intrenational comparison of hours worked

Upload: paul-maposa

Post on 14-Apr-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/30/2019 Intrenational Comparison of Hours Worked

    1/29

    Monthly Labor Review May 2009

    Comparisons of Hours Worked

    International comparisons of hoursworked: an assessment of the statistics

    A study o 13 countries reveals that measures o hours workedbased on administrative sources are relatively low while measuresbased on establishment and labor orce surveys are relatively high;thus, although ever improving, these measures cannot yet be taken

    at ace value and are useul only or broad comparisons

    Susan E. Fleck

    Susan E. Fleck is Chie, Divi-sion o Major Sector Produc-tivity, Oce o Productivityand Technology, Bureau oLabor Statistics. This articlewas prepared when she wasa supervisor in the Divisiono International Labor Com-parisons. E-mail: [email protected]

    Public commentators, the press, andgovernments are interested in thehours people work. Work hours un-

    derpin productivity measures. Te numbero hours individuals work stimulates debateon the quality o lie in an internationalcontext: do some societies live to work

    while others work to live? Te dierencesin hours worked between countries uels dis-

    cussion o economic growth, employment,and unemployment. Any comparative meas-ure between countries, however, depends ona standardization o concepts, sources, andmethods. Measuring and comparing howmany hours people spend at work acrosscountries is not an exact science, despiterecent improvements in methodology anddata coverage.

    Te recommendation rom the Inter-national Labor Organization (ILO) is touse actual hours worked, including annual

    hours actually worked, as the basis or in-ternational comparisons. Te recommen-dation to include annual hours actually

    worked was part o an updated ILO resolu-tion regarding the measurement o workingtime that was adopted at the InternationalConerence o Labor Statisticians held inthe all o 2008. Background research on

    working time and hours worked carried outby international statistical organizations

    and national statistical agencies to prepare orthe conerence has contributed to a rich debateon hours worked.

    Tis article benets rom the recent exchangeo ideas leading up to the 2008 Conerence andlooks at two data sets on hours worked. Tebetter known o the two is the Organizationor Economic Cooperation and Development(OECD) data set on average annual hours ac-

    tually worked, or all employed persons, or 30countries, published in the annual OECDEm-ployment Outlook.1 Te second data set is theBureau o Labor Statistics (BLS) underlyinghours and employment data in the annual re-port, Gross Domestic Product per EmployedPerson, which presents an international com-parison o gross domestic product (GDP) perhour worked or 13 countries. Te OECD dataset provides an explicit measure o average an-nual hours worked, while the BLS data set pub-lishes total employment and hours, rom which

    a series or average annual hours worked can bederived. Both hours-worked data series comple-ment output and productivity data published bythe respective organizations.

    Whereas data users tend to look at thenumber o average hours worked per year

    when making comparisons between countries,both BLS and OECD caution that such com-parisons are prone to error and that the dataseries best describe changes over time. Tis

  • 7/30/2019 Intrenational Comparison of Hours Worked

    2/29

    Comparisons o Hours Worked

    Monthly Labor Review May 2009

    article provides some context and explanations or thedata user on why these comparisons are raught with di-culty. It considers how concepts, sources, and methodsused to construct hours-worked data series aect analy-ses o data levels and trends. Te dierences between the

    BLS and OECD data sets discussed here highlight a majortheme o the article, namely, that the estimate o aver-age annual hours actually worked per employed person is

    just thatan estimateand it may vary with the sourcesand methods used. Nonetheless, trends are similar. Finally,the article explains why small dierences in hours workedbetween countries have little meaning, whereas large di-erences are more likely to be meaningul.

    Te countries studied are the United States, Canada,Japan, South Korea, and nine European countries: Bel-gium, Denmark, France, Germany, the Netherlands, Nor-

    way, Spain, Sweden, and the United Kingdom. Both BLSand OECD data sets depend on a variety o data sourcesand concepts used to measure and estimate hours worked.

    Te 13 countries considered here represent a wide vari-ety o developed economies. Additional data used in thisarticle come rom special studies by the OECD and theILO, as well as rom studies by researchers and nationalstatistical agencies. When time series are used, data begin

    with 1980 where available. For both data sets, pre-1991data or Germany are estimated.

    Te analysis begins with an explanation o various con-cepts and sources underpinning hours worked and o their

    uses and limitations in preparing data series on average an-nual hours actually worked. Tis explanation establishes theramework or discussing methods o estimation o averageannual hours actually worked and or describing the BLSand OECD data sets, including breaks in series. Te levelsand trends or each country are compared with the use o arank von Neumann test, to show how trends can be similar,although levels dier. With this background, the historictrends in the two data series are compared over a quarter-century whenever data are available. Furthermore, changesin the labor market that inuence hours worked, such asan expansion o part-time and womens employment, also

    will be examined. A short overview o changes in laws andnorms helps put the trends in context. Comparisons aremade between sources or the same country and betweencountries using similar methodologies. Comparisons be-tween Japan and the United States and between Norwayand Sweden highlight discrepancies in levels due to di-erences in sources and methods. Te comparisons are in-tended to provide the data user with a better understandingo the interplay among concepts, sources, and methods andhow they aect the comparisons.

    Tere are a number o explanatory actors underlyingthe dierences in hours worked across countries, such asinstitutional, legal, and policy dierences. Only institu-tional and legal actors specic to the regulation o normalhours o work will be addressed in this article; the other

    actors are beyond the scope o the analysis. Furthermore,with the recent passage o the revised ILO resolution onworking time, the concepts underlying hours worked haveexpanded to provide more detail. Tis study was preparedbeore the 2008 ILO resolution on working time was -nalized and took eect; thus, the concepts presented arebased on the original ILO resolution.

    Hours of work: concepts and sources

    Concepts. Resolutions passed by the tripartite meetingo the International Conerence o Labor Statisticiansestablish recommendations or countries to develop data

    with enough similarities to be suitable or internationalcomparisons. Te October 1962 ILO Resolution con-cerning statistics o hours o work provides guidance onconcepts and measurement relating to hours o work andon a basic ramework or collecting and analyzing data onhours. Te resolution establishes three concepts o hourso work: normal hours o work, hours actually worked,and hours paid.2 Another concept oten used in data col-lection is usual hours o work. Note that hours workedreers to measured, or actual, hours, whereas hours o

    work reers to scheduled, or planned, hours.Te box on page 5 lists the components o workingtime, based on the 1962 resolution. Items 1 through 6comprise one or more o the hours concepts mentioned inthis article. Items 7 and 8 are generally accepted as hoursnot at work.

    Normal hours o workare the maximum number o hoursworked beyond which an employer must pay an overtimepremium. Tis concept is partially addressed in item 1 inthe box. Normal hours may be xed by legislation or es-tablished by collective-bargaining agreements, dependingon the country, industry, and occupation. Te vast majority

    o countries in the world have a normal workweek o 40 ormore hours. In the United States, the normal workweekis 40 hours. In Europe, the normal workweek is usuallyless than 40 hours and ranges widely by industry or oc-cupation both within and between countries. For example,earlier this decade, the normal workweek was 29 hoursor Volkswagen production workers in Germany, but nowit is 33 hours; in France, the normal workweek has been35 hours or almost all employees or the past 10 years;and in the Netherlands, the normal workweek can be as

  • 7/30/2019 Intrenational Comparison of Hours Worked

    3/29

    Monthly Labor Review May 2009

    many as 60 hours or some workers or short periods.3Some people call normal hours o work hypothetical, inthat they measure the idealwork schedule, not the observ-able work schedule. On a practical level, employers otenarrange work schedules to keep employees hours at orbelow the normal-hour threshold, in order to avoid pay-ing overtime wage rates. Data sources or normal hours

    o work are derived rom the aorementioned legislationand collective-bargaining agreements and cover predomi-nantly employees.

    Te concept o hours actually worked encompasses allhours spent working, including overtime hours and exclud-ing absences; these are items 1 through 5 in the box.4Teconcept excludes items 6 through 8that is, hours paidbut not worked, such as paid leave, paid public holidays,and paid sick leave, as well as meal breaks and commut-ing time. As part-time work has become more prevalent,

    workers hours are less than the normal workweek, butare still counted in item 1. Although not explicitly stated

    in the resolution, hours actually worked are commonlycounted as both paid and unpaid hours at work. Data onhours actually worked are collected rom household-basedsurveys, such as labor orce surveys and time-use surveys;establishment surveys report data using other hours con-cepts, which can be adjusted to an actual-hours concept.Hours actually worked usually are reported on a personbasis (but can be adjusted to a jobs basis), account or thetotal hours individuals work on all jobs in a given reer-ence period, and generally include both persons working

    part time and persons working ull time. Yearly estimatesusually are calculated to reect a ull-year worker (that is,someone who works throughout the year).

    Te hours paidconcept is described in the 1962 resolution,but is not identied as a concept amenable to internationalcomparison. Hours paid generally include items 1 through5 in the accompanying box and exclude unpaid overtime.

