The Role of Labour The Role of Labour Market Changes in the Market Changes in the Slowdown of European Slowdown of European

Productivity GrowthProductivity GrowthIan Dew-Becker and Robert J. Ian Dew-Becker and Robert J.

GordonGordon

Presented to NorthwesternPresented to Northwestern

Macro Workshop, September Macro Workshop, September 24, 2007 24, 2007

Why the English Spelling of Why the English Spelling of “Labour”?“Labour”?

This paper is written for the October Lisbon meeting This paper is written for the October Lisbon meeting of of Economic PolicyEconomic Policy, the premier European clone of , the premier European clone of BPEA, Carnegie-Rochester, and NBER Macro Annual.BPEA, Carnegie-Rochester, and NBER Macro Annual.

They insist on English spelling and are the most They insist on English spelling and are the most draconian of all these conference series in making the draconian of all these conference series in making the authors jump through multiple hoops. We have been authors jump through multiple hoops. We have been jumping, and there are still hurdles to jump over.jumping, and there are still hurdles to jump over.

This paper takes the discussion of European This paper takes the discussion of European productivity and employment behavior, and policy productivity and employment behavior, and policy responses, to a new level, thanks to our 2006 responses, to a new level, thanks to our 2006 econ/MMSS grad Ian Dew-Beckerecon/MMSS grad Ian Dew-Becker

All references to “Europe” are to the EU-15, not EU-All references to “Europe” are to the EU-15, not EU-2525

Introduce BobbyIntroduce Bobby

This Paper is about: This Paper is about: Employment, Productivity, Employment, Productivity,

Causation, and PolicyCausation, and Policy Slowdown in productivity growth in Europe post-1995 Slowdown in productivity growth in Europe post-1995

compared to UScompared to US Revival in hours growth in Europe post-1995 compared Revival in hours growth in Europe post-1995 compared

to USto US Which way is the causation? TFP shocks to employment Which way is the causation? TFP shocks to employment

or labor market shocks to labor productivity?or labor market shocks to labor productivity? Primary emphasis on Primary emphasis on heterogeneity heterogeneity within Europe within Europe

post-1995.post-1995. Before 1995, most countries had negative growth in H/N and Before 1995, most countries had negative growth in H/N and

rapid growth in productivity (Y/H). After 1995 country rapid growth in productivity (Y/H). After 1995 country performance split to a much wider dispersion of outcomes.performance split to a much wider dispersion of outcomes.

We track the negative productivity-employment We track the negative productivity-employment correlation across time periods, countries, types of correlation across time periods, countries, types of workers, and industries.workers, and industries.

Symbols to RememberSymbols to Remember

Key IdentityKey IdentityY/N ≡ Y/H * H/N ≡ Y/H * H/E * E/NY/N ≡ Y/H * H/N ≡ Y/H * H/E * E/N

Our paper starts from the identity and Our paper starts from the identity and then asks, can we identify the two-way then asks, can we identify the two-way causation between Y/H and E/Ncausation between Y/H and E/N

We neglect H/E because there’s been no We neglect H/E because there’s been no major turnaroundmajor turnaround

Perhaps Our Most Perhaps Our Most Important Contribution is Important Contribution is

to Policy Debatesto Policy Debates Europe for 20 years has had higher Europe for 20 years has had higher

unemploymentunemployment Lower hours per capitaLower hours per capita Slowdown in productivity growth post-95Slowdown in productivity growth post-95 Europe wants to change it all with Europe wants to change it all with

political reforms – some reforms to raise political reforms – some reforms to raise productivity, others to raise employmentproductivity, others to raise employment

Our big point: they can’t have it both Our big point: they can’t have it both ways, most reforms will move productivity ways, most reforms will move productivity and employment in opposite directionsand employment in opposite directions

New Criterion for Choosing New Criterion for Choosing PoliciesPolicies

Policies we examine include the tax Policies we examine include the tax wedge, employment protection wedge, employment protection legislation (EPL), product market legislation (EPL), product market regulation (PMR), average replacement regulation (PMR), average replacement rate (ARR), and union densityrate (ARR), and union density

Most of these move Y/H and E/N in Most of these move Y/H and E/N in opposite directionsopposite directions

Consider EPL vs. ARR. They have Consider EPL vs. ARR. They have different consequences for the different consequences for the government budgetgovernment budget

Where does the Value Where does the Value of Leisure Come In?of Leisure Come In?

By using Y/N as a metric of welfare in By using Y/N as a metric of welfare in policy evaluation, we would appear to be policy evaluation, we would appear to be neglecting the value of leisure by those neglecting the value of leisure by those who are not workingwho are not working

Yet, as our initial data charts show, the Yet, as our initial data charts show, the post-1995 turnaround in European hours post-1995 turnaround in European hours consists largely of E/N, not of H/Econsists largely of E/N, not of H/E

This post-1995 transition is not about short This post-1995 transition is not about short weeks and long vacationsweeks and long vacations

This is about increased LFPR, people This is about increased LFPR, people (primarily women) moving from home (primarily women) moving from home production to market production. Leisure production to market production. Leisure is not involved (Freeman-Schettkat, 2005)is not involved (Freeman-Schettkat, 2005)

Before Previewing our Before Previewing our Approach, Let’s Look at Approach, Let’s Look at

State-of-Art of the Current State-of-Art of the Current LiteratureLiterature To keep this straight, remember two To keep this straight, remember two

dimensions: pre/post 1995 and European hours dimensions: pre/post 1995 and European hours vs. productivityvs. productivity

About the post-1995 productivity slowdown, About the post-1995 productivity slowdown, most of the best research has come from van most of the best research has come from van Ark’s data-intensive Groningen NL groupArk’s data-intensive Groningen NL group

But it has not considered the role of employment But it has not considered the role of employment changes in pushing productivity growth changeschanges in pushing productivity growth changes

In its many papers, vintage 2001 it diagnosed In its many papers, vintage 2001 it diagnosed the entire US vs EU difference as concentrated the entire US vs EU difference as concentrated in two industries, trade and finance. The in two industries, trade and finance. The difference is in the use of ICT, not the difference is in the use of ICT, not the production of ICT equipment.production of ICT equipment.

