Estimating Procedures

Because statistical concepts and methods vary from country to country, international comparisons of statistical data can be misleading. BLS attempts to derive meaningful comparisons by selecting a conceptual framework for comparative purposes; analyzing foreign statistical series and selecting those which most nearly match the desired concepts; and adjusting statistical series, where necessary and feasible, for greater intercountry comparability.

Productivity and unit labor costs

The Bureau of Labor Statistics constructs trend indexes of manufacturing labor productivity (output per hour), hourly compensation costs, and unit labor costs from three basic aggregate measures--output, total labor hours, and total compensation. The hours and compensation measures refer to employees (wage and salary earners) in Korea, Taiwan, Belgium, Denmark, and Italy. For all other countries, the measures refer to all employed persons, including employees, self-employed persons, and unpaid family workers. For all of the countries, the term "hours" refers to hours worked.

Output: In general, although some variations are present, the measures relate to total manufacturing as defined by the International Standard Industrial Classification (ISIC). However, the measures for France refer to mining and manufacturing less energy-related products, and the measures for Denmark include mining and exclude manufacturing handicrafts from 1960 to 1966.

For most countries, the output measures are value added in manufacturing from the national accounts. However, output for Japan prior to 1970 and for the Netherlands prior to 1960 are indexes of industrial production. The national accounts measures for the United Kingdom are essentially identical to their indexes of industrial production.

Manufacturing output data for the United States from 1977 forward are the gross product originating (value added) measures prepared by the Bureau of Economic Analysis (BEA) of the U.S. Department of Commerce. U.S. gross product originating is a chain-type annual-weighted series. Comparable manufacturing output data are not available prior to 1977.

The gross product originating series differs from the manufacturing output series that BLS publishes in its news releases on quarterly measures of U.S. productivity and costs. The quarterly measures are on a sectoral output basis rather than a value added basis. Sectoral output is gross output less intra-sector sales and transfers.

A value added concept has been used for the international comparisons series because the data are more readily available in the countries' national accounts, whereas sectoral output would require a complex estimation procedure. Also, although BLS has determined that sectoral output is the correct concept for U.S. measures of single factor productivity (output per hour), there are other considerations that may make value added a better concept for international comparisons, such as differences among countries in the extent of vertical integration.

The historical real output series for many of the other countries are estimated using fixed price weights, with the weights updated periodically (for example, every 5 or 10 years).  Measures of real output also may differ among countries because of different approaches to estimating the prices of high-technology products like computers and, in general, of products that undergo rapid quality change.

Labor input: The total hours measures are developed from statistics of manufacturing employment and average hours. The total hours series used for France (from 1970 forward), Norway, Sweden, and Canada are series published with the national accounts.  For the other countries, BLS constructs estimates of hours from various sources.

For the Republic of Korea and Taiwan, BLS currently publishes only measures of unit labor costs and its components--output and total compensation.  Total hours, and consequently productivity, are not computed for Korea and Taiwan because BLS has not yet developed adequate labor input series.

Compensation (Labor Cost):  The compensation measures are from the national accounts. Compensation includes employer expenditures for legally required insurance programs and contractual and private benefit plans, in addition to all payments made in cash or in-kind directly to employees. For Canada, France, and Sweden, compensation is increased to account for other important taxes on payroll or employment. For the United Kingdom, compensation is reduced between 1967 and 1991 to account for subsidies. When data for the self-employed are not available, total compensation is estimated by assuming the same hourly compensation for self-employed and employees. Real compensation for the U.S. is derived with the Consumer Price Index research series (CPI-U-RS).

Current indicators: The measures for some recent years may be based on current indicators of manufacturing output (such as industrial production indexes), employment, average hours, and hourly compensation, until national accounts and other statistics used for the long-term measures become available.

Trade-weighted measures: Since the economies covered differ greatly in their relative importance to U.S. trade in manufactured products, BLS constructs trade-weighted measures of unit labor costs to take account of these differences. The trade weights used were derived by rescaling a series covering 21 economies that was developed by the International Monetary Fund. These weights are based on disaggregated 1980 trade data for manufactured goods and take account of both bilateral trade and the relative importance of "third country" markets. Two summary measures are constructed: "Competitors" indexes, which are the trade-weighted geometric averages of the indexes for competitor economies, and relative indexes, which are the ratios of the U.S. indexes to "competitors" indexes.

Hourly compensation costs

Measures of hourly compensation costs are prepared by BLS in order to provide a better basis for assessing international differences in employer labor costs. Comparisons based on the more readily available average earnings statistics published by many countries can be very misleading. National definitions of average earnings differ considerably; average earnings do not include all items of labor compensation; and the omitted items of compensation frequently represent a large proportion of total compensation.