    Hours paid also include item 6: holidays, vacation, and sickleave. Depending on the terms o the employment contract,items 7 and 8meal breaks and commuting timealsomay be included in the hours-paid concept. Wide varia-tions across countries persist regarding how workers arepaid or holidays and nonwork time, particularly sick leave.

    Tese dierences are the primary reason that internationalcomparisons o hours paid are not made.

    Usual hours o work are not addressed in the 1962 ILOresolution on hours, but are included in the 2008 resolu-tion. Usual hours o work are hours that are typical o acertain length o time, such as a day, a week, or a month. 5

    Te concept encompasses the same components as hoursactually worked, but reers only to regularly scheduledhours. Data on usual hours o work commonly reer tothe usual work schedule during a week or month and aremost commonly collected rom household surveys. Someestablishment surveys collect data on contractual hours,

    which are usual hours o work expected to be ullled un-der individual employment agreements. Tese contractualhours are analogous to normal hours under collective-bar-gaining agreements.6

    1. Hours actually worked during normal periods o

    work.

    2. ime worked in addition to hours worked dur-ing normal periods o work and generally paid at higherrates than normal rates (overtime).

    3. ime spent at the workplace on work such aspreparation o the workplace, repairs and maintenance,preparation and cleaning o tools, and preparation oreceipts, timesheets, and reports.

    4. ime spent at the workplace waiting or standingby or such reasons as lack o work, breakdown o ma-

    Components of working time

    chinery, and accidents, or time spent at the workplace

    during which no work is done, but or which payment ismade under a guaranteed employment contract.

    5. ime corresponding to short rest periods at theworkplace, including tea and coee breaks.

    6. Hours paid or, but not worked, such as paid an-nual leave, paid public holidays, and paid sick leave.

    7. Meal breaks.

    8. ime spent on travel rom home to work and romwork to home.

  • 7/30/2019 Intrenational Comparison of Hours Worked

    4/29

    Comparisons o Hours Worked

    Monthly Labor Review May 2009

    Sources o hours data. A number o sources are used tocapture the hours concepts described in the previous sec-tion. For each hours concept, certain sources o data arepreerred over others because they provide a better meas-ure o the concept. In the context o creating a comparable

    international measure o average annual hours worked,each source has its benets and drawbacks. Te chie is-sues to address in determining the best concept and sourceo hours to use in estimating average annual hours workedare (1) how well the data collected capture the concept ohours actually worked and (2) what additional data sourc-es have to be used to create the annual estimate, becauseo either measurement issues or coverage issues. Te mainconcern is whether the source covers detailed industries,all types o workers, and the total economy.

    1. Administrative data sources. Data on normal hours owork are available through administrative data sources.Te primary purpose o such data oten is to manageprograms, not to collect statistics. Administrative data arecollected by social programs, ministries, or local, regional,and national governments. In addition to covering legisla-tion or collective-bargaining agreements on normal hours,administrative data may cover the use o public services(such as registering in employment ofces or being paidsick leave), labor code enorcement, or tax collection. Ad-ministrative data also provide inormation on hours not

    worked, particularly in countries where paid leave is cen-

    trally administered, such as Sweden and Norway.Te advantage o an administrative source or data onnormal hours is its potentially wide population coveragein those countries with large numbers o employees work-ing under collective-bargaining agreements. Europeancountries have high rates o union coverage and, in somecases, have passed legislation that extends the benetsagreed upon in collective-bargaining contracts to work-ers who are not union members.7 Tese countries collectlarge amounts o data in administrative databases becausethey have active social programs and wide-ranging laborregulations. Still, administrative data rom collective-bar-

    gaining agreements, though a common source o data onnormal hours or dierent occupations, industries, and re-gions, are not the only source: establishment surveys, suchas those conducted in France, also may provide inorma-tion on normal hours o work.

    O course, there are limitations on administrative dataas a source o inormation on hours. First, the wide rangeo administrative data on job or labor conditions thatprovides inormation on normal hours may exclude some

    workers, such as part-time workers, workers not covered

    by collective-bargaining agreements, and the sel-em-ployed. For example, in France, small and medium-sizedbusinesses together account or one-ourth o employees,but those employees are not subject to the general limita-tion o a normal 1,600-hour work year. Tus, i normal

    hours were to be the basis o an annual measure o hoursactually worked or all employed, the additional hours

    worked by employees in small and medium-sized busi-nesses would be excluded.8 Also, administrative data arecollected by job and not by person, so additional inorma-tion would be required to account or multiple jobholdersi hours worked were to be estimated by person.

    Because o limitations on concepts and data sourceso normal hours, estimates o annual hours worked basedon these sources are likely to be undercounted. Normalhours do not provide a total-economy measure o hours

    worked without adjustments that expand coverage toall employed persons and all industries. Te nature othe data sourcescollective-bargaining agreements andother sources o regulated normal hoursguaranteesthat overtime hours worked are not counted. Tus, es-timates o hours actually worked will be biased down-

    wards. As an example, some countries covered in the BLSand OECD data sets base their measure o average annualhours worked on normal hours and deduct all paid annualleave and allowable sick leave. Tis estimation techniqueundercounts hours.

    2. Survey-based data. Survey-based data have an ad-vantage over administrative data covering normal hourso work, in that surveys provide reports o hours actually

    worked by individuals and count persons employed or jobs.Data are reported rom either individuals or businesses ontheir actual labor market behavior, not on their expectedbehavior. Labor orce surveys collect data on weekly ordaily actual or usual hours worked (or both). Establish-ment surveys generally collect either weekly or monthlyhours data on an hours-paid concept. Advantages andlimitations exist with the data provided by each o thesetypes o surveys.

    a. Household surveys. Data on actual or usual hoursworked are collected rom household surveys such as labororce surveys and time-use surveys, the latter being moreirregular and with a smaller sample size. Data on hoursactually worked and usual hours o work are reported onan employed-person basis and account or the total hoursindividuals, including both ull- and part-time workers,

    work on all jobs in the reerence period.Te two major advantages o labor orce survey data

  • 7/30/2019 Intrenational Comparison of Hours Worked

    5/29

    Monthly Labor Review May 2009

    are the ability to report hours actually worked, includingpaid and unpaid overtime, and the broad coverage o theemployed. Te concept o hours actually worked capturesthe variability and irregularity o the number o hours aperson works and does not work in a given week or other

    period, and it can account or shortened workweeks, over-time hours, holidays, sick leave, and vacation. O course,the concept o usual hours o work also captures paid andunpaid overtime, as long as the overtime hours are a regu-lar part o the work schedule. Te problem is that usualhours o work do not uctuate as much as hours actually

    worked and do not capture that variability, because theyexclude irregular hours not worked, irregular overtime,and short-time work (temporary reductions in the regular

    workweek). Regarding coverage o the employed, the na-ture o a labor orce survey is to reach into all households

    with all types o workers. Tus, labor orce surveys providecoverage o the sel-employed and unpaid amily workers,both o whom are excluded in data on normal hours o

    work.Tere are a couple o limitations, however, to using labor

    orce survey data or comparisons o hours worked. First,data collection that is not ongoing (that is, discontinuousdata collection) can aect the accuracy o data on bothhours actually worked and hours not worked. Because othis problem, European Union member countries recentlyhave moved toward ongoing data collection; hence, theirestimates o average annual hours actually worked are

    based on 52 weeks o the year. But most other developedcountries collect data on a discontinuous, albeit regular,basis. By its nature, discontinuous data collection, such asone week a month or one week a quarter, does not accountor unexpected irregularities in hours worked and hoursnot workedor example, hours not worked on holidays,in bad weather, or because o school closings. Adjustmentsare made to account or hours not worked, but these ad-

    justments themselves are variable across countries, withina country, and across years, as well as by region or even oc-cupation and industry. It is likely that, as labor orce sur-

    veys in the European Union and elsewhere expand cover-

    age to all months o the year and all weeks o a month, andas questions and data collection on hours actually workedand hours not worked become more precise, some o theseinconsistencies will diminish.

    A second common concern regarding labor orce surveysis the issue o reliability. Labor orce surveys depend onrespondent recall and proxy responses; accordingly, surveyrespondents oten do not reliably report their own hours

    worked and hours not worked, because they are relying onaulty memory, and neither do proxy respondents report

    such hours reliably, because they lack inormation aboutthe intended respondent. In essence, in a labor orce surveyhours actually worked are not observed, but are reported,and people can orget the hours they actually worked.