The Prescott Literature is The Prescott Literature is about Employment and about Employment and

UnemploymentUnemployment Prescott’s famous conclusion, Prescott’s famous conclusion, allall of the decline in H/N of the decline in H/N

in EU relative to US was due to high tax wedge in in EU relative to US was due to high tax wedge in EuropeEurope Prescott did not notice the post-1995 turnaround in both taxes Prescott did not notice the post-1995 turnaround in both taxes

and H/Nand H/N He has not shown that the tax wedge explains He has not shown that the tax wedge explains bothboth the pre-95 the pre-95

decline in EU hours and the post-95 recoverydecline in EU hours and the post-95 recovery He does not control for other policy/institutional variablesHe does not control for other policy/institutional variables

Alesina and others have doubted his claims based on Alesina and others have doubted his claims based on calibration rather than econometric evidencecalibration rather than econometric evidence

Much other empirical work (Blanchard-Wolfers, Davis-Much other empirical work (Blanchard-Wolfers, Davis-Henrekson) has data ending in mid-1990s and doesn’t Henrekson) has data ending in mid-1990s and doesn’t notice or explain the turnaroundnotice or explain the turnaround

Our work moves beyond this by splitting pre/post 95 Our work moves beyond this by splitting pre/post 95 and by using the best data and econometric framework and by using the best data and econometric framework from OECD’s Bassanini and Duval. But they don’t look from OECD’s Bassanini and Duval. But they don’t look at turnaround either.at turnaround either.

The Dimensions of Our The Dimensions of Our MoveMove

Beyond the Previous Beyond the Previous LiteratureLiterature Previous Groningen literature studies EU Previous Groningen literature studies EU

productivity slowdown at the industry level but productivity slowdown at the industry level but does not tie it to effects of policy on does not tie it to effects of policy on employment. employment.

Prescott literature looks only at low European Prescott literature looks only at low European H/N pre-1995 and its link with taxes. Other H/N pre-1995 and its link with taxes. Other Prescott-related literature looks at other Prescott-related literature looks at other controls but does not quantify the role of taxes controls but does not quantify the role of taxes and other controls on the post-1995 employment and other controls on the post-1995 employment turnaround. turnaround.

Only a few papers have addressed the direction Only a few papers have addressed the direction of causation between employment and of causation between employment and productivity, and no previous paper in the productivity, and no previous paper in the tradeoff literature has explicitly studied the twin tradeoff literature has explicitly studied the twin post-1995 turnaroundspost-1995 turnarounds

Further Summary of Our Further Summary of Our New ContributionsNew Contributions

We Create new Data Aggregates Using Chain-We Create new Data Aggregates Using Chain-Weighted MethodsWeighted Methods EU-15 into four country groups (N, A-S, C, M)EU-15 into four country groups (N, A-S, C, M) 57 Industries into 12 industry groups for total economy 57 Industries into 12 industry groups for total economy

including GHI (Government, Households, Institutions)including GHI (Government, Households, Institutions) Cuts dimensionality from 855 to 48Cuts dimensionality from 855 to 48

We are among the first to use the new EU-We are among the first to use the new EU-KLEMS data set, released in 3/07. KLEMS data set, released in 3/07.

We have new data and econometric results We have new data and econometric results across countries, time, input types (capital vs. across countries, time, input types (capital vs. labor), age-sex groups, and industries labor), age-sex groups, and industries

All address the same question, how much did All address the same question, how much did policy and institutional changes in the labor policy and institutional changes in the labor market cause the twin post-1995 turnarounds in market cause the twin post-1995 turnarounds in productivity and employment growth? productivity and employment growth?

Outline of the PaperOutline of the Paper

Part 2, graphs and tables on the basic Part 2, graphs and tables on the basic macro data, including the USmacro data, including the US

Part 3, regressions of E/N on Part 3, regressions of E/N on policy/institutional variablespolicy/institutional variables

Part 4, the demographics of employmentPart 4, the demographics of employment Part 5, effects of employment shifts on Part 5, effects of employment shifts on

productivity growth (model and productivity growth (model and estimates)estimates)

Part 6, industry analysisPart 6, industry analysis

Preview of the Data Charts Preview of the Data Charts You Will SeeYou Will See

First set of charts, EU vs. USFirst set of charts, EU vs. US Y/HY/H H/NH/N Y/NY/N

Next charts, nonimportance of H/ENext charts, nonimportance of H/E Recall H/N = H/E * E/NRecall H/N = H/E * E/N

Third set of charts, allowing for Third set of charts, allowing for growth in capital (K/H) and the growth in capital (K/H) and the resulting implication for total factor resulting implication for total factor productivity (TFP).productivity (TFP).

Top Fig 1, the GreatTop Fig 1, the GreatPost-1995 Labor Post-1995 Labor

Productivity Growth Productivity Growth TurnaroundTurnaround

0

1

2

3

4

5

6

1970 1975 1980 1985 1990 1995 2000 2005

Per

cent EU-15 Y/H

US Y/H

What to Notice About LPWhat to Notice About LP The EU Slowdown is steady and continuousThe EU Slowdown is steady and continuous The US post-1995 revival is looking The US post-1995 revival is looking

increasingly temporaryincreasingly temporary We created the US trend from quarterly data We created the US trend from quarterly data

through 2007, not just the annual data through through 2007, not just the annual data through 2004 used by EU-KLEMS2004 used by EU-KLEMS

The fact that the US trend is turning around The fact that the US trend is turning around is important for interpretations of what is important for interpretations of what caused the post-1995 US revivalcaused the post-1995 US revival

That’s a separate paper. Today we That’s a separate paper. Today we primarily look inside Europe and exclude primarily look inside Europe and exclude the US from the employment and the US from the employment and productivity regressionsproductivity regressions

