productivity - concept & trends

Productivity Concepts

and Trends

Andrew Sharpe text 11/27/02 2:15 PM Page 29

31

INTRODUCTION

The issue of productivity and the relat-ed issue of innovation continue to behigh on the public policy agenda.

There is especially strong interest among pol-icy-makers in the social aspects of produc-tivity. The objective of this second issue ofThe Review of Economic Performance and SocialProgress is to examine the two-way linkagesbetween productivity and various measures ofsocial progress in Canada. The purpose of thispaper is to provide a succinct, non-technicaloverview of the productivity issue, includingdiscussion of productivity concepts, measure-ment issues, trends and prospects. Such infor-mation may serve as useful background for thepapers in this volume.

The paper is divided into six parts. Thefirst part looks at the reasons why productivityis important. The second discusses key produc-tivity concepts, including the link between pro-ductivity and welfare. The third provides sometheoretical perspectives on productivity growth.The fourth part briefly examines productivitymeasurement issues and their relevance for theproductivity debate. The fifth part presents the

key productivity trends and developments thathave taken place in Canada and other developedcountries. Finally, the sixth part briefly dis-cusses the prospects for productivity growth.

WHY IS PRODUCTIVITY IMPORTANT?

Productivity is the relationship betweenthe output of goods and services and the inputsof resources, human and non-human, used inthe production process, with the relationshipusually expressed in ratio form. Both outputsand inputs are measured in physical volumesand thus are unaffected by price changes.Multiplying quantities of the various outputsand inputs by the price each has commandedin a base year yields the comparable or constantprice values that can be added up to providemeasures of aggregate output and input.1 Theratios may relate to the national economy, toan industry, or to a firm or even a plant.Output growth that exceeds growth in meas-ured inputs — that is to say, an increase in theratio of output to inputs — is what analystsmean when they say productivity is increasing.

TH

E R

EV

IEW

OF

EC

ON

OM

IC P

ER

FO

RM

AN

CE

AN

D S

OC

IAL P

RO

GR

ES

S | 2

00

2

Productivity Concepts,Trends and Prospects:

An Overview

Andrew Sharpe

Andrew Sharpe text 11/27/02 2:15 PM 1

Productivity growth is the most impor-tant source of long-term economic growth.From 1946 to 2001, real GDP per hour growth— the productivity of labour — accounted for66 percent of real GDP output growth in thebusiness sector in Canada, the remaining 34 per-cent being growth in total hours worked — aninput that itself was growing rapidly (Table 1).

Over the long term, increasing produc-tivity is the only way to increase the standardof living, defined as real GDP per capita.Growth in per capita income can result from:increases in the employment-total populationratio, reflecting increased labour force partici-pation, lower unemployment or a larger shareof working-age population; or improved termsof trade. But these sources of income growthare unsustainable in the long run, as they haveupper bounds (except possibly for the terms oftrade). Productivity growth, on the otherhand, is not constrained by the size of the pop-ulation or other factors, and its growth is, at

least in principle, sustainable through tech-nological advances.

Thus, trends in productivity are thekey determinant of long-run trends in bothabsolute and relative living standards. Thefall-off in real income growth in Canada andother developed economies since 1973 is adirect result of slower productivity growth.The decline in Canada’s living standards inthe 1990s relative to those in the UnitedStates is largely attributable to our weakerlabour-productivity growth (Sharpe 2001).Slower increases in the amount of outputeach worker produces mean that there isslower growth in the output or income thatcan be shared among the total population.

The magnitude of the productivitygrowth estimates that economists debate —almost always below 1 percent for the aggre-gate economy — may seem small or even triv-ial to non-economists. But small differencesmatter, and the implications for society of a

32

Andrew Sharpe

TABLE 1

Productivity Trends in the Business Sector, 1961-2001

Average annual rates of change

Hourly TotalLabour Labour Unit Real Real

Real Number Average Hours Real GDP Compen- Compen- Labour Consumer ProducerGDP of Jobs Hours Worked per Hour sation sation Cost Wage Wage

1946-1973 5.05 1.72 -0.73 0.98 4.03 7.51 8.54 3.38 3.90 3.421973-1981 3.52 2.71 -0.65 2.04 1.43 11.09 13.34 10.61 1.28 1.131981-1989 3.18 1.97 0.05 2.02 1.13 5.52 7.65 4.28 0.22 0.881989-2001 2.83 1.34 -0.07 1.26 1.56 3.14 4.42 0.80 0.85 1.341989-1995 1.39 0.19 -0.25 -0.06 1.47 2.70 2.65 0.26 0.04 0.651995-2001 4.29 2.49 0.10 2.59 1.65 3.57 6.23 1.34 1.67 2.041946-2001 4.07 1.82 -0.46 1.35 2.68 6.76 8.18 3.96 2.31 2.261973-2001 3.13 1.91 -0.20 1.70 1.40 6.04 7.83 4.52 0.79 1.15

Notes: The growth rate of the Number of Jobs plus the growth rate of Average Hours gives the growth rate of HoursWorked. The growth rate of Hours Worked plus the growth rate of Hourly Compensation gives the growth rate ofTotal Compensation. The growth rate of Real GDP subtract the growth rate of Hours Worked gives the growth rateof Real GDP per Hour. The growth rate of Total Compensation subtract the growth rate of Real GDP gives thegrowth rate of Unit Labour Cost. Real Consumer Wage is defined as Hourly Compensation deflated by CPI and RealProducer Wage is defined as Hourly Compensation deflated by the GDP deflator.Source: Aggregate Productivity Measures, Statistics Canada, August 2002.


1-percent, as opposed to a 3-percent, trendproductivity growth rate are huge. Based onthe mathematical rule of 72, a 1-percent pro-ductivity growth scenario means that it willtake 72 years, or three generations, for realoutput — and hence income — per worker todouble. In contrast, under a 3-percent pro-ductivity scenario it would take only 24 years,or one generation, for real income to double.Even moving from a 1-percent to a 2-percenttrend productivity growth world — a distinctpossibility, as discussed later in the paper —cuts in half (to 36 years) the time needed todouble living standards.

There is, of course, much more to lifethan productivity and the real income growthit generates, as even economists realize. Theeconomic well-being and quality of life of thepopulation — much broader concepts thanGDP per capita — are determined by manyfactors, of which productivity is only one. Afocus on productivity does not mean thateconomists consider these other determinantsof well-being and quality of life unimportant.Economists study productivity because it iscrucial for real income growth and importantfor improving economic well-being and qual-ity of life, or at least its material aspects. Theyalso believe that a better understanding of pro-ductivity trends and determinants can lead tothe development of public policies and pri-vate-sector actions that will serve to improveproductivity performance.

CONCEPTUAL ISSUES RELATED TO PRODUCTIVITY

This section reviews a number of pro-ductivity concepts essential to an under-standing of the productivity debate.

Partial Versus Total FactorProductivityA fundamental distinction is made

between partial and total productivity measures.The former relate output to only one input,most often labour or capital, although interme-diate goods or raw materials also regularly fig-ure in some compilations of inputs, even thoughit is recognized that other inputs have con-tributed to output. Labour productivity is thebest-known partial productivity measure. Thelatter relates output to a combination of inputs,such as capital and labour. They are known astotal-factor or multifactor productivity meas-ures and represent the growth in output notaccounted for by input growth.

The most readily available and widelyused measure of productivity is labour pro-ductivity, the ratio of output to some meas-ure of labour input (employment or hours).This term sometimes creates confusion, as itcan be seen to imply that the level of labourproductivity or the rate of growth of labourproductivity is attributable solely to theeffects of labour. In fact, labour productivityreflects the influence of all factors that affectproductivity, including capital accumula-tion, technical change and the organizationof production. While the intensity of laboureffort obviously does affect labour produc-tivity, it is generally significantly less impor-tant than the amount of capital a worker hasto work with or the level of productiontechnology.

The concept of total or multifactor pro-ductivity has been developed to measure thecontribution of all factors of production to pro-ductivity growth. The rates of growth of allinputs are weighted to yield one growth ratefor the combined inputs. Total factor produc-tivity (TFP) growth is defined as the growth

33

Productivity Concepts, Trendsand Prospects: An Overview


rate of output minus the growth rate of thecombined inputs (just as labour-productivitygrowth equals output growth minus labourinput growth). As the growth rate of the capi-tal stock is generally greater than that ofemployment (and hence the capital-labourratio is rising), the growth rate of TFP (usinglabour and capital as inputs) is generally lessthan the growth rate of labour productivity.This situation arises from the fact that thegrowth rate of the combined inputs of capitaland labour exceeds that of labour alone.

