The Decoupling of Median Wages from Productivity in OECD Countries

Transcription

1 The Decoupling of Meian Wages from Prouctivity in OECD Countries Cyrille Schwellnus, Anreas Kappeler an Pierre-Alain Pionnier OECD 1 ABSTRACT Over the past two ecaes, aggregate labour prouctivity growth in most OECD countries has ecouple from real meian compensation growth, implying that increasing prouctivity is no longer sufficient to raise real wages for the typical worker. This article provies a quantitative escription of ecoupling in OECD countries over the past two ecaes, with the results suggesting that it is explaine by eclines in both labour shares an the ratio of meian to average wages (a partial measure of wage inequality). Labour shares have ecline in about two thirs of the OECD countries covere by the analysis. However, the contribution of labour shares to ecoupling is smaller if sectors are exclue for which labour shares are riven by changes in commoity an asset prices (primary an housing sectors) or by imputation choices (non-market sectors). The ratio of meian to average wages has ecline in all but two of the OECD countries covere by the analysis an appears to reflect isproportionate wage growth at the very top of the wage istribution rather than stagnating meian wages. The causes of these evelopments will be analyse in follow-up research. In the long run, raising prouctivity is the only way to raise living stanars, with real wages being the most irect mechanism through which the benefits of prouctivity growth are transferre to workers. Over the past two ecaes, however, aggregate labour prouctivity growth in most OECD countries has ecouple from real meian compensation growth. 2 Increasing prouctivity no longer appears to be sufficient to raise real wages for the typical worker, suggesting that there is a role for public policies to support a broaer sharing of the benefits of prouctivity gains in the economy. This article analyses the extent of ecoupling of wages from prouctivity growth in OECD countries over the past two ecaes. It analyses 1 Cyrille Schwellnus an Anreas Kappeler are economists in the Economics Department of the OECD. Pierre- Alain Pionnier is an economist of the Statistics Directorate of the OECD. The authors woul like to thank Anrea Bassanini, Orsetta Causa, Antoine Goujar, Mikkel Hermansen, Alexaner Hijzen, Yosuke Jin, Christine Lewis, Catherine Mann, Giuseppe Nicoletti, Rory O Farrell, Nicolas Ruiz, Jean-Luc Schneier, Paul Schreyer. Peter van e Ven, an two anonymous referees for helpful iscussions an suggestions. The support of Sarah Michelson in putting together the ocument is gratefully acknowlege. The statistical ata for Israel are supplie by an uner the responsibility of the relevant Israeli authorities. The use of such ata by the OECD is without prejuice to the status of the Golan Heights, East Jerusalem an Israeli settlements in the West Bank uner the terms of international law. s: cyrille.schwellnus@oec.org; anreas.kappeler@oec.org; an pierre-alain.pionnier@oec.org. 2 Real compensation growth is base on the value ae eflator. I N T E R N A T I O N A L P R O D U C T I V I T Y MON I T O R 44

2 whether evelopments at the macro level mainly reflect changes in labour shares or changes in wage inequality. Existing stuies have mainly focuse on the Unite States (Bivens an Mishel, 2015) an Canaa (Sharpe et al., 2008), fining that in these countries there has been substantial ecoupling of real meian wages from labour prouctivity over the past three ecaes. The only recent cross-country stuy (Uguccioni an Sharpe, 2017) fins that there are large cross-country ifferences in ecoupling of real meian wages from prouctivity. The main contributions of this article are to (i) provie evience on ecoupling for the broaest possible range of OECD countries an (ii) to aress a number of measurement issues that are likely to bias estimates of ecoupling. The analysis shows that for the covere OECD countries as a whole, total-economy ecoupling over the perio is explaine by eclines in both total-economy labour shares an the ratio of meian to average wages (a partial measure of wage inequality). These eclines are fully accounte for by pre evelopments. Excluing sectors for which labour shares are riven by changes in commoity an asset prices or for which labour shares are riven by imputation choices (primary, housing an non-market sectors) lessens the contribution of labour shares to ecoupling. For a number of countries, eclines in total-economy labour shares reflect increases in housing rents, which are relate to increases in house prices. For commoity-proucing countries, eclines in total-economy labour shares largely reflect increases in commoity rents. These are, in turn, relate to price increases on global markets on which national policies have limite leverage. While labour shares have ecline significantly in about two thirs of the analyse OECD countries covere in this article, all but two countries have experience significant eclines in the ratio of meian to average wages over the past two ecaes. The increase in wage inequality as measure by the ecoupling of meian from average wage growth appears to reflect isproportionate wage growth at the very top of the wage istribution. While wage growth at the 90th percentile (top 10 percentile) of the wage istribution has been similar to growth at the meian, average wage growth for the top 1 per cent has exceee growth at the meian by a multiple. This article is organise as follows. The first section escribes the conceptual framework for ecomposing macro-level ecoupling into contributions from labour share an wage inequality evelopments, an provies escriptive evience for the covere OECD countries. Section 2 investigates the role of the primary, housing, an non-market sectors as well as capital stock epreciation in total-economy labour share evelopments. Sector-level ata on wage inequality for the sample of OECD countries covere by the analysis are not available so that no such analysis can be conucte for the wage inequality component. Section 3 nonetheless provies a more isaggregate perspective on wage inequality evelopments by analysing the role of isproportionate wage growth of top earners. Section 4 conclues. Macro-level Decoupling: Overview Framework Conceptually, macro-level ecoupling between real compensation growth of the meian worker an labour prouctivity growth can be ecompose into the growth ifferential between average compensation an labour prouctivity an the growth ifferential between meian an average compensation. Using the notation per cent X to enote the per cent growth rate of X, macro-level ecoupling in this article is efine as follows: 45 N U M B E R 32, S P R I N G 2017