Notice how the EU Notice how the EU Turnaround in H/N cancels Turnaround in H/N cancels

out the Y/H Slowdownout the Y/H Slowdown

-3

-2

-1

0

1

2

3

4

1970 1975 1980 1985 1990 1995 2000 2005

Perc

ent

US Y/NEU-15 Y/N

US H/N

EU-15 H/N

Comments about H/N Comments about H/N and Y/Nand Y/N

Importance of expressing everything per capitaImportance of expressing everything per capita Average EU population growth 0.7 percent per year Average EU population growth 0.7 percent per year

slower than USslower than US EU Growth in H/N strongly negative pre-1995, EU Growth in H/N strongly negative pre-1995,

US strongly positiveUS strongly positive Falling level of H/N in Europe is what Prescott Falling level of H/N in Europe is what Prescott

and others have been trying to explainand others have been trying to explain Productivity and employment turnarounds cancel Productivity and employment turnarounds cancel

out. Growth in Y/N almost equal 1980-2005out. Growth in Y/N almost equal 1980-2005 EU 1.92 percent per year, US 1.97 percent per yearEU 1.92 percent per year, US 1.97 percent per year

But EU is at only 70-75 percent of US level and is But EU is at only 70-75 percent of US level and is not catching upnot catching up

Yes, H/N Growth Yes, H/N Growth Turnaround, but most of Turnaround, but most of

this was E/N not H/Ethis was E/N not H/E

-1

-0.5

0

0.5

1

1.5

2

1970 1975 1980 1985 1990 1995 2000 2005

Per

cent

EU-15 E/N

US E/N

Contrast the Lack of Contrast the Lack of TurnaroundTurnaround

in EU H/E (compare to 0 in EU H/E (compare to 0 line)line)

-2

-1.5

-1

-0.5

0

0.5

1

1970 1975 1980 1985 1990 1995 2000 2005

Per

cent

US H/E

EU-15 H/E

Implications of E/N Implications of E/N Turnaround and Lack of Turnaround and Lack of

H/E TurnaroundH/E Turnaround The regression analysis relates Y/H The regression analysis relates Y/H

growth to E/N, not H/Ngrowth to E/N, not H/N This fits together with the previous This fits together with the previous

econometric literature in which the econometric literature in which the dependent variable tends to be dependent variable tends to be unemployment or employment, not unemployment or employment, not hourshours

Focus on E/N makes it easy to Focus on E/N makes it easy to disaggregate by age-sex groupsdisaggregate by age-sex groups

Data Problem: Capital Data Problem: Capital input growth, EU-KLEMS input growth, EU-KLEMS

vs. Jorgenson-Stirohvs. Jorgenson-Stiroh

0

1

2

3

4

5

6

7

1980 1985 1990 1995 2000

Per

cent

EU-15 (KLEMS) K/H

US (KLEMS) K/H

US (Stiroh)

Counterpart is a Different Counterpart is a Different StoryStory

about US TFP Growth in about US TFP Growth in 1980s1980s

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

1980 1985 1990 1995 2000

Perc

ent

EU-15 (KLEMS) TFP

US (KLEMS) TFP US (Stiroh) TFP

Turn to Table 1, the basicTurn to Table 1, the basicTurnaround Story Turnaround Story

Turnaround differenceTurnaround difference 1995-2006 growth minus 1970-95 growth1995-2006 growth minus 1970-95 growth EU growth minus US growthEU growth minus US growth

Turnarounds in opposite directionsTurnarounds in opposite directions EU-US -2.20 for Y/HEU-US -2.20 for Y/H EU-US +1.99 for H/NEU-US +1.99 for H/N EU-US -0.19 for Y/NEU-US -0.19 for Y/N

Disproportionate role of Med countriesDisproportionate role of Med countries All turnarounds fall by ¼ when Meds All turnarounds fall by ¼ when Meds

excludedexcluded

Increased Post-1995 Increased Post-1995 HeterogeneityHeterogeneity

Standard deviations rise post 1995Standard deviations rise post 1995 Y/H from 0.63 to 1.00Y/H from 0.63 to 1.00 H/N more than doubles 0.46 to 1.02H/N more than doubles 0.46 to 1.02 Y/N more than doubles 0.55 to 1.27Y/N more than doubles 0.55 to 1.27

Range of Y/N growth post 1995Range of Y/N growth post 1995 Italy 1.18 to Ireland 6.17Italy 1.18 to Ireland 6.17 Greece and Spain #2 and #3Greece and Spain #2 and #3 Sharp contrast Spain vs. ItalySharp contrast Spain vs. Italy

Surprise: Three of four EU groups have Surprise: Three of four EU groups have faster Y/N growth than US post-1995faster Y/N growth than US post-1995

Population shares 5, 17, 49, and 29Population shares 5, 17, 49, and 29

Table 2, add capital and Table 2, add capital and TFPTFP

Shorter period, 1980-2004Shorter period, 1980-2004 Identity: Y/H growth ≡ capital Identity: Y/H growth ≡ capital

deepening + deepening + TFP growthTFP growth

US no turnaround CD so Y/H = TFPUS no turnaround CD so Y/H = TFP EU K/H turnaround almost as EU K/H turnaround almost as

negative as E/N turnaround is negative as E/N turnaround is positive. Sliding down the LD curvepositive. Sliding down the LD curve

Table 2 for the Four Table 2 for the Four Country GroupsCountry Groups

Key question. Did capital growth Key question. Did capital growth respond to faster employment growth?respond to faster employment growth? Compare turnaround in K/N to E/NCompare turnaround in K/N to E/N

Nordic, strong employment, relatively Nordic, strong employment, relatively strong Y/H, but weak capital strong Y/H, but weak capital

Anglo-Saxon (94% UK). Strong K/N Anglo-Saxon (94% UK). Strong K/N response to E/N, no turnaround capital response to E/N, no turnaround capital deepeningdeepening

Continental (49% of EU). K/N response Continental (49% of EU). K/N response 1/3 of E/N turnaround.1/3 of E/N turnaround.