Hourly compensation is defined as (1) all payments made directly to workers -- pay for time worked (basic time and piece rates plus overtime premiums, shift differentials, other premiums and bonuses paid regularly each pay period, and cost-of-living adjustments), pay for time not worked (such as for vacations and holidays), seasonal or irregular bonuses and other special payments, selected social allowances, and the cost of payments in kind -- before payroll deductions of any kind, and (2) employer expenditures for legally required insurance programs and contractual and private benefit plans (such as retirement plans, health insurance, unemployment insurance, and family allowances). In addition, for some countries, compensation is adjusted for other taxes on payrolls or employment (or reduced to reflect subsidies), even if they do not finance programs that directly benefit workers, because such taxes are regarded as labor costs. 

The BLS definition of hourly compensation costs is not the same as the International Labor Office (ILO) definition of total labor costs. Hourly compensation costs do not include all items of labor costs. The costs of recruitment, employee training, and plant facilities and services -- such as cafeterias and medical clinics -- are not included because data are not available for most countries. The labor costs not included account for no more than 4 percent of total labor costs in any country for which the data are available.

The total compensation measures are computed by adjusting each country's average earnings series for items of direct pay not included in earnings and for employer expenditures for legally required insurance, contractual and private benefit plans, and other labor taxes or subsidies. For the United States and other countries that measure earnings on an hours-paid basis, the figures are also adjusted in order to approximate compensation per hour worked.

Earnings statistics are obtained from establishment surveys of employment, hours, and earnings for most countries. Adjustment factors are obtained from periodic labor cost surveys and interpolated or projected to nonsurvey years on the basis of other available information for most countries. The information used for interpolation and projection includes annual tabulations on employer social security contribution rates provided by the International Studies Staff of the U.S. Social Security Administration, information on contractual and legislated fringe benefit changes from ILO and national labor bulletins, and statistical series on indirect labor costs. For other countries, adjustment factors are obtained from surveys or censuses of manufactures or from reports on fringe benefit systems and social security. For the United States, the adjustment factors are special calculations for international comparisons based on data from several surveys.

The statistics are also adjusted, where necessary, to account for major differences in worker coverage; differences in industrial classification systems; and changes over time in survey coverage, sample benchmarks, or frequency of surveys. Nevertheless, some differences in industrial coverage remain and, with the exception of the United States, Canada, and several other countries, the data exclude very small establishments (fewer than 5 employees in Japan and fewer than 10 employees in most European and some other countries). For the United States, the methods used, as well as the results, differ somewhat from those for other BLS series on U.S. compensation costs.

Special estimation procedures have been used for some countries because of incomplete data. Hourly earnings are computed from daily or monthly earnings using estimates of standard hours worked for Israel. Earnings for production workers are estimated from all-employee earnings for Japan (1990 to present), Korea (1975-84), New Zealand (all years), and Israel (1978 to present). For Singapore, hourly compensation costs are estimated for the years since 1985 using the trend in average weekly earnings and benefit costs of production workers or average compensation per employee from the national accounts. For Italy, both hourly earnings and additional compensation adjustment factors are obtained from periodic labor cost surveys and interpolated or projected to nonsurvey years using the trend in contractual wage rates and in benefit entitlements and costs. The adjustment factors for additional compensation relate to all employees for Mexico, Japan, Korea, Taiwan, France, Ireland, the Netherlands, Portugal, Spain, and the United Kingdom; they are constants or the midpoints of constant ranges for Hong Kong, Israel, and New Zealand.

Hourly compensation costs are converted to U.S. dollars using the average daily exchange rate for the reference period. The exchange rates used are prevailing commercial market exchange rates as published by either the U.S. Federal Reserve Board or the International Monetary Fund.

Trade-weighted measures: Because the economies covered differ in their relative importance to U.S. trade in manufactured products, BLS constructs trade-weighted measures comprising all of the foreign economies covered and selected economic groups to take account of these differences. The trade weights used to compute the average compensation cost measures for selected economic groups are relative importances derived from the sum of U.S. imports of manufactured products for consumption (customs value) and U.S. exports of domestic manufactured products (free along side (f.a.s.) value) in 1992 for each country or area and each economic group.

The trade data used to compute the weights are U.S. Bureau of the Census statistics of U.S. imports and exports converted to an industrial classification basis from data initially collected under the Harmonized Tariff Schedule commodity classification system.

The trade-weighted average rates of change are computed as the trade-weighted arithmetic average of the rates of change for the individual countries or areas; the trade-weighted average hourly compensation costs are computed as the trade-weighted arithmetic average of cost levels for the individual countries or areas. Rates of change derived from the trade-weighted average hourly compensation cost levels need not be the same as the trade-weighted average rates of change.

For these comparisons, BLS adjusts each country’s published data, if necessary, to provide measures approximately consistent with U.S. definitions. Although precise comparability may not be achieved, these adjusted figures provide a better basis for international comparisons of labor force, employment, and unemployment than the figures regularly published by each country.

