WHAT HAPPENED TO THE KNOWLEDGE ECONOMY? ICT, INTANGIBLE INVESTMENT, AND BRITAIN S PRODUCTIVITY RECORD REVISITED. and. Gavin Wallis

roiw_344 686..716 Review of Income and Wealh Series 55, Number 3, Sepember 2009 WHAT HAPPENED TO THE KNOWLEDGE ECONOMY? ICT, INTANGIBLE INVESTMENT, AND BRITAIN S PRODUCTIVITY RECORD REVISITED by Mauro Giorgio Marrano HM Treasury Jonahan Haskel* Imperial College Business School, CEPR, and IZA and Gavin Wallis HM Treasury and Universiy College London Despie he apparen imporance of he knowledge economy, U.K. macroeconomic performance appears unaffeced: invesmen raes are fla, and produciviy has slowed. We invesigae wheher measuremen issues migh accoun for his puzzle. The sandard Naional Accouns reamen of mos spending on knowledge or inangible asses is as inermediae consumpion. Thus hey do no coun as eiher GDP or invesmen. We ask how reaing such spending as invesmen affecs some key macro variables, namely, marke secor gross value added (MGVA), business invesmen, capial and labor shares, growh in labor and oal facor produciviy (TFP), and capial deepening. We find: (a) MGVA was undersaed by abou 6 percen in 1970 and 13 percen in 2004; (b) insead of he business invesmen/mgva raio falling since 1970 i has been rising; (c) insead of he labor share being fla since 1970 i has been falling; (d) growh in labor produciviy and capial deepening has been undersaed and growh in TFP oversaed; and (e) TFP growh has no slowed since 1990 bu has been acceleraing. 1. Inroducion According o common debae, he Knowledge Economy is all around us. Think anks and commenaors argue ha developed counries have no fuure in a globalized economy unless hey specialize in knowledge-inensive aciviies. Whole goods and occupaions, many based on ICT (Informaion and Communicaion Technology), ha were almos unheard of even five years ago, proliferae Noe: Financial suppor for his research comes from HM Treasury and he ESRC/EPSRC Advanced Insiue of Managemen Research, gran number RES-331-25-0030. This work, designed o simulae discussion and criical commen, was carried ou a CeRiBA, he Cenre for Research ino Business Aciviy a he ONS (www.ceriba.org.uk) and conains saisical daa from ONS, which is Crown copyrigh and reproduced wih he permission of he conroller of HMSO and Queen s Priner for Scoland. The use of he ONS saisical daa in his work does no imply he endorsemen of he ONS or HM Treasury in relaion o he inerpreaion or analysis. This work uses research daases ha may no exacly reproduce Naional Saisics aggregaes. We are graeful for commens and inpu from Angus Armsrong, Tony Clayon, Carol Corrado, Jennifer Greenslade, Charles Hulen, Chris Kelly, Nick Oulon, Dave Ramsden, Dan Sichel, Sally Srinivasan, and Ken Warwick. We are also graeful o Graeme Chamberlain, Sumi Dey-Chowdhury, Peer Goodridge, and Joe Robjohns for help wih daa. All opinions and errors in his paper are hose of he auhors alone and no heir affiliaed insiuions. *Correspondence o: Jonahan Haskel, Imperial College Business School, Tanaka Building, Souh Kensingon Campus, London SW7 2AZ, UK (j.haskel@ic.ac.uk). 2009 The Auhors Journal compilaion 2009 Inernaional Associaion for Research in Income and Wealh Published by Blackwell Publishing, 9600 Garsingon Road, Oxford OX4 2DQ, UK and 350 Main S, Malden, MA, 02148, USA. 686

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 (hink of ipods, SaNavs and Search Engine Programmers). Pressure groups exol boh he virues and conribuion of he U.K. design and creaive indusries. Where has all his shown up in U.K. economic performance? The raio of nominal invesmen o nominal GDP has sayed more or less where i was since he 1950s, which is hard o square wih he percepion ha firms are invesing in knowledge asses in he eeh of a echnological revoluion. The share of profis in GDP has also remained quie sable, so if firms are invesing more, his is no being refleced in addiional profis. U.K. produciviy performance over he 1990s remains a puzzle. According o exising sudies U.K. produciviy growh deerioraed afer 1995. 1 This is in conras o he U.S. experience, where here was a well-documened produciviy acceleraion associaed wih he ICT invesmen boom. 2 The U.K. also experienced an ICT invesmen boom in he lae 1990s, so why he produciviy slowdown? Two possible answers could explain hese puzzles. The firs is ha he invesmen in and/or he impac of he knowledge economy is in fac much less han popular discussion would sugges. The second is ha he impac is hidden by measuremen problems. This paper examines he second view for he U.K., building on previous work by Oulon and Srinivasan (2003a), Basu e al. (2003) and Oulon and Srinivasan (2005). 3 Work by Oulon and Srinivasan (2003a, 2005) examined a number of measuremen issues. They incorporaed sofware ino oupu; measured capial as capial services, no capial asses; buil he capial daa from a disaggregaed level; and measured labor qualiy raher han jus hours. Basu e al. (2003) specifically looked a wheher ICT measuremen could explain missing U.K. produciviy growh in he 1990s. The produciviy growh slowdown sill remained in all hese sudies and he auhors argued ha i was likely due o unmeasured invesmen in organizaional capial. We build on his work by using all hese elemens in our daa bu also adding inangible asses (where we hink of spending on inangible asses as describing spending on a broad range of knowledge-ype asses, such as R&D, sofware, organizaional capial, ec; deails below). Since one of our asses is organizaional capial, given he previous work, i is of ineres o see if incorporaing his can change he picure of U.K. produciviy performance in he lae 1990s. Whils here have been some sudies for he U.S. on he impac of inangibles on GDP, such as Corrado, Hulen, and Sichel (2005, 2006) and Nakamura (1999, 2001, 2003) here has no been anyhing for he U.K. We refer o Corrado, Hulen and Sichel as CHS from here onwards. In his paper we follow he cenral observaion in hese U.S. papers ha in pracice, spending on mos knowledge asses is, in accouning erms, like spending on oher inangible asses, such as sofware. 1 O Mahony and de Boer (2002), Oulon and Srinivasan (2003a) and Oulon and Srinivasan (2005). 2 Oliner and Sichel (2000), Jorgenson and Siroh (2000a, 2000b), Siroh (2002) and Gordon (2003). 3 The measuremen problems addressed in his paper are no problems wih he exising U.K. Naional Accouns, which coninue o be compiled based on inernaionally agreed definiions. The focus of his paper relaes o wha he impac would be of exending he agreed definiion of capial asses o include a broader range of inangible asses. As such, he oupu and produciviy esimaes should no be regarded as correcions o exising Naional Accouns raher as adjusmens o Naional Accouns daa for he wider definiion of inangible capial we adop. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 687