Core of the Turnaround: Core of the Turnaround: the Medsthe Meds

Biggest turnarounds for Y/H, E/N, K/H, Biggest turnarounds for Y/H, E/N, K/H, capital deepening and TFPcapital deepening and TFP

Virtually K/N response to big E/N Virtually K/N response to big E/N turnaroundturnaround

Overall, Meds shared with Nordic a Overall, Meds shared with Nordic a disappointing investment performancedisappointing investment performance

Anglo-Saxon and Continental had Anglo-Saxon and Continental had substantial positive K/N response to E/N substantial positive K/N response to E/N turnaroundturnaround

Later we compare capital response to Later we compare capital response to predictions of a calibrate modelpredictions of a calibrate model

Regression Analysis of Regression Analysis of E/NE/N

Here we inherit a large literatureHere we inherit a large literature Ratio of EU/US H/N declined by 48 Ratio of EU/US H/N declined by 48

percent from 1960-95. percent from 1960-95. For Prescott, it is all the tax wedgeFor Prescott, it is all the tax wedge For others, it is a varying mix of taxes, For others, it is a varying mix of taxes,

regulations, politics, unions regulations, politics, unions There is plenty of data and regression There is plenty of data and regression

specifications to choose fromspecifications to choose from But so far nobody has But so far nobody has usedused the regression the regression

coefficients to interpret the post-1995 E/Ncoefficients to interpret the post-1995 E/N

What tax wedge elasticityWhat tax wedge elasticityshould we expect?should we expect?

Prescott’s calibrated elasticity is -0.92Prescott’s calibrated elasticity is -0.92 Alesina Alesina et al. et al. argue argue a priori a priori not > -0.4 to -0.45not > -0.4 to -0.45

Econometric estimatesEconometric estimates Davis-Henrekson -0.22 (only four years Davis-Henrekson -0.22 (only four years

ending in 1995)ending in 1995) Bassanini-Duval, best data, best specificationBassanini-Duval, best data, best specification

--0.30 for males, -0.50 for females0.30 for males, -0.50 for females No regressions for both sexes together, which we No regressions for both sexes together, which we

needneed

Results in Table 3Results in Table 3

Changes from B-DChanges from B-D They include 20 countries, with US, Canada, Japan, They include 20 countries, with US, Canada, Japan,

whereas we want just EU-15whereas we want just EU-15 We include ages 15-64, they did prime-age, youth, We include ages 15-64, they did prime-age, youth,

and elderly separately (we do separate age-sex and elderly separately (we do separate age-sex regressions later)regressions later)

We weight the regressions by population (to do We weight the regressions by population (to do otherwise gives Luxembourg the same importance as otherwise gives Luxembourg the same importance as Germany)Germany)

Run regressions separately for both sexes, males, Run regressions separately for both sexes, males, femalesfemales

We fill in some missing data to extend the sample We fill in some missing data to extend the sample period back from 1982 to 1978period back from 1982 to 1978

What to Notice in Table What to Notice in Table 33

Three columns for both and females, two Three columns for both and females, two for malesfor males

Cols (1) and (6) have numerous other Cols (1) and (6) have numerous other controls that are not computed for mencontrols that are not computed for men

Then come results without the other Then come results without the other controls, and with and without time controls, and with and without time effects. All results have country fixed effects. All results have country fixed effectseffects

Big impact of omitting time effects. Why?Big impact of omitting time effects. Why?

Big Time Effects Big Time Effects

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

1979 1984 1989 1994 1999

Men

Women

Both

Interpretation of ResultsInterpretation of Results The time-trend-like pattern of the female The time-trend-like pattern of the female

time coefficients suggests a cultural time coefficients suggests a cultural shift favoring more employment of shift favoring more employment of femalesfemales Same thing happened in the US between Same thing happened in the US between

1965 and 19801965 and 1980 Our baseline results are those with the Our baseline results are those with the

time effects, columns (2), (4), and (7)time effects, columns (2), (4), and (7) Tax wedge coefficients close to those of Tax wedge coefficients close to those of

B-D, higher than those of Davis-B-D, higher than those of Davis-HenreksonHenrekson

Interpretation of Other Interpretation of Other CoefficientsCoefficients

Output gap should have a positive Output gap should have a positive coefficient, others should have a coefficient, others should have a negative coefficientnegative coefficient

For both, only tax wedge and ARR are For both, only tax wedge and ARR are significantsignificant

For men, same, plus PMR with wrong For men, same, plus PMR with wrong signsign

For women, tax wedge, EPL, and ARR For women, tax wedge, EPL, and ARR are significant with correct signsare significant with correct signs

How Much of the Post-1995 How Much of the Post-1995

E/N Turnaround can we E/N Turnaround can we explain?explain? Figure 4 plots the actual values and Figure 4 plots the actual values and

two predicted valuestwo predicted values The equation’s predicted value with the The equation’s predicted value with the

actual values of the explanatory variablesactual values of the explanatory variables The counterfactual prediction that holds The counterfactual prediction that holds

constant at 1995 levels the tax wedge, constant at 1995 levels the tax wedge, EPL, ARR, PMR, and union density EPL, ARR, PMR, and union density variablesvariables

Separately for four country groups, both Separately for four country groups, both sexes taken togethersexes taken together

Figure 4, Four Country Figure 4, Four Country GroupsGroupsNordic

-0.65

-0.6

-0.55

-0.5

-0.45

-0.4

-0.35

Log(

E/N

)

Actual

Predicted (83-03)

Fixed Levels post-'95

Continental

-1.1

-1

-0.9

-0.8

-0.7

-0.6

-0.5

1983 1988 1993 1998 2003

Log(E

/N)

Mediterranean

1983 1988 1993 1998 2003

-1.1

-1

-0.9

-0.8

-0.7

-0.6

-0.5

Anglo-Saxon*

-0.65

-0.6

-0.55

-0.5

-0.45

-0.4

-0.35

Interpretation of Figure Interpretation of Figure 44

Predicted lies on top of actuals, Predicted lies on top of actuals, reflecting Rreflecting R22=1.0=1.0 Same for B-D, 0.99 to 1.0Same for B-D, 0.99 to 1.0

Most important result is that Most important result is that policy/institutional variables can explain policy/institutional variables can explain the post-1995 turnaround everywhere the post-1995 turnaround everywhere but in Cont groupbut in Cont group

Which variables contribute the Which variables contribute the explanation in each country group? (not explanation in each country group? (not shown in tables)shown in tables)

Contributions by Country Contributions by Country GroupGroup

Nordic: decline in PMR and union Nordic: decline in PMR and union densitydensity

Anglo-Saxon: decline in tax wedge and Anglo-Saxon: decline in tax wedge and union density (Thatcherism?)union density (Thatcherism?)