    Nonetheless, past concerns over respondent error in

    labor orce surveys seem to be less o a problem thanpreviously thought.9 Te advent o time-use surveys hasled to research that sheds light on comparisons betweenshort-term recall o hours worked and longer term recallused in household surveys. For example, comparisons be-tween the 1998 Canadian Labor Force Survey and imeUse Survey ound that, overall, average numbers o hours

    worked are similar between the two surveys.10 One U.S.study showed that time-use survey responses accuratelyreect hours worked when the data are collected in ornear the reerence period, but that hours are reported at alevel 5 percent lower when data are collected during later

    weeks.11 Concerns remain over proxy responses.Finally, a more theoretical concern regarding the use o

    hours data rom labor orce surveys in productivity com-parisons is the need to convert the data rom a nationaleconomy concept to a domestic economy concept consist-ent with national accounts measures.12 In small countries,such as Belgium, where residents cross national bordersto work, employment data rom the household, or labororce, survey may not be a corresponding measure o thoseemployed in a countrys production o output, thus aect-ing the corresponding hours measure.

    b. Establishment surveys. Data on hours paid are col-lected rom establishment surveys. Te purpose o suchsurveys is to collect data on hours, earnings, number oemployees, compensation, and other labor characteristicso rms and their workers. Establishment surveys have atleast three advantages. First, the data are deemed reliable,because they are extracted rom payroll inormation andare considered more precise than data based on individualrecall.13 Second, industry coverage and classication alsoare deemed reliable. Tis is because establishment surveydata oten are collected at a detailed industry level, gener-

    ally complement national accounts output data, and thusalso complement industry productivity analysis. Finally, insome countries, such as the United States, establishmentsurvey sample rames are much larger and cover ar more

    workers than labor orce surveys can cover.Te limitations on establishment survey data or hours

    measures are at least ourold. First, the concept o hourspaid typically does not report hours actually worked.Rather, it includes hours paid and worked, such as theregular workweek and paid overtime; and hours paid, but

  • 7/30/2019 Intrenational Comparison of Hours Worked

    6/29

    Comparisons o Hours Worked

    Monthly Labor Review May 2009

    not worked, such as paid vacation, sick leave, and ma-ternity leave. Second, both the practice and reporting othe collection o data on hours paid dier widely acrosscountries, making comparisons difcult. In some coun-tries, such as Norway, benets or sick leave or maternity

    leave are paid by a government or a union, so the hours-paid data rom establishment survey sources exclude thesebenets; in other countries, such as the United States, paidsick leave is a benet oered by many employers, so it iscounted as hours paid. It is difcult to account or thesedierences in creating comparative measures o hourspaid between countries. Tird, survey coverage is limitedto employees, and only to certain types o employees. His-torically, establishment survey data have been collected onproduction workers and have excluded supervisory, tem-porary, or part-time employees. Only in the recent pasthave establishment surveys expanded their coverage toinclude supervisory employees. Needless to say, data onsel-employed and unpaid amily workers must be oundto complement establishment survey data on employees.Fourth, in establishment surveys, industry coverage, al-though complementary to data ound in national accounts,may not be representative o all industries. Te ocus odata collection by establishment surveys always has beenthe manuacturing sector, although countries have beenexpanding coverage to include the service sector.

    Without adjustment, hours-paid data rom establish-ment surveys do not provide a total-economy measure o

    hours actually worked that covers all employed persons inall industries. Depending on the adjustment, the estimatemay over- or underestimate hours actually worked: on theone hand, hours-paid data that are not adjusted or paidleave will overstate the estimate o hours actually worked;on the other hand, hours-paid data that are adjusted tothe hours-worked concept by means o administrative orlegislative leave data may understate hours worked i theadjustment assumes that employees take all leave that isoered them.

    Tese concepts and sources o hours worked are thebuilding blocks or the analysis in the next section, which

    addresses issues related to constructing a series o averageannual hours actually worked and examines two data setsrom the BLS and the OECD.

    Estimating and comparing hours actually worked

    In recent years, statistical reporting and measurementhave ocused on how to create comparable series o aver-age annual hours actually worked. Te reasons are two-old. First, i hours worked are to be used as a compara-

    tive quality-o-lie indicator, they are best measured overa year, to reect vacation time and other absences rom

    work. Second, demand has grown or measures o annualhours in order to estimate an economys total productivity.Average annual hours actually worked per capita provides

    a broad measure o labor utilization, broken down intothree components in a recent OECD study: the intensive,or individual, component o average annual hours actually

    worked per employed person, the extensive, or econo-mywide, component o the employment-population ratio,and a demographic actor.14 Unless otherwise stated, therest o this article considers instead the narrower, inten-sive, measure oaverage annual hours actually worked peremployed personthat is, the hours o labor that workersactually put in on the job.

    In 2003, the 17th general report by the InternationalConerence o Labor Statisticians highlighted the need torevise existing international recommendations on hoursactually worked during short as well as longer reerenceperiods and suggested that such measures be broad-ened to cover all persons in employment, including thesel-employed, by extending the content o each o thedening categories o working time to include all worksituations, such as irregular, seasonal, work at home, andunpaid work.15 Furthermore, the report suggests the de-

    velopment o an international denition oannual hourso work that allows or alternative estimation proceduresthat take into account variations in the type and range o

    national statistics o working time.16

    Tis section looks at the methodologies used to preparemeasures o average annual hours actually worked per em-ployed person and the data sets underlying the publishedmeasures. Te analysis begins with an overview o theconcepts and sources used in the BLS and OECD data sets,ollowed by a comparison o dierences in the estimateso average annual hours actually worked per employedperson in each data set, or each o the 13 countries exam-ined. A statistical test comparing trends between the twodata sets shows that the trends diverge or only 3 o the13 countries examined: the United States, France, and the

    Netherlands. Te analysis undertaken supports the per-spective o the statistical organizations that hours data arebest analyzed as trends and not as levels.

    Data sources and country methodology. As countries movetoward adopting a national accounts ramework to meas-ure labor input, or hours worked, concepts across countriesare becoming more consistent. It is the source o data andthe methodology used, rather than the concepts employed,that are at the heart o the comparability issue.

  • 7/30/2019 Intrenational Comparison of Hours Worked

    7/29

    Monthly Labor Review May 2009 9

    As Gerard Ypma and Bart van Ark attest in their2006 analysis o the OECD/Eurostat country survey onemployment and hours or national accounts, a countrysdata sources and data priorities determine the methodol-ogy that the country uses to prepare an estimate o hours,

    employment, and, eventually, average annual hours actuallyworked per employed person. Te direct method o estima-tion is based on sources that capture hours actually worked,

    whereas the component method is used to convert normal,paid, or usual hours worked to an hours-actually-workedconcept.17

    Exhibits 1 and 2 together provide a snapshot o the BLSand OECD data sets through 2006, the concept o hours,the sources o hours and employment, andwhere inor-mation was availablethe adjustments to concepts madeor each data set.18 Ypma and van Arks analysis gives detail

    where inormation is lacking. Te general term nationalaccounts concept o hours worked reers to the 1993 Sys-tem o National Accounts measure o labor inputs, which inturn reers to the ILO resolution on hours actually worked.19Individual countries may adopt measures that include anynumber o original sources and related concepts o hoursand employment, and, as necessary, may subsequently ad-

    just them to expand coverage to all employed persons, toconvert measures o paid, normal, or usual hours to hoursactually worked, or to include industrial sectors that areotherwise excluded rom a survey.20

    An important detail o the two tables is the unit o meas-

    ure o hours. Whether that unit o measurethat is, the aver-age annual hours actually workedis applied per employedperson, per job, or on the basis o ull-time equivalentscre-ates dierences between levels o data. Because one personcan hold more than one job, the average hours worked peremployed person will be greater than the average hours

    worked per job. Te concept o ull-time equivalent work-ers consolidates hours worked by part-time workers into ameasure o hours that approximates the hours worked bya ull-time employed person working a normal workweek.Average annual hours actually worked per ull-time equiva-lent worker will be greater than average annual hours actu-

    ally worked per employed person. Average annual hours ac-tually worked per employee are estimated when data or thesel-employed are not available or are difcult to integrateinto the calculations. Average annual hours actually workedper employee are generally lower than those per employedperson, because the sel-employed work longer hours thanemployees. Tis comparison o two data sets highlights howresults dier, even or the same country, i a dierent sourceo data or unit o measure is used. Eight o the 13 countrieshave major dierences in their data sources or methods.

    BLSdata set. In the ace o continued interest in broadmeasures o productivity based on hours worked, a 2007BLS report began to publish international comparisons oGDP per hour worked, as well as GDP per employed per-son.21 Te underlying data on total hours and total employ-

    ment are collected rom national sources, where available.Te report covers 16 countries, but data on hours workedcover only 13 o the 16, all 13 o which are discussed inthis article. Eorts are being made to extend coverage toAustralia as well. Data or Germany have a break in 1991;data or earlier years are estimates based on the ormer

    West Germanys hours and employment. Other breaks inseries include a 1997 break or Canada due to changes inclassication. Te years covered or Japan and the Nether-lands begin at 1996 and 1995, respectively.

    Sources and concepts o data on hours are available indetail only or some countries. Te BLS report publishesan aggregate, rather than average, measure o annual hours

    worked. Te underlying source data used to calculate av-erage annual hours actually worked in the BLS data setare most commonly total-hours-worked measures, avail-able rom national accounts, and total employment mea-sures, usually estimated rom national labor orce surveysor available rom national accounts. Data series or threecountriesJapan, South Korea, and Belgiumare publishedas average hours worked. Japan and Belgium publish averageannual hours worked in the national accounts and OECDproductivity database, respectively.22 South Koreas average

    annual hours worked are calculated rom average weeklyhours worked, based on the labor orce survey. Four othercountries hours-worked data are derived partially rom labororce surveys. For the United Kingdom, total hours are basedon labor orce survey data whereas total employment comesrom national accounts. For the United States, Canada, andthe Netherlands, labor orce surveys are the source o totalemployment data, adjusted, where necessary, to account orthe Armed Forces. otal hours data or the United States andCanada are based on establishment and labor orce surveys.