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 Spending on angible asses has a long measuremen radiion; i is par of invesmen and herefore also par of GDP. However, spending on inangible asses, wih a few excepions, is reaed as inermediae expendiure. In consrucing GDP herefore, spending on R&D for example, is reaed like spending on elecriciy, i.e. i is assumed no o be invesmen and so produces no asse a he end of he period. As an inermediae i does no appear in eiher invesmen or GDP daa. 4 The upcoming revision o he Sysem of Naional Accouns (SNA) will include a recommendaion o capialize R&D, bu his is only one ype of knowledge asse. This paper ries o answer he following quesion: wha are he consequences for a range of macroeconomic variables, including produciviy, of reaing spending on inangibles as invesmen raher han as inermediae expendiure? We do his as follows. Firs, following he approach of CHS, we assemble daa on spending for a range of inangible asses. These are wider han he usual R&D and sofware and include design, spending on repuaion, and human capial. Using hese daa we analyze he relaive quaniies of differen ypes of expendiure and how hey have changed over ime. We believe his quesion of ineres since we have lile informaion on wheher expendiure on sofware is more or less han expendiure on design, raining, ec. and also on how such expendiure has changed over ime. Second, we look a he consequences for overall business invesmen and marke secor gross value added, henceforh MGVA. 5 We believe his o be of ineres since some have conjecured ha ignoring inangibles explains why he radiionally-measured (i.e. focused on angible capial) business nominal invesmen o MGVA raio has remained so fla in he U.K. despie he percepion ha he underlying condiions for invesmen have been so favorable in recen years. We also hink i of imporance if he level of MGVA has been sysemaically undersaed. Third, we look a he consequences for marke secor labor produciviy. Since he level of MGVA rises, he level of labor produciviy rises. Bu labor produciviy growh will only rise if he level of MGVA rises increasingly over he period; so he answer o his quesion is no as clear. The reason why labor produciviy rises is ha here is an exra inpu in he economy, namely an inangible capial sock, in addiion o he angible capial sock. Thus we hink a quesion of ineres is: how much has he change in he inangible capial sock conribued o produciviy growh along wih oher inpus? The sysemaic mehod of answering his quesion is via growh accouning and so we exend previous U.K. growh accouning sudies by including inangible capial. Noe ha no only has MGVA and inpu changed, bu he facor shares will have changed as well, since he exra MGVA due o inangibles is mached on he income side by exra income o capial. 4 Michael Mandel s (2006) Business Week aricle for example describes he reamen of Apple by naional income accounans as hey coun each ipod wice: when i arrives from China, and when i sells. Tha, in effec, reduces Apple one of he world s greaes innovaors o a reseller of impored goods. 5 Owing o he difficuly of measuring produciviy and inangible spending in he governmen secor we focus on he marke secor raher han he whole economy. Hence our esimaes are based on marke secor GVA raher han GDP. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 688

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 I is worh poining ou ha we presen esimaes of MGVA, labor produciviy growh and growh accouning boh wih and wihou inangibles. Thus he reader unineresed in inangible asses can herefore obain, we hope, some ineresing daa from he paper. In paricular, all our esimaes are consisen wih he 2006 Blue Book and give daa up o 2004. 6 Our main findings are as follows. Firs, nominal business invesmen in inangible asses in 2004 is abou equal o nominal business invesmen in angibles (each is around 15 percen of MGVA). Of ha inangible invesmen, around 50 percen of he oal is on economic compeencies (invesmen in repuaion, human and organizaional capial), 35 percen on innovaive propery (mainly scienific and non-scienific R&D and design), and 15 percen on compuerized informaion (mainly sofware). Since 1970, nominal invesmen in inangibles has grown from abou 6 percen of nominal MGVA o abou 15 percen. Ineresingly, he paerns of growh look remarkably like hose in he U.S., alhough he U.K. has a bi less R&D invesmen and a bi more invesmen in design. Second, accouning for inangibles makes a considerable difference o measured MGVA and he share of ha oupu accouned for by invesmen and paymens o capial and labor. The level of nominal MGVA rises by abou 13 percen in 2004. 7 The share of nominal invesmen and paymens o capial also rise. Third, accouning for inangibles affecs labor produciviy growh (LPG) and oal facor produciviy growh (TFPG) in he marke secor. Wihou inangibles, we confirm previous work ha LPG and TFPG boh slow down beween 1990 95 and 1995 2000. 8 We also documen a furher sligh slowdown in 2000 04 in LPG bu a speeding up in TFPG. Wih inangibles, he picure changes ineresingly. Firs, boh LPG and TFPG accelerae beween 1990 95 and 1995 2000, suggesing ha he U.K. economy was building sufficien inangibles during ha period such ha heir omission affeced he produciviy saisics in imporan ways. Second, even wih inangibles, he pos-2000 slowdown in LPG sill remains. There are of course a hos of measuremen problems in esimaing invesmen in inangible asses. We have herefore assessed he robusness of our findings o differen measuremen mehods and assumpions. In summary, our major findings (on he shares of nominal invesmen and wages and LPG and TFPG) are robus o changes in hese measuremen mehods and assumpions. The ouline of he res of his paper is as follows. In he nex secion, we se ou how inangibles affec MGVA and growh. Secion 3 describes he daa used o ry o measure he impac of reaing inangible spending as invesmen (in our case he impac on business invesmen and marke secor gross value added), and Secion 4 oulines our growh accouning approach. Secion 5 presens our growh accouning resuls and Secion 6 concludes. 6 The Blue Book is he annual publicaion of a balanced se of U.K. Naional Accouns. 7 This is less han he share of inangible invesmen in MGVA because some inangible invesmen (some sofware, mineral exploraion and copyrigh coss) is already included in MGVA (abou 2 percen). I is slighly more han he figure in Giorgio Marrano and Haskel (2006), bu ha repored inangible invesmen as a proporion of GDP. 8 O Mahony and de Boer (2002), Oulon and Srinivasan (2003a) and Oulon and Srinivasan (2005). 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 689