Continental: decline in EPL and PMRContinental: decline in EPL and PMR Mediterranean: Mainly PMR, some EPLMediterranean: Mainly PMR, some EPL Tax wedgeTax wedge

Mattered only for Anglo-SaxonMattered only for Anglo-Saxon Other groups, taxes didn’t change, Med Other groups, taxes didn’t change, Med

countries tax coefficient close to zerocountries tax coefficient close to zero

Demographics of Demographics of EmploymentEmployment

First, a shift-share analysis of E/N First, a shift-share analysis of E/N turnaroundturnaround Was the increase in E/N in each age-sex Was the increase in E/N in each age-sex

groupgroup Or did E/N composition move toward groups Or did E/N composition move toward groups

with already-high E/N?with already-high E/N? Employment and share effects sum to the Employment and share effects sum to the

total change in E/N over a time periodtotal change in E/N over a time period Figure 5 shows the results by the four Figure 5 shows the results by the four

country groupscountry groups

12 Age-Sex, 4 Country 12 Age-Sex, 4 Country GroupsGroupsNordic

15-24

25-34

35-44

45-54

55-64

65+

15-24

25-34

35-44

45-54

55-64

65+

Employment Effect

Share Effect

Men

Women

Anglo-Saxon

15-24

25-34

35-44

45-54

55-64

65+

15-24

25-34

35-44

45-54

55-64

65+

Continental

-0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02 0.025

15-24

25-34

35-44

45-54

55-64

65+

15-24

25-34

35-44

45-54

55-64

65+

Mediterranean

-0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02

15-24

25-34

35-44

45-54

55-64

65+

15-24

25-34

35-44

45-54

55-64

65+

Interpretation of Figure Interpretation of Figure 55

Nordic and Anglo-Saxon share effects Nordic and Anglo-Saxon share effects reflect aging of the populationreflect aging of the population

Increase in 55-64 could also result from Increase in 55-64 could also result from higher retirement ageshigher retirement ages

Employment effects in Cont group were Employment effects in Cont group were smaller than in Nordic and A-Ssmaller than in Nordic and A-S

Employment effects are substantially Employment effects are substantially larger for Med group, especially for larger for Med group, especially for prime-age femalesprime-age females Empl effect 90%, female empl effect 70% of Empl effect 90%, female empl effect 70% of

totaltotal

E/N Regressions for Age-E/N Regressions for Age-Sex Groups in Table 4Sex Groups in Table 4

Tax wedge effects are larger for females Tax wedge effects are larger for females than males in all age groups except youth than males in all age groups except youth

Output gap coefficients are much larger for Output gap coefficients are much larger for youthyouth

EPL has large negative effects for youth EPL has large negative effects for youth and elderlyand elderly

ARR has significant negative coefficients, ARR has significant negative coefficients, large for several groupslarge for several groups

Alternating signs for PMRAlternating signs for PMR Strong union effects above age 45 Strong union effects above age 45

Changes in Aggregate Changes in Aggregate Experience,Experience,

a Channel from a Channel from Employment to ProductivityEmployment to Productivity No direct data on experienceNo direct data on experience Standard approach, “potential Standard approach, “potential

experience”experience” Age – years of education – 5Age – years of education – 5

But this misses inflow of inexperienced But this misses inflow of inexperienced workers in a country like Spainworkers in a country like Spain

We build on Wasmer (2001), who We build on Wasmer (2001), who assumes workers are a random sample assumes workers are a random sample from the population each yearfrom the population each year

Our Alternative ApproachOur Alternative Approach Wasmer’s randomness is implausibleWasmer’s randomness is implausible We assume instead that new entrants have We assume instead that new entrants have

no previous work experience.no previous work experience. See the example of how we calculate See the example of how we calculate

experience in Box A on p. 41experience in Box A on p. 41 100 people born 1970100 people born 1970 E/N jumps from 40 to 80 percent on 1/1/95E/N jumps from 40 to 80 percent on 1/1/95 On that date, half have 10 years experience, the On that date, half have 10 years experience, the

other half have zeroother half have zero Equation (3) on p. 18 takes into account Equation (3) on p. 18 takes into account

depreciation, increases of employment, and depreciation, increases of employment, and the retirement of the most experienced the retirement of the most experienced workersworkers

Figure 6: Calculated Figure 6: Calculated Experience Experience

by Country Groupby Country Group

9.5

10

10.5

11

11.5

12

12.5

1975 1980 1985 1990 1995 2000 2005

Nordic

Mediterranean

Anglo-Saxon

Continental

Interpretation of Figure Interpretation of Figure 66

Experience increased in all groups but Meds, Experience increased in all groups but Meds, reflecting their influx of new workers reflecting their influx of new workers starting in late 1980sstarting in late 1980s Experience growth gap between Cont and Meds Experience growth gap between Cont and Meds

can explain 0.5 percent per year of productivity can explain 0.5 percent per year of productivity growth gap post-1995growth gap post-1995

Rise in experience goes in opposite direction Rise in experience goes in opposite direction from slowdown in productivity growthfrom slowdown in productivity growth

Table 5 reports detailsTable 5 reports details Germany-Spain experience gap an explain ¾ of Germany-Spain experience gap an explain ¾ of

actual Germany-Spain productivity gapactual Germany-Spain productivity gap

Figure 7: Scatter of Figure 7: Scatter of ΔΔ Exp Exp on on ΔΔ(E/N)(E/N)

Δ(Exp.)= 0.14 - 0.06*Δ(E/N)

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

-2 0 2 4 6 8 10 12 14

Δ(E/N)

Δ(E

xper

ience

)

