possibility of general slowdown or country-specific catch-up effects in the rate of growth of labour efficiency. We turn now to our model and its results.
53 pages
50 KB – 53 Pages
PAGE – 1 ============
NBER WORKING PAPER SERIESINTERNATIONAL COMPARISON OF THESOURCES OF PRODUCTIVITY SLOWDOWN 197’3Š1982John HelliwellPeter SturmGerard SalouWorkingPaper No. 1165NATIONALBUREAU OFECONOMIC RESEARCH1050Massachusetts AvenueCambridge, MA 02138September 19814The research reported here is part of the NBER’s research programin International Studiesandproject in Productivity (WorldEconoiy). Any opinions expressed are those of the authorsandnotthose of the National Bureau of Economic Research.
PAGE – 2 ============
NBER Working Paper #1465September 1984International Comparison of the Sources ofProductivity Slowdown 1973-1982ABS1RACThis paper uses an integrated model of aggregate supply to analyze thepost-1973 slowdownin productivitygrowth in the seven major OECD economies.Factor substitution, unexpected demand changes, profitability, and inventorydisequilibrium all contribute to the explanation, which is based on athree-factor nested aggregate production function, including energy, andpostulating Harrod-neutral disembodied technical progress. The model is firstapplied separately to the seven countries assuming constant (thoughcountry-specific) rates of technical progress. This model provides empiricalevidence that this rate of progress has in fact slowed down for several of thefaster-growing countries, even after adjusting for factor substitution andcyclical factors. The model is therefore re-estimated, and the sources ofproductivity decline recalculated, on the hypothesis that rates of efficiencygrowth in other countries are converging to those in the United States.John HelliwellPeter Sturm and Gerard SalouDepartment of EconomicsO.E.C.D.University of British ColumbiaGrowth Studies DivisionVancouver, B.C.2, rue Andre PascalCanada V6T lY2F-750l6 ParisFrance(064) 228 4953
PAGE – 3 ============
INThRNATIONAL (DMPARISON OF ThE SOURCES OF WODUCTIVITYSLOWDOWN 1973-19821.Introductjon*What happened to productivity in the 1970s? Why has post-1973 growthin output per employed person in the major OECD economies been so much slowerthan it was between 1962 and 1973? These are key questions in economichistory, and their answers have important implications for economic policy.Among the many papers devoted to this topic, analysing the problem either atan aggregate level or for disaggregated industries, most have concluded thatthere is a large residual to be explained, or have had to use rough guessesand often inconsistent methods to allocate the blame among a large variety ofpossible causes. Our study is based on the assumption that enough post-1973history is now on record to permit using a more consistent and systematicapproach, in which a comparable yet fairly simple analytic framework isapplied to data from a number of countries. Errors in the explanationgenerated by this framework will show the size of the remaining puzzlerequiring appeal to other models or additional influences.In our analysis of the post-1973 slowdown in labour productivity growthin the seven major OECD economies, we apply an aggregate three-factor model ofproduction behaviour that treats factor substitution, unexpected demandchanges, deviations from desired inventory levels and profitability in anintegrated and consistent manner. We proceed in two stages in a way thatillustrates two distinct elements of comparative macroeconomics: the first isthe use of comparable theory and data to assess productivity growth in anumber of countries, and the second is the development of an internationalframework to explain the inter-country differences and possible futureevolution of long-term productivity growth.