A key issue in TFP measurement is theweighting of these inputs. Under competi-tive conditions, the current dollar-incomeshare of the factor of production — labourincome for hours worked and interest, grosscapital income (profits and depreciation) forthe capital stock — is normally consideredthe relative contribution of the factor to out-put and consequently used to weight the fac-tor to produce an index of total input, or thegrowth rate of the index. When markets arenot competitive, as in the case of monopolies,the weighting issue is much more complex.

The meaning of TFP is also controver-sial. Some economists interpret it as a meas-ure of overall technical change, others as ameasure of disembodied technologicalchange — that is, technical change that isnot embodied in new machinery and equip-ment — while still others argue that TFP isin no way a measure of technological change(Lipsey and Carlaw 2000).

It is incorrect to say that TFP is a supe-rior or preferred measure of productivitycompared to labour productivity, as the twoconcepts serve different purposes. For thoseinterested in how efficiently all factors of pro-duction are used in the production process,TFP is the relevant productivity measure

since it takes into account the productivityof factors of production other than labour,such as capital, intermediate goods andenergy. For those interested in the potentialof the economy to raise the standard of liv-ing, labour productivity is the relevant pro-ductivity measure: it tells us how muchoutput or income is produced by each work-er and, when combined with the total num-ber of workers, how much total income thereis to be distributed among the population.

Output Per Worker VersusOutput Per HourLabour input can be measured either in

terms of the average annual number of workersor in terms of the total number of hours workedin a year. The latter is the more appropriate forlabour productivity since it represents a moreprecise measure of labour input than personsemployed. One should always specify whichconcept of labour productivity is being used.The growth rates of output per worker and out-put per hour may differ when there is a changein the hours worked over time. Indeed, histor-ically the large fall in average working time hasmeant that output per hour has grown signifi-cantly faster than output per worker.

International productivity comparisonscan also differ greatly when annual hours perworker vary across countries. American work-ers put in more hours annually than workersin many European countries. Therefore, pro-ductivity measures based on output perworker portray US productivity levels in amuch more favourable light than measuresbased on the more relevant output per hour.For example, in 2001 Norway’s GDP perperson employed was 81.5 percent of that inthe United States on the basis of output perperson employed, but 110.6 percent on the

34

Andrew Sharpe


basis of output per hour — a difference of29.1 percentage points. The Netherlands alsoshows a large difference (28.4 points) betweenthe two productivity measures, from 73.4 per-cent of the US level for output per personemployed to 101.8 percent for output perhour worked (see Tables 3 and 5 below).

Productivity Levels VersusGrowth RatesA second important distinction is that

between productivity levels and growth rates.The former refers to the output per unit ofinput at a given point. For example, in theyear 2001 the level or value of output per hourin the business sector in Canada was $30.06,expressed in constant 1992 prices. The latterrefers to the percentage change in levels ofoutput per hour, expressed in constant prices,between two points in time. An examplewould be the 20.4-percent increase in labourproductivity between 1989 and 2001, whenoutput per hour was $24.97. One often hearsthe complaint that Canada’s productivity ispoor. This could be in reference to a lowaggregate productivity level, to a low produc-tivity growth rate, or both. Commentatorsshould always specify whether they are refer-ring to levels or growth rates, as the implica-tions can differ significantly.

International comparison of productiv-ity levels requires that levels expressed in adomestic currency be converted into a com-mon currency. This conversion can be doneusing either market exchange rates orexchange rates based on purchasing powerparities (PPPs) — that is, the exchange ratethat equalizes the price of a basket of goodsand services between two countries. For accu-rate comparison, it is imperative that PPPsbe used, although the development of reli-

able PPPs is a complex matter, particularlyat the industry level.2 A range of PPPs, pro-duced by different agencies and researchers,has resulted in a wide range of estimates forlevels of relative international productivity.

The Cyclical Behaviour ofProductivityThe short- to medium-term movement

of productivity is determined by two influences— an underlying productivity trend and acyclical component. Over the long term, thecyclical component is offsetting, with cyclicalupturns cancelling out cyclical downturns sothat actual productivity growth tends to con-verge on trend growth. Actual productivitygrowth between cyclical output peaks providesan approximation of trend productivity,although the trend may also be influenced byaverage capacity utilization over the cycle anddifferences in capacity utilization at the peaks.

The short-term behaviour of labourproductivity is explained by lags in theadjustment of labour input to changes inoutput. If labour input adjusted simultane-ously with changes in output, productivitygrowth would always be at trend. Lags in theadjustment of labour input, both employ-ment and total hours worked, are caused bya number of factors, including firms’ unful-filled expectations concerning demand con-ditions, the existence of overhead labour thatis relatively invariant to output levels, and atendency for firms to hoard skilled labour indownturns so as not to lose their investment.

For the reasons outlined above, the rateof change in output per worker tends to movein a procyclical pattern, declining below trendin downturns and rising above trend in recov-eries. The rate of change in output per hourshows a slightly more dampened procyclical

35



movement, as it is easier to adjust averageweekly hours through short-time or overtimethan to adjust employment levels. Total factorproductivity, which includes the capital stockas well as labour as an input, exhibits evengreater procyclical variation in movementthan output per worker because of the fixityof the capital input.

The cyclical behaviour of productivityhas two implications. First, one should notextrapolate long-term productivity trendsfrom short-term developments. For example,with the Canadian economy entering a periodof weak growth in 2001, due to falling aggre-gate demand, slower productivity growth canbe expected for cyclical reasons. This does notmean that long-term productivity growth hasnecessarily deteriorated, as any productivityshortfall now can be recovered later in thecycle. Second, to minimize the impact of cycli-cal influences on productivity, growth ratesshould be calculated at comparable points inthe cycle, preferably on a peak-to-peak basis.

THEORETICAL PERSPECTIVES ONPRODUCTIVITY GROWTH

Economic theory advances in stages.First, a simple framework based on highlyrestrictive and often unrealistic assumptionsis developed. Then, these assumptions aregradually eliminated as the model attemptsto incorporate more elements of reality. Thedevelopment of the theory of economic andproductivity growth from the 1950s to the1990s has conformed to this pattern.

The modern study of economic growthand long-run productivity growth dates fromthe 1950s when Robert Solow, MosesAbramovitz and Dale Jorgenson identified the

basic inputs of a growing economy as labour,capital and technology. Solow (1957), in awidely cited article, concluded that techno-logical change, not labour and capital, wasresponsible for most economic growth.However, he did not measure the contributionof technological change to economic growthdirectly, but rather measured it as a residualafter the contribution of labour and capital hadbeen calculated. Solow characterized this resid-ual as “a measure of our ignorance.” In theSolow model, technological change was exoge-nous, or “manna from heaven,” although thistreatment of technology was not meant to betaken literally but rather was intended as anabstraction, to simplify and facilitate themodel’s focus on long-term growth.

Solow’s theoretical framework for theanalysis of economic growth served as the basisfor the development, by Edward Denison(1962), of a growth-accounting framework thatattributed economic growth to a number ofsources, including increases in the education ofthe labour force, the contribution of capital, theshift of resources from low-productivity endeav-ours to the mainstream of the modern economy,gains from knowledge and economies of scale.

The limitations of both the neoclassi-cal, or Solow, growth model and growth-accounting methodology in explaining thegrowth process — in particular their inabil-ity to account for the post-1973 productivi-ty slowdown — have in recent years led tothe development of more sophisticated mod-els of economic growth by such economistsas Paul Romer. A key feature of many ofthese models is their emphasis on knowledgeas the driving force of productivity growth.

Romer (1990) points out that “the neo-classical assumptions of diminishing returnsto increasing investment and perfect compe-

36

Andrew Sharpe


tition placed the accumulation of new tech-nologies at the centre of the growth processand simultaneously denied the possibilitythat economic analysis could have anythingto say about this process.” In other words,while early versions of growth theory con-vincingly demonstrated the importance ofstudying technology, the aggregate macro-economic models used left little room for theanalysis of the sources of invention or inno-vation, new and improved products orprocesses, or organizational or structuralchange (Landau et al. 1996).

In recent years, the basic neoclassicalmodel has been enhanced and expanded uponin at least five broad areas (Landau et al. 1996).These developments reflect the elimination ofmany of the model’s restrictive and unrealis-tic assumptions.

> Neoclassical growth theory assumedthat all firms behaved in the same man-ner in their effort to maximize profits. Itis now widely recognized that while theprofit motive is still important, behav-iour can differ greatly among firms.Economists interested in economicgrowth are now exploring such issues ashow firms learn from experience, howgood management differs from badmanagement, how firms differ in theirmeans of gathering and transmittinginformation internally, and how firmscompete in international markets.

> The neoclassical model also assumed per-fect competition. This is a particularlyunrealistic assumption for a growthmodel, because in a world characterized byperfect competition firms have no incen-tive to undertake research and develop-ment, since they can sell at the marketprice all they can produce. Such a model

also assumes away the important real-world issue of the appropriability of thegains from technical progress. Many mod-els of economic growth now assumemonopolistic competition and give explic-it treatment to patents as a mechanism forinfluencing the appropriability of thegains from technical progress.