    Te source o data or the remaining countries is total hoursworked and employment based on national accounts.

    O the countries included in the BLS series, the aver-age hours worked are on an employed-person basis orall but Japan, Norway, Spain, and the United Kingdom.Data on hours worked or Japan reer to employees andexclude the sel-employed. Data or Norway are on a ull-time equivalent basis, and data or Spain and the UnitedKingdom are on a jobs basis.

    OECD data set. Once a year, the OECD Employment Outlookpublishes data on average annual hours actually worked per

  • 7/30/2019 Intrenational Comparison of Hours Worked

    8/29

    Comparisons o Hours Worked

    10 Monthly Labor Review May 2009

    United States 1950 None Establishment Labor orce Hours paid, Labor orce Data on Divide total Persurvey survey with survey Armed Forces hours by total employed

    adjustment employment personto hoursworked

    Canada 1961 1997, Labor orce Establishment National Labor orce No more Divide total PerNAICS survey survey accounts survey known hours by total employed

    sources employment personJapan 1996 None National No more National No inorma- No inorma- No inormation Per

    accounts known accounts tion available tion available available employeesources

    South Korea 1980 None Labor orce No more Average No inorma- No inorma- Average weekly Persurvey known hours worked, tion available tion available hours 52 employed

    sources by week person

    Belgium 1970 None Administra- No more National No inorma- No inorma- No inorma- Pertive data known accounts tion available tion available tion available employed

    sources person

    Denmark 1966 None National Administra- National National No more Divide total Peraccounts tive data accounts, accounts known hours by total employed

    based on sources employment personnormal hours

    France 1970 None National No more National National No more Divide total Peraccounts known accounts accounts known hours by total employed

    sources sources employment person

    Germany 1960 1991 National No more National National No more Divide total Peraccounts known accounts accounts known hours by total employed

    sources sources employment person

    Netherlands 1995 None National No more Volume o Labor orce Data on Divide total Peraccounts known person-hours survey Armed Forces hours by total employed

    sources worked employment person

    Norway 1970 None National No more Man-hours National No more Divide total Full-timeaccounts known accounts known man-hours equivalent

    sources sources worked by totalemployment

    Spain 1979 None National No more No inorma- National No more Divide total Per jobaccounts known tion available accounts known hours by total

    sources sources jobs

    Sweden 1980 None National No more No inorma- National No more Divide total Peraccounts known tion available accounts known hours by total employed

    sources sources employment person

    UnitedKingdom 1971 None Labor orce No more No inorma- National No more Divide total Per job

    survey known tion available accounts known hours by totalsources sources jobs

    1 Te national accounts concept o hours worked is hours actually worked, unless otherwise noted.

    Exhibit 1. BLS concepts, sources, and methods, 13 countries

    Methodologyused to create

    average annualhours actually

    worked

    Country Beginning

    year

    Breaks in

    series

    Primarysource ofdata on

    total hoursworked

    Othersources

    of data ontotal hours

    worked

    Hoursconcept

    used insource data1

    Primarysource

    of data ontotal

    employment

    Othersources

    of data ontotal

    empoyment

    Unit ofmeasure

  • 7/30/2019 Intrenational Comparison of Hours Worked

    9/29

    Monthly Labor Review May 2009 11

    United States 1950 None Establishment Labor orce Hours paid, Establishment Labor orce (otal hours/ Persurvey survey with adjust- survey survey total employed

    ment tohours worked

    Canada 1961 1997, Labor orce Establishment National No inorma- No inorma- Direct measure Per jobNAICS survey survey accounts tion available tion available o average actual

    hours worked,with

    adjustments orweeks not covered

    and holidays

    Japan 1970 None Establishment Labor orce Hours Establishment Labor orce OECD Per jobsurvey survey worked survey survey estimates

    South Korea 1980 None National No other National National No more OECD Peraccounts known accounts accounts, known estimates employedbased on sources based on sources person

    labor orce labor orcesurvey survey

    Belgium 1983 None Labor orce Administrative Usual hours No inorma- No inorma- OECD Persurvey data worked tion available tion available estimate, employed

    accounting or person

    underreportingo time notworked and

    public holidays

    Denmark 1970 None National Administrative National National No more OECD Peraccounts data accounts accounts known estimates employed

    sources person

    France 1970 None Administrative Establishment National No inorma- No inorma- French Perdata and labor accounts, tion available tion available national employed

    orce surveys based on accounts personhours oered

    Germany 1991 1991 data Administrative Labor orce National No inorma- No inorma- German Perseries begin data survey accounts, tion available tion available national employed

    based on accounts personnormal hours

    Netherlands 1987 2002, 2003, Labor orce Administrative Usual hours No inorma- No inorma- OECD Per OECD survey data worked tion available tion available estimate, employed

    estimates accounting or personunderreporting

    o time notworked and

    public holidays

    Exhibit 2. OECD concepts, sources, and methods, 13 countries

    Methodologyused to create

    average annualhours actually

    worked

    Country Beginning

    year

    Breaks in

    series

    Primarysource ofdata on

    total hoursworked

    Othersources

    of data ontotal hours

    worked

    Hoursconcept

    used insource data1

    Primarysource

    of data ontotal

    employment

    Othersources

    of data ontotal

    empoyment

    Unit of

    measure

    personemployment) multiple-

    jobholder rate

  • 7/30/2019 Intrenational Comparison of Hours Worked

    10/29

    Comparisons o Hours Worked

    12 Monthly Labor Review May 2009

    employed person. Te data are based on the OECD produc-tivity database. Data on hours worked are converted, wherenecessary and possible, to employed persons rom jobs.Some data or theEmployment Outlookhours series are basedon sources that dier rom the productivity database. TeOECD data set covers 30 countries and provides estimates oaverage annual hours actually worked per employed person(that is, all those employed, including the sel-employed andunpaid amily workers) and per employee (that is, excludingthe sel-employed and unpaid amily workers).23 Te years

    covered or Belgium and the Netherlands begin at 1983 and1987, respectively.Compared with the BLS data set, the OECD data set

    provides slightly more metadata, because the organizationcollects and processes a questionnaire on national accountsrom national statistical agencies o member countries. Tehours concept used with the OECD data set is consistent

    with national accounts or 7 o the 13 countries in the dataset. (See exhibit 2.) Te countries or which data sourcesare derived not solely rom national accounts include the

    United States, Japan, Belgium, the Netherlands, Spain, andthe United Kingdom. For the United States, both hours andemployment are taken rom the BLS major sector productiv-ity measures. Data or Japan are measured primarily by anestablishment survey and are OECD estimates. Estimates oaverage annual hours actually worked or Belgium and theNetherlands are developed rom the European Union labororce survey, using usual hours o work and adjusting orhours not worked. Data or Spain are based on hours actu-ally worked, as well as usual hours o work or those deemed

    not at work in the labor orce survey. Te data or the UnitedKingdom are based completely on the labor orce survey, butare compatible with national accounts concepts.

    More inormation on the OECD data set is availablerom Ypma and van Arks analysis o 2004 hours-workeddata based on the OECD/European Union national ac-counts questionnaire.24 South Korea and the United King-dom are the only two countries or which the dara sourceis solely the labor orce survey. Te United States, Canada,and Japan are categorized as using primarily survey (both

    Norway 1962 None Establishment Labor orce National No inorma- No inorma- Norwegian Full-timesurvey survey and accounts tion available tion available national equivalents

    administrative accountsdata

    Spain 1977 1987, Labor orce Establishment Actual and No inorma- No inorma- Spanish Full-timechange in survey survey usual hours tion available tion available statistical equivalents

    survey worked institute

    Sweden 1950 1996, Labor orce Establishment National No inorma- No inorma- Swedish Perchange in survey survey accounts tion available tion available national employeddata source accounts person

    UnitedKingdom 1970 1984, 1992, Labor orce No more Actual hours Labor orce No more Average hours Per

    change in survey known worked survey known actually employeddata source; sources sources worked 52 person

    1994,include

    NorthernIreland;1995,

    change in

    1 Te national accounts concept o hours worked is hours actually worked, unless otherwise noted.

    Exhibit 2. ContinuedOECD concepts, sources, and methods, 13 countries

    Methodologyused to create

    average annualhours actually

    worked

    Country Beginning

    year

    Breaks in

    series

    Primarysource ofdata on

    total hoursworked

    Othersources

    of data ontotal hours

    worked

    Hoursconcept

    used insource data1

    Primarysource

    of data ontotal

    employment

    Othersources

    of data ontotal

    empoyment

    Unit of

    measure

    method

  • 7/30/2019 Intrenational Comparison of Hours Worked

    11/29

    Monthly Labor Review May 2009 1

    labor orce and establishment) data and not administra-tive data. Te third category is split between the countriesthat use survey data more than administrative datasuchas Norway, Spain, and Swedenand those which use pri-marily administrative data, supplemented by labor orce and

    establishment survey datasuch as Denmark, France, andGermany. For Belgium and the Netherlands, OECD pre-pares an estimate o average hours actually worked basedon the labor orce survey.