3 Y P Decoupling per Y cent L W per me cent P Y (1) where Y enotes nominal value ae, P Y enotes the value ae price, L enotes hours worke an W me enotes the nominal meian wage. The first term on the right-han-sie is labour prouctivity growth an the secon term is real meian wage growth in terms of the value ae price. By aing an subtracting real average wage growth per cent( W avg P Y ) equation (1) can be re-written as follows: Decoupling per cent Wavg P Y per cent Y PY L per cent Wme P Y + (2) where the first term in square brackets enotes the growth ifferential between labour prouctivity an the real average wage an the secon term in square brackets enotes the growth ifferential between the real average an the real meian wage. The growth ifferential between labour prouctivity an the real average wage can be approximate as per cent( ( W avg L) Y), i.e. the per cent ecline in the labour share. The per cent Wavg P Y growth ifferential between the real average an the real meian wage can be re-written as per cent( W avg W me ), i. e. t h e p e r c e n t increase in the ratio of the average to the meian wage. A high ratio of the average to the meian wage typically reflects high compensation at the top of the wage istribution, so that it can be interprete as a partial measure of wage inequality. In this article, compensation an value ae are eflate by the same value ae price inex 3 so that ecoupling between real average compensation an labour prouctivity reflects eclines in labour shares. 4 Deflating compensation by a consumption eflator an value ae by the value ae eflator woul rive an aitional wege between meian wage growth an prouctivity growth (Uguccioni an Sharpe, 2017). This wege is largely riven by countries' external terms of trae since the consumption eflator inclues importe goos whereas the value ae eflator inclues only omestic prouction. 5 For the countries covere by the analysis as a whole, the growth ifferential between real wages base on a consumption eflator an the value ae eflator has been limite an epens on whether the Final Consumption Expeniture (FCE) eflator from the national accounts or the Consumer Price Inex (CPI) is use in the analysis (Appenix Chart A1). 6 However, for a number of commoity-importing countries, real wages base on a consumption eflator woul have grown less than real wages 3 Note that the value ae price inex is ifferent from the GDP price inex. GDP inclues taxes less subsiies on proucts whereas value ae oes not. Value ae is thus a more relevant concept to stuy the relation between labour prouctivity an wages. 4 Felstein (2008) argues that wages an value ae shoul be eflate by the same output price inex, as the basic economic relation is between nominal wages an the marginal revenue prouct of labour. 5 Despite the exclusion of this wege, the analysis here oes cover the effects on the labour share an wage inequality of changes in the terms of trae. Only the wege between the consumption an value ae eflator per se is exclue from the analysis. 6 Differences between the FCE eflator an the CPI mainly reflect the treatment of impute rents of home owner-occupiers. While both actual an impute rents are inclue in househols' final consumption expeniture for all countries, impute rents are not inclue in the basket of goos an services unerlying the CPI for a number of countries. The FCE eflator is therefore more comparable across countries t h a n t he C P I. S ee A p p en ix C h art A1 at: ht t p : / / w ww. c sls.ca/ipm/32/ Schwellnus_Kappeler_Pionnier%20Appenix.pf I N T E R N A T I O N A L P R O D U C T I V I T Y MON I T O R 46