> The neoclassical model assumes that thesecrets of technical progress are availableto all. The implication is that produc-tivity levels in all countries will con-verge on that of the technological leader,as each country avails itself of the tech-nological knowledge. This assumptionignores the obvious point that the socialability to gain technological advantagevaries greatly among nations, which iswhy productivity levels have not con-verged. Putnam (2001) has developedthe concept of “social capital” as a factorof production to explain internationalvariation in growth rates and produc-tivity levels.

> The neoclassical model assumes that allindustries are equally important. Butsome economists now argue that certainindustries may be more important tolong-run productivity growth than oth-ers because they yield a greater rate ofsocial return through externalities (e.g.,the information technology sector) ormay exhibit increasing returns to scale.

> An implication of the early growththeory is that the long-term steady-state rate of growth is determined bythe rate of technical progress and pop-ulation growth and is independent ofthe rate of saving and investment. Butrecent research suggests that higherrates of accumulation and investment

37



can increase productivity growth, thatthere is no steady-state rate of growth andthat the inputs in the growth process actindependently. For example, Boskin andLau (1992) find that the higher the cap-ital stock, the greater the ability of tech-nology to increase productivity,because most technology is embodiedin capital goods.

Multi-faceted Determinants orSources of ProductivityBuilding on the recent theoretical

developments reviewed above, a large litera-ture has developed aimed at deriving theirimplications for public policy. In Canada,recent contributions in this area includeHarris (1999) and Sharpe (1998).

Based on a review of the cross-countrygrowth literature, Harris (1999) identifies threeproximate drivers (the Big Three) of productivi-ty growth: investment in machinery and equip-ment; education, training and human capital;and openness to trade and investment. In addi-tion, he notes that once one moves from theproximate determinants to the indirect linkages,productivity growth can be influenced by a largenumber of factors. His compendium of potentialindirect productivity determinants includesinnovation (both product and process), diffusionof technology (national and international), spa-tial agglomeration (e.g., Silicon Valley), externaleconomies of scale at the industry level, govern-ment consumption (negative), managementpractices, public infrastructure (positive), incomeinequality (negative), high taxes (negative), smallfirms (negative), labour market flexibility (posi-tive), exchange rate stability (positive) and lowinflation (positive).

Sharpe (1998) identifies the followingseven determinants of productivity growth:

> The rate of technical progress, determinedby the rate of developing new productand process innovations and the pace ofdiffusing those innovations.

> Investment in physical capital such as machin-ery and equipment and structures. Themore capital a worker has to work with,the greater the output he can produce. Itis estimated that 80 percent of technicalchange is embodied in new capital equip-ment, particularly machinery. Withoutgross investment, technical progresswould be all but impossible. Hence,physical investment is essential for pro-ductivity growth.

> The quality of the workforce, including aver-age educational, training and experiencelevels. Literacy and numeracy skills aswell as technical skills are essential if anindustry is to benefit from technicaladvances and make effective use ofmachinery.

> The size and quality of the natural resourcebase. For example, the high level of out-put per hour in Alberta reflects the con-centration of the oil and gas industry inthis province and the high value added(which includes economic rent) per work-er generated by the industry.

> Industrial structure and intersectoral shifts,since the aggregate level of labour pro-ductivity is a weighted average of indus-try labour productivity levels, where theweights are the labour input shares.

> The macroeconomic environment or aggregatedemand conditions defined by the size ofthe output gap and the relationshipbetween actual and potential outputgrowth. Prolonged periods of insufficientdemand can have a negative long-termeffect on productivity growth.

38

Andrew Sharpe


> The microeconomic policy environment, broad-ly defined as the policies that affect behav-iour at the firm level, including tradepolicy, tax policy, industrial policy, com-petition policy, and policies on privatiza-tion, intellectual property, regulationand foreign ownership.There is still considerable uncertainty

about the drivers of productivity. The relativeand absolute contributions made by the differ-ent determinants may vary over time and acrossspace. Many of the factors in productivitygrowth are interrelated and may act in synergy.

Productivity and UnemploymentLabour productivity is technically defined

as the relationship between output and theemployed labour force. It ignores the unem-ployed and others outside the labour force whowould like to have a job but do not. From thisperspective, conventional productivity meas-ures do not represent an appropriate indicatorof the efficient allocation or uses of labour froma societal perspective. Rising productivity canand sometimes does coexist with high or evenrising unemployment, although one couldargue that in such a situation societal produc-tivity is not rising. It is very unproductive orinefficient to have a large number of workersproducing zero output.

One way to deal with this issue is todevelop a productivity measure that defineslabour input as inclusive of the employed andunemployed. Such a measure expresses thesocial relationship between the labour resourcesthat society has available for production andactual output, in contrast to the economic rela-tionship between the labour resources actuallyused in production and output. Chart 1 showstrends in Canada from 1976 to 2001 for eco-nomic productivity, defined as output per per-

son employed, and social productivity, definedas output per labour force participant. Not sur-prisingly, the level of social productivity isabout 7-10 percent below that of economic pro-ductivity over the period, reflecting the addi-tion of the unemployed to the denominator andno change in the numerator. The recessions ofthe early 1980s and early 1990s producedmuch larger declines in social productivity thanin economic productivity. From a societal pro-ductivity perspective, economic downturnshave a very negative effect on overall produc-tivity of the labour force.

Productivity, Economic Well-Being and HappinessProductivity growth can contribute to

greater economic well-being. One approach

39


1997

$

40000

45000

50000

55000

60000

65000

70000

2001

1996

1992

1996

1991

1981

1986

1976

GDP per person in the labour force

GDP per worker

CHART 1

Economic Productivity (Output Per PersonEmployed) vs. Social Productivity (Output PerPerson in the Labour Force) in Canada

Source: Labour Force Historical Review 2001(R) CD-ROM,Statistics Canada Cat. No. 71F0004XCB; and GDP datafrom CANSIM II v3860085, June 3, 2002.


to the measurement of economic well-beingis the Index of Economic Well-being devel-oped by the Centre for the Study of LivingStandards (Osberg and Sharpe 1998, 2002a,2002b) based on Osberg (1985). This indexis based on four components of economicwell-being: consumption flows; non-financialstocks of wealth; equality; and economicsecurity in terms of low risk of unemploy-ment, financial distress due to illness, single-parent poverty and poverty in old age. Sharpe(2002b) demonstrates how real income growtharising from productivity gains can lead toincreased private and public consumption,higher stocks of capital, lower poverty andgreater economic security.

Despite the importance of productivityfor real income growth, one should retain asense of perspective on the productivity issue.Just because productivity can contribute tohigher levels of economic well-being, it doesnot necessarily follow that it should be thetop social priority.

Two points are relevant in this regard.First, in poor countries productivity growthis absolutely crucial to raise the material stan-dard of living to an acceptable level andreduce absolute poverty. In contrast, Canadais already a rich country with high livingstandards for the vast majority of the popu-lation. Increased productivity leads to high-er consumption levels and greater economicwell-being, but it may do little for subjectivewell-being or happiness. Studies have foundthat after a certain income level has beenachieved in rich countries, further realincome growth can have little if any addi-tional impact on happiness for the overallpopulation (Easterlin 1974, 1995). Moneycannot buy happiness, at least not in the longrun. Since the goal of public policy is to

increase the overall well-being, not just theeconomic well-being, of the population, pro-ductivity should not be sold as a panacea forsociety’s ills.

Second, productivity can provide the basisfor potential increases in a number of the com-ponents of economic well-being, such as equali-ty and economic security. But there is nomechanism whereby higher productivity growthautomatically translates into less income inequal-ity or lower poverty, as in the case of higher realwages leading to greater private consumption.For example, growing wage inequality may pre-vent low-skilled workers from benefiting fromproductivity growth. Government action maybe needed to eliminate poverty and decreasesocial inequality.

MEASUREMENT ISSUES

Statistical agencies do not gather pro-ductivity statistics directly from economicagents. Rather, they construct productivitymeasures from data on inputs and outputs.Indeed, almost the entire body of economicstatistics collected by statistical agencies —data on output, employment, prices, invest-ment, raw materials, inventories — is usedin the compilation of productivity statistics.An examination of the reliability of produc-tivity statistics thus becomes, in effect, anexamination of the reliability of much of thesystem of economic statistics.

Figure 1 is a schematic representationof the basic data requirements of productivi-ty statistics, or the building blocks of pro-ductivity measurement. At the extreme leftis the productivity ratio, defined as the ratioof real output to input. This ratio may be apartial productivity measure, such as labour

40

Andrew Sharpe


productivity, where real output is related toonly one input, or a multifactor or total fac-tor productivity measure where an index ofreal output is related to an index of more thanone input. Inputs in addition to labour thathave been included in multifactor productiv-ity calculations are capital, including bothfixed capital and inventories, and intermediategoods, including raw materials and energy.