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 2. Model We follow CHS in seing ou he following model. Suppose here are hree goods produced, a consumpion good, wih real oupu volume C and price P C, a angible invesmen good, I wih price P I, and an inangible invesmen good N wih price P N, where he subscrip denoes ime. 2.1. Inangibles Treaed as Inermediaes Suppose firs ha he inangible invesmen good is regarded as an inermediae. The angible capial sock K is assumed o accumulae according o he perpeual invenory model (1) K = I + ( δ K) K 1 1 wih depreciaion rae d K (assumed consan over ime). Then we can wrie he producion funcion for each secor and, assuming facors are paid heir marginal produc and he producion funcion is homogenous of degree one, he money flows for each secor as follows: N N L K (2) ( a ) Inangible secor: N = F ( L, K, ) ; P N = P L + P K N, N, N, I ( b ) Tangible secor: I I L = F ( LI,, KI,, NI,, ) ; P I = P L I, + P K K N I, + P NI, ( ) = ( ) = + C c Consumpion secor : C F LC,, KC,, NC,, ; P C C P L LC, K N P KC, + P NC, where he superscrips N, I, and C refer o he hree secors. So, for example, in equaion (2a), he lef-hand side producion funcion in he inangible secor says ha he oupu of inangibles is produced by labor in he secor and angible capial in he secor. The righ-hand side equaion says ha wih facors paid heir marginal producs, he value of he inangibles produced equals he reurns o labor and angible capial used in ha secor. Since inangibles are assumed o be inermediaes, he producion funcions in equaions (2b) and (2c) for he angible and consumpion secor show ha he volume of inangible oupu is simply an inpu ino he producion of angible and consumpion goods (we omi oher inermediaes which similarly ne ou). Since hey are inermediae inpus, inangibles do no appear in oal oupu which can be wrien 9 N, (3) Q C I L K P Q = P C + P I = P L + P K 9 This equaion shows he equaliy of GDP on he expendiure side (consumpion plus invesmen) and income side (rewards o he non-inermediae facors labor and capial). On he producion side, value added in he C, I and N secors are, respecively, he value of consumpion less paymens o inangibles used in he consumpion secor (he inermediae good), he value of invesmen less paymens o inangibles in he invesmen secor, and he value of inangibles. Adding hese up gives economy value added as he value of consumpion plus invesmen, which wih facors being paid heir marginal produc is equal o wages and capial paymens in all hree secors. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 690

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 where he prime indicaes he case where inangibles are reaed as inermediae expendiure and L = L N + L I + L C and K = K N + K I + K C. 2.2. Inangibles Treaed as Capial Now suppose ha he inangible invesmen good is reaed as capial. Then as well as he angible capial accumulaion, inangible capial sock, R also accumulaes according o (4) R = N + ( δr) R 1 1 where R depreciaes a rae d R. The producion funcion and money flows for each secor can be wrien N N L (5) ( a ) Inangible secor: N = F ( LN, KN, RN, ) ; P N = P L K R P K + P R +,,, N, N, N, I I L K R ( b ) Tangible secor : I = F ( L, K, R, ) ; PI = PL + P K + PR I, I, I, C C L ( c ) Consumpion secor: C = F LC, KC, RC, ; P C P L K R P K + P R,,, C, C, C, I, I, I, ( ) = + Noe ha in conras o equaion (2) he sock of inangible capial, R, raher han inangible oupu, appears as an inpu in he producion funcions, and he paymens o ha sock, P R, appears in he paymen equaions raher han paymen R for he enire used up inermediae oupu. The corresponding oupu ideniy now N includes he value of oupu of he inangible good on he producion side, P N, R and he paymens o he sock of inangibles, P R, on he income side (6) Q C I N L K R P Q = P C + P I + P N = P L + P K + P R where he oal oupu of he inangible good N = N N + N I + N C and he inangible sock is R = R N + R I + R C. Thus he following poins are worh noing. Firs, oupu is increased under Q Q he second approach from P Q o P Q. Second, he invesmen rae increases I Q I N Q from PI P Q o ( PI + P N) PQ and he labor share falls from L Q L Q P L P Q o P L P Q, where he labor share is he proporion of oal income paid o labor. Finally, o undersand he implicaions for TFPG, we may wrie a growh accouning relaion from he producion funcions above: (7) L K ( a) ΔlnTFP = ΔlnQ s Δln L s Δln K L K R ( b) ΔlnTFP = ΔlnQ s ΔlnL s ΔlnK s ΔlnR 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 691

