefit. In this case a single rate is calculated for a group of benefits that cannot be separately identified and the same rate is then applied to each of them in the questionnaire. In theory this should not cause a significant problem to the final results even if it means that rates are incorrect at the detailed level. Take, for example, the case where data do not allow two different unemployment benefits to be distinguished so that a single tax rate is determined for all recipients of these benefits. Both are potentially liable to taxes on income but, in practice, one is means-tested and paid only to persons on a very low income who are unlikely to pay any taxes on their income so that the real AITR for that benefit ought to be zero or close to it. The second is a wage-related benefit and many of the recipients will have some income from employment during the year and be liable to pay taxes, part of which will then be attributed to the amount of unemployment benefit received. Calculating a single tax rate on the total amount of benefits paid out will result in a rate that is correct for the total of unemployment benefits but too high for the means-tested benefit and too low for the wage-related benefit. Since there is no intention at the current time to publish data at the detailed level it does not particularly matter that estimates of net benefits will be incorrect for the individual benefits. It may, however, have an impact on other levels of aggregation. In this case, the net benefits and effective tax rates at the function level will not be affected but the values for means tested and non-means tested benefits at the function level will be impacted and so will the same aggregates across all types of benefit. The issue becomes still more problematic when benefits from different functions are grouped together. In this case, the incorrect values at detailed level will impact on results at function level. The extent to which the function-level results are affected depends on the proportion of liable benefits within a function that are affected by this aggregation, their share of the total size of the aggregate (both of which can be seen from the gross data) and the difference in the aggregate rate applied and the real rate for each benefit in the affected functions (which cannot be determined). Nevertheless, it is thought that in the majority of cases the impact of calculating levy rates for groups of benefits which span DOC 2 _ EU Synthesis report.doc\alphametrics Ltd.\20 November 2009 4/8
functions will be limited. Moreover, the number of cases where data are calculated by scheme or by groups of schemes which include benefits from multiple functions are limited because benefits tend to be grouped by a particular purpose and therefore mostly apply to just one function and only occasionally to two or maybe three functions. Table 3 shows the extent to which social benefits have been grouped together for the calculation of levy rates. In this table (and the tables that follow in this document) the column scope refers to the proportion of liable benefits that is affected. The category of Large then refers to cases where most liable benefits are affected, Medium refers to cases where a significant part of liable benefits was affected and cases where an insignificant proportion of liable benefits was affected are categorised as Small. Ten of the twenty-seven countries that have benefits liable to income taxes use some grouping of benefits that cross functions, though in four of these cases the grouping affects only small proportions of benefits and in one of the large cases less than 10% of all benefits are taxable so the scope for error is small. Twelve countries use grouping of benefits within a function for at least part of their tax rate data whilst seventeen use detailed data. Note that the categories are not mutually exclusive because a different approach may be applied to different sets of benefits depending on the available data and, to some extent, in the added value of making a more detailed assessment. The potential problem of cross-function groupings applies to a lesser extent to data on social contribution rates with only six countries providing AISCRs on this basis for some of their data. Table 3: Grouping of rates for AITR and AISCR calculation Rates calculated at detailed level Rates calculated for groups of benefits across functions Rates calculated for groups of benefits within functions Scope AITRs Total AISCRs Total Large DK, AT, CZ, EE, HU*, IE, LV, MT, NO, PL, RO, SE, IS 13 DK, EE*, HU*, IE, IT*, LV*, PL, PT*, IS Medium DE, LU 2 CY, DE, LU, NL, FI 5 Small FI, IT 2 0 Large EL, DE, FI, LT*, BE, 6 DE, FI 2 NL Medium 0 BE 1 Small MT, IT, LV, SI 4 ES, LU, CY 3 Large CY, UK, ES, FR, IT, 7 CY, EL, UK, BE, ES, FR, 7 PT, SI*, NL Medium LU 1 LU 1 Small DK, IE, SE, BE 