Either of two real output measures can beused to construct productivity indexes — realvalue added and real gross output. The formerdefines output as the total incomes of the factorsof production (basically, labour and capital) inan industry, sector or economy. The latterdefines output as the physical output producedby an industry, sector or economy. At the indus-try or sectoral level, real gross output comprises

41


FIGURE 1

The Building Blocks of Productivity Measurement

Productivityratio (real

output/input)

real valueadded (double

deflationmethod)

real grossoutput

total hours

real capitalstock

realintermediate

goods(includingenergy)

real grossoutput

realintermediate

goods

real output

current dollar grossoutput

gross output deflators

current dollarintermediate goods

intermediate goodsdeflators

current dollar grossoutput

gross output deflators

employment

average weekly hours

current dollar capitalstock

capital stock deflators

current dollarintermediate goods

intermediate goodsdeflators

input


real value added and real intermediate goods. Atthe aggregate level in a closed economy, realgross output is equivalent to real value added asintermediate goods are netted out.

The most appropriate output concept ofindustry productivity when labour or labour andcapital are included as inputs is real value added.Use of real gross output may bias the resultsbecause of substitution, in the productionprocess, between intermediate goods and labouror capital.3 On the other hand, the most appro-priate output concept when intermediate goodsare included as an input is real gross output.

Real value added is calculated througha double-deflation procedure whereby realintermediate goods are subtracted from realgross output. Real gross output is calculatedthrough the deflation of current-dollar grossoutput by gross output deflators. Real inter-mediate goods are calculated in a similar man-ner, from current-dollar intermediate goodsand intermediate goods deflators.

Turning to the input side, labour input,most appropriately measured as total hoursworked, is determined by employment andactual average weekly hours. The real capitalservices arising from the capital stock (fixedcapital and sometimes inventories) are derivedfrom current-dollar capital stock estimates andcapital stock deflators.

From the above discussion, five basicbuilding blocks of productivity measurementcan be identified: estimates of labour input,including both employment and averageweekly hours; estimates of current-dollar cap-ital stock; estimates of current-dollar inter-mediate goods; estimates of current-dollargross output; and estimates of product priceindices. These product price indices are, inturn, used to derive deflators for gross output,capital stock and intermediate goods.

Productivity statistics are plagued by anumber of measurement problems, the mostimportant of which are outlined below.

Price Indices, QualityAdjustment and HedonicsPrice indices for goods and services are

crucial for deflating the current value of out-put to produce real output and hence produc-tivity estimates. But quality changes in goodsand services over time must be integrated intoprice indices if true changes in real output areto be captured.4

The Panel to Review ProductivityStatistics (1979) identifies three types of qual-ity change. Type 1 is the change in the quanti-ty of costly resources used to produce a product,such as the addition of a remote-control deviceto a television set. Type 2 occurs when a tech-nological innovation raises the quality of aproduct without any increase in currentresource inputs, such as when new models ofcomputers have more memory and greater pro-cessing ability but cost the same or less thanthe models they replace. Type 3 quality changerefers to any design change in durable goodsthat results in higher or lower operating costs,holding constant both the quantity of servicesprovided by the good and the wages and pricesof the inputs used in its operation. An exam-ple is the redesign of an engine to improve fuelefficiency.

Until the 1980s, statistical agenciesmade adjustments for Type 1 quality changebut largely ignored Types 2 and 3. Since then,there has been growing recognition of theimportance of these latter types of qualitychange, as represented by computers and moreenergy-efficient consumer durables, respec-tively, and attempts to adjust for them. Themost common method of adjustment is known

42

Andrew Sharpe


as hedonics. This involves the application of astatistical regression to the different models ofa given type of product available in two ormore years, where the dependent variable is theprice of each model and the independent vari-ables are its measured characteristics.

The application of hedonics has pro-duced very large decreases in the quality-adjusted price indices for computer hardwareand, to a lesser degree, telecommunicationsequipment, leading to enormous increases inreal output and hence productivity growth inthese sectors. Indeed, the computer hardwaresector has accounted for a highly dispropor-tionate share of output and productivitygrowth in the United States and to a lesserdegree in Canada, where the sector is lessimportant. In principle, productivity-growthcomparisons across countries can be greatlyaffected by differences in the use of hedonicsby national statistical agencies. In reality, thisfactor does not appear to account for significantinternational differences in aggregate produc-tivity growth rates, although sectoral growthrates can be affected (Pilat 2001).

Non-marketed OutputA key requirement for the development of

productivity estimates is that output be meas-ured independently of inputs. If output is meas-ured by the quantity of inputs, productivitygrowth will by definition be zero. In sectorswhere output is not marketed, it is not possibleto deflate the nominal value of output to producereal output and hence productivity estimates.This means that there are no reliable estimatesof productivity growth for these sectors — pri-marily public administration and the publiclyfunded components of the education and healthsectors. It is therefore best to exclude these sec-tors from aggregate productivity measures.

For this reason, the business sector is themost appropriate category for analysing produc-tivity trends at the aggregate level and the sectorfor which official productivity statistics are pro-duced. The wide availability of data for totalemployment and real GDP does mean that pro-ductivity estimates for the total economy are oftenreferred to, although, because of the lack of meas-ured productivity growth in the non-marketedsectors, these estimates have a downward bias.

It is in theory possible to develop pro-ductivity growth estimates for the non-busi-ness sector by measuring, in physical units,the output of the sector. Possible physicalindicators include the number of graduatesof the education system, the number of pro-cedures performed in hospitals and the num-ber of cheques processed by a governmentoffice. But such indicators may represent onlypart of the output of the sector and, moreimportantly, may exhibit significant qualitychanges over time. The development of reli-able productivity growth estimates for thenon-marketed sector is still in its early stages.

The Underground EconomyThe issue of the underground economy

often arises in discussions of productivitytrends. It is pointed out that the underestima-tion of real output because of unrecorded under-ground activity, not offset by a commensurateunderestimation of inputs, will produce adownward bias to productivity level estimates.Estimates of the size of the underground econ-omy vary widely. The most authoritative (andlowest) estimate is that by Statistics Canada(1994), which found that the undergroundeconomy represented around 3 percent of GDPin Canada in 1992. A key reason for the smallsize of the underground economy relative toGDP is that Statistics Canada is aware, through

43



various sources, that many transactions in cer-tain sectors, such as construction, are not report-ed to the tax authorities; the agency cantherefore make imputations for this unreport-ed economic activity in the national accounts.

It should also be noted that it is not theexistence of the underground economy per sethat produces bias in productivity growth rates,but rather changes in its relative size over time.If the size of the underground economy remainsstable, productivity levels may be underesti-mated but productivity growth rates will beunaffected. Of the many measurement issuesfacing national accountants and productivityanalysts, bias associated with the undergroundeconomy is certainly not the most serious.

Conceptual Problems in theDefinition of OutputIn certain industries in the business or

marketed sector, the definition of what actual-ly constitutes output poses conceptual prob-lems that affect productivity estimates. Forexample, is the output of the banking sectorthe intermediation function the banks serve (asproxied by the value of the spread betweenwhat the banks earn and what they pay out ininterest, net of expenses), or is it the services pro-vided by the sector (number of accounts main-tained, number of cheques processed, convenienceprovided by ATMs, etc.)? Other industries withconceptual problems include insurance, gam-bling and brokerage houses.

These conceptual issues are gradually beingworked out, with the result that productivity esti-mates for these industries are becoming more reli-able. For example, statistical agencies havechanged the definition of output in the bankingsector, from the first definition noted above to thesecond, with the result that measured productiv-ity growth in the sector has increased.

Quality Adjustment of InputsA key issue in productivity research is

whether inputs such as labour and capital shouldbe adjusted for quality changes, just as outputis adjusted. Statistical agencies certainly produceand release unadjusted estimates of labour andcapital inputs. They also often adjust inputs forquality changes in the compilation of produc-tivity estimates, particularly TFP estimates.

With quality adjustment, quality improve-ments increase the growth rate of the input andhence its contribution to output. This means thatthe size of the residual or TFP is reduced, shed-ding more light on the sources of growth. This isconsidered by many to be the main advantageof adjustment. The advantage of non-adjust-ment is that the conceptual and methodologi-cal difficulties inherent in adjustment areavoided and the productivity numbers are easi-er to interpret and understand.

