Comment on I-T-vA, Comment on I-T-vA, “Market Services “Market Services

Productivity Across Europe Productivity Across Europe and the US”and the US”

Robert J. GordonRobert J. GordonNorthwestern University & NBERNorthwestern University & NBER

NBER Summer InstituteNBER Summer InstituteCambridge MA, July 19, 2007Cambridge MA, July 19, 2007

The Circle of DiscussionsThe Circle of DiscussionsBetween the Two PapersBetween the Two Papers

BvA discusses IDB-RJG on European BvA discusses IDB-RJG on European Productivity Slowdown at SF Fed in 11/05Productivity Slowdown at SF Fed in 11/05

Nordhaus discusses IDB-RJG, version #3 at Nordhaus discusses IDB-RJG, version #3 at NBER exactly one year ago (7/20/06), not NBER exactly one year ago (7/20/06), not exactly favorable, “waiting for the seventh exactly favorable, “waiting for the seventh draft”, quoting Galbraith.draft”, quoting Galbraith.

We’re working on draft #6, and now I’m We’re working on draft #6, and now I’m here to discuss I-T-vA on the same topichere to discuss I-T-vA on the same topic

Today’s paper was presented at Today’s paper was presented at EPEP Frankfurt 4/07, ours will be at Frankfurt 4/07, ours will be at EP EP Lisbon Lisbon 10/0710/07

Today’s paper hereafter “ITVA” for shortToday’s paper hereafter “ITVA” for short

One Year Ago Today One Year Ago Today WDN Quoted Galbraith, 7WDN Quoted Galbraith, 7thth Draft Draft

I don’t quote Galbraith, but some of you know I don’t quote Galbraith, but some of you know that I am prone to quoting Broadway song lyricsthat I am prone to quoting Broadway song lyrics

Broadway shows: something in common with ZviBroadway shows: something in common with Zvi• Sondheim “Merrily we go along” in DCSondheim “Merrily we go along” in DC• Custom-made tape of “She Loves Me”Custom-made tape of “She Loves Me”

One of my faves is the ultimate calendar song, One of my faves is the ultimate calendar song, every June 1 I must find an excuse to quote from every June 1 I must find an excuse to quote from “June is Busting out all Over” “June is Busting out all Over”

Rodgers & Hammerstein, Rodgers & Hammerstein, Carousel Carousel opened on opened on 4/19/45, set in southern Maine (Think of this as 4/19/45, set in southern Maine (Think of this as the Barbara Fraumeni song, southern Maine)the Barbara Fraumeni song, southern Maine)

Did it in Stockholm on 6/1/06, similar climateDid it in Stockholm on 6/1/06, similar climate

Usually I quote from the Chorus, Usually I quote from the Chorus, Today from the Verse, about this Today from the Verse, about this

PaperPaper

March went out like a lion, awakin’ the March went out like a lion, awakin’ the water in the baywater in the bay

Then April cried and stepped aside and Then April cried and stepped aside and along came pretty little Mayalong came pretty little May

May was full of promises, but she didn’t May was full of promises, but she didn’t keep them quickly enough for somekeep them quickly enough for some

And a crowd of doubtin’ Thomases was And a crowd of doubtin’ Thomases was predicting that the summer’d never comepredicting that the summer’d never come

This is the “May” paperThis is the “May” paper

The EU-KLEMS data base is a monumental The EU-KLEMS data base is a monumental achievement, DWJ might even say achievement, DWJ might even say “magisterial”“magisterial”

This paper has 32 pages of text & tables, This paper has 32 pages of text & tables, plus a 28-page data appendix and tables. plus a 28-page data appendix and tables. The data appendix is magnificent.The data appendix is magnificent.

Thus the paper, like May, is full of Thus the paper, like May, is full of promises but it doesn’t keep those promises but it doesn’t keep those promises. This discussion will be about promises. This discussion will be about the high hopes of the data appendix and the high hopes of the data appendix and the disappointment of the main text.the disappointment of the main text.

Credits for this discussionCredits for this discussion are joint with . . . are joint with . . .