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 where equaion (7a) shows he expression for TFPG =DlnTFP in he case where inangibles are expensed and equaion (7b) where hey are capialized and he shares of each facor are denoed wih an s. 10 As he equaions show, he effec of including inangibles on TFPG is ambiguous. Whils he level of oupu has risen, he growh rae may or may no rise depending on he growh rae of real inangible invesmen. So he effec on DlnQ is ambiguous. In addiion, he capializaion of inangibles means ha (he growh in) an addiional inpu has o be included as a deerminan of growh. Thus we have more capial asses accouning for DlnQ so ha TFPG may rise or fall. Noe finally ha he shares differ beween equaions (7a) and (7b) since boh oupu and he paymens o capial differ. N The exra oupu from now including inangible oupu (wih value P N )is mirrored by he paymens o he exra facor of producion, namely he inangible capial sock. Since i is a par of capial, his increases he overall paymens o capial. Also, he producion funcions make clear ha he inangible inpu is he volume of inangible spending in he firs case and he (flow of services from he capial) sock of inangible capial in he second. This means ha he income flows have been evaluaed using he renal raes of labor, and angible and inangible capial services. 3. Inangible Spending and he Invesmen and Labor Shares Here we discuss our daa on inangible spending, invesmen,and he overall invesmen and labor shares. We follow CHS in idenifying hree main inangible asse classes. 11 Firsly, compuerized informaion (mainly sofware); secondly, innovaive propery (mainly scienific and non-scienific R&D); and finally, firm compeencies (company spending on repuaion, and human and organizaional capial). 3.1. Spending on Inangible Asses Table 1 shows our choice of inangible asses, heir daa sources, he expendiure figures for 2004, he proporion of he expendiure assumed o be invesmen (following he assumpions of CHS), he percenage of oal inangible invesmen, heir deflaors, and depreciaion raes. In his secion we will limi our aenion o 10 The shares are he paymens o each facor as a share of oal paymens o all facors. Toal paymens add up o oupu, which of course consiss of paymens o wo facors when inangibles are inermediaes and paymens o hree facors when expensed; hus he shares are differen beween (7a) and (7b). 11 CHS have an exensive discussion of why invesmen in hese inangibles should be reaed as capial. They also discuss wha fracion of measured inangible expendiure should be regarded as invesmen and hereby capialized. Finally, one migh ask if he lis of inangible caegories is exhausive. This is discussed in Giorgio Marrano and Haskel (2006). I does, for example, accord wih he caegories used by he U.K. Compeiion Auhoriies in calculaing inangible asses for he purpose of compeiion analysis. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 692

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 TABLE 1 Inangible Asses Type of inangible invesmen 2004 Cross Secion Time Series Toal Spending bn, 2004 Proporion of Spending Considered as Invesmen % of Toal Inangible Invesmen, 2004 Deflaor (1) (2) (3) (4) (5) (6) (7) (8) Depreciaion Rae Compuerized informaion (1) Compuer sofware ONS esimaes 2004 1970 ONS daa 21.59 1 16.7 ONS deflaors 0.33 (2) Compuerized daabases Included in our sofware esimaes (3) Toal 21.59 17.7 Innovaive propery (4) Scienific R&D Curren expendiure on R&D from BERD. R&D in compuer indusry subraced 2004 1981 BERD 1980 1970 Backcas using he growh rae in he Annual Absrac of Saisics 12.4 1 10.2 Implied marke secor GVA deflaor (5) Mineral exploraion Naional Accouns 2004 1970 ONS daa 0.4 1 0.3 Implied marke secor GVA deflaor (6) Copyrigh and license coss (7) Oher produc developmen, design and research (8) New produc developmen coss in he financial indusry Naional Accouns 2004 1970 ONS daa 2.4 1 2 Implied marke secor GVA deflaor 20% of all inermediae purchase by Financial Services indusry, ONS daa. Inermediae purchases reduced by purchases of adverising, sofware, consuling and design. 2003 1992 20% of inermediae consumpion of he financial secor (SIC 65, 67 I-O 100,102). Source: Inpu Oupu Analysis 1991 1970 Backcased using he growh rae of he urnover of he secor Banking, finance, insurance business services, leasing from he Blue Book (afer consrucing a consisen ime series wih he Blue Book daa) 1 6 1 4.9 Implied marke secor GVA deflaor 0.2 0.2 0.2 0.2 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 693

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 TABLE 1 (coninued) Type of inangible invesmen 2004 Cross Secion Time Series Toal Spending bn, 2004 Proporion of Spending Considered as Invesmen % of Toal Inangible Invesmen, 2004 Deflaor (1) (2) (3) (4) (5) (6) (7) (8) Depreciaion Rae (9) New archiecural and engineering designs (10) R&D in social science and humaniies Esimaed as half of he oal urnover of he archiecure and design indusry SIC 74.2, ABI daa. Turnover reduced by purchases of adverising, sofware, consuling. Includes also urnover of specialiy design aciviies SIC 74842 muliplied by 2 o consider also own accoun Esimaed as wice indusry revenues of social science and humaniies R&D indusry 2004 1995 50% of he urnover of he indusry SIC 74.2, source ABI published daa. 1994 1992 50% of he urnover of he indusry SIC 74.2, source Service Secor Review. 1991 1985 Backcased using he growh rae of he urnover of archiecs and engineers as published in Business Monior. 1984 1979 Backcased above using he growh rae of urnover of SIC 74.2 1979 1970 Backcased using he growh rae of he urnover of he secor Banking, finance, insurance business services, leasing from he Blue Book (afer consrucing a consisen ime series wih he Blue Book daa) 2004 1995 Two imes he urnover of he SIC 73.2. Source: ABI 1994 1992 Two imes he urnover of SIC 73.2. Source: Service Secor Review 1991 1986 Backcas using he growh rae of he urnover of Research and Developmen services as in Business Monior. Assumpion: The Business monior survey alhough differen capures he same rends of he secor. 1985 1981 Backcas using he growh rae of R&D BERD. Assumpion: R&D Berd (mainly scienific) and R&D in human sciences grew a he same rae 1981 1970 Back cas using he growh rae of R&D as in he Annual Absrac of Saisics. 18 1 14.7 Implied marke secor GVA deflaor 0.3 1 0.28 Implied marke secor GVA deflaor (11) Toal 39.5 32.4 0.2 0.2 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 694

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 (12) Brand equiy (13) Adverising expendiure Toal spending on adverising as repored by Adverising Associaion, less expendiure on classified ads (14) Marke research Twice revenues of he marke and consumer research indusry as repored in ABI Economic compeencies 2004 1970 Adverising Associaion daa 14 0.6 6.9 Implied marke secor GVA deflaor 2004 1995 Twice he urnover of indusry 74.13 source ABI 1994 1992 Turnover of he secor from Service secor review 1991 1970 Backcas using Adverising Associaion growh rae of urnover. Assumpion: he urnover of he marke research indusry grew a same rae as he one of he adverising indusry 4.5 0.6 2.2 Implied marke secor GVA deflaor (15) Toal 18.5 1 9.1 (16) Firm-specific human capial NESS05, a similar survey of employer provided raining, adjused o consider privae secor expendiure and all UK 1970 2004 Backcas using rends in wage coss and he indusrial srucure of he workforce o exrapolae he resuls of he NESS05 survey 28.8 1 23.7 Implied marke secor GVA deflaor 0.6 0.6 0.4 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 695