The Importance of StatisticalRevisionsStatistical agencies revise, on a regular and

periodical basis, the economic series they pro-duce. As productivity estimates draw upon awide range of economic data, including estimatesof employment, hours, nominal output, pricesand capital stock, they are subject to frequent —and often significant — revisions. Indeed, theserevisions are the scourge of productivity analysts,but a necessary evil since the most recent datamust be used. Unfortunately, the revision of pro-ductivity data can result in the rewriting andreinterpretation of productivity trends.

Two examples illustrate this point. In May2001, Statistics Canada released its AggregateProductivity Measures data which showed thatoutput per hour in the business sector advancedat a 1.2-percent average annual growth ratebetween 1995 and 2000, a performance character-

44

Andrew Sharpe


ized as weak by productivity analysts. Later thatsame month, Statistics Canada released new esti-mates of the national accounts using, for the firsttime, the Fisher chain index and capitalizingsoftware expenditures. These changes boostedproductivity growth by a very significant 0.5 per-centage points, to 1.7 percent per year for thesame period, and forced productivity analysts tochange their characterization of productivitygrowth over this period.

In July 2001 the US Bureau of LaborStatistics revised its estimates on business-sectoroutput per hour based on new national accountsdata from the Bureau of Economic Analysis.Instead of increasing 2.8 percent per year over the1995-2000 period, as originally reported earli-er in the year, productivity growth was reviseddownward to 2.4 percent. This indicated thatthe acceleration in productivity growth was lessthan previously believed.

PRODUCTIVITY TRENDS ANDDEVELOPMENTS

This section of the paper highlights anumber of the developments that have charac-terized productivity growth in the post-warperiod in OECD countries and in Canada. Theinternational trends discussed are: the post-1973 productivity slowdown, the post-warproductivity convergence phenomenon, laggingproductivity growth in certain service sectors,and the post-1995 productivity growth acceler-ation in the United States. The Canadian trendsexamined are: the relative decline of Canada’sproductivity performance, the growing Canada-US manufacturing productivity gap and sec-toral productivity trends.

Three distinct productivity trends or styl-ized facts can be identified in the post-war

period for the United States and two for otherdeveloped economies, including Canada. From1945 to 1973, developed countries experienceda golden age of productivity growth, withlabour-productivity growth advancing at a rateof 3 percent or more per year. After 1973, vir-tually all developed countries entered a periodof slower productivity growth. The failure ofproductivity to pick up in the first half of the1990s despite the introduction of informationtechnologies led observers to coin the term“productivity paradox.” Since 1995, the UnitedStates has been in a period of much strongerproductivity growth, resolving the productivi-ty paradox as least for that country.

The Post-1973 ProductivitySlowdownThe most important productivity devel-

opment in the post-war period has been theslowdown in labour and total factor productiv-ity growth, a phenomenon that affected virtu-ally all industrial countries and most industries.According to official Statistics Canada esti-mates, growth in output per hour in the busi-ness sector fell by nearly two-thirds, from4.0 percent per year in the 1946-73 period to1.4 percent in the 1973-2001 period (Table 1and Chart 2). Growth in output per hour in thebusiness sector averaged 1.4 percent per year in1973-81, fell slightly to 1.1 percent in 1981-89 and rose to 1.6 percent after 1989.

The post-1973 productivity slowdownaffected most sectors of the Canadian economy(Table 2). Of the 10 one-digit SIC industriesfor which official data are available, eight expe-rienced significantly lower growth in outputper hour after 1973 (agriculture; fishing andtrapping; logging and forestry; mining, quar-rying and oil wells; manufacturing; trans-portation and storage; communications and

45



other utility industries; and retail trade). Thetwo exceptions were construction and whole-sale trade, both of which have seen an improve-ment in productivity growth since 1973.

After more than 20 years of debate, thereis still no consensus among economists on thecauses of the productivity slowdown. The viewthat appears to be gaining the largest numberof adherents is that the slowdown reflected theebbing or withering away of the impact of thehistorically unprecedented factors that cametogether to boost productivity growth in theimmediate post-war period (e.g., the shift of theworkforce out of low-productivity agriculture,increased international trade, rapid capital accu-mulation and diffusion of the stock of tech-nologies and know-how built up butunexploited during the Great Depression andthe Second World War). The productivity

experience in the post-1973 period in NorthAmerica can be seen as a return to the long-runhistorical trend of around 1.5 percent per year.

The implications of the productivityslowdown are well recognized by government.For example, in 1994 the federal Departmentof Finance (1994) released the document A NewFramework for Economic Policy (the Purple Book),which states: “At the root of the economicproblem has been the failure of productivity toincrease at the rates that prevailed during thepost-war years to the mid-1970s” (p. 15).

Post-War ProductivityConvergence in OECD CountriesThe United States has been the world

technological leader in the post-war period,with the highest level of productivity amongindustrial countries. On the other hand, it hasexperienced (until recently) one of the slowestrates of productivity growth. Economistsbelieve this is not an accidental situation butrather reflects the dynamics of international pro-ductivity growth. Technological catch-up orconvergence is seen as the major reason whymost OECD countries experienced faster pro-ductivity growth than the United States in thepost-war period.

Through technological catch-up, low-productivity countries have the potential ofenjoying rapid (although declining) produc-tivity growth until their productivity levelsbegin to converge on that of the leader. Indeed,a number of countries in the developed worldhave effectively exploited this potential in thepost-war period. The average unweighted levelof output per hour worked in OECD countriesexcluding the United States went from 44 per-cent of the US level in 1950 to 83 percent in2001 (see Table 5 below).

46

Andrew Sharpe

Perc

enta

ge

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

1995

-200

1

1989

-199

5

1989

-200

1

1981

-198

9

1973

-198

1

1973

-200

1

1946

-197

3

CHART 2

Output Per Hour in the Business Sector,Canada (Average Annual Rates of Change)

Source: Aggregate Productivity Measures, StatisticsCanada, August 2002. For 1989 onwards, data are fromthe quarterly Aggregate Productivity Measures series,CANSIM II v1409153, July 2001.


The convergence hypothesis is based onfour advantages productivity laggards mayexploit (Abramovitz and David 1996).(1) These countries can make use of state-of-the-art technology produced by the technologicalleader. (2) Because these countries have low cap-ital-labour ratios, the marginal product of cap-ital is high. (3) Less developed countries haveconsiderable opportunities to shift resources outof low-productivity activities. (4) These coun-tries can benefit from economies of scale as theirmarkets grow.

However, there is no mechanism bywhich the productivity levels of poor countriesautomatically converge on that of the leader.Indeed, outside the industrial countries, therehas been little convergence towards US pro-ductivity levels, with the exception of a numberof countries in East Asia. Persistent nationalcharacteristics can inhibit laggard countries fromexploiting the advantages of backwardness.These include poverty of natural resources;

small domestic markets; barriers to trade;forms of economic organization or systems oftaxation that reduce rewards for effort, enter-prise or investment; and deeper elements ofnational culture that limit responses of peopleto economic opportunities. Throughout theThird World, deep-rooted political constraintsimposed on social capability have preventedconvergence, but when these constraints areremoved, as has happened in East Asia, thepotential for convergence can be realized.

Lagging Productivity Growth inCertain Service IndustriesIn general, productivity growth in the

service sector has lagged behind that in thegoods sector, although a number of service-sector industries, such as communicationsand trade, have posted respectable rates ofproductivity growth.5 Service-sector produc-tivity growth has been held back by thefinancial, insurance and real estate sector and,

47


TABLE 2

Trends in Labour Productivity, Output Per Hour Worked by Industry, Canada, 1961-2000

Average annual rates of change in output per hour

1961-1973 1973-1981 1981-1989 1989-2000 1989-1995 1995-2000

Business Sector 3.8 1.6 1.1 1.6 1.5 1.8Agriculture 5.9 6.6 2.4 4.6 4.3 5.0Fishing & Trapping 2.6 -0.4 -3.0 -0.7 -1.4 0.2Logging & Forestry 3.9 1.9 3.4 -0.1 -2.0 2.2Mining, Quarrying& Oil Well 6.1 -6.1 3.1 1.5 3.5 -0.8

Manufacturing 4.2 1.8 2.4 2.3 3.3 1.0Construction 0.5 3.9 -0.4 -0.6 -0.5 -0.7Transportation& Storage 5.2 -0.5 2.7 1.8 2.1 1.4

Communication & other Utility Industries 5.8 3.1 1.7 2.3 1.3 3.5

Wholesale Trade 2.3 1.6 4.6 1.6 1.2 2.1Retail Trade 3.6 1.3 1.1 2.0 0.3 4.0

Notes: Statistics Canada has updated the business sector and manufacturing series to 2001 consistent with the regu-lar annual revision. All other industries will be updated later in the fall of 2002 on the basis of the North AmericanIndustry Classification System.Source: Aggregate Productivity Measures, Statistics Canada, August 2002.


more importantly, by community, personaland business services. Indeed, Sharpe et al.(2002) show that the education and healthsectors recorded negative measured labour-productivity growth in both Canada and theUnited States in the 1990s, which draggeddown aggregate productivity growth.