Ian Dew-Becker, current co-author on EU Ian Dew-Becker, current co-author on EU productivity growth and former RA (2003-05)productivity growth and former RA (2003-05)• Thanks for Ian’s spontaneous comments on the I-T-vA Thanks for Ian’s spontaneous comments on the I-T-vA

paper and on an initial version of my commentspaper and on an initial version of my comments Robert Krenn, my current RA at NorthwesternRobert Krenn, my current RA at Northwestern

• Following in Ian’s path three years youngerFollowing in Ian’s path three years younger• Robert Inklaar has been very kind in answering RK’s Robert Inklaar has been very kind in answering RK’s

pesky data questions over the past few days.pesky data questions over the past few days. Both are products of NU’s elite MMSS program, Both are products of NU’s elite MMSS program,

where (among other things) they learn to live and where (among other things) they learn to live and breathe STATA as sophomoresbreathe STATA as sophomores

And you’ll see at the end the amazing things that And you’ll see at the end the amazing things that RK has produced from EU-KLEMS in the last 48 RK has produced from EU-KLEMS in the last 48 hours (from Minneapolis)hours (from Minneapolis)

The Big Achievement of EU-The Big Achievement of EU-KLEMSKLEMS

These authors deserve high praise These authors deserve high praise for their role in creating EU-KLEMSfor their role in creating EU-KLEMS

As DWJ said in opening this As DWJ said in opening this “productivity week,” EU-KLEMS is a “productivity week,” EU-KLEMS is a monumental achievementmonumental achievement

Now we can get it all off the web, Now we can get it all off the web, and indeed this discussion will show and indeed this discussion will show you a bit of what is available beyond you a bit of what is available beyond the tables in today’s I-T-vA paperthe tables in today’s I-T-vA paper

Some of the EU-KLEMS Some of the EU-KLEMS Achievements: Turn First to Achievements: Turn First to

the I-T-vA Data Appendixthe I-T-vA Data Appendix Data provided for output and inputs not in the simple 1957 Solow-Data provided for output and inputs not in the simple 1957 Solow-

Kendrick framework. Instead, achieves the Jorgenson-Griliches Kendrick framework. Instead, achieves the Jorgenson-Griliches research agendaresearch agenda

Data to adjust for labor and capital qualityData to adjust for labor and capital quality• Data also distinguish between gross output, II, and VAData also distinguish between gross output, II, and VA

Uniform as possible for 25 EU countries, US, Canada, Japan, KoreaUniform as possible for 25 EU countries, US, Canada, Japan, Korea Easy to use, just try to download anything and it pops out in ExcelEasy to use, just try to download anything and it pops out in Excel Biggest achievements:Biggest achievements:

• Harmonization (see below)Harmonization (see below)• Incredible task of pulling together labor composition and relative Incredible task of pulling together labor composition and relative

compensation measures, allowing comprehensive measures of labor compensation measures, allowing comprehensive measures of labor qualityquality

• Imposition of consistent industry-by-industry PPPsImposition of consistent industry-by-industry PPPs• Locating data to make distinction between GO, II, and VALocating data to make distinction between GO, II, and VA

Further Comments Further Comments on the Data Appendixon the Data Appendix

About p. 33, Zvi didn’t call the services “unmeasurable,” he About p. 33, Zvi didn’t call the services “unmeasurable,” he called them “hard to measure”. There’s a differencecalled them “hard to measure”. There’s a difference

Harmonization is a monumental achievement of EU-KLEMSHarmonization is a monumental achievement of EU-KLEMS• Wonderful Groningen innovation: Imposed in computer pricesWonderful Groningen innovation: Imposed in computer prices• Also imposed on depreciation ratesAlso imposed on depreciation rates• But what about output measurement in finance, business But what about output measurement in finance, business

services?services?• ABC distinction: finance, bus svcs 46% “C” but Hotels and ABC distinction: finance, bus svcs 46% “C” but Hotels and

restaurants 73% “A” gradesrestaurants 73% “A” grades This is important because of EU-US LP difference in finance This is important because of EU-US LP difference in finance

and business servicesand business services No sense from paper about whether US does a different job, No sense from paper about whether US does a different job,

no accounting for “A B C” for USno accounting for “A B C” for US What role of movement from “C” to “A” in post-1995 What role of movement from “C” to “A” in post-1995

change in productivity growth behavior?change in productivity growth behavior?

