Harmonized Household Budget Survey how to make it an effective supplementary tool for measuring living conditions

Transcription

1 Harmonized Household Budget Survey how to make it an effective supplementary tool for measuring living conditions Andreas GEORGIOU, President of Hellenic Statistical Authority Giorgos NTOUROS, Household Surveys Section Abstract Economic crises highlight the need for effective social surveys that accurately measure in a comparable way living conditions in various parts of the European Union. Eurostat s requirements on the Household Budget Surveys (HBS) comparability and harmonization between states are not as demanding as those for other household surveys. No EU HBS regulation exists with provisions on survey design, survey characteristics, data transmission, dissemination and publication. In this context, the paper proposes the continuation of the discussion on harmonization and comparability (sample size, frequency, reference periods, concepts and definitions, core variables, preventing non response, COICOP-HBS) in order to render the survey an effective tool for measuring inequalities and living conditions at European level using as the basic variable on consumption, while also assuring comparability with the EU SILC instrument. Cohesion among the indicators resulting from the surveys can only be assured by using a common framework for the HBS at European level. Introduction The recent economic crisis has highlighted the need for statistical tools that permit adequate monitoring of the effects of economic fluctuations on income/consumption distribution and inequalities, social exclusion and quality of life over time. At EU level, the EU SILC instrument is the reference point for comparative statistics on income distribution and social exclusion, being fully harmonised across member states. With the information that is collected, structural indicators for social cohesion are calculated and systematic statistics are produced on income inequalities, inequalities in households living conditions, poverty and social exclusion.

2 The EU Household Budget Survey is a set of independent national multi-purpose sample surveys conducted by the National Statistical Offices of member states, collecting information, from a representative sample of households, on the composition of households, household members employment status, living conditions and, mainly, focusing on household members on goods and services as well as on their income. Despite Eurostat s efforts, the HBS is still not fully harmonised. The main purpose of the HBS is to determine in detail the pattern of household s in order to revise the Consumer Price Index. Moreover, the HBS is the most appropriate source for: (i) the completion of the available statistical data for the estimation of total private consumption, (ii) the study of households s and their structure in relation with households income and other economic, social and demographic characteristics, (iii) analyzing the longitudinal changes in the living conditions of the households, (iv) studying the relationship between households purchases and receipts in kind, (v) studying low-income limits in the different socio-economic categories and population groups, and (vi) studying the changes in the nutritional habits of the households. For a number of reasons, the EU-SILC and the HBS are two surveys that are very valuable statistical sources for the study of many social phenomena, such as inequality, poverty, social exclusion, etc. The HBS is the only household survey collecting information on both actual and imputed consumption (in a resolution of about 900 different codes /items or services), housing amenities, consumer durables and socioeconomic (occupational, demographic, educational, etc.) characteristics of households and their members. The combination of these variables also allows the investigation of the determinants of child poverty and the identification of specific risk groups, which allows inter alia the possibility to measure absolute poverty or to estimate inflation by household characteristics. On the other hand, the EU-SILC, up to now, regarding the calculation of indicators, only provides information on income in cash and not on income in kind. Since 2007, information on income in kind is also being collected, however, it is not yet being used, as reliability and comparability tests have not yet been completed. DGINS

3 It should be noted that each one of these two surveys has both advantages and disadvantages concerning the identification of vulnerable groups. Nevertheless, the HBS has an advantage as, apart from information on purchases, it also collects information on imputed income. This is particularly important for the exact identification of groups characterized as poor. Obviously, the needs of a household residing in a rented home are greater than those of a household, of the same demographic type, residing in a home it owns. Similarly, the needs of a household consuming part of its own agricultural production are less than those of a household, of the same demographic type and same disposable income, which is obliged to cover its needs exclusively through purchases. This paper intends to discuss the main issues that can affect comparability of the HBS at EU level in order to render the survey an effective tool for measuring inequalities at European level using as the basic variable for consumption. 1. Efforts towards the harmonization of the HBS I In statistics, comparability is the main dimension of data quality. According to Verma (2006) Comparability means the extent to which the results for different countries can be put together, compared, and interpreted in relation to each other and against common standards. An assessment of how far such comparability has been achieved in practice requires us to examine the data and procedures both from the input and the output sides. Comparability is achieved via the harmonisation of objectives, concepts, definitions, classifications, variables and coverage of the population. From the 1980s, Eurostat (1980, 1993, 1997, 2003, 2005, 2009) made efforts towards harmonization on HBS of objectives, concepts, definitions, classifications, variables and coverage of the population. Despite Eurostat s efforts, the HBS is still not fully comparable among members states. Eurostat and member states could usefully work together more intensively on harmonization of the HBS regarding sample size, frequency, reference periods, non-response, COICOP-HBS etc. The working groups that will be set up should work according to the model applied for EU LFS and EU SILC. DGINS