Another Counterfactual Another Counterfactual ExerciseExercise

Regarding post-1995 Regarding post-1995 TurnaroundTurnaround Table 6 reports calculated post-1995 Table 6 reports calculated post-1995

change in experience on two assumptionschange in experience on two assumptions Actual employment growthActual employment growth No employment growthNo employment growth

Look at Meds, esp. Spain: big negative Look at Meds, esp. Spain: big negative employment effects on experienceemployment effects on experience Paper, p. 20. Experience can explain 2/3 of Paper, p. 20. Experience can explain 2/3 of

Spain vs. Italy productivity growth differenceSpain vs. Italy productivity growth difference Otherwise, effects are small, with neg Otherwise, effects are small, with neg

effects in IRE and LUX, positive in DNKeffects in IRE and LUX, positive in DNK

Part 5, Two Approaches toPart 5, Two Approaches toLink Productivity Growth to Link Productivity Growth to

E/N GrowthE/N Growth

A calibrated structural modelA calibrated structural model Given the behavior of employment, was are the Given the behavior of employment, was are the

implications for productivity growth and implications for productivity growth and investment?investment?

As interesting for its errors as for its As interesting for its errors as for its predictions: which aspects of European predictions: which aspects of European outcomes are surprising?outcomes are surprising?

Reduced form regressions of productivity Reduced form regressions of productivity change on employment change, using IV to change on employment change, using IV to control for simultaneitycontrol for simultaneity

The Structural ModelThe Structural Model

Described in words in Box B on pp. 42-43Described in words in Box B on pp. 42-43 Described in equations in Technical AppendixDescribed in equations in Technical Appendix Two changes to the basic Ramsey modelTwo changes to the basic Ramsey model

Two types of agents (men and women) Two types of agents (men and women) differentiated by skill and taste for market work.differentiated by skill and taste for market work.

Allows us to trace effects of societal changes favoring Allows us to trace effects of societal changes favoring more market work by womenmore market work by women

Novelty: introduce experience into a DSGE modelNovelty: introduce experience into a DSGE model Lump all policy/institutional changes into a Lump all policy/institutional changes into a

single parameter, the distaste for worksingle parameter, the distaste for work

Effects of a Change inEffects of a Change inDistaste for WorkDistaste for Work

Short-run EffectsShort-run Effects Decline in K/H ratioDecline in K/H ratio Decline in productivityDecline in productivity Output risesOutput rises Investment risesInvestment rises

Long-run effectsLong-run effects K/H rises back toward its initial levelK/H rises back toward its initial level Experience risesExperience rises These two channels allow productivity to These two channels allow productivity to

recover after its initial declinerecover after its initial decline

Results in Table 7Results in Table 7

Take E/N as given, calculate implied Take E/N as given, calculate implied turnarounds in growth of Y/H and K/Nturnarounds in growth of Y/H and K/N

Note: actual values are slightly Note: actual values are slightly different than in Table 2, since this different than in Table 2, since this starts in 1985 rather than 1980starts in 1985 rather than 1980

Any effects of a turnaround in technical Any effects of a turnaround in technical progress will not be reflected in the progress will not be reflected in the predicteds, rather in the errorspredicteds, rather in the errors

Important Results in Important Results in Table 7Table 7

LP turnaround:LP turnaround: All predicted values have correct signAll predicted values have correct sign Big shortfalls in actual vs. predicted for Cont & MedBig shortfalls in actual vs. predicted for Cont & Med

K/N turnaroundK/N turnaround Med shortfall in K/N explains a further 0.14 of their Med shortfall in K/N explains a further 0.14 of their

Y/H shortfall (0.33 * -0.42)Y/H shortfall (0.33 * -0.42) Col (7) shows total predicted Y/H turnaroundCol (7) shows total predicted Y/H turnaround Overall, K/N helps slightly to improve RMSE of Overall, K/N helps slightly to improve RMSE of

predictions, substantially for Meds, wrong direction predictions, substantially for Meds, wrong direction for A-Sfor A-S

Note overall prediction nails the Nordics almost Note overall prediction nails the Nordics almost exactlyexactly

RHS of Table 7: RHS of Table 7: PredictionsPredictions

Prediction of Y/H and K/N 2005-2010 if Prediction of Y/H and K/N 2005-2010 if employment level remains constant at 2005employment level remains constant at 2005

Intuition: if Med E/N stops growing, it will stop Intuition: if Med E/N stops growing, it will stop pushing down on Y/H growth so their Y/H pushing down on Y/H growth so their Y/H growth should be fastergrowth should be faster

That’s what Table 7 shows – major future That’s what Table 7 shows – major future improvement for Meds, smaller for Nordic and improvement for Meds, smaller for Nordic and A-SA-S Implies a future convergence in EU productivity Implies a future convergence in EU productivity

growthgrowth But this could be an implausible prediction for But this could be an implausible prediction for

employment. Why should Med E/N stop employment. Why should Med E/N stop growing?growing?

Reduced-form RegressionsReduced-form RegressionsExplaining Productivity Explaining Productivity

GrowthGrowth We merge two strands of literatureWe merge two strands of literature

Employment explained by Employment explained by policy/institutional variablespolicy/institutional variables

Productivity growth explained by Productivity growth explained by employment growth: tracing the labor employment growth: tracing the labor demand curvedemand curve

The model showed two elements to The model showed two elements to the Y/H vs. E/N tradeoffthe Y/H vs. E/N tradeoff The effect on K/HThe effect on K/H The effect through experienceThe effect through experience

Turn to Table 8 for the Turn to Table 8 for the ResultsResults

Table 8 starts with naïve regressions and Table 8 starts with naïve regressions and gradually becomes more sophisticatedgradually becomes more sophisticated

Column (1) includes only E/N growth and the Column (1) includes only E/N growth and the change in the output gapchange in the output gap Coefficient on E/N -0.79Coefficient on E/N -0.79 No difference made by fixed effects in col. (2) No difference made by fixed effects in col. (2)

except better fitexcept better fit McGuckin-van Ark claim tradeoff is McGuckin-van Ark claim tradeoff is

temporary, they add longer lags to show temporary, they add longer lags to show turnaroundturnaround Not true in col. (3)Not true in col. (3)

How to control for How to control for simultaneity?simultaneity?

