The Industry Origins of the US-Japan Productivity Gap

Transcription

1 KEO Disussion Paper No. 105 The Industry Origins of the US-Japan Produtivity Gap Dale W. Jorgenson (Harvard University) Koji Nomura (Keio University) February 3, 2007 Revised in June 18, 2007 Abstrat This paper presents a omparison of total fator produtivity (TFP) levels between the U.S. and Japan for the period and alloates the gap to individual industries. We arefully distinguish the various onepts of purhasing power parity (PPP) and measure them within the framework of a U.S.- Japan bilateral input-output table. We also measure industry-level PPPs for apital, labor, energy, and materials inputs and output for 42 industries ommon to the U.S. and Japan, based on detailed estimates for 164 ommodities, 33 assets, inluding land and inventories, and 1596 labor ategories. The U.S.-Japan produtivity gap shrank during three deades of rapid Japanese eonomi growth, The Japanese manufaturing setor ahieved parity with its U.S. ounterpart by the end of the period. With the ollapse of the Japanese eonomi bubble at the end of the 1980s, the U.S.-Japan produtivity gap reversed ourse and expanded to 79.5 perent by This an be attributed to rapid produtivity growth in the IT-produing industries in the U.S. during the late 1990s and the sharp aeleration of produtivity growth in the IT-using industries in the U.S. during Wholesale and Retail Trade emerged as the largest ontributor to this gap, aounting for 25.1 perent of the lower TFP of the Japanese eonomy. JEL Codes: C82, D24, E23. Keywords: Purhasing Power Parity, Investment, Produtivity, Growth.

2 I. Introdution The growth of the Japanese eonomy from 1960 to 1990 far outstripped that of other industrialized eonomies and inspired the admiration of the world! Japanese GDP grew at 10 perent per year during , nearly uadrupling the size of the Japanese eonomy. Starting from only 25.5 perent of U.S. GDP per apita in 1960, the Japanese eonomy made a major leap to 56.7 perent of the U.S. level in Rapid growth of the Japanese GDP at 4.5 perent per year through 1990 nearly doubled the size of the Japanese eonomy again, bringing Japanese GDP per apita to 78.3 perent of the U.S. The rapid losing of 3.7 perent per year in the U.S.-Japan gap in GDP per apita during was ahieved by 2.1 perent annual growth in relative input per apita and 1.6 perent annual redution in the TFP gap. Japanese input per apita was a very respetable 48.9 perent of the U.S. level in 1960, but more than doubled from , reahing 78.4 perent of the U.S. level in Japanese input per apita ontinued to inrease rapidly, rising to 91.2 perent of the U.S. level in The initial U.S.-Japan TFP gap of 52.4 perent in 1960 had losed to only 86.1 perent in After the Plaza Aord of 1985 and the bursting of the Japanese bubble eonomy at the end of the 1980s, the gap in GDP per apita between Japan and the U.S. reversed ourse, opening by 0.9 perent annually through Japanese GDP per apita fell to 71.2 perent of the U.S. level in 2000 and 69.5 perent in 2004, below the relative levels of the mid-1980s. Growth of input per apita in Japan slowed onsiderably, falling to 87.2 perent of the U.S. level in 2000 and remaining almost unhanged at 87.6 perent in Japanese total fator produtivity, relative to the U.S., fell from 86.1 in 1990 to 81.7 in 2000 and 79.5 in 2004, refleting the sharp aeleration in U.S. TFP growth after 1995 and the more modest reovery of Japanese TFP growth. The onvergene of Japanese eonomy to U.S. levels of output and TFP has been analyzed in a number of earlier studies Jorgenson, Kuroda, and Nishimizu (1987), Jorgenson and Kuroda (1990), van Ark and Pilat (1993), Kuroda and Nomura (1999), Nomura (2004, Ch.4), and Cameron (2005). As desribed in this paper, Japanese TFP for the manufaturing setor was euivalent to its U.S. ounterpart in This paper also investigates the industry origins of the produtivity gap between the U.S. and Japan during the period of onvergene, , and the period of divergene, One of the most important ontributions of this paper is to measure detailed purhasing power parities (PPP) between the U.S. and Japanese industries. The various onepts of PPP are arefully distinguished and implemented within the framework of a U.S.-Japan bilateral input-output table. 1 We measure the industry-level PPPs for apital, labor, energy, and materials inputs and output for 42 1 An alternative approah to onstruting PPPs at the industry level is presented by Inklaar and Timmer (2007), Eonomi Systems Researh, this issue. 1