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 TABLE 1 (coninued) Type of inangible invesmen 2004 Cross Secion Time Series Toal Spending bn, 2004 Proporion of Spending Considered as Invesmen % of Toal Inangible Invesmen, 2004 Deflaor (1) (2) (3) (4) (5) (6) (7) (8) Depreciaion Rae (17) Purchased Daa on revenues of managmen consuling indusry from Managemen Consuling Assocaion (MCA). To obain he privae secor expendiure we applied he privae secor/oal expendiure of he MCA o he grossed up oal of he indusry (sill provided by he MCA) (18) Own accoun Esimaed as 20% of value of execuive ime using ASHE daa on wages in execuive occupaions, excluding sofware occupaions 2004 1992 MCA daa for 2004 adjused o cover jus he privae secor backcased using growh rae of urnover of SIC 7414 excluding PR source: ABI from 1995 2004 and using Service Secor Review for 1992 94 1991 1985 Backcased using growh rae of urnover of caegory managemen consulan source Business Monior 1984 1979 Backcased using growh rae of he oal business services as published in Business Monior 1979 1970 Backcased using daa above using urnover of he secor Banking, finance, insurance business services, leasing from he Blue Book (afer consrucing a consisen ime series wih he Blue Book daa) 2004 1975 Managers earnings 2004 backcased using ASHE and NESPD (consrucing a consisen ime series) 1974 70 Backcas using highes average wage from STAN Organizaional srucure 7 0.8 4.6 Implied marke secor GVA deflaor 15.3 1 12.5 Implied marke secor GVA deflaor (19) Toal 22.3 17.1 (20) Toal 69.6 49.9 (21) GRAND TOTAL 130.8 100 0.4 0.4 Noes: Column (1) shows he ype of inangible invesmen. Column (2) indicaes he daa sources used for he cross secion esimaes of inangible spending in 2004. Column (3) shows he daa sources for he ime series esimae. Column (4) shows he expendiure for each asse for 2004. Column (5) indicaes he percenage of he expendiure assumed as invesmen. Column (6) shows for each inangible asse he percenage of oal inangible invesmen i represens. Column (7) indicaes he deflaor used and Column (8) he depreciaion rae. ONS is he Office for Naional Saisics. BERD is Business Enerprise Research and Developmen. ABI is Annual Business Inquiry. ASHE is Annual Survey Hours and Earnings. STAN is he srucural analysis daabase of he OECD. SIC is Sandard Indusrial Classificaion. NESS05 is he Naional Employer Skills Survey 2005. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 696

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 he firs six columns. The ype of inangibles, column 1, and he daa sources for he cross secion, column 2, are exensively described in Giorgio Marrano and Haskel (2006). 12 3.2. Cross Secion Resuls To give some idea of he scale of expendiures, column 4 ses ou expendiure on each asse for 2004. This is hen convered o invesmen using he fracion in column 5. Column 6 hen shows he fracions of oal inangible invesmen each row accouns for. The following poins are worh noing. Firs, around 50 percen of oal invesmen is on firm spending on repuaion, human and organizaional capial (economic compeencies). Abou 35 percen is on innovaive propery and 15 percen on compuerized informaion. Second, according o hese numbers, invesmen in R&D is jus one par of invesmen in knowledge asses. In fac, R&D invesmen is less han invesmen in sofware, for example. 13 Third, he bigges single figure is raining invesmen. 3.3. Time Series Resuls Figure 1 shows he ime series for nominal inangible invesmen for he aggregaed caegories as a share of adjused nominal MGVA. Figure 1b shows he ime series for he U.S. from CHS (2006) while Figure 1a reproduces he esimaes for he U.K. They are cumulaive chars so ha ha op line shows he share of oal inangible invesmen in inangible-adjused MGVA. 14 The lowes line shows he share of brand equiy and he line above ha shows he share of brand equiy plus he firm-specific resources. Thus he gap beween he lines is he share of each caegory of invesmen. A number of poins are worh making. Firs, he oal line shows he growing imporance of nominal inangible invesmen in he economy, rising from around 6 percen of MGVA in he 1970s o 13 percen in 2004 (6 percen o 15 percen of unadjused MGVA, which includes some sofware, i.e. as currenly measured in he Naional Accouns). Second, all invesmen ypes have risen, wih he excepion of brand equiy, which is more or less fla. The mos marked increases are for compuerized invesmen and firm-specific resources. These wo groups show herefore he bigges increases in he share of overall inangible invesmen. 3.4. Labor Share The labor share is calculaed as he raio beween compensaion of employees and he sum of labor compensaion and capial compensaion, he laer called operaing surplus in he U.K. Naional Accouns (in urn, for he whole 12 There are wo minor changes wih respec o Giorgio Marrano and Haskel (2006). All he daa are now consisen wih he 2006 Naional Accouns and in he asse new archiecural and engineering designs we include also wice he urnover of he SIC 74842 Specialiy designs aciviies (around 4bn in 2004). 13 Alhough he wo ypes of spending migh have quie differen poenial spillovers. 14 Tha is, he denominaor is GVA in he marke secor, adjused for he presence of inangible invesmen. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 697

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 15 (a) U.K. 12 Compuerized informaion Non-scienific R&D 9 Scienific R&D 6 3 Firm-specific resources Brand Equiy 0 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 (b) U.S. Figure 1. Inangible Invesmen (percenage of oupu) Noes: The figures show he ime series for inangible invesmen for he aggregaed caegories as a share of oupu. For he U.K. oupu is marke secor GVA adjused o include all inangibles. For he U.S. oupu is non-farm business oupu. The firs char (a) shows he ime series for he U.K. while he second (b) shows he ime series for he U.S. I is a cumulaive graph so ha ha op line shows he share of oal inangible invesmen in inangible-adjused marke secor GVA. The lowes line shows he share of brand equiy and he line above ha shows he share of brand equiy plus he firm-specific resources. Thus he gap beween he lines is he share of each caegory of invesmen. Brand Equiy includes adverising and marke research. Firm specific resources includes firm specific human capial and organizaional srucure. Scienific R&D includes scienific R&D and mineral exploraion. Non-scienific R&D includes copyrigh and licences coss, new produc developmen coss in he financial indusry, new archiecural and engineering design, and R&D in social science and humaniies. Compuerized informaion includes sofware. Source: U.K. daa our calculaions. U.S. daa CHS (2006). 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 698