4 base on the value ae price inex irrespectively of the precise measure of the consumption eflator use in the analysis (Appenix Table A1). 7 Data Sources an Definitions The growth ifferential between labour prouctivity an real average compensation in this article is irectly compute from national accounts ata. Labour prouctivity is compute as the ratio of real gross value ae at factor cost to the number of hours worke while average compensation is compute as the ratio of real compensation to the number of hours worke in the economy. Real gross value ae at factor cost is obtaine by eflating nominal gross value ae at factor cost by the corresponing value ae eflator. Total compensation is compute as the sum of the compensation of employees an the compensation of the selfemploye, which is impute by assuming that hourly compensation of the self-employe an of epenent employees is the same at the level of iniviual inustries (see Appenix). The compensation of employees encompasses remuneration in cash an in kin an inclues employees' an employers' social contributions. Real compensation is obtaine by eflating nominal compensation by the same value ae price inex use to eflate nominal value ae at factor cost. Value ae at factor cost, compensation of employees, employment an eflators are source from the OECD Annual National Accounts atabase. The growth ifferential between average an meian compensation is approximate by the growth ifferential between gross average an meian wages, with gross wages being efine as compensation excluing employers' social contributions. The approximation is imprecise if evelopments in employers' social contributions iffer for the meian an average workers. However, more precise ata are unavailable since national accounts o not report istributional statistics. 8 Meian an average wages are source from the OECD Earnings Database that compiles ata on gross wages of full-time workers from a variety of sources, incluing househol, labour force an enterprise surveys. Gross wages encompass remuneration in cash an in kin, incluing regular payments, irregular supplements an employee social contributions. They exclue stock options, severance payments, cash government transfers, transport subsiies an employers' social contributions. Definitions are not fully consistent across countries, with ata referring to weekly or monthly wages for most countries but to hourly or annual wages for some others. 9 The labour share is efine as the ratio of total nominal labour compensation to value ae at factor cost. Given that nominal value ae is expresse at factor cost, i.e. net of taxes less subsiies on prouction, value ae can be fully ecompose into total labour compensation, incluing an impute labour compensation to self-employe workers, an total gross operating surplus (GOS), incluing the part of the mixe income of self-employe workers consiere as GOS. Aggregate wage inequality is approximate by the ratio of meian to average wages while top income inequality is approximate by the ratio of meian wages of full-time employees to the average wage of the top 1 per 7 See Appenix Table A1 at: 8 In the OECD countries covere by the analysis, employers' social contributions account for aroun 20 per cent of total compensation. 9 Ieally, meian an average wages woul be base on the istribution of hourly wages of both part-time an full-time workers. However, focusing on full-time workers has the avantage that the wage istribution is not affecte by changes in the share of part-time workers when only the istribution of weekly or monthly wages is available. 47 N U M B E R 32, S P R I N G 2017

5 Chart 1: Macro-level Decoupling in OECD Countries, , (1995 = 100) Panel B: Total Economy Excluing Primary Panel A: Total Economy Housing, an Non-Market Sectors ee eee eee e e W ZZ Note: Unweighte average of 24 OECD countries for Austria, Belgium, Germany, Finlan, Hungary, Japan, Korea, Unite Kingom; for Australia, Spain, France, Italy, Polan, Sween; for Czech Republic, Denmark; for Canaa, New Zealan; for Norway, Unite States; for Irelan; for Netherlans; for Israel; for Slovak Republic. In Panel A, all series are eflate by the total economy value ae price inex. In Panel B, all series are eflate by the value ae price inex excluing the primary, housing an non-market sectors. The sectors exclue in panel B are the following (ISIC rev. 4 classification): (1) Agriculture, Forestry an Fishing (A), (2) Mining an quarrying (B), (3) Real estate activities (L), (4) Public aministration an efence, compulsory social security (O), (5) Eucation (P), (6) Human health an social work activities (Q), (7) Activities of househols as employers (T), an (8) Activities of extraterritorial organizations an boies (U). 1. "Wage inequality" refers to total economy ue to ata limitations. Source: OECD National Accounts Database, OECD Earnings Database. ee eee eee e e cent of income earners from the Worl Wealth an Income Database (Alvareo et al., 2016). Results For the OECD countries covere in this article as a whole, there has been significant ecoupling of real meian wages from prouctivity over the past two ecaes as real meian wages have grown at a lower average rate than labour prouctivity (Chart 1). Base on the total economy measure, meian compensation woul have been aroun 8 per cent higher than observe in 2013 if it ha perfectly tracke labour prouctivity since Base on the measure excluing the primary, housing an the non-market sectors, ecoupling implies a 5 per cent loss in compensation for the meian worker over the perio The ecoupling of real meian wages from labour prouctivity for the covere OECD countries as a whole reflects both eclines in labour shares an increases in wage inequality. In line with previous stuies on ecoupling (Bivens an Mishel, 2015; Uguccioni an Sharpe, 2017), this article uses as a starting point compensation an value ae in the total economy (Chart 1, Panel A). This measure of ecoupling suggests similar contributions of eclines in labour shares an increases in wage inequality to ecoupling. However, the total economy inclues sectors for which labour shares are largely etermine by fluctuations in commoity an asset prices, such as the primary an I N T E R N A T I O N A L P R O D U C T I V I T Y MON I T O R 48

6 housing sectors, or for which labour shares are riven by imputation choices, such as the nonmarket sector. Labour share fluctuations in these sectors may have ifferent istributional implications from those in the prouction sector. Once the primary, housing an the nonmarket sectors are exclue from the analysis, the contribution of the labour share to ecoupling becomes smaller than the contribution of wage inequality (Chart 1, Panel B). While real meian wages have ecouple from labour prouctivity in the majority of countries (15 of 24) covere by the analysis, there have been large cross-country ifferences, both in the extent of ecoupling an the relative contributions of labour shares an wage inequality (Table 1). Among large OECD countries, there was significant ecoupling in Germany, Japan an the Unite States. In these countries the relative contributions of labour Table 1: Cross-country Differences in Macro-level Decoupling in OECD Countries, Annualise growth rates; excluing primary, housing an non-market sectors 3URGXFWLYLW\ 5HDODYHUDJH FRPSHQVDWLRQ 5HDOPHGLDQ FRPSHQVDWLRQ 'HFRXSOLQJ Australia Austria Belgium Canaa Czech Republic Denmark Finlan France Germany Hungary Irelan Israel Italy Japan Korea Netherlans New Zealan Norway Polan Slovak Republic Spain Sween Unite Kingom Unite States OECD G Note: See note to Chart 1 for country an year coverage. OECD an G-7 averages unweighte. Source: OECD National Accounts Database, OECD Earnings Database. 49 N U M B E R 32, S P R I N G 2017