PAGE – 4 ============
-3-In the first stage, we apply our model separately to comparable sets ofdata for each of the countries, and use the results to disentangle the effectsof energy price changes from those of unexpected changes in demand andprofitability in explaining each country’s post-1973 slowdown in the rate ofgrowth of output per employed person. The model used has a nested long-termproduction function in which capital and energy are combined in a vintage CESinner function, and this bundle then enters a Cobb-Douglas outer function withefficiency units of labour.At the first stage of the research long-runtechnical progress is taken to be Harrod-neutral, occurring at a constantannual rate.Actual production is determined by a behavioural utilizationrate equation in which the ratio of actual output to that determined by theproduction function (with given technical progress and normal utilizationrates for employed factors) is explained by unexpected changes in finaldemand, abnormal profitability, and the discrepancy between actual and targetinventory levels. For all countries there is strong evidence that changes indemand and profitability have led to important declines in the rate ofutilization of employed factors, and hence in the rate of growth of measuredtotal factor productivity and output per employed person, between 1973 and1982. The post-1973 factor mix has also changed substantially, to use morelabour and less energy, thus leading to further reductions in the rate ofgrowth of output per employed person.The first stage of our research also showed large internationaldifferences in the underlying rate of Harrod-neutral technical progress, withevidence of a slow-down in its rate of growth, even after adjusting for factormix and cyclical factors, in those countries with initially lower but rapidlyrising levels of labour efficiency.We also found evidence of somesignificant increases in labour income as a share of total factor payments in
PAGE – 5 ============
-4-countries where the cost of labour had risen faster than the rate ofHarrod-neutral technical progress, thus suggesting that the elasticity ofsubstitution between labour and the capital-plus-energy bundle may not be ashigh as assumed by the use of a Cobb-Douglas outer function.These results led us to a second stage of research in which the initialmodel was further developed, on the basis of the first stage results, toinclude an explanation of the international transmission and convergence oftechnical progress.As a separate development, we have also applied thetwo-level production structure using the CES form at both levels, andincluding simultaneous estimation of the derived factor demand equations(Helliwell, Sturm and Salou, 1984).In this paper, we shall concentrate onthe convergence model using the Cobb-Douglas outer function, with referenceswhere appropriate to the two-level CES results. Thismarks a second distinctphase of comparative international macroeconomics, where the uniform modellingand comparison of production in national economies leads to a more generalframework in which increasing trade in goods and services, and especiallyinformation and technology, tends toward international convergence of rates ofgrowth of labour efficiency. Since the United States shows the highest leveland the slowest rate of growth in Harrod-neutral technical progress of all ofthe seven countries (except for the United Kingdom), and does not showevidence of decline from the 1960s to the 1970s and l980s (after adjusting forchanges in cyclical position), its experience is used to approximate the rateof growth of technical progress to which other countries are converging.
PAGE – 6 ============
Š5-The logic of our approach suggests that we present the first stageresults in some detail, including the evidence that led us to proceed to thesecond stage. This will permit the second-stage results to be presented moreconcisely, since their underlying framework and basic pattern are similar, andwill permit a clearer view of what is at stake in moving from a constant to avariable rate of technical progress.2.cal and Theoretical_BackgroundTable1 shows the average annual growth of output per employed person,GDP per capita, and changes in various input and factor price ratios, in eachof the seven major OECD countries between 1973 and 1982, with the growth ratesbetween 1962 and 1973 given for comparison.The average annual growth ofoutput per employed person fell by 2.5 percentage points from the pre-1973 tothe post-1973 period.In all cases the average growth was less than half asIhighover the 1973-1982 period as over the 1962-73 period.In the UnitedStates and Canada, the countries with the lowest rates of growth of outputper employed person 1962-1973, there was almost no growth in output peremployed person from 1973 to 1982. Reference to statistics for productivitygrowth in industrialised countries over a century or more cautions that thetwenty-five years of high productivity growth prior to 1973, especiallyoutside North America, may be the exceptional performance (1).Thus anysatisfactory explanation of the post-1973 slowdown that implies a return to1962-73 average growth rates of output per employed person should also beconsistent with the permanent transition to these growth rates from thoseexperienced in the first half of the century.
PAGE – 8 ============
Š7-.Our approach has some similarities with that of Bruno in hiscomparative studies of productivity slowdown in manufacturing (1984) and inthe private economy as a whole (1982).Like us, he assumes an underlyingthree-factor model of production, and permits short-term deviations of actualoutput from normal output. He does, however, not use his production structureto define a normal output series, and hence is not able to separate cleanlythe effects of factor substitution from those of factor utilization. From hiscross-section equation, he concludes (1982, p. 99) that about half of theslowdown in private-sector productivity growth was due to import priceincreases and half to demand slowdown.His estimated import price effectcombines the influence of factor substitution with the impact of profitabilityon utilization which is separately estimated in our model.Sylos-Labini (1984) explains productivity growth in Italy and theUnited States in terms of output growth, changes in the wage rate relative tothe price of machinery, and current and lagged investment rates. The firstterm probably captures mainly cyclical effects, while the wage ratio probablycombines some factor substitution with some profitability effects.Bytreating the production decision and the definition of potential output in amore explicit way, we are able to disentangle these separate influences moreclearly, and then later to use our integrated framework to test for thepossibility of general slowdown or country-specific catch-up effects in therate of growth of labour efficiency.We turn now to our model and its results.