4 DK, IE 2 9 IE: categorisation is based on the most likely scenario given the data provided Types of grouping are not mutually exclusive because some countries use different methods for different sets of benefits * Portion of TOTAL benefits subject to levies: <10% As already alluded to above, the grouping of benefits can also be a potential source of inaccuracies in the final results when the group combines benefits of different types, particularly when benefits that are means-tested are combined with non-means-tested benefits. Often, means-tested benefits are not taxed and even when they are potentially liable to taxation, the effective tax rates are low because the benefits, by definition, are paid to those with low incomes so that few taxes are paid. In fact, just over half of the countries which apply any form of levy to social protection benefits (14 out of 27) do not apply any levies to means-tested benefits. Table 4 shows the number of cases where levy rates were calculated for groups of benefits that include both means-tested and non-means-tested benefits. Tax rates (AITRs) may be affected in four countries out of a potential twenty-seven that apply taxes to social benefits whilst three of a potential nineteen are affected for social contribution rates (AISCRs). Again, in terms of total net social protection benefits this issue should not cause any error and the problem only arises in terms of analysis by type of benefit (means-tested/non-means-tested). Grouping of benefits is liable to over estimate the taxation of means-tested benefits and underestimate the taxation of non-means-tested benefits. DOC 2 _ EU Synthesis report.doc\alphametrics Ltd.\20 November 2009 5/8
Table 4: Distinction between means tested and non means-tested Scope AITRs Total AISCRs Total Means-tested and non meanstested Large AT, EL, UK, FI 4 EL, UK, FI 3 benefits not Medium 0 0 distinguished Small 0 0 Finally, in terms of incomplete breakdowns, there are some cases where data do not distinguish between benefits which are liable to taxes and those which are not. In other words, a tax or social contribution rate has been derived for the total value of a group of benefits including some that are not liable to any levies and the same rate has been applied to each of the benefits in the group. Again, the final value of net benefits is unaffected but analysis of the proportion (by value) of benefits liable or not liable to levies will be inaccurate. Probably in future this problem can be overcome by adjusting the overall rate according to the share of liable/non-liable and applying this higher rate only to the liable benefits and zero to the non-liable benefits. Only one country was seriously affected by this when calculating AITRs and AISCRs (Table 5). Note that this point does not concern cases where data is unavailable to calculate the taxation on specific benefits (see next section). Table 5: Distinction between taxable and non-taxable benefits Taxable and Non-taxable benefits are undistinguished Scope AITRs Total AISCRs Total Large EL 1 EL 1 Medium DE 1 DE 1 Small UK 1 UK 1 Coverage In terms of coverage the net benefits data collection should first collect tax and social contribution rates for all social benefits that are liable to these levies and, secondly, collect tax rates that fully reflect the full impact of the fiscal system. The data collected should refer to 2005. Coverage of liable benefits In a number of countries the rates of taxes and/or social contributions applicable to certain benefits could not be estimated due to the lack of a suitable data source (Table 6). Only one country (Greece) faced this constraint to such an extent that the value of net social protection benefits currently calculated is potentially overestimated to any significant extent (Table 6). This issue affected tax rates only. A few countries reported that tax and/or social contribution rates were incomplete for a small proportion of benefits but in none of these cases are the omissions likely to have any significant impact on final estimates of net expenditure. Table 6: Availability of necessary sources of data Scope AITRs Total AISCRs Total Insufficient data to calculate Significant EL 1 0 certain rates Insignificant BE, ES, NO 3 BE, ES 2 Coverage of fiscal system A small number of countries reported that difficulties in providing data that take full account of the fiscal system. Two issues in particular arose. Firstly, four countries could not take into account in the AITRs any post year-end tax adjustments such as rebates in case of overpayment of taxes paid in advance (e.g. through PAYE). A further five countries (out of a potential twenty-seven that apply some taxation to benefits) were unable to include in the AITRs the impact of other non-universal tax system adjustments, including fiscal benefits that would be in the scope of the enlarged approach to measuring net benefits. Of this total of nine DOC 2 _ EU Synthesis report.doc\alphametrics Ltd.\20 November 2009 6/8