7 Chart 2: Changes in Gross an Net Labour Shares in OECD Countries, Percentage Points Note: Three-year averages starting an ening in inicate years. OECD an G7 refer to unweighte averages for the relevant countries inclue in the figure for Australia, Canaa, France, Korea, Latvia, Mexico, Portugal; for Chile; for Unite Kingom; for New Zealan. Source: OECD National Accounts Database. shares an wage inequality iffere significantly. For instance, in the Unite States aroun 40 per cent of overall ecoupling (0.5 percentage points of 1.25 percentage points) is explaine by eclines in labour shares while this factor explains virtually all ecoupling in Japan. In a number of other OECD countries, real meian wages have grown at similar or even higher rates than labour prouctivity. These countries inclue a number of large countries, such as France, Italy an the Unite Kingom, where labour shares have increase an wage inequality has remaine broaly constant or increase only moestly over the perio. Dissecting Labour Share Developments Several recent stuies have emphasise that istributional an policy implications of labour share changes epen on the inclusion of capital epreciation an housing rents in value ae (Rognlie, 2015; Brigman, 2014). This section provies an in-epth analysis of labour share evelopments, incluing for OECD countries for which overall ecoupling cannot be compute because ata on the wage istribution are unavailable. 10 Gross or net labour shares? Even though most analyses of labour shares are base on gross value ae, only value ae 10 The labour share analysis is base on National Accounts ata only. Therefore, the country sample an time coverage changes compare to the overall ecoupling analysis, which also makes use of Labour Force Surveys. Notably, the labour share analysis inclues aitionally the year 2014 an the following countries: Estonia, Greece, Latvia, Lithuania, Luxembourg, Portugal an Slovenia. The labour share analysis also changes the time coverage for a number of countries. For instance, the labour share analysis for Norway covers , instea of ; the labour share analysis for Slovak Republic covers , instea of For further etails see Footnotes to Chart 1 an Table 2. I N T E R N A T I O N A L P R O D U C T I V I T Y MON I T O R 50

8 net of capital consumption is available for compensation of workers an capital owners once prouctive capital has been restore to its preprouction level. 11 From an income istribution perspective, it may therefore be more appropriate to base labour shares on net rather than gross value ae (Brigman, 2014; Rognlie, 2015; Cho et al., 2017). For the analyse OECD countries an the G7 countries as a whole, evelopments in gross an net labour shares over the perio have been similar (Chart 2). This is consistent with Rognlie (2015, Figures 1 an 2) who shows that average net an gross labour shares of G7 countries iverge before 1975 but evolve similarly thereafter. However, for some countries there have been large ifferences between net an gross labour share evelopments. There is little empirical evience in the national accounts that ifferences between the evolution of gross an net labour shares are relate to longer-term technological evelopments. The increase of aroun 2 percentage points in the average value ae share of capital epreciation for the analyse OECD countries over the past two ecaes is commonly attribute to the substitution of rapily epreciating ICT capital for more slowly epreciating traitional equipment (Appenix). However, the increase in the share of ICT capital in the total capital stock in volume terms (Appenix Chart A3) has been offset by the ecline in relative prices so that the substitution of ICT equipment for other types of equipment cannot explain the increase in the value ae share of epreciation, which is measure at current prices (Appenix Chart A4). In fact, the share of ICT capital in the total capital stock at current prices has remaine broaly constant or has even Chart 3: Relationship between the Change in Depreciation Share an Change in the Output Gap in OECD Countries, Percentage point changes e n 'Z >s ll Ze l ^se /^> ^s< / <KZ &/E h EKZ ^t &Z,hE > WZ h^ /Z> E> h^ E> E< >hy 'Z WK> h :WE n Note: The ratio of epreciation to gross value ae is expresse in current prices for Korea, Portugal, Sween an Unite Kingom; for New Zealan. Source: OECD National Accounts Database, OECD Economic Outlook Database. 11 Analyses base on gross labour shares inclue Karabarbounis an Neiman (2014); Pionnier an Guietti, (2015); an OECD (2012). 51 N U M B E R 32, S P R I N G 2017