4 2. Factors affecting comparability Comparability should be ensured for all aspects of the HBS survey. However, in the present manuscript we will discuss the main factors affecting comparability among member states, those being definitions, minimum and achieved sample size, frequency, reference and recording periods, non response and the classification of goods and services. Significant steps towards harmonization have been made by using common definitions for household, household members, and reference person concepts. Moreover, efforts are being made for the harmonization of the rest of the core variables (such as age, de facto marital status, household composition, labour status, status and occupation in employment, economic sector, highest level of education, net monthly income, etc.) from 2010 onwards across all countries. One of the primary efforts towards strengthening data comparability between countries is to determine a minimum sample size in each country. According to Eurostat (HBS quality report, round 2005) the achieved HBS sample sizes do not comply with this requirement: for instance, the sample sizes in France, Spain, United Kingdom, ones of the biggest EU countries, are relatively low (respectively 10240, 8881 and 6785 households) compared to the achieved sizes in less populated countries like, for instance, Austria (8400), Portugal (10403) or Romania (33066). The differences in sample sizes between the HBS and the EU-SILC in a given country can be large and these differences vary widely across member states. As shown in Table 1 below, in the Netherlands and in Sweden the sample size of the HBS is 16.4% and 33.9% of the EU SILC respectively, while in Germany the HBS sample size is 398% of the EU SILC sample size and in Portugal 225%. Such a distribution of differences is likely to damage comparability at EU level. It is thus obvious that there is a need for harmonized and representative minimum sample sizes for countries in order to improve comparability. DGINS

5 Table 1: Achieved sample sizes (in households) HBS 2005/EU-SILC 2005 Country HBS 2005 EU-SILC 2005 HBS/EU-SILC (%) The Netherlands (NL) Sweden (SE) Finland (FI) Denmark (DK) Germany (DE) Portugal (PT) Poland (PL) Greece (GR) Another issue is that the HBS data are being collected with irregular frequency, unlike other household surveys. Twenty (20) countries conduct annual surveys and 8 countries collect data every five years. Many users have argued though that the practical usefulness of the data could be enhanced by increasing the frequency of the survey (e.g., every one or two years) and, thus possibly, being able to handle (or interpret) better zero-measurements, which are present in a large number of households. An appropriate convenient frequency may, thus, be annual. Expenditure and income variables refer to a specific start and end date, called the reference period. In order to reduce non-sampling errors and difficulties in recalling the relevant details, various reference periods are used in the survey, according to the frequency of the types of incurred by the households or the received income. The HBSs do not have harmonized recall periods. The different recall periods can affect the quality of the data expected from countries. In 2003, ELSTAT tested two questionnaires by conducting a pilot survey in a sample of 100 households. In the first questionnaire, the reference period of footwear was the last month and in the second questionnaire a 3-month period. The estimates of total footwear differed by 15 per cent between the two different types of questionnaires. The estimates were higher for the long period (Table 2). DGINS

6 Table 2. Average footwear by different reference periods Reference period: Reference period: Difference % Good and services 1 month 3 month Footwear for men Footwear for women Footwear for children (3 to 13 years) and infants (0 to 2 years) Another example from the above-mentioned pilot survey concerns olive oil, rice, wine, potatoes and pasta consumption, for which different reference periods were used: - The fourteen (14) days of the survey for the daily and - Both, the last twelve (12) months prior the end of the household survey (including the 14 days of the survey) and the fourteen (14) days of the survey for the daily. Table 3: Average of some food items by different reference periods Goods and services Reference period: 14 days Reference period: last 12 months Difference % Olive oil Rice Wine Potatoes Pasta As a consequence of these findings, there has since been special handling in the Greek NSI of large food purchases, such as olive oil, potatoes, rice, pasta, cheese, dried nuts, maize, other cereals, jams and marmalades, sugar, dried vegetables, wine etc. Reference period is considered to be a significant factor for the harmonization and comparability in question. DGINS

7 Duration of recording period also varies significantly among member states. Two weeks is common practice for most of the countries, one month is the recording period for Luxembourg, Germany and Spain, while one week is the practice in Portugal and one year is used in Belgium. A one-month recording period appears to be the most appropriate, since households nowadays shop less frequently. In any case, any change in the recording period of should be combined with the implementation of innovations in the data collection procedure, as the change in the recording period itself may increase nonresponse. The non-response rate is very high. The response rate in Greece is 60.3% while the mean response rate at EU level is around 60%. There are, however, important variations between countries: from 5.9% in Belgium and 41.9% in Austria to 88.9% in Cyprus and 90.3% in Romania. The non-response rate can be attributed to refusal, non-contact, inability to cooperate due to long-term illness, inability to cooperate due to language problems, etc. Despite the efforts of all member states, such as using advance letters, or accepting responses in a mode desired by respondents e.g. , improving the quality of the survey framework, providing better survey publicity by the media, using experienced interviewers, who can explain the survey s importance, etc, response rate is falling over time and it is our strong belief that innovations in data collection are needed to reduce nonresponse. The presentation made by Norwegian statisticians in Eurostat s Task Force on preventing non-response in March 2010, by using electronic data collection (via e-cards) from the stores would be a good practice to follow. According to this approach, all commodities will be registered with their price and amount in today's registers. Detailed commodity data will be sent over the Internet to a server at Statistics Norway every time one of the sampled households is making a purchase. Participating households are equipped with a Smart Card (ID). Similar examples also exist in HICP and PPP. As mentioned above, the main purpose of the HBS is to determine in detail the household pattern in order to revise the Consumer Price Index, and moreover, the HBS is the most appropriate source for the completion of the available statistical data and the estimation of total private consumption. But there is no consistency and comparability among COICOP used in HBS, HCIP and PPS. DGINS