countries that reported incomplete coverage of the fiscal system in AITRs, six included adjustments as in the fiscal benefit section in order to correct the final estimates of net expenditure. Some of these were later removed because the value of fiscal benefits reported covered the whole of the relief and not only the part related to relief on taxes paid on benefits. In other words the amounts reported relate to the enlarged approach and not the restricted approach. Table 7: Inclusion of tax system adjustments Countries Total Post tax year adjustments not included in AT, LT, LV, PT 4 rates Other tax system adjustments not included in EE, FR, PL, SK, BE 5 rates Alternatively included under fiscal benefits AT, EE, FR, PL, SK, BE 6 In a few cases the AITRs and AISCRs provided are calculated from an incomplete tax base either the tax base used assumes benefit income only, or the tax base includes some non-taxable income. The extent to which this impacts on AITRs and AISCRs is not known but in one case taxes are imposed on a small part (<10%) of social protection benefits so the impact cannot be great. Social contribution rates are affected in only two countries. Finally, two countries were unable to take account of additional relevant taxes which can apply to some benefit income for example taxes on income applied at regional or local level. This may mean that the final estimates of net expenditure are overstated for these countries. The extent of the overstatement is not known but assumed to be small. Table 8: Tax base and levies types Discrepancies in tax base (i.e. income from non-benefit income not included, non taxable income included) AITRs Total AISCRs Total BE, IS, SI*, PL 4 BE, IS 2 Additional levy types not included BE, NL 2 0 * Portion of TOTAL benefits subject to levies: <10% Observation period In Italy the primary data source is biennial and covers even years. The data provided for the pilot data collection therefore refer to 2006. It has been suggested that estimation models could be developed to provide annual results (Table 9). Table 9: Frequency of availability of necessary source data Countries Total Data not available every year IT 1 CONCLUSIONS Despite a number of methodological difficulties faced by countries in the pilot collection of data on net social protection benefits something that is to be expected given the complexity of the issue all countries managed to provide data. One questionnaire is known to have incomplete coverage (Greece) and the methodological approach of a couple of countries 4 needs to be reviewed before a full data collection is implemented. But in general it appears that the data are more or less complete and 4 Social contribution rates provided by the Netherlands for some benefits include employer s contributions. Tax rates for Poland were calculated assuming benefit income only. DOC 2 _ EU Synthesis report.doc\alphametrics Ltd.\20 November 2009 7/8
plausible. It has to be concluded, therefore that the first criterion for continuing the project that the data collection is feasible is positive for a large majority of countries. Systematic validation of the data provided has shown that they are not perfect Table 10 summarises the main problems observed. However, many of the issues related to the grouping of benefits to calculate AITRs and/or AISCRs introduce inaccuracies only at the most detailed level, which it is not planned to publish, and at secondary levels, the extent of which may well be small. Overall the final estimates of net expenditure should mostly be unaffected. Issues of incomplete coverage are also minor in most cases and taking into account final post year-end tax adjustments or similar would not have any major impact on the overall results. We therefore conclude that in terms of the second criterion regarding the quality of the data collected is also positive for the majority of countries. Table 10: Methodological practices which apply to a significant proportion of benefits Theme Countries Total Identification Rates calculated for groups of benefits across functions Lack of distinction between means-tested and non means-tested benefits Lack of distinction between taxable and non-taxable benefits Coverage EL, DE, FI, BE, NL 5 AT, EL, UK, FI 4 EL 1 Insufficient data to calculate certain rates EL 1 Post tax year adjustments not included in rates AT, LT, LV, PT 4 Other tax system adjustments not included in rates EE, FR, PL, SK, BE 5 Discrepancies in tax base (i.e. incomes not included, non tax base income included) BE, IS, PL 3 Additional levies not included BE, NL 2 DOC 2 _ EU Synthesis report.doc\alphametrics Ltd.\20 November 2009 8/8