9 ecline for OECD countries (Appenix Chart A5). 12 There is more support in the ata for the hypothesis that the share of epreciation in gross value ae is highly counter-cyclical, which implies that net labour share evelopments are largely riven by the business cycle rather than structural evelopments. The relationship between changes in the share of epreciation in value ae an changes in output gaps appears to be negative (Chart 3). Greece, for instance, experience the largest wiening of the output gap over 2007 an 2014 an is the country in the sample for which the share of epreciation in value ae increase most. The increase in the value ae share of epreciation appears to mainly reflect cyclical evelopments rather than a long-term structural change riven by the long-term ecrease in ICT prices. In sum, the business cycle affects gross value ae much more than capital consumption, thus implying that the value ae share of epreciation is highly counter-cyclical. This makes it ifficult to separate structural changes which are the main focus of this article from cyclical changes in the net labour share. Consequently, the remainer of the article focuses on gross labour shares. Total-economy labour shares or labour shares excluing the primary, housing an non-market sectors? 13 The ecline in the total-economy labour share observe in many OECD countries may partly be riven by evelopments in specific inustries for which there are significant conceptual an measurement issues. For instance, total-economy labour shares are partly riven by evelopments in housing rents. Although the typical worker may actually benefit more from increases in housing rents than from other forms of capital income, the overwhelming part of housing rents ens up in gross operating surplus (i.e. capital income) in the national accounts. Given that the labour share in the housing sector is well below the labour share of the total economy, an increase in the share of housing to total value ae puts ownwar pressure on the total-economy labour share (Box 1). A further issue with total-economy labour shares is that labour share evelopments are partly riven by commoity price evelopments an by imputation choices in the non-market sector (Table 1). For countries with large primary sectors (agriculture, forestry, fishing, mining an quarrying as well as extraction of oil an gas), evelopments in total-economy labour shares are largely riven by evelopments in commoity prices; when commoity prices increase, aggregate profits rise without commensurate increases in aggregate wages. 14 In Norway, for instance, where the oil an gas sec- 12 See Appenix at: 13 This article uses inustry accounts an imputes labour compensation of the self-employe at the inustry level rather than following the approach of Rognlie (2015) an Karabarbounis-Neiman (2014) of using the non-financial corporations' institutional account without correction for the self-employe. As in Rognlie (2015) an Karabarbounis-Neiman (2014). Pionnier an Guietti (2015) have shown that in the national accounts of some countries self-employe workers are allocate to the non-financial corporations' institutional sector, thereby affecting levels an trens of non-financial corporations' labour shares. 14 The ecline in the aggregate labour share partly reflects a change in inustry composition: as commoity prices increase, the share of the mining sector - for which the labour share is low - in total value ae increases. I N T E R N A T I O N A L P R O D U C T I V I T Y MON I T O R 52

10 Box 1: Have Increase Housing Rents Contribute to Declines in Labour Shares? For a number of countries, increases in housing rents contribute to eclines in total-economy labour shares (Box Chart 1). Between 1995 an 2014, the share of the housing sector in total value ae increase by more than 4 percentage points in Greece, Italy, Latvia, Portugal an Spain, an by more the 2 percentage points in the Czech Republic, Finlan, Israel an Unite Kingom (Appenix Chart A6). Housing value ae consists of rents pai by tenants to lanlors an impute rents of homeowners which are both inclue in the national accounts. Since the share of this value ae istribute as labour compensation is low or non-existent (employment in the housing sector mainly correspons to real estate agents an employees of corporations engage in renting activities), the overwhelming part of housing value ae ens up in gross operating surplus (i.e. capital income) in the national accounts. Given that rents an house prices are highly correlate, a house price boom typically raises the total-economy capital share. Box Chart 1: Change in Labour Shares of the Total Economy an Total Economy Excluing Housing in OECD Countries, Percentage points Note: Three-year averages starting an ening in inicate years. OECD an G7 refer to unweighte averages for the relevant countries inclue in the figure for Australia, France, Korea an Portugal; for New Zealan; for Canaa; for Unite Kingom; for Irelan an Unite States. Source: OECD National Accounts Database. The istributional consequences of increases in housing rents may be ifferent from increases in capital income in the prouction sector of the economy. Housing wealth is more equally istribute in the population than prouctive capital so that increases in housing rents can be seen as an inirect channel through which income is transmitte to the typical worker (Murtin an Mira 'Ercole, 2015; Sierminska an Megyesi, 2013). Increases in housing rents an their istribution across workers raise a set of public policy issues unrelate to prouct an labour markets that are the main focus of this article. Increases in housing rents coul, for instance, be aresse by public policies irectly targeting the housing market, in particular by loosening overly restrictive lan-use regulations. This woul have the ouble benefit of raising workers' access to homeownership an limiting rent increases for tenants. 53 N U M B E R 32, S P R I N G 2017

11 Chart 4: Changes in the Total Economy Labour Share with an without the Primary, Housing an Non-market Sectors in OECD Counties, Percentage points Note: Three-year averages starting an ening in inicate years. OECD an G7 refer to un-weighte averages for the relevant countries inclue in the Figure for Australia, France, Korea an Portugal; for New Zealan; for Canaa; for Unite Kingom; for Irelan an Unite States. Source: OECD National Accounts Database. tor is large, the non-housing labour share ecline by aroun 5 percentage points over the perio , but it increase by aroun 1 percentage point when agriculture, mining an non-market sectors are exclue as oil prices increase over the perio covere by the analysis (Table 1). 15 Moreover, national accounting conventions for the non-market sector may bias evelopments in labour shares. Value ae in the non-market sector is equal to the sum of wage compensation an capital consumption, which artificially implies limite variation over time. 16 Declines in labour shares have typically been smaller (an increases larger) when housing, the primary sector (agriculture an mining) an the non-market sector are exclue from the analysis (Chart 4). The primary, housing an nonmarket sectors represent about one thir of total value ae on average across OECD countries. Moreover, changes in the labour share of both the total economy an in this narrower aggregate have not been uniformly negative. For about two thirs of the analyse OECD countries, labour shares ecline between 1995 an 2014 while they increase for the remaining thir. This fining is consistent with Cho et al. (2017) who also conclue that there has been a small ecline in the average gross labour share of 23 OECD countries over the last 20 years, but with substantial heterogeneity across countries. In their sample, gross labour shares ecline in 14 countries, whereas they increase in the remaining 9 countries. 15 Since profits of the Norwegian mining sector partly flow into a sovereign wealth fun benefiting future generations of workers, the ecline in the total-economy labour share overstates the extent to which value ae is appropriate by capital. 16 The finance sector is inclue in the analysis. Excluing the finance sector woul only have a marginal effect on labour share evelopments for most countries, the exception being Luxembourg for which the labour share woul increase by an aitional 2 percentage points if the finance sector were exclue. I N T E R N A T I O N A L P R O D U C T I V I T Y MON I T O R 54