A technology shock could cause a A technology shock could cause a negative correlation between productivity negative correlation between productivity and employmentand employment

We follow Bourles and Cette (2005) by We follow Bourles and Cette (2005) by using IVusing IV List of instruments bottom Table 8List of instruments bottom Table 8

Column (4), E/N coeff drops from -0.84 to Column (4), E/N coeff drops from -0.84 to

-0.64 and -0.62 with country fixed effects -0.64 and -0.62 with country fixed effects col (5)col (5)

Effects of the Policy Effects of the Policy VariablesVariables

Add in the policy variablesAdd in the policy variables You would expect coeff on E/N to become smaller You would expect coeff on E/N to become smaller

because previously policy effects were working because previously policy effects were working through E/Nthrough E/N

You would expect positive coefficients on policy You would expect positive coefficients on policy variablesvariables

Final results in column (7)Final results in column (7) E/N coeff drops to -0.42E/N coeff drops to -0.42 ARR, EPL strong effects, PMR marginal effectARR, EPL strong effects, PMR marginal effect Note: Three-year changes in policy variables are Note: Three-year changes in policy variables are

included in list of instruments in columns (6) and included in list of instruments in columns (6) and (7), see Table 10 below(7), see Table 10 below

Accuracy of Predicted post-Accuracy of Predicted post-19951995

Productivity Growth Productivity Growth TurnaroundsTurnarounds Nordic, big missNordic, big miss

Intuitively, they managed big employment gains Intuitively, they managed big employment gains which would be predicted to reduce productivity which would be predicted to reduce productivity growth, but they had a minimal turnaroundgrowth, but they had a minimal turnaround

A-S, dominated by UK. Big miss for IrelandA-S, dominated by UK. Big miss for Ireland Continental, quite close, better than modelContinental, quite close, better than model Mediterranean residuals are same as the Mediterranean residuals are same as the

modelmodel Country group predictions RMSE of 0.54, Country group predictions RMSE of 0.54,

compared to 0.41 for modelcompared to 0.41 for model

The Combined Effects of The Combined Effects of PolicyPolicy

and Institutions on and Institutions on Productivity GrowthProductivity Growth

We have seen some policies push We have seen some policies push employment down and productivity upemployment down and productivity up

Table 10 shows the first-stage Table 10 shows the first-stage coefficients in the productivity coefficients in the productivity regressionsregressions The policy variables are entered as three-The policy variables are entered as three-

year changesyear changes All policy variables have the correct sign, All policy variables have the correct sign,

all but EPL are highly significantall but EPL are highly significant

Direct and Indirect Direct and Indirect EffectsEffects

Policy variables have two effects on Policy variables have two effects on productivityproductivity Direct effect through coefficients in Table 8Direct effect through coefficients in Table 8 Indirect effect in reducing employment in Table Indirect effect in reducing employment in Table

10, which in turn raises productivity in Table 810, which in turn raises productivity in Table 8 If a policy raises Y/H and reduces E/N, it may If a policy raises Y/H and reduces E/N, it may

have little or no impact on Y/Nhave little or no impact on Y/N Simulations of a unit standard deviation Simulations of a unit standard deviation

shock that occurs over a five-year periodshock that occurs over a five-year period For instance, One stdev shock to union density is For instance, One stdev shock to union density is

15.6 percentage points15.6 percentage points Coefficients on these variables have no direct Coefficients on these variables have no direct

economic meaning because they are index economic meaning because they are index numbersnumbers

Simulated Effects on E/NSimulated Effects on E/NEmployment

-12

-10

-8

-6

-4

-2

0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Percent

Perc

ent

Union Density

Tax Wedge

PMRARR

EPL

Simulated Effects on Simulated Effects on ProductivityProductivity

Productivity

0

1

2

3

4

5

6

7

8

9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Percent

Perc

ent

Union Density

Tax Wedge

PMR

ARR

EPL

Simulated Effects on Simulated Effects on Output per CapitaOutput per Capita

Output per Capita

-3.5

-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Percent

Perc

ent

Union Density

Tax Wedge

PMR

ARR

EPL

Summary of Simulations Summary of Simulations on Y/Non Y/N

What Figure 8 shows:What Figure 8 shows: ARR and EPL provide a free lunch. Boost ARR and EPL provide a free lunch. Boost

productivity more than employment productivity more than employment declines, resulting in a rise in Y/Ndeclines, resulting in a rise in Y/N

Union density is unambiguously badUnion density is unambiguously bad Taxes and PMR don’t seem to matter muchTaxes and PMR don’t seem to matter much

SurprisesSurprises That all five policy/inst variables matter for That all five policy/inst variables matter for

Y/HY/H There is no long-run bounce-back effectThere is no long-run bounce-back effect

The Y/H and E/N The Y/H and E/N TurnaroundsTurnarounds

at the Industry Levelat the Industry Level This is the first paper to produce these This is the first paper to produce these

datadata The EU-KLEMS data base only became The EU-KLEMS data base only became

available in March 2007available in March 2007 Double-aggregation by industry and Double-aggregation by industry and

country groupcountry group Table 11 presents the results for the US, Table 11 presents the results for the US,

the EU excluding the Meds, and the Medsthe EU excluding the Meds, and the Meds Top half the raw growth rates, bottom the Top half the raw growth rates, bottom the

contributions (nominal output shares)contributions (nominal output shares)

US vs EU, Groningen hasUS vs EU, Groningen hasEmphasized Trade and Emphasized Trade and

FinanceFinance In Table 11 these account for roughly In Table 11 these account for roughly

half of the US-EU difference. Adding half of the US-EU difference. Adding business services brings the amount business services brings the amount explained to 95 percentexplained to 95 percent Large literature on US retailing, big boxes, Large literature on US retailing, big boxes,