For example, between the 1989 cycli-cal peak and 1998, output per hour in theservice sector fell 1.3 percent per year, afterfalling 0.5 percent per year during the 1980s.This decline in the absolute level of produc-tivity affected all industries within the sec-tor: accommodation, food and beverage (-1.9percent per year); amusement and recreationservices (-1.4 percent); business services (-1.1 percent); health services (-1.0 percent);education and related services (-0.6 percent);and personal, household and other services (-0.5 percent). As the service sector accountsfor more than one quarter of total hoursworked in the business sector, this develop-ment exercised a significant downward influ-ence on total business-sector productivity.

One explanation for the slower produc-tivity growth in the service sector is the greaterinherent difficulty of increasing productivityin certain service industries. For example, thenon-tangible nature of services limits the pos-sibilities for mechanization, while the one-to-one personal nature of many services, such ashealth care — where output depends on inter-action with the user — makes standardizationdifficult. A second explanation is that officialmeasures of service-sector output have a seri-ous downward bias, greater than in the goodssector. Indeed, productivity growth in the serv-ice sector, if properly measured, may not beinferior to that in the goods sector.

Specific problems in the measurement ofreal output and hence productivity in market-

ed service industries include: conceptual diffi-culties in the definition of output in sectorssuch as banking and insurance; improvementsor deterioration in quality of output that arenot captured by the price indices; absence ofappropriate service-sector data for productivitymeasurement (data coverage is much better forgoods industries); difficulties incorporatingcompletely new services into existing priceindices; and the extreme heterogeneity of trans-actions in certain service industries, such aslegal and health services, which makes pricesystems non-linear and not directly linked towhat is received by the customer.

Post-1995 Acceleration in USProductivity GrowthSince 1995, productivity growth has

picked up significantly in the United States.Between 1995 and 2001, output per hour inthe business sector advanced at a 2.4-percentaverage annual rate, up nearly a full percentagepoint from the 1.5 percent of the 1989-95 peri-od. This development has been taken by manyas prima facie evidence of a new economy char-acterized by higher trend productivity growthbased on information technologies (IT).Research has shown that the productivity pick-up, while greatest in the IT-producing sector,has also spread to the IT-using industries,including those in the service sector, support-ing the view that IT is now having a pervasiveimpact on productivity (Stiroh 2001). Someeconomists, such as Robert Gordon (2000),argue that a significant component of this accel-eration is transitory, related to the cyclical fac-tors and the investment boom of the second halfof the 1990s, and that productivity growth willbe slower in the medium term. Others, such asMartin Baily (2002), believe that most of theacceleration is of a permanent nature.

48

Andrew Sharpe


Data prepared by the Centre for the Studyof Living Standards on Canada’s aggregatelabour productivity performance show that thegreater deterioration relative to the UnitedStates was particularly abrupt in the second halfof the 1990s. In 1976, GDP per personemployed in Canada was 88.1 percent of that inthe United States. In 1995, it was 85.0 percent.By 2001, however, it had fallen to 82.5 percent(Table 6 and Chart 3). The relative declinebetween 1995 and 1999 reflects the accelerationof productivity growth in the United States, asCanadian productivity growth has not fallen inabsolute terms.

TABLE 3

Relative GDP Per Person Employed inOECD Countries

US=100 in all years

1950 1973 1989 1995 2001

Australia 76.1 73.5 76.6 80.2 79.5Austria 36.1 70.7 82.0 79.1 78.6Belgium 63.2 82.1 98.6 100.5 94.2Canada 91.4 92.1 88.0 85.7 79.7Denmark 65.8 73.5 76.6 80.8 77.8Finland 37.8 59.4 72.2 77.5 76.3France 47.2 77.8 91.0 91.2 84.6Unified Germany n/a n/a 80.8 81.5 76.5

West Germany 53.2 83.7 94.7 90.2 n/a

Greece 26.4 58.1 62.8 59.7 61.4Ireland 37.5 51.3 75.6 82.6 90.2Italy 39.9 67.9 80.1 85.0 77.6Japan 19.4 59.1 74.5 74.2 69.9Netherlands 77.0 94.5 84.7 80.6 73.4New Zealand n/a 79.6 66.0 64.3 57.7

Norway 52.3 64.6 77.6 86.8 81.5Portugal 21.6 45.8 49.6 52.2 49.5Spain 25.4 60.0 80.3 84.3 74.5Sweden 58.9 70.0 69.0 73.9 71.6Switzerland 85.5 93.2 83.5 78.2 73.6Turkey 11.8 21.1 27.1 29.5 28.8UK 65.9 66.3 71.8 74.0 70.9US 100.0 100.0 100.0 100.0 100.0

Source: Groningen Growth and Development Centre &The Conference Board, June 13, 2002. www.eco.rug.nl/GGDC/index-dseries.html

Unlike the United States, the other indus-trial countries show no evidence of a post-1995acceleration in productivity growth. In Canada,for example, business-sector output per houradvanced at a 1.7-percent average annual ratebetween 1995 and 2001 — barely above the1.5 percent experienced over the 1989-95 peri-od. With the acceleration in productivity growthin the United States in the second half of the1990s, the productivity leader forged ahead ofthe followers and increased the productivity gap.This represents a situation of productivity diver-gence, in contrast to the productivity conver-gence of the pre-1995 period.

Canada’s Relative Productivity DeclineFrom an international perspective, Canada

has suffered a relative deterioration in its produc-tivity performance in recent years. Data have beencompiled by the Groningen Growth andDevelopment Centre at the University ofGroningen in the Netherlands. In 1973, Canadaranked second out of 22 OECD countries interms of output per person, with 92.1 percent ofthe output per person employed, relative to thatof the United States, the leader (Table 3). By2001, Canada had fallen to fifth place, at 79.7percent, behind Belgium, France, Ireland,Norway and of course the United States. In GDPper capita, Canada also fell from second to fifthover the period (Table 4). Canada’s relative declinein terms of output per hour was even greater —from second to 13th place — because of the fewerhours worked per year in most European coun-tries (Table 5). Canada’s relative productivitydecline largely reflects the pick-up of productiv-ity growth in Europe, where productivity levelsconverged towards, and in a number of cases —Belgium, France, the Netherlands and Norway— surpassed US levels on a per-hour basis.

49



The factors behind Canada’s relative pro-ductivity decline are vigorously debated.Certainly, from an accounting perspective, thedeterioration in its productivity performance inthe IT sector compared to that in the UnitedStates can explain the lion’s share of the declinein recent years (Rao and Tang 2001).

The Widening Canada-USManufacturing Productivity GapCanada’s relative performance in manu-

facturing productivity has been equally poor.Since 1981, Canada has had by far the weak-est productivity growth among the G7 coun-

tries in this sector. Growth in output per hourhas averaged 2.1 percent per year, comparedto the G7 unweighted average of 3.3 percent(Table 7 and Chart 4). In the period 1995 to2000, Canada’s productivity performance waseven worse, averaging 0.9 percent per year com-pared to the G7 average of 2.8 percent.

These developments have produced awidening gap in Canada-US manufacturingproductivity, most notably in the second halfof the 1990s. Manufacturing output per hourfell from 87.5 percent of the US level as recent-ly as 1993 to a low of 67.3 percent in 2001 (seeTable 6). Manufacturing productivity has been

50

Andrew Sharpe

TABLE 5

Relative GDP Per Hour Worked in OECDCountries

US=100 in all years

1950 1973 1989 1995 2001


West Germany 48.6 85.3 106.0 108.2 n/a

Greece 24.6 51.8 60.1 57.1 59.0Ireland 33.3 44.3 71.7 82.9 99.2Italy 44.1 70.4 87.6 95.7 88.7Japan 21.5 54.4 68.6 73.5 72.6Netherlands 77.4 104.0 109.0 108.7 101.8New Zealand n/a n/a 68.2 66.4 61.3



TABLE 4

Relative GDP Per Capita in OECD Countries

US=100 in all years

1950 1973 1989 1995 2001


West Germany 54.5 89.6 89.2 81.9 n/a

Greece 22.1 50.7 48.2 46.0 47.2Ireland 38.1 43.5 49.9 61.5 82.1Italy 38.5 67.1 73.0 72.9 69.1Japan 20.2 68.8 78.2 80.3 72.9Netherlands 62.9 78.9 72.9 75.1 75.1New Zealand 88.8 75.3 61.2 59.8 55.8




largely driven by trends in IT-producing indus-tries such as manufacturers of electronic andother electrical equipment and industrialmachinery and equipment in the United States.This sector is relatively more important in theUnited States than in Canada and, in addition,has experienced a faster growth rate (Chart 5).This situation alone accounts for the wideninggap in Canada-US manufacturing productivity(Sharpe 1999; Rao and Tang 2001).6

Variance in Sectoral Productivity Levels and Growth7

Productivity levels and growth ratesvary greatly across industries. Industry-leveldifferences in output per hour reflect a num-ber of factors, including differences in capi-tal intensity of production, quality of humancapital, the existence of resource rents andcompetitive conditions. Industry-level dif-ferences in productivity growth rates reflectthe pace of technological change in the sec-tor, the ability to mechanize production, thepace of investment in both physical andhuman capital and competitive conditions.