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 economy, his adds up o nominal GDP, subjec o some minor ax/subsidy and saisical adjusmens). One problem in calculaing his is he reamen of he income of he self-employed, whose income, ermed mixed income, migh be considered a combinaion of labor and capial income. I is included, in he U.K. marke secor daa, wih operaing surplus. This booss he capial share and hus poenially booss he fracion of MGVA growh ha is capial deepening. 15 We decided o spli mixed income ino labor and capial income. One way of doing his is o use he Labor Force Survey o esimae he pay of he selfemployed by assuming hey earn he same wage as equivalen (in erms of age, skills, ec.) employed workers. This mehod produces an esimae of selfemployed labor income equal o 98 percen of mixed income. The oher mehod is o simply calculae he capial labor raio from operaing surplus afer excluding mixed income and hen apply his same raio o he mixed income componen. This mehod produces an esimae of self-employed labor income equal o on average around 65 percen of mixed income. We seled on he final opion and as a consequence, he labor share is higher han in he case in which he mixed income is lef in operaing surplus. Figure 2 shows he ime series for he labor share in he boh he U.K. and he U.S. excluding and including inangibles. Figure 2b shows he labor share from 1970 o 2003 for he U.S. and Figure 1b he series for he U.K. A number of poins are worh noing. Firs, he levels of he labor shares are similar, wih he 1970 level being abou 68 percen excluding sofware and inangibles and 63 percen including hem (he U.S. figures are 71 percen and 66 percen). Second, he overall rend in boh counries is fla when excluding inangibles and falling when including hem. If anyhing he rend is smooher in he U.S., reflecing he U.K. change from he peak in he mid 1970s (well known union push) and he rend downwards o he early 1980s. 3.5. Invesmen Share Figure 3 shows nominal invesmen as a percenage of MGVA for he U.K. and U.S. including hree cases: radiional Naional Accouns excluding sofware; including sofware (his is no shown in he U.S. graphs); and including all inangibles (where in he U.K. daa he MGVA denominaor excludes sofware, includes sofware, and includes all inangibles respecively). There are wo major findings. Firs, in boh counries, wihou inangibles, he nominal invesmen share is fla or a lile bi decreasing and is a similar amoun (around 15 percen). Second, once we include inangibles i increases in levels, o around 25 percen by he end of he period; he rend is upwards. 4. Growh Accouning To implemen he growh accouning se ou in Secion 2, we proceed as follows. Firs, we measure labor inpu L as employee hours. Second, we 15 Indeed calculaed on his basis he U.K. capial share is abou 10 percenage poins above he non-farm business U.S. capial share. None of he inernaional comparisons of labor shares ha we could find gave his kind of difference. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 699

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 80 (a) U.K. 75 70 65 60 55 Including inangibles Including sofware Excluding sofware 50 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 (b) U.S. Figure 2. Labor Shares (percenage of oupu) Noes: The figures show he ime series for he labor share in he boh he U.K. and he U.S. excluding and including inangibles. The firs char (a) shows he labor share from 1970 o 2004 for he U.K., while he second char (b) shows he labor share from 1950 o 2003 for he U.S. For he U.K., oupu is marke secor GVA adjused o be consisen wih he amoun of inangibles included. For he U.S., oupu is non-farm business oupu. Source: U.K. daa our calculaions. U.S. daa CHS (2006). express MGVA and capial in per employee hour erms. Third, in pracice he qualiy of labor likely varies and so we disinguish beween employee hours, L and qualiy adjused employee hours, L QA. Thus our growh accouning expressions are: L QA K (8) ( a) ΔlnTFP = Δln( Q L) s Δln( L L) s Δln( K L). L QA K R ( b) Δ lntfp = Δln Q L s Δln L L s Δln K L s Δln R L ( ) ( ) ( ) () ( ) 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 700

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 28 (a) U.K. 26 24 22 Including inangibles 20 18 Including sofware 16 14 12 Excluding sofware 10 8 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 (b) U.S. Figure 3. Nominal Invesmen Shares (percenage of oupu) Noes: The figures show invesmen shares as a percenage of marke secor GVA for he U.K. and non-farm business oupu for he U.S. in hree cases: radiional Naional Accouns excluding sofware, including sofware (his is no shown in he U.S. graph), and including all inangibles. In each case he marke secor GVA for he U.K. is he appropriae version for he amoun of inangibles included. Source: U.K. daa our calculaions. U.S. daa CHS (2006). A number of poins are worh noing regarding equaion (8). Firs, he shares are averages of shares over which he ime difference is aken, so ha (8) is a Tornquis index number. Second, he share of capial is defined as one minus he share of labor. This is accurae if here are consan reurns o scale a he overall economy level, bu is clearly an area where beer measuremen would be helpful. Third, since here are in pracice many capial asses (for angibles, plan, buildings, vehicles, and compuer hardware; for inangibles, sofware, R&D, ec.) he DlnK and DlnR erms have o be consruced o incorporae hese many ypes. This is done following Oulon and Srinivasan (2003a), who in urn follow Jorgenson and 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 701