ILO Guidelines: The International Labor Office has established guidelines for the measurement of labor force, employment, and unemployment. These guidelines can be accessed at the following web site: http://www.ilo.org/public/english/120stat/res/ecacpop.htm

The ILO guidelines have become standards for many countries; consequently, definitions used in labor force surveys are now broadly similar in outline and spirit if not in all of their details. The guidelines facilitate cross-country comparisons because they serve to draw countries toward a common conceptual framework. In some respects, however, the ILO guidelines are vague or provide options. Also, the United States Current Population Survey (CPS) does not follow all of the ILO guidelines. For these reasons, adjustments sometimes must be made to adapt the labor force data of other countries more closely to U.S. concepts.

Cross-country differences: Differences in the concept of employment are few, relating mainly to treatment of the Armed Forces, unpaid family workers, and layoffs. For unemployment, however, there are more numerous conceptual and methodological differences across countries, sometimes requiring significant adjustments to the official foreign country data for greater comparability with U.S. concepts.

The varying interpretations of unemployment derive from different national circumstances and needs. Countries generally have very good reasons for their own interpretations of, or deviations from, the ILO guidelines. Further, some countries choose to derive their official unemployment figures from administrative data on employment office registrations or benefits claims, while others, such as the United States, derive their official data from labor force surveys. Although not explicitly stated by the ILO, it is well recognized that labor force surveys are the desirable source for international comparisons of unemployment. In most countries, such surveys cover the entire noninstitutional population of working age, broadly follow the ILO standard definitions, and produce data that can be used for adjustment purposes.

Administrative data are not suitable for international comparisons, because they do not cover all persons who may be unemployed and because administrative regulations differ greatly across countries. Administrative data, however, are useful in international comparisons in that they provide monthly and quarterly trend data on unemployment for countries that conduct labor force surveys only annually. How these data are used is discussed in the section below on data sources.

The BLS comparisons are based on data from labor force surveys that are adjusted for greater comparability with U.S. concepts, if necessary. BLS has identified the following conceptual areas where foreign survey-based unemployment concepts may differ from those used in the U.S. survey: (1) The measurement and meaning of "active job search" and "current availability" on the part of unemployed persons; (2) the treatment of persons on layoff; (3) whether persons waiting to start a new job must seek work to be counted as unemployed; and (4) the treatment of student jobseekers.

Data sources: The statistics for 7 of the 10 countries regularly studied – the United States, Canada, Australia, Japan, Italy, Sweden, and since 1992, the United Kingdom — are obtained from monthly or quarterly labor force surveys. With the exception of the United Kingdom, these surveys are used directly by BLS to derive both annual and monthly (only quarterly for Italy) estimates of the labor force and unemployment. The U.K. quarterly survey relates to Great Britain only, except for the second (spring) quarter which also includes Northern Ireland. The U.K. Office of National Statistics provides BLS with data adjusted to ILO concepts and that include Northern Ireland for all quarters and months.

Adjustment factors for France, Germany (former West through 1990 and unified after 1990), the Netherlands, and, prior to 1992, the United Kingdom are derived from annual labor force surveys (biennial surveys in the Netherlands prior to 1983) conducted in the spring. The unemployment figures from the spring surveys are adjusted to an annual average basis using monthly administrative data on the number of registrants (or claimants for benefits, in the case of the United Kingdom) at public employment offices. The labor force figures from these annual surveys are also adjusted to an annual average basis using various short-term indicators. For Germany (unified), France, and the Netherlands the data are obtained from the Statistical Office of the European Communities’ (EUROSTAT) annual labor force survey rather than from the national surveys. These data may be adjusted further by BLS for comparability with U.S. concepts. BLS makes monthly and quarterly estimates of unemployment rates for all the countries covered except for the Netherlands.

Impact of adjustments: No adjustments are made to the national published data for Australia, since that country’s concepts are very close to those of the United States. Slight adjustments are made to the published data for Japan, France, Italy (since 1993), and the Netherlands. A more substantial adjustment was made to the Italian data prior to a definitional change made in October 1992 that drew the data closer to U.S. concepts with regard to job search criteria.

More significant adjustments are made to the national published unemployment rates for Canada, Sweden, Germany, and the United Kingdom. For Canada, four adjustments are made. The main one in terms of impact is the exclusion of so-called passive jobseekers from the unemployed. These are persons who conduct their search for work merely by reading newspaper ads. Such individuals are included in the unemployed in Canada, but are excluded in the United States. The overall impact of the adjustments lowers the annual Canadian unemployment rate by a little less than 1 percentage point in recent years.