3 industries ommon to the U.S. and Japan. These estimates are based on more detailed estimates of a ommon lassifiation of 164 ommodities, 33 assets, inluding land and inventories, and 1596 labor ategories. In order to trae the U.S.-Japan produtivity gap to its origins at the industry level we use two data bases for the U.S. and Japan. These have a losely omparable struture and employ similar national aounting onepts. The U.S. data base is onstruted by Jorgenson, Ho, Samuels and Stiroh (2007) and extends the database of Jorgenson, Ho, and Stiroh (2005) bakward to 1960 and forward to We also extend the Japanese data base presented by Jorgenson and Nomura (2005) to Our ommon system of industrial lassifiation for the U.S. and Japan distinguishes the IT-produing setors Computers, Communiations Euipment, and Eletroni Components from other industries. Setion II we present our methodology and data soures for measuring PPPs. Using PPPs for inputs and output of eah industry, we ompare relative levels of output, inputs, and produtivity in Setion III. We aggregate these results to obtain relative levels of GDP, apital and labor inputs, and produtivity for the U.S. and Japanese eonomies. Finally, we alloate the produtivity gap between the U.S. and Japan by industry over the period We find that the produtivity gap for manufaturing had disappeared by 1990 but re-emerged after For non-manufaturing the gap was sizable throughout the period. Setion IV onludes the paper. The detailed methodology for our U.S.- Japan produtivity omparisons is presented in the Methodologial Appendix. II. Measuring Purhasing Power Parities We estimate the purhasing power parities (PPPs) for apital, labor, energy, and materials (KLEM) inputs by industry. In setion i we introdue the ommon U.S.-Japan industry lassifiation used in this paper. We introdue our measurement of the detailed PPPs for 164 ommodities, based on a variety of onepts and a multitude of data soures in Setion ii(a). Setion ii(b) desribes the industrylevel PPP for output and intermediate inputs for energy and materials. The industry-level PPPs for apital and labor inputs are presented in setions iii and iv, respetively. i. Common Industry Classifiation Our data for produtivity in Japan are based on Jorgenson and Nomura (2005). These data draw on the Keio Eonomi Observatory (KEO) database maintained at Keio University in Tokyo. We employ the U.S. data presented by Jorgenson, Ho, Samuels, and Stiroh (2007). These extend the database desribed in detail by Jorgenson, Ho, and Stiroh (2005). The two databases have a losely omparable struture and employ similar national aounting onepts. The ost of apital is used in measuring apital inputs and imputing the value for non-market prodution of apital servies by household and government setors. 2