PAGE – 9 ============
-8-3.The Two-LevelProduction Structure and the Influence of EngçThebasicproduction structure employed in the first stage of ourresearch is explained in detail in the technical annex,It consists of athree factor nested aggregate production function.The inner CES functioncombines capital and energy into a vintage capital-energy bundle which iscombined with efficiency units of labour in an outer Cobb-Douglas function todefine potential output.Estimates of the relevant production functionparameters are presented in Table 2. Energy is singled out for treatment as aseparate factor of production, in addition to capital and labour, since itsrelative price moved so differently in the pre-1973 and post-1973 periods,leading to changes in output per employed person that need to be disentangledfrom the other causes of changes in labour productivity.Since energy istreated as a separate factor of production, the output concept has to beenlarged to include net energy imports (2).Our choice of a nested structure with capital and energy combined inthe inner function follows earlier research conclusions (e.g. Berndt and Wood1979) that the separability assumptions implied by this nesting are moreplausible than the alternatives.We restricted our choice to productionfunction structures that can be represented explicitly by their primary forms,as well as by their dual cost functions, so that we are able to have anexplicit factor-based measure of potential output, and hence can decomposeoutput per employed person into two key ratios:Q/NE =(Q/QSv).(QSVINE)(1)Where:Q= actualoutput,NE=numberof employed persons, andQSV=outputdefined by the production function, with all quantitiesof employed factors operating at normal utilization rates.
PAGE – 10 ============
-9-.We employ a flexible vintage structure for the capital and energybundle. Some vintage structure is likely to be required since much capitalequipment and many buildings are designed to embody a certain energyrequirement. However, it would not be appropriate to apply a rigid vintagemodel, in which energy requirements were held fixed at their initial levelsuntil the capital is scrapped, because there is much in the way of adjustmentand retrofitting that can be done to change the energy use of the existingcapital stock. We therefore designed our vintage model so that the degree ofretrofitting should be a parameter to be estimated simultaneously with thelong-term elasticity of substitution between capital and energy in the CESbundle. The distribution parameters in the inner CES bundle are derived fromthe assumption that the actual and desired energy/capital ratios are equal onaverage over the ainple period, and by scaling the mean of KEV to have thesame mean as K, so that the derived CES dual cost index for capital and energy:PKE=(b.PK +c5.PE)11’5)(2)measures the cost (after vintage effects have been worked out) of owning andproviding energy for one unit of the capital stock.For any given pair ofvalues for the retrofitting parameter (LU in Table 2) and the elasticity ofsubstitution (s) between capital and energy, equation 8 in the technical Annexdefines the vintage energy requirement EV (3). Using EV as a predictor of E,Table 2 shows the maximum likelihood pairs of Ri and s for each of the sevencountries. The long-term elasticity of substitution ranges from .6 to .95(the highest value considered).If those values seem high, it should beremembered that they refer to energy as it is priced for final users,including all taxes and distribution charges, and that those prices have movedless than proportionately in response to changes in world crude oil prices,
PAGE – 11 ============
-10-even in those countries where world crude oil price changes were passed onimmediately to final users.There is wider variation in values for theretrofitting parameter, which in some countries is not very preciselydetermined statistically, and in any case is likely to get confused with thenumerous non-price policies that several countries adopted to speed theconservation of energy, and especially of crude oil, following the world oilprice shocks of 1973-74 and 1979-80. These latter policies may, for example,have combined with the increasing availability of natural gas service toexplain the estimated high speed of the U.K. response.Examination of the various country panels in Figure 1 shows why theelasticity of substitution and the retrofitting parameter both tend to behigh, and also why there is a trade-off between RI and s in the likelihoodsurfaces for several countries. In all seven countries there were matchingdecreases in relative energy prices and in capital/energy ratios prior to1973, with the trends reversed thereafter.How big were the effects of energy price changes on factorsubstitution, and hence on the rate of growth of output per employed person?The answer to this question depends on the properties of each country’sproduction function. So far we have explained the method for obtaining theparameters of the inner CES function, and for defining the quantity and priceof the bundle of capital plus energy. The higher price of energy leads to asubstitution of capital for energy and also to a rise in the price index forcapital plus energy, leading to substitution of labour for the bundle ofcapital and energy. The size and speed of this substitution depend on theparameters of the outer function and on the speed with which actual employmentand the capital stock respond to changes in desired factor proportions.
50 KB – 53 Pages