    Comparison acrossBLSandOECD data sets. Te next sectioncompares the data on average annual hours actually workedper employed person between the BLS and OECD datasets.25 In preparation or that analysis, note that dierencesin data arise because o dierences in sources, concepts,coverage, and units o measure. For Denmark, France or19902002, Germany or 1991 onward, Norway, and Swe-den, data sources in each data set are the same. For Canada,

    Japan, South Korea, Denmark, and the Netherlands, aver-age hours are higher in the BLS data set than in the OECDdata set. For France in earlier years, and or Belgium andSpain, the OECD estimates are higher than the BLS esti-mates. For the United States, Germany in earlier years, andthe United Kingdom, average annual hours worked are notconsistently higher or lower in either data set.

    Te dierences between the data sets or the UnitedStates and Japan are difcult to pinpoint, given that cover-age, sources, and methodology dier between data sets or

    both countries. Dierences in units o measure aect thedierent levels among the data sets or Canada, Spain, andthe United Kingdom. For Belgium, South Korea, and theNetherlands, the contrast between the BLS and OECD datasets or each country is due to the source o the data: admin-istrative or survey based; the administrative-data adjustmentor time not worked aects comparisons or two o the threecountries, and the use o normal hours aects the third.

    Te country-by-country comparison to be presentedhighlights how data sources, measurement methods, andunits o measure matter. Te dierences can be catego-rized as ollows:

    Administrative sources reporting normal hours owork result in lower estimates o average annualhours actually worked than do data rom surveys.

    Among surveys, data that are primarily rom establish-ment surveys using usual hours or paid hours workedproduce lower estimates than do data that are primarilyrom labor orce surveys; data rom labor orce surveysmay overstate hours reported, due to proxy reporting.

    Adjustments to exclude hours not worked may over-

    estimate time not worked and lower estimates ohours worked.

    Units o measurement can aect the levels o hoursworked that are reported.

    1. More similar than diferent: Denmark, France, Germany,Norway, and Sweden. Te Nordic countries covered, aswell as Germany, and France or some years, use the samedata source in both the BLS and OECD data sets and dieronly slightly or not at all across years. For Denmark, averageannual hours actually worked or both data sets are rom thecountrys national accounts and run parallel to each other.In 1980, average annual hours per employed person wereabout 1,650 or both data sets; by 2006, they had allen to1,577 (OECD) and 1,608 (BLS). (See chart 1, top panel.)

    Te 30-hour dierence between data sources is likely dueto dierences in rounding or method o calculation.

    For France, the source or both data sets is the Frenchnational accounts. From 1980 to 1989, dierences are notlarge, averaging about 2 to 4 days a year or any given year.(See chart 1, middle panel.) Te two data sets yield identi-cal results or 19902002 and diverge only minimally or200306. Te BLS methodology o linking data rom di-erent sources with similar concepts or the period beore1990 creates slight dierences between the two data sets.

    For Germany, both data sets use that countrys nationalaccounts rom 1991 orward. Te data sources are identi-cal, and so are the series on average annual hours actually

    worked. Hours worked in 2006 were among the lowest thatyear o all the countries studied. Te 1,436 average annualhours worked is the equivalent o working 36-hour weeksonly 9 months o the year. (See chart 1, bottom panel.)

    For both Norway and Sweden, national accounts datawere used to prepare estimates o hours worked or both datasets. Nonetheless, the sources o the two countries dataad-ministrative sources and the labor orce surveycreate theappearance that there are large dierences in the Norwegianand Swedish labor markets when hours measures are com-pared. In Norway, hours worked were listed as 1,580 in 1980and had allen to 1,400 by 2006. (See chart 2, top panel.) In

    Sweden, hours worked were near 1,500 in 1980 and 1981;peaked in 1999; returned to 1,580, an increase equivalent to2 weeks o work, by 2002; and mostly held steady since then,coming in at 1,583 in 2006. (See chart 2, bottom panel.)

    Tis dierence between Sweden and Norway will be exam-ined more careully in the next section.

    2. United States and Japan: countervailing diferences. Tedata sets dier or the United States and Japan. Te di-erences, however, are so varied that it is difcult to pin-

  • 7/30/2019 Intrenational Comparison of Hours Worked

    12/29

    Comparisons o Hours Worked

    1 Monthly Labor Review May 2009

    Chart 1. Average annual hours actually worked, all employed persons, Denmark, France, and Germany,19802006

    OECD

    BLS

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    Hoursworked

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    Denmark

    Germany

    1,900

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    OECD

    BLS

    France1,900

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    OECD

    BLS

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    Hoursworked

    Hoursworked

    Hoursworked

    Hoursworked

    Hoursworked

  • 7/30/2019 Intrenational Comparison of Hours Worked

    13/29

    Monthly Labor Review May 2009 1

    Chart 2. Average annual hours actually worked, all employed persons, Norway and Sweden, 19802006

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    Norway 1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    BLS, OECD1

    Hoursworked

    Hoursworked

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    BLS, OECD1

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    Sweden

    1 For Norway, BLS and OECD data are identical or every year except 199 and 1999, or which they difer by 1 hour. For Sweden, BLSand OECDdata are identical.

    Hoursworked

    Hoursworked

  • 7/30/2019 Intrenational Comparison of Hours Worked

    14/29

    Comparisons o Hours Worked

    1 Monthly Labor Review May 2009

    point how they might produce dierences in time trends.U.S. estimates o hours are produced by the BLS Divisiono Major Sector Productivity and are based on hours-paid data rom an establishment survey on production

    workers, adjusted to an hours-worked measure by means

    o the labor cost index and urther adjusted to accountor industries and categories o workers not otherwise in-cluded, as well as sel-employed and unpaid amily work-ers, based on the U.S. Current Population Survey.26 Teestimates cover the total economy. Te OECD uses aggre-gate employment data based on the same methodologyto create a data series o average annual hours actually

    worked and then adjusts rom a jobs to an employed-person basis. Te BLS, by contrast, uses employment datarom the national labor orce survey, adjusted to includemilitary employment. Te dierences between the levelso hours published in the OECD and BLS data sets reectthe historically dierent trends in U.S. employment asmeasured by establishment and labor orce surveys. Teoverall dierence between the two data sets lies in thesource o employment data and the underlying dier-ences between the two surveys.27

    In the case o Japan, the OECD series on average hoursactually worked is estimated rom Japans establishmentsurvey or employees and includes labor orce survey dataon the sel-employed. Te BLS data set is based on thenational accounts data or employees rom 1997 onward.Using the categories o dierences outlined earlier, labor

    orce survey data are expected to produce higher ratesthan national accounts data based on administrative orestablishment survey data. But or Japan, the OECD hoursseries based on the labor orce survey is lower, on aver-age, than the BLS hours series based on national accounts.Further complicating matters is the act that hours or allthe employed would be expected to be lower than hoursor employees, given the nature o sel-employment.However, that expectation is not borne out in the case othe two data sets on Japan: the employee data rom thenational accounts trend higherthan the OECD data on allemployed persons rom the labor orce. Only in the case

    o units o measure does the direction o the dierencehold. Data on hours worked are on a per-job basis or theOECD and a per-person basis or BLS. Tis is the only oneo three dierences that explains why hours-worked dataare higher or the BLS data set. Chart 3 shows the averageannual hours actually worked by all employed persons, orthe United States and Japan.

    3. Canada, Spain, and the United Kingdom: units o meas-ure matter. In these three countries, the unit o measure,

    among other things, drives the dierences between thedata sets. For Canada, the BLS data series is based on ameasure o hours per employed person, whereas the OECDdata series is based on a measure o hours per job. Allother things being equal, average hours actually worked

    per employed person are higher than average hours actu-ally worked per job. Also or Canada, the two data setsuse the same source or hours-worked data, but dierentsources or employment data. Te source oOECD data isthe Canadian national accounts, which combine establish-ment and labor orce survey data; by contrast, the sourceoBLS data is an employment series or employed personsrom the labor orce survey. Te BLS gure is higher orall years, partly because o the dierence in sources andpartly because the unit o measurement is employed per-sons rather than jobs.

    For Spain, the BLS hours series draws rom nationalaccounts data based partially on the countrys labor orcesurvey and reported on a per-job basis. Te OECD dataset uses a data series estimated by the national statisticalinstitute, is based on actual and usual hours rom the la-bor orce survey, and adopts a ull-time-equivalent unit omeasure. Tese dierences create two nearly parallel dataseries, with the BLS series, on the per-job basis, at a lowerlevel than the OECD series. ogether, the source and theunit o measure or Spain explain why the BLS data setshows lower levels than the OECD data set.