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 Griliches (1967), by noing ha he heoreically correc capial measure in a producion funcion is he services ha capial provides ino oupu. In urn he services for each ype of capial can be measured by he renal paymens ha a profi-maximizing firm would pay were i rening is capial. Since in pracice firms rarely do his bu buy he capial asse for a price p A and hen use i over is lifeime, he marke-clearing renal paymen for an asse B (where B can be angible or inangible asses), p B, can be derived as (9) [ ( )] = B A A A A pi = Ti ri pi, 1+ δi pi pi pi, 1, B K, R where T is a ax adjusmen and r is he rae of reurn on he asse. 16 This equaion holds for each ype of capial i. The relaion beween his and he DlnK and DlnR erms in (8) can be derived as follows. Firs, he overall level of profi in he economy, P, is, by definiion, he overall paymens o capial, which is he sum of all renal paymens o each capial ype. This can be wrien as (10) Π n K m R p i i K = 1 i i= n+ 1 i = + p R i where here are n angible asses and n + 1om inangible asses. Second, he overall volume index of capial services can be shown o be a share-weighed average of all he asse-specific DlnK and DlnR erms: (11) K Δln K p K Π Δln K n i= 1 i, i i R ΔlnR p R Π ΔlnR = ( ) m = ( i, ) i= n+ 1 i i where he shares are he flow of renal paymen for each asse as a share of oal renal paymens (P). 17 As an empirical maer, we have o ake a number of seps. Firs, we do no have informaion on ime-varying depreciaion raes and so se hem consan over ime. Second, we do no have informaion on asse-specific raes of reurn, r i. In a compeiive marke, r i will equalize across asses. If we assume his hen we have wo equaions (9) and (10) in wo unknowns, namely r and p K which we can solve for. The economic inuiion of his is ha since we know he overall paymen o capial, P in he economy, from Naional Accouns, we can solve for he unobserved asse-specific renal prices ha would ensure ha all paymens o capial asses added up o P. Third, in line wih he Tornquis mehod above, he weighs in (11) are he ime-averaged weighs over which he difference is aken. To summarize, we herefore implemen growh accouning in he following seps. (i) Collec a ime series of nominal invesmen in inangible and angible asses, deflae o ge real invesmen series, and build a real capial sock using he perpeual invenory mehod (see equaions (1) and (4)). (ii) Recalculae MGVA o include inangibles (see equaion (6)). (iii) Adjus he operaing surplus P for MGVA (see equaion (6)). (iv) Build a Hall/Jorgenson capial services measures of 16 The derivaion of equaion (10) is se ou in Oulon and Srinivasan (2003a). 17 By conras he wealh sock, which is ofen presened as a measure of capial, is he shareweighed sum of capial socks, where he shares are he asse prices. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 702

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 TABLE 2 Tangible Asses Asse Type Time Series Deflaor Depreciaion Rae (1) (2) (3) (4) (1) Compuer hardware Naional Accouns consisen invesmen series (see Wallis, 2009 for deails) (2) Buildings Naional Accouns invesmen series. Consisen wih 2006 Blue Book. Ne sock esimaes based on Wallis (2009) (3) Plan and machinery Naional Accouns invesmen series. Consisen wih 2006 Blue Book. Ne sock esimaes based on Wallis (2009) (4) Vehicles Naional Accouns invesmen series. Consisen wih 2006 Blue Book. Ne sock esimaes based on Wallis (2009) ONS delflaor (for 1983 70 backcased using growh raes of NIESR s hardware deflaor) Naional Accouns capial sock deflaors Naional Accouns capial sock deflaors Naional Accouns capial sock deflaors 0.4 (Naional Accouns) 0.025 (BEA) 0.13 (BEA) 0.25 (BEA) Noes: Column (1) shows he asse ype. Column (2) indicaes he daa sources for he ime series. Column (3) shows he deflaor used and column (4) he depreciaion rae. BEA is Bureau of Economic Analysis. NIESR is Naional Insiue of Economic and Social Research. all capial inpus, ensuring he asse renal paymens are consisen wih he adjused operaing surplus (see equaions (9) o (11)). (v) Build a qualiy adjused labor index o measure L QA (see equaion (8)). (vi) Underake growh accouning. The nex secions describe how o do his. 4.1. Time Series of Nominal Invesmen Invesmen in inangible asses is se ou above. For angible asses, we use he daa based on Wallis (2009); see Table 2. Briefly, he daase consiss of a long back-hisory of consan price invesmen daa classified by SIC92 indusries. The asse breakdown of he invesmen series is: buildings, plan and machinery, and vehicles. In order o rea compuers as a separae asse, compuer invesmen is separaed from invesmen in plan and machinery and he associaed price deflaors adjused o accoun for his. The daa are hen aggregaed o marke secor levels. 4.2. Deflaion and Real Invesmen Series The choice of deflaors, as CHS discuss, is a difficul one. One possibiliy is o develop a price index for he paricular inangible according o he coss incurred in developing i, so ha, for example, if mos of he coss of R&D is paymens o scieniss, hen he deflaor migh be he wage of scieniss. As CHS show however, his implicily assumes ha scieniss have no increase in produciviy in he R&D process. A second possibiliy is o use he oupu deflaor. This is someimes 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 703

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 jusified in sudies of, for example, R&D, where a physical uni has lile meaning and so i is fel bes o deflae by he price of he good which presumably embodies he knowledge ha he R&D is generaing. Triple and Bosworh (2004) offer a similar jusificaion for managemen consulans. Our deflaors are se ou in more deail in Table 1. For compuer hardware, we use daa from he Office for Naional Saisics (ONS), he Bank of England and he Naional Insiue of Economic and Social Research (NIESR). They are close o each oher only in some years. ONS sops in 1984 and so we back-cas using NIESR daa. We have explored U.S. deflaors and he resuls are robus o his change. Sofware deflaors are aken as follows. For own accoun we use wages of he relevan occupaions and hen a 2.5 percen produciviy adjusmen. Whils his has he problem similar o he R&D deflaor above, i is used by he ONS and so has he benefi of being consisen wih heir pracices (which is useful in our conex since a leas par of he sofware is incorporaed ino he Naional Accouns; i is also consisen wih he U.S. reamen of sofware). For purchased sofware we use he ONS purchased sofware deflaor. For all oher inangible asses we use an implici deflaor calculaed from our nominal marke secor oupu and real marke secor oupu series. 18 Turning o he remaining angible asses we use deflaors for plan, vehicles, non-it machinery, and buildings consisen wih he U.K. Naional Accouns (see Wallis, 2009). 4.3. Capial Sock A consan depreciaion rae assumes geomeric depreciaion, he accuracy of which is of course open o quesion as well as requiring one o sele on a depreciaion rae. Given he doubs and uncerainy over his, we sele here on applying convenional assumpions abou angible asses o he accumulaion of inangible asses. Table 1 ses ou our assumed raes. For inangible asses hese are based on CHS (2006) assumpions. For angible capial we use exising Naional Accouns deprecaion raes. Our sensiiviy analysis included varying he assumpions on he inangible asse side. 4.4. Adjusing Operaing Surplus of Marke Secor The ONS publishes marke secor operaing surplus series back o 1992. We back-cas he series o 1970 using he gross operaing surplus growh raes for he whole economy. As dwellings are no modeled as par of he producive capial sock, because hey do no form par of he inpu ino producion, he par of operaing surplus aribuable o dwellings is subraced. This par of operaing surplus is measured by owner-occupied impued rens and he depreciaion of he sock of dwellings. To adjus he operaing surplus for he inangibles we simply add nominal inangible invesmen. We build hree versions of marke secor gross operaing surplus. The firs version excludes sofware invesmen already presen in he Naional Accouns. The second includes 18 The oupu series are adjused for he inclusion of inangibles. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 704