In the 1989-2000 period, agricultureenjoyed the most rapid growth in output perhour (see Table 2), at an average 4.6 percentper year, followed by communications (2.3 per-

51


TABLE 6

Productivity Levels, Canada Relative to the US

US=100 in all years

Real GDP per Output per HourEmployed Person in Manufacturing

1976 88.1 90.9 1977 89.1 92.1 1978 89.0 92.5 1979 88.7 91.4 1980 87.9 89.8 1981 86.8 89.6 1982 88.1 86.3 1983 87.2 87.5 1984 87.4 92.3 1985 87.6 91.3 1986 86.0 88.7 1987 86.5 82.7 1988 86.4 79.7 1989 85.6 82.9 1990 84.7 83.2 1991 84.0 83.1 1992 83.4 85.1 1993 84.0 87.5 1994 85.1 87.0 1995 85.0 83.2 1996 84.7 80.1 1997 83.8 79.5 1998 83.3 76.1 1999 82.6 72.3 2000 82.6 69.3 2001 82.5 67.3

Source: Output per hour series based on an estimateof 79.5 for the 1997 benchmark year by Bart van Ark,Inklaar and Timmer (2000), applied to data fromInternational Comparisons of Manufacturing Productivityand Unit Labour Costs, April 5, 2001, News, Bureau ofLabor Statistics, US Department of Labor; andAggregate Productivity Measures for Canada. Outputper worker is from the CSLS income and productivitydatabase, August 2002.

Perc

enta

ge

82

83

84

85

86

87

88

89

90

91

92

2001

1996

1991

1986

1976

1981

GDP per hour

GDP per worker

CHART 3

Labour Productivity Trends, Canada Relativeto the US, 1976-2001

Source: CSLS income and productivity database, basedon data from the National Accounts and Labour ForceSurvey for Canada and the National Income and ProductAccounts and Current Population Survey for the UnitedStates; August 2002.


cent); manufacturing (2.3 percent); retail trade(2.0 percent); transportation and storage(1.8 percent); wholesale trade (1.6 percent) andmining, quarrying and oil wells (1.5 percent).Three sectors experienced negative productiv-ity growth in the 1990s: logging and forestry(-0.1 percent per year); construction (-0.6 per-cent); and fishing and trapping (-0.7 percent).

PROSPECTS FOR PRODUCTIVITYGROWTH

As noted in the introduction to thispaper, the shift from a 1.0-1.5-percent trendproductivity world to a 2.0-2.5-percent worldwould have implications for a large number ofeconomic and social variables. Real wages andincomes would be higher, as would tax rev-

enues, allowing expansion of social programs— political circumstances permitting. If wewere all collectively richer, fewer tradeoffsbetween competing economic and social endswould be necessary.

From 1995 to 2001, the business sectorin the United States enjoyed output-per-hourgrowth of 2.4 percent per year, up 1 percent-age point from the 1973-95 period but belowthe 3.3 percent of the 1947-73 period. Themillion-dollar question for productivity analystsis whether the post-1995 acceleration is a per-manent or temporary development. Opinionsdiffer on this issue, depending in large part onhow one views the causes of the acceleration. Foreconomists such as Robert J. Gordon, whobelieve that much of the improvement in pro-ductivity growth is due to the strong cyclicalconditions in the second half of the 1990s, andwho see productivity growth narrowly concen-

52

Andrew Sharpe

TABLE 7

Growth in Output Per Hour in Manufacturing,Industrial Countries

Average annual rates of change

1981- 1989- 1989- 1995- 1981-1989 2000 1995 2000 2000

Canada 2.25 2.03 3.00 0.88 2.12US 3.25 3.71 2.93 4.64 3.51Japan 3.89 3.74 3.85 3.62 3.81France 3.74 3.93 3.76 4.14 3.85UK 5.28 2.77 3.31 2.12 3.82Italy 3.08 2.35 3.38 1.13 2.66Germany (Unified) n/a n/a n/a 2.85 n/a

G7 average 3.58 3.09 3.37 2.77 3.30

Belgium 4.13 2.91 2.63 3.25 3.42Denmark 1.02 n/a n/a n/a n/aNetherlands 4.08 2.55 3.36 1.59 3.19Norway 2.69 0.88 1.26 0.43 1.64Sweden 3.15 4.45 4.58 4.29 3.90

Note: Only the 1995-2000 average includes Germany. Dataon Unified Germany is only available beginning in 1991.Source: International Comparisons of ManufacturingProductivity and Unit Labour Costs Trends, April 2002,Bureau of Labor Statistics, US Department of Labor.

Perc

enta

ge0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

Swed

en

Fran

ceUK

Jap

anUS

Bel

giu

m

Net

h.

Ital

y

Can

ada

No

rway

CHART 4

Output Per Hour in Manufacturing, Selected OECD Countries, 1981-2000(Average Annual Growth Rates)

Source: International Comparisons of ManufacturingProductivity and Unit Labour Costs Trends, April 2002,Bureau of Labor Statistics, US Department of Labor.


trated in IT-producing industries with mini-mum productivity gains arising from invest-ment in IT-using industries, the outlook forproductivity is not rosy. They see productivi-ty growth reverting to 1.5 percent, just slight-ly above the trend of the 1973-95 period. Onthe other hand, economists such as DaleJorgenson and Martin N. Baily, who believesomething fundamental happened to changeproductivity behaviour and raise trend pro-ductivity in the mid-1990s, and believe thischange was associated with IT, the outlook forproductivity growth is more favourable,although down somewhat from the unsustain-able pace in the second half of the 1990s. Baily(2002) projects annual productivity growth in

the range of 2.2-2.7 percent for the remainingyears of this decade.

There was virtually no acceleration in pro-ductivity growth in the second half of the 1990sin Canada, at least relative to the first half of thedecade. Business-sector output per houradvanced 1.6 percent, well below that recordedin the United States. The issue for Canada iswhether it will follow the US lead and see a pick-up in productivity growth. Again, economistsare divided on the issue. For example, Sharpe andGharani (2002) project business-sector output-per-hour growth of a minimum of 2 percent peryear this decade, based on the view that Canadatends to lag behind the United States and thatthe productivity-augmenting effect of IT invest-ment will finally have a payoff, just as it did inthe United States in the late 1990s. In otherwords, technological catch-up will provide thebasis for stronger productivity growth in Canada.Wilson and Dungan (2002), on the other hand,foresee little acceleration in productivity growth,arguing that the smaller size of the IT sector inCanada will prevent this country from enjoyingUS growth rates.

CONCLUSION

This paper has presented an overview ofproductivity concepts, measurement issues, andproductivity trends and prospects in a Canadiancontext. This material provides background foran understanding of productivity issues addressedin the papers in this volume. A number of mes-sages or themes emerge. (1) Productivity is acomplex, nuanced concept with several dimen-sions, including such aspects as the differenttypes of productivity and the distinction betweengrowth rates and levels. (2) For a number of rea-sons, the measurement of productivity is fraught

53


Perc

enta

ge

2

4

6

8

10

12

14

16

18

20

2000

1996

1992

1988

1984

1980

1976

Canada electrical & electronic product industries

Canada machinery industries (excluding elec. machinery)

United States electronic and other electric equipment

United States industrial machinery and equipment

CHART 5

Selected Industries Real GDP Share inManufacturing, Canada-US Comparison

Source: Statistics Canada, GDP by Industry, and theNational Income and Product Accounts, US Bureau ofEconomic Analysis.


with problems and the methodologies chosen toconstruct productivity estimates can greatlyinfluence those estimates. (3) The determinantsof productivity growth are multi-faceted, andinclude both economic and social variables. (4) Canada’s productivity performance in recentyears has been mediocre, particularly comparedto that of the United States. (5) The interests ofall Canadians converge on the importance of pro-ductivity growth as it is the basis of sustainedreal-income growth. (6) Productivity should notbe oversold as a panacea for society’s problems.Productivity growth and the additional incomeit generates are necessary conditions, but not theonly ones, for improving the quality of life andincreasing the well-being of Canadians.