12 Table 2: Changes in Labour Shares in OECD Countries, Percentage points 7RWDOHFRQRP\ H[FOKRXVLQJ H[FOSULPDU\VHFWRU H[FOSXEOLFVHFWRU Australia Austria Belgium Canaa Czech Republic Denmark Estonia Finlan France Germany Greece Hungary Irelan Israel Italy Japan Korea Latvia Lithuania Luxembourg Netherlans New Zealan Norway Polan Portugal Slovak Republic Slovenia Spain Sween Unite Kingom Unite States OECD G Note: Three-year averages starting an ening in inicate years. OECD an G7 refer to unweighte averages for the relevant countries inclue in the Table for Australia, France, Korea an Portugal; for New Zealan; for Canaa; for Chile; for Unite Kingom; for Irelan an Unite States. Source: OECD National Accounts Database. For a number of countries, the change in the labour share is significantly more positive when the housing sector is exclue from the analysis (Table 2, Column 2). For most of these countries, incluing Greece, Italy, Spain an the Unite Kingom, this reflects house price booms in the run-up to the global crisis of that were followe by a slow ownwar ajustment of rents in the subsequent bust so that the share of rents in value ae increase over the perio For countries with large primary sectors, such as Australia, Canaa an Norway, labour share evelopments are significantly more positive when the primary sector 55 N U M B E R 32, S P R I N G 2017

13 Chart 5: Cumulate Change in Labour Share excl. Primary, Housing an Non-market Sectors in 31 OECD Countries, Unweighte average, in percentage points e & e e ee eee eee e e Note: for Austria, Belgium, Czech Republic, Germany, Denmark, Spain, Estonia, Finlan, France, Unite Kingom, Greece, Hungary, Israel, Italy, Japan, Lithuania, Latvia, Luxembourg, Netherlans, Norway, Polan, Portugal, Slovakia, Slovenia an Sween; for New Zealan; for Australia an Korea; for Canaa; for Irelan an Unite States. Source: OECD National Accounts Database. is exclue from the analysis, which reflects the tren increase in commoity prices over the perio (Table 2, Column 3). On the whole, for the OECD countries covere by the analysis the commoity price effect appears to be larger than the house price effect. Excluing the non-market sector typically amplifies changes in labour shares stemming from the remaining sectors because the labour share in the non-market sector is broaly stable (Table 2, Column 4). 17 Pre- or post-crisis evelopments? Most of the ecline in the business labour share excluing the housing an primary sectors took place before the global crisis of (Chart 5). However, labour share evelopments have been very heterogeneous across countries, with no pre-crisis ecline for the country at the thir quartile of the istribution of cumulate labour share changes an a large ecline for the country at the bottom quartile. Given that this narrowly efine labour share is not affecte by house an commoity price evelopments, the timing of the ecline an reboun suggests that the structural factors that rove own the labour share before 2005 weakene thereafter. The timing of the ecline an the reboun of the labour share is consistent with evience suggesting that the pace of expansion of global value chains associate with China's integration 17 The stability of the labour share in the non-market sector reflects to a large extent the national account convention that value ae in the government sector is equal to labour compensation plus consumption of fixe capital, so that the labour share is highly stable an aroun 1. I N T E R N A T I O N A L P R O D U C T I V I T Y MON I T O R 56