BLS adjusts the national Swedish unemployment data to add students who also sought work into the unemployed for greater comparability with U.S. concepts that include such students in the unemployed. This adjustment raises the Swedish unemployment rate by about one and one-half percentage points in recent years.

The national published unemployment rates for Germany and the United Kingdom are based on employment office data rather than labor force surveys. These administrative data overstate the unemployment rate for Germany according to U.S. concepts, and they understate the U.S.-based rate for the United Kingdom. The adjusted unemployment rates for Germany are about 3 percentage points lower than the official administrative figures, whereas the adjusted data for the United Kingdom are about 2 percentage points higher.

These different results in the direction of adjustment warrant some explanation. In Germany, administrative data on registered unemployment include a significant number of persons who have "marginal" jobs but are looking for other jobs or more work. Such persons would be classified as employed in a labor force survey. Furthermore, the official German unemployment rate is calculated as the number of registered unemployed as a percent of the civilian wage and salary labor force, which also serves to overstate the German unemployment rate in comparison with the adjusted to U.S. concepts rate that is based on the entire civilian labor force, a larger denominator. By contrast, for the United Kingdom, the official unemployment figures are based on the count of claimants for unemployment benefit. Only jobless persons are included in the claimant count, the claimant count is more restrictive than the more inclusive full registration count, and the unemployment rate is calculated as the number of claimants for benefits as a percent of the total labor force, including military personnel. All of these factors lead to understatement of the U.K. unemployment rate in comparison with U.S. concepts.

Exceptions to adjustments: The foreign country data are adjusted as closely as possible to U.S. concepts, with the exception of age limits and the treatment of layoffs for which no adjustments are made. Instead, age limits are adapted to the age at which compulsory schooling ends in each country. In the case of layoffs, it was decided that the nature of layoffs was so different across countries that no adjustment should be made on this point. For further information on the treatment of layoffs, see Joyanna Moy and Constance Sorrentino, "Unemployment, labor force trends, and layoff practices in 10 countries," Monthly Labor Review, December 1981, pp. 3-13.

In addition, for most countries no adjustments are made for deviations from U.S. concepts in the treatment of unpaid family workers, persons waiting to start a new job, and passive jobseekers. Further, no adjustments are made (except for Canada) for differences in the way current availability for work is measured. Except for the passive jobseekers in Canada, for which an adjustment is made, these "unadjusted" conceptual differences have a negligible effect on the comparisons.

For further information on comparability issues and quantification of the "unadjusted" differences, see "International Unemployment Rates: How Comparable are They?" PDF (97K) by Constance Sorrentino, Monthly Labor Review, June 2000, pp. 3-20.

Breaks in series: There are breaks in the data series for the United States (1990, 1994, 1997, 1998, 1999), Canada (1966, 1976), and the adjusted series for Canada (1976), France (1992), Germany (1983, 1991), Italy (1986, 1991, 1993), the Netherlands (1983,1988), and Sweden (1987, 1993). Canada’s data have been adjusted for the 1966 break back to 1959. Sweden’s data have been adjusted for the 1993 break back to 1987. No adjustments have been made for the other breaks in series. Reasons for the breaks are documented in the monthly release on comparative unemployment rates and in the labor force compendium available on this web site.

Real gross domestic product per capita and per employed person

BLS measures of comparative trends of gross domestic product (GDP) per capita and per employed person are based on country measures of real GDP, population, and employment. BLS level comparisons of GDP, GDP per capita, and GDP per employed person are based on converting GDP into U.S. dollars using purchasing power parities (PPPs) for benchmark years, dividing by population and the number of employed persons, and extrapolating to other years on the basis of relative changes in trend measures.

The benchmark PPPs are from the United Nations International Comparison Project (UNICP). This project has become regionalized and the PPPs for the OECD countries are calculated by the OECD and the Statistical Office of the European Communities (EUROSTAT). The benchmark figures are derived by comparing relative prices at detailed levels of expenditure (PPPs by item of expenditure) and aggregating these price relatives to derive overall PPPs for total GDP. The original volume measures of GDP from the benchmark studies are modified by BLS, where applicable, to account for subsequent revisions by countries of their national accounts. BLS also constructs PPPs for GDP for all years by applying relative changes in implicit price deflators for GDP, as measured by each country, to the benchmark PPP's.

Consumer prices

The consumer price indexes are not adjusted for comparability except to convert them to a uniform base year. National differences exist, for example, with respect to population coverage, frequency of market basket weight changes, and treatment of homeowner costs.

Background | Description of Methods | Data Sources | Estimating Procedures | Analysis and Presentation | Usage and Limitations

Last Modified Date: October 16, 2001

URL: http://www.bls.gov/FLS
Phone: (202) 691-5654
Fax: (202) 691-5679
FLS data questions: flshelp@bls.gov
Technical (web) questions: webmaster@bls.gov
Other comments: feedback@bls.gov