EU land-use regulationsEU land-use regulations Difference between EUxMed and Med Difference between EUxMed and Med

lies elsewherelies elsewhere About 70 percent in manufacturing, rest in About 70 percent in manufacturing, rest in

construction and utilitiesconstruction and utilities

Is There a TradeoffIs There a Tradeoffat the Industry Levelat the Industry Level

There is no reason to think soThere is no reason to think so Changes in policies and attitudes of Changes in policies and attitudes of

women toward work should influence all women toward work should influence all industries, not the pattern across industries, not the pattern across industriesindustries

Nevertheless, we find a negative Nevertheless, we find a negative correlation of Y/H and E/N across correlation of Y/H and E/N across industriesindustries We did this for country groupsWe did this for country groups Here we display scatter and regressions for Here we display scatter and regressions for

all countriesall countries

Figure 9: The Turnaround Figure 9: The Turnaround by Industryby Industry

-10

-5

0

5

10

15

-15 -10 -5 0 5 10 15

Y/H Turnaround (Percent)

E/N

Turn

arou

nd (P

erce

nt)

▲ = ICT or Communications

■ = Med. (non-ICT or Comm)

Table 12: Table 12: Regressions of LP Turnaround* on E/N Turnaround*

Countries

Exclude ICT and Comm.

Coefficient

T-Statistic N R2 RMSE

All No -0.45 -4.35 179 0.10 3.000

All Yes -0.54 -5.94 149 0.19 2.495

Mediterranean Only No -0.82 -4.19 36 0.34 2.920

Mediterranean Only Yes -0.83 -5.60 30 0.53 2.140

* Turnaround equals 1995-2004 average growth minus 1980-1995 average growth

Which Industries were the Which Industries were the Heroes and Culprits in TFP Heroes and Culprits in TFP

Growth?Growth? Figure 10 shows post-1995 TFP Figure 10 shows post-1995 TFP

turnaround in each country group on turnaround in each country group on the vertical axis vs. EU total on the the vertical axis vs. EU total on the horizontalhorizontal

Horizontal shows largely negative Horizontal shows largely negative turnarounds for EUturnarounds for EU

Nordic: strong in ICT manufacturingNordic: strong in ICT manufacturing A-S: strong in finance, business A-S: strong in finance, business

services, weak in ICT manufacturingservices, weak in ICT manufacturing

Fig 10aFig 10a

Nordic

Business Services

Retail/

Wholesale

ICT

Mfg.

Finance

Real Estate

Nondurable Mfg

Non-ICT Durable

Mfg.Construction/

Utilities

Trans

Ag./Mining

Communication

GHI-0.05

-0.15

0.1

0.2

0.05 0.10

45º line

Fig 10bFig 10bAnglo-Saxon

-0.10

0.05

0.15

0.25

0.05

Business

Services

Retail/

Wholesale

ICT Mfg.

Finance

Real Estate

Nondurable Mfg.

Non-ICT Durable Mfg.

Construction/Utilities

Trans Ag./Mining

Communication

GHI

45º line

Continential and MedContinential and Med

Continental look like EU (49% pop Continental look like EU (49% pop share)share) Weak in finance and bus servicesWeak in finance and bus services (Finance people want to live in London (Finance people want to live in London

not Frankfurt)not Frankfurt) MedsMeds

All but three industries underperformAll but three industries underperform Particularly bad in all three parts of Particularly bad in all three parts of

manufacturingmanufacturing

Fig 10cFig 10cContinental

Retail/

Wholesale

ICT Mfg.

Finance

Real Estate

Nondurable Mfg.

Non-ICT Durable

Mfg.

Construction/

Utilities

Trans

Ag./Mining

Communication

GHI

Business

Services

0.10

0.05

-0.05

-0.15

45º line

Fig 10dFig 10dMediterranean

Business

Services

Retail/

Wholesale

ICT

Mfg.

Finance

Real Estate

Nondurable Mfg

Non-ICT Durable

Mfg.

Construction/

UtilitiesTrans.

Ag./Mining

Communication

GHI

-0.10 0.05

0.05

-0.15

0.15

-0.25

0.10

45º line

Finally, look at US vs EUFinally, look at US vs EU

Trade, finance, business services stand Trade, finance, business services stand outout

EU does better in comm, constr & ut, EU does better in comm, constr & ut, ag & miningag & mining

One common element, US does better One common element, US does better in using computersin using computers

Heterogeneity within EU not surprisingHeterogeneity within EU not surprising SF, Boston, Austin TX have outperformed SF, Boston, Austin TX have outperformed

in US as Ireland & Finland & London have in US as Ireland & Finland & London have outperformed in EUoutperformed in EU

Figure 11: US vs. EUFigure 11: US vs. EU

EU TFP

US T

FP Business

Services

Retail/

Wholesale

ICT

Mfg.

Finance

45º line

Real Estate

Nondurable

Mfg

Non-ICT Durable

Mfg.

GHI

Construction/

Utilities

Trans.

Ag./Mining

Communication

0.05

0.35

0.25

0.05

-0.05

-0.05 0.10

Conclusions and PolicyConclusions and PolicyImplicationsImplications

Using country groups brings coherence to Using country groups brings coherence to complex patterns across 15 EU countriescomplex patterns across 15 EU countries

Big increase in heterogeneity after 1995Big increase in heterogeneity after 1995 Negative correlation between employment Negative correlation between employment

and productivityand productivity Two channels between policy variables and Two channels between policy variables and

productivity – direct and indirect through productivity – direct and indirect through employmentemployment

Negative correlation not only across Negative correlation not only across countries and time but also across industriescountries and time but also across industries

Thinking about Policy viaThinking about Policy viathe Tradeoffthe Tradeoff

Some policies improve Y/N, some Some policies improve Y/N, some have little effecthave little effect

An added consideration is in policy An added consideration is in policy effects on the government budgeteffects on the government budget Consider EPL vs. ARR. Raising ARR Consider EPL vs. ARR. Raising ARR

raises govt expenditure, EPL does notraises govt expenditure, EPL does not Raising taxes raises government Raising taxes raises government

revenue with little effect on Y/Nrevenue with little effect on Y/N An anti George Bush messageAn anti George Bush message