4 Our first step is to establish a ommon lassifiation for U.S. and Japanese industries. In our previous U.S.-Japan omparisons Jorgenson, Kuroda, and Nishimizu (1987), Jorgenson and Kuroda (1990), Kuroda and Nomura (1999), and Nomura (2004, Ch.4) we have employed a ommon industry lassifiation with about 30 industries. For this study we have developed a new 42-industry lassifiation that enables us to identify the IT-produing setors Computers, Communiations Euipment, and Eletroni Components. 2 We uantify the impat of IT prodution, as well as investment in IT euipment and software, in both the U.S. and Japan. Common lassifiations for apital and labor inputs are desribed below. The publi setor is a speial hallenge in reating a ommon industry lassifiation. This setor should be separated from private business, although it is diffiult to maintain this distintion in pratie. In both the U.S. and Japan publi setor ativities are inluded in private industries with similar tehnologial harateristis. For example, government-run power authorities are lassified as Eletri Utilities in both eonomies. We set the produtivity gap between the U.S. and Japan eual to zero for the non-market prodution of apital servies by households and the publi setor. ii. Purhasing Power Parities for Output and Intermediate Inputs (a) Elementary Level Purhasing Power Parities There are two approahes to defining PPPs. Using prodution-side data for domestially produed goods, the PPPs in produer s pries are ratios of average unit pries, eah defined as the monetary value over the physial uantity. This approah is espeially easy to implement in setors with outputs defined in physial units, for example, Eletriity. On the other hand, PPPs an be estimated from demand-side data by eliminating the wedge between produer s pries and purhaser s pries due to trade and transportation margins and taxes, taking import pries into aount. We employ a hybrid approah developed by Masahiro Kuroda, Kazushige Shimpo, and ourselves. In this paper we have revised the elementary-level estimates of Nomura and Miyagawa (1999). They provide PPPs for 164 ommodity groups, estimated from both prodution-side and demand-side prie data within the framework of a U.S.-Japan bilateral input-output table. 3 Sine a number of the key papers 2 Our new industry lassifiation eliminates important inonsistenies in our earlier studies. For example, Jorgenson and Kuroda (1990) and Kuroda and Nomura (1999) inluded Computers in Mahinery in the U.S. and Eletri Mahinery in Japan. The two mahinery industries were ompared without reognizing this differene. Nomura (2004) ombined the two mahinery industries into a single industry and avoided this inonsisteny. We provide a more detailed lassifiation that treats Computers as a separate industry. 3 Nomura and Miyagawa (1999) onsiderably revised the methodology and estimates in the preeding studies of the Japan Industrial Poliy Researh Institute (1994) Studies on New Estimates of PPP and (1996) Studies on the International Input-Output Table, both implemented as joint projets of the METI and Keio University and published in Japanese. 3

5 desribing the hybrid approah have been published only in Japanese, we first outline the methodology and then disuss the data soures for our PPPs. Our measurement of elementary level PPPs is based on the 1990 U.S.-Japan Bilateral Input- Output (I-O) Table, published by Ministry of Eonomy, Trade and Industry (METI) in We postulate a prie model desribing the relationships among produer s pries and purhaser s pries for domestially produed and imported goods. The U.S.-Japan trade struture in the 1990 Bilateral I-O Table maintains onsistent prie differenes between the two eonomies, onsidering differenes in freight and insurane rates, duty tax rates, wholesale and retail trade margins, transportation osts (railway, road, water, air, and others), and import shares of eah ommodity in the U.S. and Japan. 4 Using demand-side data for purhaser s prie PPPs for final demands, we estimate the produer s prie PPPs for domestially produed goods, based on our prie model and related parameters. Also, using produtionside data for PPPs in produers pries, we estimate PPPs for omposites of domestially produed and imported goods by inverting the prie model. One of the diffiulties in estimating PPPs in produer s pries from demand-side data is to define PPPs for imported goods. These are reuired to separate PPPs for domestially produed ommodities from omposite PPPs for ommodity groups that inlude imports. 5 Using the U.S.-Japan bilateral I-O Table, goods purhased in Japan an be separated into domestially produed goods, goods imported from the U.S., and goods imported from the Rest of the World (ROW). The purhaser s pries in Japan for goods imported from the U.S. an be linked to pries of domestially produed goods in the U.S., taking into onsideration the osts of freight and insurane reuired for shipment from the U.S. to Japan and duties levied by Japanese ustoms. Similarly, import pries in the U.S. an be linked to domesti output pries in Japan. The most omprehensive demand-side data are the Eurostat-OECD Purhasing Power Parities. Prie differenes are defined as the gaps between purhaser s pries, inluding wholesale and retail margins for final demands. These pries reflet not only the pries of domestially produed goods but also imported goods. We assume that pries of imports into Japan from the U.S. are sums of produer s pries in the U.S., freight and insurane osts, ustoms duties, margins for wholesale and retail trade, and 4 The data for freight and insurane rates and duty tax rates by ommodity are available as supplementary tables of the 1990 US-Japan Bilateral I-O Table. We used the 1992 U.S. Use Table from the Bureau of Eonomi Analysis (BEA) and the 1990 Japan Benhmark I-O Table (Ministry of Internal Affairs and Communiations, MIC) for rates of wholesale and retail margins and transportation osts by ommodity in eah of final demand and intermediate demand. These rates are aggregated in both ountries to orrespond to the 164 ommodities. 5 Trade statistis may provide rough estimates for defining PPPs for imported goods as ratios of nominal values to physial uantities. We use the trade statistis only for energy inputs like rude oil, oke, and LNG; we employ the trade statistis for both eonomies and Energy Pries and Taxes published by IEA. 4