    For the United Kingdom, the BLS and OECD data sets

    each use that countrys labor orce survey data on hours ac-tually worked. Te source o data on average hours workedper person is the same, but the source o data on employ-ment diers. Te BLS data source or employment is a na-tional accounts data series o aggregate jobs that combinesdata rom both establishment and labor orce surveys. Teemployment source or the OECD data series is solely thelabor orce survey, measured on an employed-person basis.

    Without more detailed inormation on the national ac-counts methodology, it is difcult to determine the extentto which the establishment survey data may aect thehours-worked measure. Te unit o measure does explain

    the dierence in the two trends: the trend is lower or theBLS series, which is based on jobs, than it is or the OECDseries, which is based on employed persons. Chart 4 showsthe average annual hours actually worked by all employedpersons, or Canada, Spain, and the United Kingdom.

    4. Belgium, South Korea, and the Netherlands: normal hoursand time not worked. Te inclusion o normal hours basedon administrative data to estimate time worked and to ad-

    just or time not worked also drives dierences between

  • 7/30/2019 Intrenational Comparison of Hours Worked

    15/29

    Monthly Labor Review May 2009 1

    Chart 3. Average annual hours actually worked, all employed persons, United States and Japan, 19802006

    1,900

    1,0

    1,0

    1,0

    1,20

    1,00

    1,0

    1,0

    1,0190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    OECD

    BLS

    1,900

    1,0

    1,0

    1,0

    1,20

    1,00

    1,0

    1,0

    1,0

    United States

    2,200

    2,100

    2,000

    1,900

    1,00

    1,00

    1,00190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    OECD

    BLS

    2,200

    2,100

    2,000

    1,900

    1,00

    1,00

    1,00

    Japan

    Hoursworked

    Hoursworked

    Hoursworked

    Hoursworked

  • 7/30/2019 Intrenational Comparison of Hours Worked

    16/29

    Comparisons o Hours Worked

    1 Monthly Labor Review May 2009

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    Chart 4. Average annual hours actually worked, all employed persons, Canada, Spain, and United Kingdom,19802006

    OECD

    BLSCanada

    1,0

    1,20

    1,00

    1,0

    1,0

    1,0

    1,20

    1,00

    1,0190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    1,0

    1,20

    1,00

    1,0

    1,0

    1,0

    1,20

    1,00

    1,0

    Spain2,200

    2,000

    1,00

    1,00

    1,00

    1,200

    1,000190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    OECD

    BLS

    2,200

    2,000

    1,00

    1,00

    1,00

    1,200

    1,000

    OECD

    BLS

    United Kingdom

    Hoursworked

    Hoursworked

    Hoursworked

    Hoursworked

    Hoursworked

    Hoursworked

  • 7/30/2019 Intrenational Comparison of Hours Worked

    17/29

    Monthly Labor Review May 2009 19

    data sets. Te BLS and OECD data sets show dierenttime trends or Belgium, South Korea, and the Nether-lands. Upon analysis, the BLS data series based on normalhours present a lower trend in hours worked, as in thecase o Belgium. For South Korea and the Netherlands,

    the OECD adjustments to time not worked, using normalor administrative data, create an hours-worked series thataverages 1 to 3 weeks less than the BLS series or bothcountries (except South Korea in earlier years).

    For Belgium, BLS uses the average-hours-worked seriesrom the OECD productivity database, which diers romthe OECD data set based on the Employment Outlook.

    Tese data or Belgium are based on administrative data,according to Ypma and van Ark.28 Te OECD data set, bycontrast, uses the labor orce survey to create an estimateo hours worked. Te tendency o administrative data toproduce lower estimates, by undercounting overtime andoverestimating leave time taken, explains the lower num-bers in the BLS data set or Belgiums hours relative to thenumbers in the OECD data.

    In the case o South Korea, the OECD and BLS data se-ries both use the labor orce survey as their primary sourceo data. On the one hand, the OECD estimates or SouthKorea are based on that nations labor orce survey and in-clude an adjustment downward to aggregate hours workedin the year, in order to account or time not worked, beoredividing by employment. On the other hand, the BLS es-timates or South Korea are based on published data on

    average weekly hours worked or persons at work. Te av-erage is multiplied by 52 to create a yearly average, and noadjustments are made or time not worked. Te OECDsadditional adjustment or time not worked contributes toa lower estimate o average annual hours actually workedcompared with the BLS estimate, even though the OECDunit o measure takes account o all those who are em-ployed, as opposed to the BLS employee measure.

    For the Netherlands, aggregate hours data or the BLSdata set are based on the Dutch national accounts hours-

    worked data series and employment is rom the labororce survey, adjusted to include the Armed Forces. Te

    OECD data sets estimate o average annual hours actu-ally worked is based on the labor orce surveys gure orusual hours o work and includes adjustments to time not

    worked. Te dierent sources provide dierent data series.For 2006, OECD reports 1,391 average annual hours ac-tually workedabout 2 person-weeks less than the BLSseries gure. One would expect that labor orce surveydata would produce a higher average-hours-worked se-ries. However, iOECDs adjustments to time not workedoverestimate the hours not worked, then the number o

    hours worked will be underestimated. Tis would explainthe act that data rom the BLS hours-worked series yieldhigher numbers than do data rom the OECD series basedon the labor orce survey. Chart 5 shows the average an-nual hours actually worked by all employed persons, or

    Belgium, South Korea, and the Netherlands.Both the BLS and the OECD suggest that the data

    user compare the trends over time between countries.A rank von Neumann test comparing the dierences inlevel data between the BLS and OECD data sets or eachcountry determined that the trends are similar or 10o the 13 countries examined in this article. Tat is, theonly 3 countries that show a signicant probability ohaving experienced a random degree o change betweendata sets over time were the United States, France, andthe Netherlands. Tus, or these 3 countries, there is a

    variability in the rankings which implies that the twosets o data are not drawn rom the same population,

    which in this case would be represented by the datasource. Te test results or the other countries show thatthe rankings o the dierences between the levels are notdierent rom each other, indicating that the associateddata sets exhibit trendlike eatures. Tis statistical testprovides evidence that, or the majority o the countriesexamined, the comparison made o trends over time isconsistent and useul, even when dierent sources ormethods are used.

    Comparison of hours worked and working time

    Te concept o hours worked, as addressed in this article,is a purely quantitative measure o the number o hours anindividual spends at work. Working time, by contrast, is abroader concept that encompasses quality-o-worklie issues,including the scheduling o hours o work, such as overtime,split-shits, and just-in-time exible work schedules; night

    work and weekend work; and part-time work.A cross-country comparison o hours worked or the

    13 countries examined in this article, using the OECDdata set, reects a number o institutional changes in both

    working time and hours worked. Historically, the UnitedStates pioneered reductions in working time well in ad-vance o other industrial nations, although Western Eu-rope caught up by the 1980s.29 Since then, a number ochanges in the structure o the labor market have contrib-uted urther to a reduction in working time. First, normalhours o work have declined in many developed countriesbecause o changes in laws and collective-bargainingagreements. Second, women have increasingly joined thelabor orce and work, on average, ewer hours than men.

  • 7/30/2019 Intrenational Comparison of Hours Worked

    18/29

    Comparisons o Hours Worked

    20 Monthly Labor Review May 2009

    Chart 5. Average annual hours actually worked, all employed persons, Belgium, South Korea, and theNetherlands, 19802006

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    OECD

    BLS

    Belgium

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    ,000

    2,900

    2,00

    2,00

    2,00

    2,00

    2,00

    2,00

    2,200

    South Korea

    190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    OECD

    BLS

    ,000

    2,900

    2,00

    2,00

    2,00

    2,00

    2,00

    2,00

    2,200

    Netherlands

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,20190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    BLS

    OECD

    1,00

    1,0

    1,00

    1,0

    1,00

    1,0

    1,00

    1,20

    Hoursworked

    Hoursworked

    Hoursworked

    Hoursworked

    Hoursworked

    Hoursworked

  • 7/30/2019 Intrenational Comparison of Hours Worked

    19/29

    Monthly Labor Review May 2009 21

    Finally, part-time hours worked in the growing servicesector mitigate the overtime work pattern in the relativelysmaller manuacturing sector. Each o these labor marketconditions merits discussion.

    A 2004 OECD report on working time analyzes the

    broader measure o labor utilizationaverage annualhours actually worked per capitashowing that thesehours have barely declined over the past three decades,even as average annual hours actually workedper employed

    person ell signicantly.30 Te large decline in average an-nual hours worked per worker was oset by increases inboth the employment rate (or employment-populationratio) and the share o the population that is o workingage. Te employment rate has risen as more women jointhe workorce and as older workers stay in their jobs rath-er than retire. Both women and older workers are morelikely to work ewer hours in a ull-time job or becomepart o the growing ranks o part-time workers.

    A 10-year snapshot with available data o the employ-ment-population ratio, part-time employment rate, and

    womens labor orce participation rate reects, to a lesserdegree, the 30-year trend just described. (See table 1.) In 9o the 13 countries examined, there were small increases inthe employment-population ratio. (Japan and South Koreasaw a small decline and Spain experienced a large increase.)

