April 24, 2017

Middle Class Fortunes in Western Europe

Methodology

Data source

The data for this analysis are obtained from the Cross-National Data Center in Luxembourg (LIS), a research center that harmonizes and provides access to data from government surveys and other sources for a large number of countries. Of the 12 countries included in this study, the latest available data for eight countries – Finland, Germany, Luxembourg, the Netherlands, Norway, Spain, the UK and the U.S. – are for 2013, and the latest data available for Denmark, France, Ireland and Italy are for 2010.

The survey years in the LIS data archive refer to the reference year for which income data were collected. For instance, if a survey conducted in March 2011 asked about a household’s income in calendar year 2010, the LIS archive would reference that survey of a country as a 2010 survey, not a 2011 survey.

In a more specific example, for the U.S., LIS provides access to data from the Current Population Survey (CPS) conducted by the U.S. Census Bureau. The specific CPS archived by LIS is the Annual Social and Economic Supplement (ASEC) conducted in March each year. Each ASEC survey gathers data on household income in the calendar year prior to the survey date. This means that LIS files for the U.S. for 1991, 2000, 2010 and 2013 actually come from surveys conducted in March of 1992, 2001, 2011 and 2014. The implication is that distribution of adults across income tiers in the U.S. reflects the U.S. population in 1992, 2001, 2011 and 2014 even as the income tiers are defined on the basis of earnings in 1991, 2000, 2010 and 2013. A similar pattern applies to the estimates for most other countries in this study. For the sake of convenience, this report uses the LIS survey date to refer to both the income estimates and the demographic estimates.

In the LIS data, the household income measure that is available for all countries for all years is disposable income, i.e., cash income (wages, interest, rents, pensions, etc.) and near-cash income (government assistance) less income taxes and social security contributions. This is because respondents in some countries’ surveys are asked to report only net incomes (referred to as “net” datasets in LIS parlance). Also, until the mid-1990s, LIS datasets often coded missing amounts on income with the value zero. Thus, in keeping with LIS practices, households with disposable income equal to zero are excluded from the analysis in this report.

A note on the 1991 estimates for Germany

In this report, the 1991 estimates for Germany are derived from the German Socio-Economic Panel (SOEP) conducted by the German Institute for Economic Research (DIW Berlin) in 1990. This survey collected data on the incomes of households in calendar year 1989, which are projected to 1991 in this analysis for comparison with other countries (see Methodology for details). The version of SOEP used in this report is from the Cross-National Data Center in Luxembourg (LIS). A key advantage of using the LIS database is that the data from different countries are harmonized for ease of comparison.

In the LIS database, the 1990 German Socio-Economic Panel is closest to the starting date of this report’s analysis. Although SOEP is conducted annually, the other SOEP surveys that date somewhat near to 1991 and are available through LIS were fielded in 1985 and 1995. A disadvantage of using the 1990 survey is that the sample includes households from (former) West Germany only. By contrast, the samples for 2000, 2010 and 2013 available through LIS represent post-reunification Germany. However, this difference in samples is estimated to have minimal impact on the principal findings of the analysis.

A recent report by DIW Berlin examined the change in the status of lower-, middle- and upper-income tiers in Germany from 1991 to 2013. In that study, the 1991 estimates are derived from the 1992 SOEP survey, which sampled the unified German population (as noted, this survey is not available in the LIS database). DIW Berlin finds that median household income in post-reunification Germany was little changed from 1991 to 2013, increasing 3.4% over more than two decades, compared with the estimate of 0.4% in this report. DIW Berlin also finds that the share of adults in middle-income households in Germany decreased 5 percentage points from 1991 to 2013, compared with a decrease of 6 percentage points noted in this report. (The method DIW Berlin used to identify middle-income households was also used in this report.) Changes in the shares of adults in lower- and upper-income tiers are also similar between the two reports.

A difference between the DIW Berlin study and this analysis is that DIW Berlin uses gross household income to define the middle-income tier and to measure changes in its well-being. This is because SOEP only collects data on gross household income. However, this report uses disposable household income, which is included in the LIS version of the SOEP data (estimates of disposable household income are derived from tax simulation models). This LIS addition makes it possible to draw comparisons across a larger number of countries.

Extrapolation of income

The estimates of income in 1991 and 2000 for some countries are based on projections from surveys conducted in years close to those dates. These survey years are as follows: Denmark – 1992, France – 1989, Germany – 1989, Ireland – 1987, Netherlands – 1993 and 1999, Spain – 1990, and the UK – 1999.

The projections assume that from the survey year until either 1991 or 2000, income changed at an annual rate equal to 70% of the annual rate of change in real household final consumption expenditures. Household final consumption expenditures are from national income accounts and are expressed in local currency units and per capita terms (the World Bank is the source for these data). The reason for using 70% of the rate of change rather than 100% is that national income accounts tend to overstate the level and change in consumption or income as measured by household surveys (Deaton, 2003 and Nolan, Roser and Thewissen, 2016). The extent of the overstatement will vary from country to country. This study follows the practice adopted by Birdsall, Lustig and Meyer (2013) of projecting household survey data at 70% of the rate of change in national income account measures.