6 transportation osts. We also assume that the import pries from the ROW are same as the omposite pries of domesti goods and goods imported from the U.S. Given these assumptions, we determine the two relative pries for domestially produed goods and omposite goods simultaneously within the framework of the U.S.-Japan bilateral I-O Table. Another diffiulty in estimation of PPPs from demand-side data is the absene of prie omparisons for goods not purhased as final demands. For example, the prie for an intermediate good like semiondutors plays a signifiant role in our U.S.-Japan produtivity omparisons. In order to supply the missing information, METI has arried out a Survey on Disparities between Domesti and Foreign Pries of Industrial Intermediate Inputs. By ontrast with the Eurostat-OECD PPPs, this survey targets only intermediate goods. Prie differenes are defined as purhaser s pries, inluding the differene in wholesale and (sometimes) retail margins for intermediate goods in eah eonomy. Using these data, the PPPs for domestially produed goods an be estimated on the basis of our prie model. There are relatively rih data soures for U.S.-Japan prie omparisons. In addition to the two large-sale investigations by Eurostat-OECD and METI, a number of surveys have been implemented by Japanese ministries and agenies in the 1990s. This reflets the inreasing poliy interest in U.S.-Japan prie differenes, resulting from strengthening the Japanese yen, relative to the U.S. dollar, after the Plaza Aord of We have used PPP data for onsumer goods (METI, Cabinet Offie), transportation and related servies (Ministry of Land, Infrastruture and Transport, MLIT), residential buildings and infrastruture (MLIT), mediine and medial euipment (Ministry of Health, Labour and Welfare, MHLW), barber servies, beauty treatments, leaning, and movies (MHLW), living expenses (Cabinet Offie), food and restaurants (Ministry of Agriulture, Forestry and Fisheries, MAFF), wood produts (Forestry Ageny), ellular phones and ar phones (Ministry for Internal Affairs and Communiations, MIC), whiskey (National Tax Ageny). These data are estimated for different years and different stages of demand. We have reoniled these data within a bilateral I-O framework, taking into aount the differenes in timing of the surveys. For omputer hardware and software we have used new data from the IDC survey implemented during In addition, we have investigated U.S.-Japan prie differenes for mathed models of 6 In this sample survey pries are distinguished by vendor, brand, and proessor. The results are reported in New Media Development Assoiation (2002) Survey of Prie Differenes between the U.S. and Japan for Purhasing IT Goods and Servies. Using the sample pries in this survey, the geometri average of the U.S.-Japan relative pries are 0.84 (2001.Q3) and 0.97 (2001.Q4) for desk-top PC s, 1.00 (2001.Q3) and 1.15 (2001.Q4) for lap-top PC s, 1.23 (1999) for servers, and 1.00 (2002) for database software. 5