    Te part-time employment rate grew rom a low point inSouth Korea and Spain; it ell in the United States andSweden, and it remained steady in Canada, Japan, France,

    Norway, and the United Kingdom. Te part-time employ-ment rate rose in the remaining countries. Dutch policyand legislation provide strong incentives or part-time em-

    ployment, which are reected in the act that more thana third o workers are employed part time in the Nether-lands. Te womens labor orce participation rate inchedup in all o the countries studied, except or Japan, where itell, and the Netherlands and Spain, where it rose dramati-

    cally. Nearly a tenth o women in the latter two countriesjoined the labor orce over the 10 years examined.

    In both the OECD and BLS data series, 19802006trends in average annual hours actually worked per em-ployed person broadly reect the institutional norms andlaws relating to working time in each o the 13 countriesdiscussed. Tis section next addresses some o the sig-nicant institutional and legislative changes that have oc-curred in the past 26 years in these countries.31

    Countries with high working time. O the countries ex-amined, the United States, Canada, Japan, South Koreaand the United Kingdom and Italy until recentlysharesome or all o the ollowing characteristics in their labormarket institutions and laws:

    a normal workweek o 40 hours or more no limit on maximum hours o work allowed per

    week vacation time subject to tenure in job wage or leave penalties or absence rom work limited or no legal entitlement to vacation time.

    Te United States and Japan impose no legal limit on themaximum number o hours worked per week. Regardingpaid time o, business practice in the United States varies

    Table 1.

    1996 2006 1996 2006 1996 2006

    United States ............................. ............................. ...................... .2 .1 1. 1. 9. 9.Canada.......................... .............................. ............................. ....... 9.1 . 19. 1. . 2.1Japan ............................. .............................. ............................. ....... 0.9 . 21. 2. 9. .9South Korea .......................... ............................. ........................... 9. .9 . 9. .9 0.Belgium ............................. ............................. .............................. .. .1 . 1.9 19. .0 .9Denmark ........................... ............................. .............................. .. 0. 2. 1.9 1.1 . 0.France ........................... .............................. ............................. ....... 9.1 1.2 1.2 1. . 1.1Germany ........................... ............................. .............................. .. 2.0 2.2 1.2 21.9 . 1.2Netherlands .......................... ............................. ........................... .2 2. 29. . 9. .Norway ......................... .............................. ............................. ....... 0.2 2. 21.2 21.1 .2 0.Spain ............................. .............................. ............................. ....... .9 2. . 11.1 .2 .0Sweden......................... .............................. ............................. ....... . 0. 1. 1. 9. 0.United Kingdom ........................... ............................. ................. . 0.1 2. 2. . .

    Table 1.

    Employment-population ratio

    Part-timeemployment rate

    Womens laborforce participation rate

    Three important international labor market indicators, 1996 and 2006

    Country

  • 7/30/2019 Intrenational Comparison of Hours Worked

    20/29

    Comparisons o Hours Worked

    22 Monthly Labor Review May 2009

    widely, with some businesses granting leave only ater ayears tenure, others increasing the number o leave dayswith job tenure, and about a ourth providing no paidleave at all. Japanese and South Korean labor laws dierrom business practice. Businesses are supposed to pay orovertime and to promote leave or employees. In practice,however, workers usually take vacation hours when sick,because sick leave is oten unpaid. In some cases, employ-ers penalize workers absences by deducting or not pro-

    viding bonus pay or vacation time.32 Canada, the UnitedKingdom, South Korea, and Japan require statutory paid

    vacation time or ull-time employees, while there is norequirement in U.S. law to provide vacation time, eitherpaid or unpaid. O the six countries with high workingtime, only the United Kingdom and Italy require employ-ers to pay part-time or temporary employees or their an-nual leave. Te European countries are set apart by theact that they recently adopted the European Unionsmandates on working-time restrictions.33

    Between 1988 and 1997, Japanese laws reduced normalhours o work rom 48 to 40 hours per week; between 1997and 2004, South Korea ollowed suit. (See box, this page.)

    Tere have been ew changes in labor laws in the remain-

    ing our countries during the past 25 years. In the 1990s,the United Kingdom and Italy complied with the Euro-pean Union regulations to limit working hours in 2002 and2003, respectively.

    Countries with low working time. Conditions in Belgium,Denmark, France, Germany, the Netherlands, Norway, andSweden dier rom those o the high-working-time coun-tries just described. Te aorementioned recent changesto labor laws in the United Kingdom and Italy now place

    these two countries in the low-working-time category.Tese countries share some or all o the ollowing charac-teristics in their labor market institutions and laws:

    a legal or collectively bargained workweek o less

    than 40 hours a limit on the maximum number o hours worked

    during the week and a limit on the maximum num-ber o overtime hours worked during the year

    statutory paid vacation time o a minimum o 4weeks per year or ull-time workers and proratedor part-time employees

    near-universal entitlement to statutory vacationtime

    broad coverage o collective-bargaining agreementsthat provide even more generous leave entitlementsthan those written into law.

    Revised laws regarding normal hours o work have beenimplemented throughout Europe as a result o the Euro-pean Union Directive on Working ime, which was rstintroduced in 1993 and most recently revised in 2003.34

    Tese laws (1) limit the hours that employees can workovertime throughout the year and (2) establish vacationrights o 4 weeks per year or ull-time employees, withprorated vacations or part-time employees.

    Germany, the Netherlands, Norway, and Sweden havea high share o workers covered by collective-bargaining

    agreements; these countries saw reductions in the work-week as a result o changes in those agreements in the late1980s. Te Netherlands passed national legislation in 2000that allowed employees to choose the number o hoursthey want to work. Te legislation led to a urther growthin part-time employment, which had begun to grow in the1980s.35 Te trend toward reductions in working time wascomplemented by the implementation o the EuropeanUnion Working ime Directive in member countries. Telast two o the major European countries to ratiy changesin labor laws to comply with the directive were the UnitedKingdom in 2002 and Italy in 2003.

    Te case o France is unique, because the reduction inthe normal workweek was initiated by laws, not collec-tive-bargaining agreements. A series o laws was passedbeginning in the 1990s to reduce the number o hours inthe normal workweek, with the primary purpose o de-creasing high unemployment. Te changes began with theRobien law in 1996, ollowed by the Aubry laws in 1998and 2002, eectively reducing the normal workweek rom39 hours to 35 hours.

    Te trend toward reductions in hours shows signs o

    South Koreans work longer hours per week than

    workers in many other OECD countries, despite na-tional legislation that phased in the 40-hour work-week by 2004. Te 2007 South Korean labor orcesurvey reports that nearly 60 percent o all employedpersons who were at work when the survey was takenactually worked 45 hours or more a week, whereas lessthan 30 percent worked a 36- to 44-hour workweek.Less than 15 percent o part-time employed persons

    who were at work when the survey was taken worked35 or ewer hours a week.

    South Koreas long hours worked

  • 7/30/2019 Intrenational Comparison of Hours Worked

    21/29

    Monthly Labor Review May 2009 2

    reversing in some countries. French legislation in 2003specically, the Fillon lawexcluded small businessesrom the normal maximum workweek limit o 35 hours,and urther revisions in 2007 were intended to provide

    greater exibility in scheduling hours or businesses. InGermany, since 2003 a number o collective-bargainingagreements, among them the trend-setting Volkswagenand IGMetall agreements, have seen an increase in thelength o the regular workweek (which remains under 40hours) in exchange or job security. Te trend o raising theceiling on normal hours continues today in contract bar-gaining, especially in Germany. However, hours-workedstatistics do not necessarily reect this or any other trend.(See box, this page.)

    Numerous studies o industrial relations in both thecountries with high working time and those with low work-

    ing time provide detailed inormation on the institutions,labor markets, and demographics that reinorce the quar-ter-century trends seen in the OECD and BLS data series onaverage annual hours actually worked per employed person.Among the ndings are high, but declining, hours workedin Asian countries; little change in hours worked in Anglo-phone countries, where a large share o workers continuesto work more than normal hours; and alling hours workedin European countries, because o a reduction in normaland contractual hours and rising part-time employment.36

    Comparison of Japanese and U.S. hours worked

    Pinpointing whether one countrys average hours actuallyworked are more or less than anothers or a given year orperiod is not a precise science. Te next two sections look

    at the data series or two countries whose labor marketconditions do not seem to be reected in their data: Japanand Sweden. Japans hours-worked series in both the BLSand OECD data sets show that the average hours actually

    worked by Japanese workers are on a par with those workedby U.S. workers, deying the many reerences to that coun-trys long-hours culture that have become commonplace.On the other end o the spectrum, Swedens hours workedtrended upward during the 25-year period studied, quiteunlike the trend in the other 12 countries and, in particular,quite unlike its neighbor Norway, which has similar laborpractices. An analysis o the data sources used to constructthe various time series, together with a look at alterna-tive sources, provides a urther window o understandinginto the challenges o international comparisons o dataon hours worked. Te estimates or Japan and Sweden arecompared with those or the United States and Norway,respectively, and with alternative data sources.