14 Chart 6: Contribution of Manufacturing an Services to Labour Share Developments in OECD Countries, Excluing primary, housing an non-market sectors, percentage points Note: Three-year averages starting an ening in inicate years. OECD an G7 refer to unweighte averages for the relevant countries inclue in the Figure for Australia, Korea; for New Zealan; for Canaa; for Irelan an Unite States. Source: OECD National Accounts Database. into the worl traing system which may have contribute to labour share eclines (IMF, 2017) slowe in the wake of the global crisis of (Ferrantino an Taglioni, 2014). Alternative explanations coul be the slowing pace of IT-relate technological change or the reuce scope of regulatory reforms, especially in network inustries, which appear to be two major rivers of labour share eclines (Karabarbounis an Neiman, 2014; Azmat et al., 2012). The post-2005 reboun in the labour share may partly also reflect business cycle conitions, with limite ownwar ajustment of wages relative to profits uring an in the wake of the global economic crisis. Manufacturing or services? In most of the countries examine here, changes in labour shares when primary, housing an non-market sectors are exclue reflect similar rather than iverging evelopments in manufacturing an services an a limite role of changes in inustry composition (Chart 6). 18 If labour share evelopments were entirely riven by eclines in labour shares within manufacturing which is more expose to increase trae integration than services or by a shift in inustry composition from manufacturing to services, this woul suggest globalization as the most plausible explanation of aggregate labour share evelopments. However, the similarity of evelopments in services an manufacturing oes not imply that technological change is the ultimate source of aggregate labour share evelopments as globalization may inuce technological change or isplace manufacturing workers that are then re-employe in services at lower wages. 18 This is consistent with previous stuies suggesting that labour share evelopments are overwhelmingly riven by evelopments within inustries (OECD, 2012; De Serres et al., 2001). 57 N U M B E R 32, S P R I N G 2017

15 Chart 7: Change in Ratio of Meian to Average Wages in OECD countries, Percentage points e e Note: Three-year averages starting an ening in inicate years. OECD an G7 refer to unweighte averages for the relevant countries for Chile, Czech Republic, Denmark; for Australia, Spain, France, Italy, Polan, Sween; for Norway, New Zealan; for Canaa; for Netherlans. Source: OECD Earnings Database. Dissecting Wage Inequality Developments Increases in wage inequality have contribute to aggregate ecoupling by reucing the ratio of meian to average wages in a wie range of OECD countries. The average ecline in the ratio of meian to average wages base on the OECD Earnings Database was aroun 2 percentage points over the perio , but for a number of countries, incluing the Czech Republic, Hungary, Korea, New Zealan, Polan an the Unite States, eclines in the ratio were significantly more pronounce (Chart 7). Of the analyse OECD countries only Chile, Italy an Spain bucke the tren of increasing wage inequality. The ecline in the ratio of meian to average wages appears to be overwhelmingly riven by high wage growth of top earners. Information on wages of workers at the top of the wage istribution from surveys is unreliable which reflects top-coing, sampling issues an unerreporting so that it is preferable to base wage growth of top earners on tax recors. 19 Alvareo et al. (2016) provie average wage income of the top 1 per cent of income earners, which likely overlaps with the top 1 per cent of wage earners. Accoring to these ata, which are available only for a limite number of countries, the most striking evelopment over the past two ecaes has been the ivergence of wages of the top 1 per cent of income earners from both the 90th percentile an the meian of wage earners (Chart 8). 20 Well-known explanations for increase wage inequality such as skill-biase technological change an globalization cannot plausibly account for the isproportionate wage growth at the very top of the wage istribution. Skillbiase technological change an globalization may both raise the relative eman for highskille workers, but this shoul be reflecte in broaly rising relative wages of high-skille 19 Atkinson et al. (2011), Burkhauser et al. (2012), Deaton (2005) an Ruiz an Woloszko (2016) iscuss issues with the coverage of top earners in surveys. 20 To the extent that surveys only incompletely capture wage growth at the top of the wage istribution an therefore unerestimate average wage growth, the actual ecoupling of meian from average wages may be larger than suggeste by surveys. I N T E R N A T I O N A L P R O D U C T I V I T Y MON I T O R 58

16 Chart 8: Wages Trens in OECD Countries by Group, Inex, 1995=100 e ee eee eee e e Note: Inices base on unweighte average for seven OECD countries: Australia ( ), Spain ( ), France ( ), Italy ( ), Japan ( ), Korea ( ), an Unite States ( ), for which ata on wages of the top 1 per cent of income earners are available. All series are eflate by countryspecific value ae price inices. Source: OECD Earnings Database; Worl Wealth an Income Database. workers rather than narrowly rising relative wages of top-earners. Brynjolfsson an McAfee (2014) argue that igitalisation leas to "winnertake-most" ynamics, with innovators reaping outsize rewars as igital innovations are replicable at very low cost an have a global scale. Recent stuies provie evience consistent with "winner-take-most" ynamics, in the sense that prouctivity of firms at the technology frontier has iverge from the remaining firms an that market shares of frontier firms have increase (Anrews et al., 2016). This type of technological change may allow firms at the technology frontier to raise the wages of their key employees to "superstar" levels. Conclusion This article is limite to a quantitative escription of ecoupling of real meian wages from labour prouctivity in OECD countries as well as its proximate causes, i.e. changes in labour shares an wage inequality. The crosscountry heterogeneity in these movements an the fact that wage inequality is mainly riven by high wage growth of top earners suggest that longer-term global trens such as technological change an globalization alone cannot fully account for ecoupling of wages from prouctivity. Country-specific factors, incluing public policy settings, may play a significant role in shaping the effects of global trens on labour shares an wage inequality. Further research nees to investigate the structural causes of the ecoupling of wages from prouctivity an the relation with economic policies. Country- an sector-level ata coul be use to analyse the extent to which movements in labour shares an wage inequality are relate to measures of technological change, trae integration an public policies. Of particular interest is the issue whether igitalization, eclining real investment prices an trae integration with labour-abunant countries reuce labour shares an raise wage inequality an whether public policies can play a mitigating role. Micro-level ata coul be use to analyse 59 N U M B E R 32, S P R I N G 2017