Although survey data on income are projected to 1991 and 2000 for some countries, the underlying distributions of adults across lower-, middle- and upper-income tiers are frozen at the date of the surveys. This is because the incomes of all households are projected at the same rate and the position of any single household relative to the (projected) median does not change. In France, for example, the estimates of income for 1991 are based on survey data from 1989. But the distribution of adults across income tiers still reflects the ground reality in 1989. This means that the reported change in the distribution of adults across income tiers from 1991 to 2010 actually reflects more than a 19-year change.

Conversion to 2011 prices

Income estimates in this study are presented in 2011 prices. It is desirable to convert income or consumption data to 2011 prices because the purchasing power parities (PPPs) for all countries are derived from the round of international price comparisons conducted in that year (additional detail on PPPs and their use in this study are provided below). The conversion is a matter of inflating the survey data by the change in a country’s consumer price index (CPI) from the relevant year to 2011. This calls for the averaging of CPI data over two to three years for some countries for some survey dates. For example, the 2000 survey for France was conducted from May 9, 2000, to May 6, 2001, and respondents were asked to report income in the 12 months preceding the interview. Thus, in principle, the income data in the 2000 survey span the period from May 1999 to April 2001. The CPI applied in this case, as the base for inflating to 2011 prices, is the average of the CPI levels for 1999, 2000 and 2001. The CPI data are obtained from the World Bank or the International Monetary Fund (in the case of Germany).28

Conversion to 2011 purchasing power parities

Purchasing power parities are exchange rates corrected for differences in the prices of goods and services across countries. Thus, estimates of household income in this report are expressed in purchasing power parities to allow for a more accurate comparison of the standard of living across countries. The latest available estimates of PPPs are from a round of international price comparisons conducted in 2011 by the International Comparisons Program at the World Bank. The specific PPPs used in this report are the ones that pertain to individual consumption expenditures by households.

In principle, one PPP dollar represents the same standard of living across countries. The U.S. serves as the base country for price comparisons and for currency conversions. Thus, for the U.S., one U.S. dollar equals one PPP dollar. But for Denmark, for example, the Danish krone to U.S. dollar conversion rate – 5.36 in 2011 – is different from the krone to PPP dollar rate – 8.524 for individual consumption expenditures by households. This means that due to the higher cost of living in Denmark, 8.524 Danish krones, not 5.36, are needed to obtain what $1 buys in the U.S.

In LIS, survey data are expressed in local currency units prevailing at the time. For several countries examined in this report, the survey data for 2010 and 2013 are expressed in euros, but the data from years near 1991 and 2000 are in currencies that existed before the adoption of the euro. These countries are Finland, France, Germany, Ireland, Italy, Luxembourg, the Netherlands and Spain. It was necessary in these cases to convert data from the earlier years to euros because the 2011 PPPs for them are euro-denominated. The rates used for these countries to convert local currency units to the euro are the ones reported by Eurostat.

Adjusting income for household size

Household income data reported in this study are adjusted for the number of people in a household. That is done because a four-person household with an income of, say, $50,000 faces a tighter budget constraint than a two-person household with the same income. In addition to comparisons across households at a given point in time, this adjustment is useful for measuring changes in the income of households over time, especially over the long run. In the U.S., for example, the average household size decreased from 3.1 persons in 1970 to 2.5 persons in 2015, a drop of 19%. Ignoring this demographic change would mean ignoring a commensurate loosening of the household budget constraint.

At its simplest, adjusting for household size could mean converting household income into per capita income. Thus, a two-person household with an income of $50,000 would have a per capita income of $25,000, double the per capita income of a four-person household with the same total income.

A more sophisticated framework for household size adjustment recognizes that there are economies of scale in consumer expenditures. For example, a two-bedroom apartment may not cost twice as much to rent as a one-bedroom apartment. Two household members could carpool to work for the same cost as a single household member, and so on. For that reason, most researchers make adjustments for household size using the method of “equivalence scales.”29

A common equivalence-scale adjustment is defined as follows:

Adjusted household income = Household income / (Household size)N

By this method, household income is divided by household size exponentiated by “N,” where N is a number between 0 and 1.

Note that if N = 0, the denominator equals 1. In that case, no adjustment is made for household size. If N = 1, the denominator equals household size, and that is the same as converting household income into per capita income. The usual approach is to let N be some number between 0 and 1. Following other researchers, this study uses N = 0.5.30

In practical terms, this means that household income is divided by the square root of household size – 1.41 for a two-person household, 1.73 for a three-person household, 2.00 for a four-person household and so on.31

Once household incomes have been converted to a “uniform” household size, they can be scaled to reflect any household size. The average size of a household in the U.S. is 2.5 and is between two and three in the countries in Western Europe. Thus, the income data reported in this study are computed for three-person households. That is done as follows:

Three-person household income = Adjusted household income * [(3)0.5]

It is important to note that once the household-size adjustment has been made, it is immaterial whether one scales incomes to one-, two-, three- or four-person households. Regardless of the choice of household size, the same results would emerge with respect to the trends in the well-being of lower-, middle- and upper-income groups.