    Te OECD data series or Japan shows that, or 2006,annual average hours actually worked were 1,784, a gurethat is 35 hours less than the U.S. estimate o 1,804. (Seechart 6; data beore 1996 are not available or the BLS dataset.) Over a quarter century, Japans annual average hours

    actually worked declined by 42 eight-hour workdays andthe U.S. average ell by less than 2 eight-hour workdays. Isit possible that U.S. workers now work longer hours thantheir Japanese counterparts? Further, how does one ex-plain the common practice o employees working unpaidovertime in Japan despite recent regulations restrictingovertime hours?37 Finally, what about the culture o long

    work hours as exemplied by ofcial recognition o the oc-cupational hazard o death rom overwork, a phenomenonthe Japanese call karoshi?38

    Some researchers think that the data or Japan under-count unpaid overtime and long hours o work. Evidence on

    the incidence o overtime work in Japan, shown repeatedlyin many special surveys on labor conditions, together with ahistorical comparison help interpret Japans data series. Teincidence and degree o usual overtime in Japan rom 1997through 2007 are given in table 2, which compares rangeso hours worked by persons who worked at least two-thirdso the year; these workers represent approximately 80 per-cent o employed persons.39 In all 3 years shown, 87 percentor more o these year-round employed persons workedat least a 35-hour week. However, rom 1997, the year in

    Germanys minijobs

    Germanys minijobs escape measurement. A

    growing number o people work in such jobs, alsocalled one-euro jobspositions that have a limit onthe hours that can be worked and that oer wageson which earnings are not subject to income taxesand employer taxes are reduced. Te program wasintended to create jobs or the unemployed, but em-ployed workers have taken on minijobs as second jobsbecause o the tax advantage. In 2004, minijobs ac-counted or about 12 percent o employment, and 37percent o minijobs went to people who had another

    job. Minijobs are excluded rom the administrativeramework o tax collection, so data on the hours

    worked at them and the number o jobs they generateare excluded rom hours-worked statistics (personalcommunication, Dr. Ulrich Walwei, Bundesagenturr Arbeit/Institute or Employment Research, Ger-many, April 2006).

  • 7/30/2019 Intrenational Comparison of Hours Worked

    22/29

    Comparisons o Hours Worked

    2 Monthly Labor Review May 2009

    which legislation was passed to reduce the normal work-

    week rom 44 to 40 hours, the share o persons who usuallyworked 43 or more hours per week shited slightly upward,rom 57 percent in 1997 to 61 percent in 2002. Te per-centage ell to 59 percent in 2007. Over the year, a numbero employees do not take vacation time, even though theyare entitled to it. According to one 2005 study, workerstake less than hal their vacation or the year, accumulatingan average o 18 untaken vacation days.40

    Further evidence o the undercount o hours in theOECD data set is ound in akeshi Mizunoyas research.Mizunoya uses both labor orce and establishment surveysto determine the degree to which dierent survey sources

    or Japanese data matter. His critique o the OECD annual-hours-worked data series or underreporting hours workedin Japan stems rom the type o survey that the OECD uses.Rather than using the establishment survey, as the OECDdoes, Mizunoya uses the labor orce survey or 3 years dur-ing the 1990s to account or unpaid overtime, developingan estimate o employees average annual hours actually

    worked.41 Chart 7 compares Mizunoyas estimates with theOECD annual-hours-worked data series. Te Mizunoyaestimates are greater than the OECD data or each o the

    years studied1990, 1995, and 1999increasing rom a

    240-hour to a 270-hour dierence over the decade, or theequivalent o at least 6 weeks more a year. Because, on aver-age, the sel-employed work more hours than employees,the Mizunoya estimate, based on employees, does not ullycompensate or the greater number o hours worked by thesel-employed.

    Tis example rom Japan leaves the lesson that under-standing labor markets is key to deciphering the dier-ences in data sources and explaining how those dier-ences aect comparisons.

    Swedish and Norwegian hours worked

    Te BLS and OECD data sets or Sweden and Norway areidentical, each using the data prepared by that countrysnational accounts. However, the data series or Swedenshows that average annual hours actually worked in 2006

    were the highest among countries with low working timeand were about 175 hours more than those o SwedensNordic neighbor Norway. wenty-ve years ago, Swedenshours were lower than Norways, but average annual hoursactually worked in 2006 were reported to be 1,583 or

    Chart 6. Average annual hours actually worked, all employed persons, the United States and Japan,19802006

    2,200

    2,100

    2,000

    1,900

    1,00

    1,00

    1,00190 192 19 19 19 1990 1992 199 199 199 2000 2002 200 200

    2,200

    2,100

    2,000

    1,900

    1,00

    1,00

    1,00

    Japan

    United States

    Hoursworked

    Hoursworked

  • 7/30/2019 Intrenational Comparison of Hours Worked

    23/29

    Monthly Labor Review May 2009 2

    Sweden and 1,407 or Norway. (See chart 8.) Until the1990s, hours ell in both countries, but Swedens hoursworked rose throughout the decade and remain the high-est among countries with low hours worked. By contrast,Norways hours worked show a continuously decliningtrend. Is it possible that Swedes work 5 weeks more per

    year, on average, than Norwegians? Tis seems unlikely,or a number o reasons. First, both countries have laborlaws that provide generous statutory paid leave o 5 weeksa year1 more week than that mandated by the EuropeanUnion Working ime Directiveand ull- and part-time

    workers are eligible or this leave. Second, Sweden has11 national holidays compared with Norways 9. Finally,many employees in both countries are covered by collec-tive-bargaining agreements and work less than a 40-hour

    workweek.Te similarities in labor conditions belie the act that

    the two countries economies experienced dierent levelso prosperity in the 1990s. Norways oil wealth cushioned itrom the austerity that the Swedish economy had to turn toin the 1990s. Sweden experienced a strong economic down-turn and increasing unemployment, and saw its generoussocial policies curbed throughout the decade.42 Te increasein the countrys hours worked in the 1990s is counterintui-tive: a weak economy generally contributes to a decline in

    hours worked, both individually and across the economy.Te decline in hours worked as o 2000 can be explainedby a number o changes, including continued reductions innormal hours o work through collective-bargaining agree-ments in the private sector43 and adverse eects o the ex-pansion o an already generous sick leave policy, leading toa daily rate o absence rom work o 20 percent.44 In light othese developments, Swedens average annual hours actu-ally worked appear suspiciously high.

    Te Swedish national accounts primary source o data onemployment and hours worked is the countrys labor orce

    survey. Te Norwegian national accounts data, by contrast,are based on normal hours o work reported by administra-tive data sources. Administrative data used by Norway leadto the lowest estimates o hours actually worked, whereaslabor orce surveys, such as those used by Swedens nationalaccounts, produce the highest estimates. Tese dierencesin underlying data sources make it difcult to compare thetwo countries data series. It is probable that hours actually

    worked in each country lie somewhere in between the twoseries values, but it is highly unlikely that Swedish people

    work 4 to 5 more weeks a year than Norwegians do.Using data rom similar sources and creating a simple

    methodology o comparison shrinks the dierences be-tween the two countries hours-worked gures considerably.and increases their levels as well. Harmonized labor orcesurvey data on hours actually worked per week or Norwayand Sweden are available or 2006. Because the two coun-tries labor orce surveys are continuous, one can estimateaverage annual hours actually worked by multiplying theaverage o hours actually worked per week by 52. Te labororce survey reports higher hours overall or both countriesand diminishes the dierence between them. As the ol-lowing tabulation shows, the dierence between Norwaysand Swedens average annual hours actually worked declines

    rom 4 weeks to 1 weeks when comparable data sourcesand methodologies are used:

    actually worked per employed person, 2006

    Country National accounts European Unionlabor orce survey

    Norway ......... ............... 1,407 1,817Sweden ......... ............... 1,583 1,872

    Tese examples highlight how dierences in concepts and

    Table 1.

    All year-round employed persons ............................. ........................ ,,000 0,,100 1,1,100

    Less than 1 ............................. ............................. .............................. .............. .9 1.0 1.2121 ............................. .............................. ............................. ........................ 1.9 2. 2.922 ............................. .............................. ............................. ........................ .0 .1 .12 ............................. .............................. ............................. ........................ .1 29.0 29.0 ............................. .............................. ............................. ........................ 1.0 12. 12.2 ............................. .............................. ............................. ........................ 1. 1. 1.199 ............................. .............................. ............................. ........................ 1.9 19. 1.90 or more ............................. ............................. .............................. .............. 10. 1.1 1.

    Table 2.

    1997 2002 2007

    NOTE: Year-round employed persons are those who work more than 200days per year.

    Percent distribution of weekly hours worked by year-round employed persons, Japan, 1997, 2002,

    and 2007

    SOURCE:Employment Status Survey, Statistics Bureau, Management andCoordination Agency, Government o Japan.

    Weekly hours worked

    Average annual hours

  • 7/30/2019 Intrena