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 his invesmen and adds also he revision o Naional Accouns sofware esimaes presened in Chamberlin and Chesson (2006). The hird includes all inangibles. Concerning labor compensaion, he ONS publishes marke secor labor compensaion series back o 1992. We back-cas he series using he growh rae of he OECD esimaed wage bill. 4.5. Recalculaing MGVA o Include Inangibles The ONS publishes a ime series back o 1992 of MGVA a curren prices. We wish o remove privae residenial propery from our daa. So o be consisen wih he reamen of operaing surplus above, we subrac acual and impued renal from housing from he MGVA daa. We back-cas he series using he growh rae of he sum of he marke secor labor compensaion and operaing surplus. To adjus for he inangibles we simply hen add he nominal invesmen in inangibles (noe no spending bu invesmen) o nominal MGVA, ensuring ha we do no double coun any inangibles already included (such as some sofware and mineral exploraion). Regarding he real marke secor growh rae, he ONS publishes ime series ha go back a leas o 1970. We adjused he real growh rae for inangibles devising an index of changes in real adjused MGVA as a share-weighed change of real MGVA and real inangible invesmen, wih he weighs being he share of each expendiure caegory in overall GVA. As for he gross operaing surplus we build hree versions of MGVA. The firs version excludes sofware invesmen already presen in he Naional Accouns. The second includes his invesmen and adds he planned revision o he sofware esimaes. The hird includes all inangibles. 4.6. Capial Services To do his we use mehod described above (equaions (9), (10) and (11)). We smooh he rae of reurn and he capial gain erm by aking a hree-year moving average. All raes of reurn are posiive, bu for some years in he middle 1970s he building renal raes were negaive. We se hem equal o he neares posiive rae. 4.7. Qualiy Adjused Labor Index We use here he Bank of England index ha adjuss hours for educaion, gender and age (see Bell e al., 2005), kindly provided o us by Nick Oulon and Sally Srinivasan. I migh be fel ha including boh raining spending by firms and he labor qualiy/composiion adjusmen double couns he conribuion of skills. Concepually, we wish o separae ou he par of skills ha are invesed in by he firm (since ha would be par of marke secor inangible invesmen) and ha par invesed in by he sae. Under Becker-ype assumpions, invesmen by firms, which is only in firm-specific raining, will no raise worker wages since workers canno use hese skills ouside he firm. General raining will raise worker wages. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 705

Review of Income and Wealh, Series 55, Number 3, Sepember 2009 Now, he adjusmens in he labor qualiy/composiion erm muliply worker ypes by heir wages relaive o a base worker ype. If firms are only funding firm-specific raining, hen since marke wages are unaffeced, such marke wage adjusmens will no double coun labor qualiy. One problem wih his approach is if he effec of experience in he qualiy adjusmen is in par due o he firm-specific raining ha older workers have received earlier in heir career. If so, his would mean double couning for a leas par of he experience effec. 5. Growh Accouning Resuls 5.1. Overall In he resuls ha follow we use he following convenions. Growh in capial services and labor qualiy are Tornquis indices as is growh in MGVA. The averages repored are 100 imes he arihmeic averages of year-on-year Tornquis growh raes (e.g. 2000 04 is he average of 2000 01, 2001 02, 2002 03, and 2003 04). TFP growh is he residual and he capial and labor shares add o 1. Our growh accouning decomposiions sar in 1979 wih our inangible capial daase equal o zero in 1970. However, due o he period we are mos ineresed in being he 1990s, and our daa being of beer qualiy from 1990 onwards, our analysis focus on 1990 onwards. This also allows us o ignore any iniial condiions problems in associaion wih he inangible capial sock. 19 We underook wo main checks on he daa. Firs, he ONS Blue Book 2006 does no growh accouning bu does include some sofware in oupu. Thus we generaed MGVA daa excluding all sofware, including jus sofware, and including all inangibles. We checked our daa ha included sofware agains he ONS daa and found he growh raes very close. 20 Second, Oulon and Srinivasan have underaken a major indusry-level sudy ha includes sofware in boh heir oupu daa and heir capial services daa (Oulon and Srinivasan, 2005). These resuls were up o 2000 and were consisen wih he 2002 ONS Blue Book. The 2002 Blue Book daa had limied coverage of sofware, so a major conribuion of Oulon and Srinivasan (2005) was o add in sofware o boh oupu and capial services. In recen unpublished worked, hey use daa o 2003, consisen wih he 2005 ONS Blue Book, again incorporaing sofware. A change here is ha ONS have revised heir employee-hours daa o be consisen wih he 2001 populaion census; Oulon and Srinivasan have revised heir daa accordingly. We use heir labor hours and qualiy measure, ha hey kindly supplied us. This allows us o make a beer comparison of our baseline resuls, wihou inangibles, wih heirs. 5.2. Growh Accouning Resuls, 1990 2004 Table 3 shows he growh accouning resuls for 1990 2004. We look a his period o compare he resuls wih Oulon and Srinivasan (2005) and o explore a 19 The angible capial sock is based on a very long run of invesmen daa, back o he 1800s in some insances, so here are no iniial condiions problems o deal wih. 20 The Blue Book 2006 includes somewha less sofware han we do. For example, for 2004, our daa is abou 21bn while in he Blue Book 2006 is abou 11bn. 2009 The Auhors Journal compilaion Inernaional Associaion for Research in Income and Wealh 2009 706