8 Eurostat proposes one COICOP for HICP/PPP/HBS. According to Eurostat (minutes of Working Group 2011 on Living Conditions), this project is still in progress and the internal team has to discuss more comments received from the member states. The greater level of detail may cause problems to the selection of data, but the construction of one common classification for HICP/PPP/HBS will bring higher levels of consistency and comparability, and users will be familiar with only one core classification. 4. Comparison of poverty indicators using SILC and HBS This section provides a comparative analysis of EU-SILC and HBS results concerning poverty indicators. For the comparison of some poverty indicators derived from the Greek HBS 2005 and EU- SILC 2005 micro-data, we use the methodology of EU SILC 1 as closely as possible. We follow the same methodology that Eurostat has used for comparison of these indicators in EU-HBS quality reports. The surveys of 2005 have been used, because the sample sizes are approximately the same. More specifically: In the case of EU SILC, for each household, the household net monetary income was divided by the equivalized household size and the result was given to each household member as an estimate of the personal income. The indicators were calculated at individual level using this personal income as well as the personal sample weights. 2 In the case of HBS, the database was used to produce results at the level of the individual by replicating the household records according to household size (per capita). Specifically, each household monetary from HBS was divided by the equivalized household size and the result was given to each household member as an estimate of the personal. The indicators were calculated at 1 According to the methodology for measuring poverty, the poverty line is calculated with its relative concept (poor in relation to others) and is defined at 60% of the median total equivalized disposable income of the household, using modified OECD equivalized scale. Equivalent size refers to the OECD modified scale which gives a weight of 1.0 to the first adult, 0.5 to other persons aged 14 or over who are living in the household and 0.3 to each child aged under 14. As total equivalized disposable income of the household is considered total net income (that is income after deducting taxes and social contributions) received by all household members. 2 Quality report of the Household Budget Surveys 2005 DGINS

9 individual level using this personal as well as the personal sample weights In the following tables, we can see the extent of poverty concerning alternative age groups after having evaluated two HBS and one EU SILC variables, as follows: (a) The distribution of the total equivalent disposal income in cash (only the company car is included) from EU SILC. (b) The distribution of the total equivalent monetary from HBS and (c) The distribution of the total equivalent from HBS Table 1. At-risk-of-poverty threshold At-risk-of-poverty threshold monetary income ,40 Table 2. At risk of poverty by age groups. % Age groups monetary income Total Table 3. Income quintile share ratio S80/S20 Income quintile share ratio S80/S20 monetary income DGINS

10 Table 4. Dispersion around the at-risk-of-poverty threshold. % Threshold monetary income 40% of median % of median % of median % of median Table 6. At-risk-of-poverty rate, by household type. % Household type monetary income One adult younger than 64 years One adult older than 65 years Single parent at least 1 dependent child Table 5. Relative at-risk-of-poverty gap. % Relative at-risk-ofpoverty gap monetary income DGINS

11 Table 6. Child Poverty by age groups.% Age monetary income There are differences among the HBS-monetary and EU-SILC monetary income indicators. In most cases above, the HBS indicators are lower. In this light, HBS indicators provide an important view, which at least complements any information and analysis that is derived from the EU-SILC income indicator. 4. Conclusions Although, the analysis presented is not exhaustive, focusing only on certain aspects of the methods and measurement of the HBS, the need for harmonization among member states appears unquestionable. Eurostat and member states could usefully work together more intensively on harmonization of the HBS regarding sample size, frequency, reference periods, non-response, COICOP etc. Methodological workshops and task forces on specific issues could usefully be organized by Eurostat in the near future. DGINS

12 References (1) Verma V., Gabilondo L. (1993). Family Budget Surveys in the EC: Methodology and Recommendations. Luxembourg: Statistical Office of the European Community, Series 3E. (ISBN ). (2) Verma V.,Issues in data quality and comparability in EU-SILC Eurostat and Statistics Finland Conference, Helsinki, 6-7 November, (3) Minimum sample sizes for the next HBS, DOC LC-HBS/08/09/EN, Working Party, Living Conditions, June 2009, Eurostat Luxembourg. (4) Eurostat (1980) Methodology of surveys on Family Budgets, Luxembourg. (5) Eurostat (1997) Household Budget Survey in the EU: Methodology and recommendations for harmonization, Luxembourg. (6) Eurostat (2003) Household Budget Survey in the EU: Methodology and recommendations for harmonization, Luxembourg. (7) Eurostat (2011), Minutes of Working Group of Living Conditions, Luxembourg DGINS