NOTES

The author would like to thank Someshwar Rao,Daniel Schwanen, France St-Hilaire, an anonymousreferee and participants in the 25-26 January 2002IRPP-CSLS authors’ workshop for comments on anearlier version of the paper. He would also like tothank Jeremy Smith for excellent research assistance.The paper draws on earlier work by the author in theproductivity area, including Sharpe (1998, 2002a),Rao and Sharpe (2002), and Osberg and Sharpe(1998, 2002a and 2002b).

1 It should be noted that with the recent adoption byStatistics Canada of chain-Fisher indexes, thecomponents of real GDP no longer add up exactly toreal GDP.

2 The construction of PPPs requires comparisons ofprices across countries. Internationally consistentsurveys on the prices of goods and services inexpenditure categories have been carried out by theOECD on a regular basis, so estimates of PPPs forGDP and consumer expenditure are available.However, there are no surveys of product prices, soestimates of PPPs for industry output are muchharder to compile.

3 Of course, if materials in addition to labour andcapital are used as inputs, the bias disappears.

4 Nordhaus (1997) provides a fascinating account ofthe history of the price of light, showing that on aquality-adjusted basis it has experienced an

enormous long-term decline. When the quality-adjusted price of light is integrated into priceindexes, he finds, over the 1800-1992 period livingstandards have increased by between 40 (low-biasassumptions) and 190 (high-bias assumptions) timesinstead of the conventionally estimated factor of 13.The implications of quality adjustment for thequantification of trends in living standards are verygreat.

5 See the volumes edited by Griliches (1992) andDiewert et al. (1999) for papers on productivitytrends in a number of service industries.

6 For discussion of additional factors affecting theCanada-US manufacturing productivity gap, seepapers from the CSLS conference on the Canada-USmanufacturing productivity gap posted atwww.csls.ca under past events.

7 In July 2002, the Centre for the Study of LivingStandards updated its comprehensive productivitydatabase on the basis of the North AmericanIndustry Classification System (NAICS), replacingearlier estimates based on the 1980 StandardIndustrial Classification (SIC). Using StatisticsCanada data on labour input, capital stock andoutput data, this database provides estimates oflabour productivity levels (both output per workerand output per hour), capital productivity levels andTFP indexes for the years 1976-2001 inclusive forCanada (1976-98 on a 1980 SIC basis and 1987-2001 on a NAICS basis) and 1984-2001 for the 10provinces (1984-98 on a 1980 SIC basis and 1997-2001 on a NAICS basis), giving as much industrydisaggregation as confidentiality rules permit. Thisdata base (www.csls.ca) is freely accessible to thepublic.

REFERENCES

Abramovitz, M., and P.A. David. 1996. “Convergence andDeferred Catch-up: Productivity Leadership and theWaning of American Expectionalism.” In The Mosaicof Economic Growth, ed. R. Landau, T. Taylor, and G.Wright. Stanford, CA: Stanford University Press.

Baily, M.N. 2002. “The New Economy: Post Mortem orSecond Wind?” Journal of Economic Perspectives 16(2):3-22.

Boskin, M., and L.J. Lau. 1992. “Capital. Technology andEconomic Growth.” In Technology and the Wealth ofNations, ed. N. Rosenberg, R. Landau, and D.Mowery. Stanford, CA: Stanford University Press.

Denison, E. 1962. The Sources of Economic Growth in theUnited States and Alternatives Before Us. Washington:Brookings Institution.

54

Andrew Sharpe


Department of Finance. 1994, October. A NewFramework for Economic Policy. Ottawa: Departmentof Finance.

Diewert, E., A. Nakamura, and A. Sharpe. 1999. ServiceSector Productivity and the Productivity Paradox, SpecialIssue of the Canadian Journal of Economics, April.

Easterlin, R.A. 1974. “Does Economic Growth Improvethe Human Lot? Some Empirical Evidence.” InNations and Households in Economic Growth, ed. P.A.David, and M.W. Reder. New York: AcademicPress, pp. 89-125.

—— 1995. “Will Raising the Incomes of All Increase theHappiness of All?” Journal of Economic Behaviour andOrganization 27(1):35-48.

Gordon, R. 2000. “Does the ‘New Economy’ Measure Upto the Great Inventions of the Past?” Journal ofEconomic Perspectives 14(4): 49-74.

Griliches, Z. ed. 1992. Output Measurement in the ServiceSectors. NBER Studies in Income and Wealth, Vol.56. Chicago: University of Chicago Press.

Harris, R.G. 1999, December. “Determinants of CanadianProductivity Growth: Issues and Prospects.”Discussion paper #8. Ottawa: Industry Canada.

Landau, R., T. Taylor, and G. Wright. 1996.“Introduction.” In The Mosaic of Economic Growth, ed.R. Landau, T. Taylor, and G. Wright. Stanford, CA:Stanford University Press.

Lipsey, R.G., and K. Carlaw. 2000. “What Does TotalFactor Productivity Measure?” InternationalProductivity Monitor 1:31-40. Available: www.csls.ca

Nordhaus, W. 1997. “Do Real Output and Real WageMeasures Capture Reality? The History of LightSuggests Not.” In The Economics of New Goods, ed. T.Bresnahan, and R.J. Gordon. NBER. Chicago:University of Chicago Press.

Osberg, L. 1985. “The Measurement of Economic Well-Being,” In Approaches to Economic Well-Being, ed. D.Laidler. Research Vol. 1, MacDonald RoyalCommission. Toronto: University of Toronto Press.

Osberg, L., and A. Sharpe. 1998. “An Index of EconomicWell-Being for Canada.” Research report 99-1,Applied Research Branch, Strategic Policy. Ottawa:Human Resources Development Canada.

—— 2002a. “The Index of Economic Well-Being.”Indicators: The Journal of Social Health, 1(2): 24-62.

—— 2002b. “An Index of Economic Well-Being forSelected OECD Countries.” Review of Income andWealth, September.

Panel to Review Productivity Statistics. 1979. Measurementand Interpretation of Productivity. Washington, D.C.:National Research Council.

Pilat, D. 2001. “Productivity Growth in the OECD Area:Some Recent Findings.” International ProductivityMonitor 3:32-43. Available: www.csls.ca

Putnam, R.D. 2001. Bowling Alone: The Collapse andRevival of American Community. New York:Touchstone.

Rao, S., and A. Sharpe, eds. 2002. Productivity Issues inCanada. Calgary: University of Calgary Press.

Rao, S., and J. Tang. 2001. “The Contribution of ICTs toProductivity Growth in Canada and the UnitedStates in the 1990s.” International Productivity Monitor3:3-18. Available: www.csls.ca

Romer, P. 1990. “Endogenous Technical Change.” Journalof Political Economy 98:S71–S102.

Sharpe, A. 1998. Productivity: Key to Economic Success.Centre for the Study of Living Standards for theAtlantic Canada Opportunities Agency. Ottawa andMoncton, NB: CSLS and ACOA. Available:www.csls.ca

—— 1999. “What Do the Canada-US ProductivityNumbers Mean?” Policy Options May:29-34.

—— 2001. “Determinants of Trends in Living Standardsin Canada and the United States, 1989-2000.”International Productivity Monitor 2:3-10. Available:www.csls.ca

—— 2002a. “Recent Productivity Developments in theUnited States and Canada: Implications for theCanada-US Productivity and Income Gaps.”International Productivity Monitor 4:3-14. Available:www.csls.ca

—— 2002b. “The Contribution of Productivity toEconomic Well-Being.” In Productivity Issues inCanada, ed. S. Rao, and A. Sharpe. Calgary:University of Calgary Press.

Sharpe, A., and L. Gharani. 2002. “Trend Productivityand the New Economy.” In Productivity Issues inCanada, ed. S. Rao, and A. Sharpe. Calgary:University of Calgary Press.

Sharpe, A., S. Rao, and J. Tang. 2002. “Perspectives onNegative Productivity Growth in Service SectorIndustries in Canada and the United States.” Paperpresented at the Services Productivity Workshop,Brookings Institution, Washington, D.C.

Solow, R. 1957. “Technical Change and the AggregateProduction Function.” Review of Economics andStatistics 39(3):312-320.

Statistics Canada. 1994. The Size of the UndergroundEconomy in Canada. Cat. 13-603. Ottawa: StatisticsCanada.

Stiroh, K. 2001. “Is IT Driving the US ProductivityRevival?” International Productivity Monitor 2:31-36.

55



Van Ark, B., R. Inklaar, and M. Timmer. 2000. “TheCanada-US Manufacturing Productivity GapRevisited: New ICOP Results.” Groningen Growthand Development Centre.

Wilson, T., and P. Dungan. 2002. “Productivity in the NewEconomy.” In Productivity Issues in Canada, ed. S. Rao,and A. Sharpe. Calgary: University of Calgary Press.

56

Andrew Sharpe


productivity - concept & trends

Business

productivity of labour

productivity issue

hour growth

weaker growth

real income growth

productivity price values

analysts productivity

growth rate of average