Problems with EU-KLEMSProblems with EU-KLEMS Authors drop 5 countries and reduce the sample Authors drop 5 countries and reduce the sample

to EU-10 instead of EU-15to EU-10 instead of EU-15 We have found that this peeling off of the We have found that this peeling off of the

“rejected 5” is unnecessary“rejected 5” is unnecessary• Ireland has some missing data in sub-industries before Ireland has some missing data in sub-industries before

1995, no problem to patch this1995, no problem to patch this• Sweden industry 322 telecom production, Sweden industry 322 telecom production, negative VA in negative VA in

2001, dropped after 20022001, dropped after 2002 How could Ericson have negative VA??How could Ericson have negative VA?? We solve not by throwing out Sweden but by having a We solve not by throwing out Sweden but by having a

slightly broader definition of ICT production in Sweden then slightly broader definition of ICT production in Sweden then in other countries (30t33)in other countries (30t33)

No attention to Diewert point about net vs. gross, No attention to Diewert point about net vs. gross, despite imposition of harmonization of despite imposition of harmonization of depreciationdepreciation

Four Criticisms at the Outset, about Four Criticisms at the Outset, about substance & expositionsubstance & exposition

DWJ and Co-authors (J-H-S-S) on Monday DWJ and Co-authors (J-H-S-S) on Monday were right that we’re interested in the were right that we’re interested in the entire economy, no just NFPB or in this entire economy, no just NFPB or in this paper just in “Market Services” (MS), a paper just in “Market Services” (MS), a subset of NFPBsubset of NFPB

Other Sectors:Other Sectors:• OP = Other Private = ag, mining, mfg, OP = Other Private = ag, mining, mfg,

construction, utilitiesconstruction, utilities• GHI = govt, households, institutionsGHI = govt, households, institutions

1995 Share of Total Economy VA1995 Share of Total Economy VAin OP, MS, and GHIin OP, MS, and GHI

0 10 20 30 40 50 60

GHI

Mkt Services

Other Private

EU

USUS

What the “Market Services” What the “Market Services” Orientation OmitsOrientation Omits

EU: 32.0 percent in OP, 23.6 in GHIEU: 32.0 percent in OP, 23.6 in GHI US: 26.7 percent in OP, 24.9 in GHIUS: 26.7 percent in OP, 24.9 in GHI Note: US has a higher share in MS as has Note: US has a higher share in MS as has

been widely noted (Europe home been widely noted (Europe home production vs. US market production)production vs. US market production)

Surprise: US has a slightly higher share in Surprise: US has a slightly higher share in GHIGHI

Focus on MS ignores cross-country Focus on MS ignores cross-country differences in ICT production and ICT use differences in ICT production and ICT use in OPin OP

Does Omission of OP and GHIDoes Omission of OP and GHIMatter?Matter?

Henceforth, the word “turnaround” Henceforth, the word “turnaround” will refer to growth 1995-2004 minus will refer to growth 1995-2004 minus growth 1978-1995growth 1978-1995

EU-US turnaround is primarily in EU-US turnaround is primarily in market servicesmarket services

But a second theme is the increase in But a second theme is the increase in within-EU heterogeneitywithin-EU heterogeneity

Here much of the action is in OP, not Here much of the action is in OP, not just in MSjust in MS

Second Initial Complaint,Second Initial Complaint,Refusal to Provide EU averagesRefusal to Provide EU averages

Look at Tables 2, 3, 4, 5, 6Look at Tables 2, 3, 4, 5, 6 The “averages” are unweighted The “averages” are unweighted

across countriesacross countries The “averages” include US so don’t The “averages” include US so don’t

allow a contrast of EU vs. USallow a contrast of EU vs. US Yet the EU-10 weighted averages are Yet the EU-10 weighted averages are

right there on EU-KLEMS web site, right there on EU-KLEMS web site, waiting for the discussant to provide waiting for the discussant to provide the missing lines in these tablesthe missing lines in these tables

Paper’s Table 2 IncludingPaper’s Table 2 IncludingEU-10 Averages: MS Nominal VA EU-10 Averages: MS Nominal VA