17 the transmission of prouctivity gains to wages at the firm level, in particular whether macrolevel ecoupling reflects changes in the composition of firms or changes within firms an the role of public policies in these evelopments. References Alvareo F., A. B. Atkinson an S. Morelli (2016) "The Challenge of Measuring UK Wealth Inequality in the 2000s," Fiscal Stuies, Vol. 37, No. 1, pp Anrews D., C. Criscuolo an P. Gal (2016) "The Global Prouctivity Slowown, Technology Divergence an Public Policy: A Firm Level Perspective," OECD Economics Department Working Paper for WP1, ECO/CPE/ WP1(2016)26. Atkinson, A., T. Piketty, an E. Saez (2011) "Top Incomes in the Long Run of History," Journal of Economic Literature, Vol. 49, No. 1, pp Azmat, G., A. Manning an J. Van Reenen (2012) "Privatization an the Decline of Labour's Share: International Evience from Network Inustries," Economica, No. 79/315, pp Bivens, J. an L. Mishel (2015) "Unerstaning the Historic Divergence Between Prouctivity an a Typical Worker's Pay," EPI Briefing Papers, No. 406, Economic Policy Institute, Washington. Brigman, B. (2014) Is Labor's Loss Capital's Gain? Gross versus Net Labor Shares, Bureau of Economic Analysis, mimeo, October. Brynjolfsson, E. an A. McAfee (2014) The Secon Machine Age - Work, Progress, an Prosperity in a Time of Brilliant Technologies, WW Norton & Company. Burkhauser, R., S. Feng, S. Jenkins, an J. Larrimore (2012) "Recent Trens in Top Incomes Shares in the Unite States: Reconciling Estimates from March CPS an IRS Tax Return Data," Review of Economics an Statistics, Vol. XCIV, pp Cho, T., S. Hwang an P. Schreyer (2017) "Has the Labour Share Decline? It Depens," OECD Statistics Working Papers No. 2017/01, OECD Publishing, Paris. Deaton, A. (2005) "Measuring Poverty in a Growing Worl (or Measuring Growth in a Poor Worl)," Review of Economics an Statistics, Vol. LXXXVII, pp De Serres, A., S. Scarpetta an C. e la Maisonneuve (2001) "Falling Wage Shares in Europe an the Unite States: How Important is Aggregation Bias?" Empirica, Vol. 28, pp Felstein, M. S. (2008) "Di Wages Reflect the Growth in Prouctivity?" NBER Working Paper, No , National Bureau of Economic Research. Ferrantino, M. an D. Taglioni (2014) "Global Value Chains in the Current Trae Slowown," Worl Bank Economic Premise, No IMF (2017) Unerstaning the Downwar Tren in Labor Income Shares, in Chaper 3 in Worl Economic Outlook April. International Labour Organisation (2015a) The Labour Share in G20 Economies, ILO, IMF, OECD, WB, Report prepare for the G20 Employment Working Group Antalya, Turkey, February. International Labour Organisation (2015b) Income Inequality an Labour Income Share in G20 countries: Trens, Impacts an Causes, ILO, IMF, OECD, WB, Report prepare for the G20 Employment Ministeres Meeting an Joint Meeting with the G20 Finance Ministers, Ancara, Turkey, 3-4 September. Karabarbounis, L. an B. Neiman (2014) "The Global Decline of the Labour Share," Quarterly Journal of Economics, Vol. 129, No. 1, pp Murtin, F. an M. Mira 'Ercole (2015) "Househol Wealth Inequality Across OECD Countries: New OECD Evience," OECD Statistics Brief, No. 21, OECD Publishing, Paris. OECD (2012) Labour Losing to Capital: What Explains the Declining Labour Share? Chapter 3 in OECD Employment Outlook, Publishing, Paris. Pionnier, P-A. an E. Guietti (2015) "Comparing Profit Shares in Value-Ae in Four OECD Countries: Towars More Harmonize National Accounts," Statistics Directorate Working Papers, No. 61, OECD Publishing, Paris. Rognlie, M. (2015) "Deciphering the Fall an Rise in the Net Capital Share," Brookings Papers on Economic Activity, March. Ruiz, N. an N. Woloszko (2016) "What Do Househol Surveys Suggest About the Top 1% Incomes an Inequality in OECD Countries?" OECD Economics Department Working Papers, No. 1265, OECD Publishing, Paris. Sharpe, A., J-F. Arsenault an P. Harrison (2008) "Why Have Real Wages Lagge Labour Prouctivity Growth in Canaa?" International Prouctivity Monitor, No. 17, Fall, Sierminska, E. an M. Megyesi (2013) "The Distribution of Wealth Between Househols," Research note, No. 11, European Commission. Sharpe an Uguccioni (2017) "Decomposing the Prouctivity-Wage Nexus in Selecte OECD Countries, ," International Prouctivity Monitor, Vol. 32, Spring. I N T E R N A T I O N A L P R O D U C T I V I T Y MON I T O R 60