Share in 1980, 1995, 2004Share in 1980, 1995, 2004

0 5 10 15 20 25 30 35 40 45 50

2004

1995

1980EU

US

Turnaround is Different in MSTurnaround is Different in MSvs. Total Economyvs. Total Economy

0 0.5 1 1.5 2 2.5 3 3.5

1995-2004

1978-1995

1995-2004

1978-1995

EU

US

Market Services

Tot

Total Economy

Third Complaint: They Don’t Third Complaint: They Don’t Provide Growth Rates of of Labor Provide Growth Rates of of Labor

and Capital Qualityand Capital Quality Yet these can be easily calculated from their data base.Yet these can be easily calculated from their data base. We didn’t have time to calculate these for today’s We didn’t have time to calculate these for today’s

discussiondiscussion We also didn’t have time to calculate anything about capital We also didn’t have time to calculate anything about capital

deepening or TFP growthdeepening or TFP growth Capital & labor quality are hiding inside Appendix Table A2Capital & labor quality are hiding inside Appendix Table A2

• Turns out from Table A2 that changes in cap and labor Turns out from Table A2 that changes in cap and labor composition are minor compared to shifts in II share of GO, not composition are minor compared to shifts in II share of GO, not discussed in body of paper at alldiscussed in body of paper at all

• Employment vs. hours, columns 1 vs. 2 of Table A2 is a straw Employment vs. hours, columns 1 vs. 2 of Table A2 is a straw person. person. EveryoneEveryone already measures productivity as output per already measures productivity as output per hour, not output per person.hour, not output per person.

What’s worse, the alternative MFP concepts in Table A2 What’s worse, the alternative MFP concepts in Table A2 refer only to 1995-2004, not to the refer only to 1995-2004, not to the turnaround turnaround before and before and after 1995after 1995

Fourth Complaint:Fourth Complaint:No Country Sub-aggregatesNo Country Sub-aggregates

They have 10 EU countries listed They have 10 EU countries listed alphabetically, along with USalphabetically, along with US

Are there interesting ways to group the 10 Are there interesting ways to group the 10 EU countries?EU countries?

Not for them, but yes for usNot for them, but yes for us We’ve given up on Tigers-TortoisesWe’ve given up on Tigers-Tortoises We’ve adopted standard geographical We’ve adopted standard geographical

breakdown: Nordic, Anglo-Saxon, Other, breakdown: Nordic, Anglo-Saxon, Other, and Mediterraneanand Mediterranean• Citation: Pissarides et al, Citation: Pissarides et al, Women at Work, Women at Work,

Oxford Univ Press 2005Oxford Univ Press 2005

Substantive Analysis in Paper:Substantive Analysis in Paper:“Growing Role ICT and HK”“Growing Role ICT and HK”

The best part is their inclusion of the “technology gap” in The best part is their inclusion of the “technology gap” in the regressions, the traditional “convergence” term that we the regressions, the traditional “convergence” term that we also include and is always included in the Barro / Sala-i-also include and is always included in the Barro / Sala-i-Martin literature.Martin literature.

Otherwise, This part of the paper is dismayingOtherwise, This part of the paper is dismaying No link with the DWJ formulation of labor qualityNo link with the DWJ formulation of labor quality They use the LEVEL of college share as a determinant of They use the LEVEL of college share as a determinant of

the GROWTH RATE of MFPthe GROWTH RATE of MFP This is crazy. This would predict that the US would have This is crazy. This would predict that the US would have

had the fastest productivity growth 1973-95had the fastest productivity growth 1973-95 Plus they fail to run multivariate regressions, they do HK, Plus they fail to run multivariate regressions, they do HK,

ICT, and regulations ICT, and regulations separately!separately!

The Entire Analysis of Human The Entire Analysis of Human Capital is MisconceivedCapital is Misconceived

Go back to the DWJ concept of “labor Go back to the DWJ concept of “labor quality,” this is the quality,” this is the difference in the difference in the growth rate of effective labor input vs. raw growth rate of effective labor input vs. raw labor inputlabor input

It is a growth rate conceptIt is a growth rate concept It grows when the share of college grads It grows when the share of college grads

grows, it has nothing to do with the LEVEL grows, it has nothing to do with the LEVEL of educational attainmentof educational attainment

Why didn’t the US have the fastest Why didn’t the US have the fastest productivity growth in the 70s and 80s productivity growth in the 70s and 80s when it was furthest ahead in educ when it was furthest ahead in educ attainment?attainment?

Regression AnalysisRegression Analysis

Not Motivated by a ModelNot Motivated by a Model Confusion between Levels and Rates of Confusion between Levels and Rates of

ChangeChange No attempt at Multivariate Regression, No attempt at Multivariate Regression,

why should ICT, Human Capital, and why should ICT, Human Capital, and Regulations be entered one-at-time Regulations be entered one-at-time instead of together?instead of together?

ICT section like HK section is plagued by ICT section like HK section is plagued by running level variables when rate of running level variables when rate of change variables are what matterchange variables are what matter

Now it’s time to switch colors and look at some new results

with EU-KLEMS What we did

Aggregated VA across countries and industries with nominal Tornquist VA weights

Aggregated hours with hours weights Created 12 industry sub-aggregates including

OP, MS, and GHI Created Four EU-country groups This is supposed to demonstrate how wonderful

are the EU-KLEMS data, how easy to use. But, wow, it is a lot of work. All those missing

numbers, what to do about them

Conclusions #1Conclusions #1 With a Wonderful new Data Base, don’t With a Wonderful new Data Base, don’t

limit your analysis to 45% of the economylimit your analysis to 45% of the economy Enormous praise for the huge opportunity Enormous praise for the huge opportunity

of the research community to analyze this of the research community to analyze this prize of 2007prize of 2007

Pull out the labor quality and capital Pull out the labor quality and capital quality changes explicitly across countries quality changes explicitly across countries and talk about themand talk about them

Data worries about harmonization of Data worries about harmonization of computer prices and depreciation and not computer prices and depreciation and not of measures of finance and business of measures of finance and business servicesservices

Conclusions #2Conclusions #2

Please in the future don’t forget these Please in the future don’t forget these basic guidelines for an analysisbasic guidelines for an analysis• Don’t leave out more than half of the economyDon’t leave out more than half of the economy• Don’t omit EU-10 averages when comparing US Don’t omit EU-10 averages when comparing US

to EUto EU• Don’t omit the basic contribution of the DWJ Don’t omit the basic contribution of the DWJ

approach, the growth rates of labor and capital approach, the growth rates of labor and capital qualityquality

• Don’t list EU countries alphabetically, assemble Don’t list EU countries alphabetically, assemble them into plausible sub-aggregatesthem into plausible sub-aggregates

Conclusions #3Conclusions #3

Don’t mix up level and rate of Don’t mix up level and rate of change effectschange effects• Why should EU-KLEMS adhere so closely Why should EU-KLEMS adhere so closely

to DWJ-ZG 1967 and then go off the to DWJ-ZG 1967 and then go off the deep end in confusing levels from deep end in confusing levels from growth ratesgrowth rates

• Why should the LEVEL of human capital Why should the LEVEL of human capital have anything to do with MFP growth?have anything to do with MFP growth?

• Same for ICT useSame for ICT use

Conclusions from Our LookConclusions from Our Lookat EU-KLEMSat EU-KLEMS

Yes, market services are at the heart Yes, market services are at the heart of the EU-US turnaroundof the EU-US turnaround• But it’s not like van Ark in 2001, it’s not But it’s not like van Ark in 2001, it’s not

80% trade and 20% financial. It’s now 80% trade and 20% financial. It’s now trade and business services, do we know trade and business services, do we know how to measure BS?how to measure BS?

But market services are not at the But market services are not at the heart of heterogeneity within Europeheart of heterogeneity within Europe

What’s Going on within Europe?What’s Going on within Europe?

Only Anglo-Saxon shows dominance of MSOnly Anglo-Saxon shows dominance of MS Other country groups (Nordic, OE, Med) Other country groups (Nordic, OE, Med)

have the big divergence in OP, not in MShave the big divergence in OP, not in MS Much still to analyze across the four Much still to analyze across the four

country groups – ICT production, ICT use, country groups – ICT production, ICT use, capital deepening, MFPcapital deepening, MFP

Stay tooned for all that in Lisbon on Stay tooned for all that in Lisbon on October 20, 2007!October 20, 2007!