7 representative prepakaged software and personal omputers that an be purhased on the internet. 7 Based on sample pries for omputer hardware and software, the Japanese prie is almost the same as the U.S. prie; however, the Japanese prie is somewhat higher than the U.S. prie for servers. We have onverted these pries to base-year PPPs for 1990, using the prie indexes for eah omponent from our databases for the U.S. and Japan, desribed in Jorgenson and Nomura (2005). (b) Industry-Level Purhasing Power Parities After determining the elementary level PPPs, we have estimated five PPPs for eah of 164 ommodities: (1) a produer s prie PPP for domestially produed goods, exluding net indiret taxes, (2) a produer s prie PPP for omposite goods sold to households, (3) a produer s prie PPP for omposite goods sold to industry, (4) a purhaser s prie PPP for omposite goods sold to households, and (5) a purhaser s prie PPP for omposite goods sold to industry. 8 For the purpose of bilateral omparisons of produtivity, the PPP estimates (1), (3), and (5) are used as the elementary-level PPPs for output, intermediate inputs, and the auisition of investment goods, respetively. We aggregate the 164 elementary level PPPs into the 42-industry PPPs for output, using the translog index in Euation (15) of the Methodologial Appendix. The weights are the average shares of eah industry s output in the two eonomies in the 1990 U.S.-Japan Bilateral I-O Table. Similarly, industry-level PPPs for intermediate inputs are translog indexes, using the average shares as weights. 9 The relative pries for non-market prodution in the government and household setors are set eual to unity in order to make the U.S.-Japan produtivity gap eual to zero in these setors. iii. Purhasing Power Parities for Capital Inputs (a) Purhasing Power Parities for Investment In the Methodologial Appendix, we define elementary-level pries for apital inputs. The unit prie of apital input of asset i in eonomy, K Q, ij is defined in Euations (4) and (5) in the Appendix. We summarize the formulation for the prie of apital input as: (1) Pˆ i = P, K, A, f ˆ i i 7 Based on our investigation on sample pries for Mirosoft Windows XP, Exel, Word, PowerPoint, and Offie and Adobe Arobat, the geometri average relative prie is 1.09 in November If alternative data for prie disrepanies are available for a ommodity, regardless of the stage of demand, we finally have to hoose one PPP. We follow Nomura and Miyagawa (1999) in hoosing the PPP that provides the best approximation to prodution ativities in the U.S. and Japan by omparing aggregate measures of input oeffiients evaluated at onstant pries. 6

8 where ˆ represents the unit ost for auisition of one dollar s worth of assets. The oeffiient A P, i f i is an annualization fator that transforms the ost of auisition into the prie of apital servies. Defining the PPP for the ost of auisition as desribed in ii(b), the PPP for apital input is defined as: (2) f =. J K i A PPP i PPP U i fi The key to measuring the PPP for apital input is the relative value of annualization fator and the PPP for the auisition of assets. Our first step in measuring PPPs for apital inputs is to onstrut a ommon asset lassifiation for the U.S. and Japan. We define a 33-asset lassifiation, onsisting of 29 tangible assets, two intangible assets (mineral exploration and software), inventories, and land. To measure PPPs for the auisition of eah asset, we onstrut translog indexes of the purhaser s prie PPPs for the omposite goods by industry. These indexes are based on our estimates for elementary level PPPs for the 164 ommodities desribed above. The PPPs for auisition of inventories are assumed to be the average of PPPs for auisition of tangible assets, exept for buildings and onstrution. The differene in land pries between the U.S. and Japan has a substantial impat on the PPPs for apital inputs, as pointed out by Nomura (2004, Ch.3). Nomura indiates that Japan s auisition prie of land for ommerial and industrial uses was 9.1 times higher than that in the U.S. in The prie for apital auisition in Japan is 2.9 times higher than that in the U.S. in 1990 if we inlude land in apital input, but only 24 perent higher in Japan if land is exluded. We employ the higher estimates, inluding the PPP for land. It has been negleted in our previous measurement of TFP gap between the U.S. and Japan, although the apital input of land has been ounted in TFP measures in both ountries. (b) Purhasing Power Parities for Capital Input The final step in measuring PPPs for apital inputs is to determine the relative value of the annualization fators between the U.S. and Japan for eah asset and eah industry. A novel feature of our data sets for the U.S. and Japan is that the annualization fators are measured on the basis of omparable formulations of the prie of apital input, assuming asset-speifi revaluations for all assets and endogenous rates of return for eah industry. Tax onsiderations are a key omponent of the pries of 9 In our omparison, we treat all inputs of energy purhased by the energy onversion setors Petroleum Refining, Eletriity, and Gas Supply as materials inputs, not energy inputs. 7

9 apital inputs. 10 The measurement of apital servies is desribed by Jorgenson, Ho, and Stiroh (2005, Ch.5) for the U.S. and Nomura (2004, Ch.3) and Jorgenson and Nomura (2005) for Japan. The annualization fators are estimated for 59 assets in 36 industries in the U.S. and 103 assets in 47 industries in Japan. The estimates are aggregated into measures for the 33-asset U.S.-Japan ommon asset lassifiation in eah industry. Inluding land as a apital input, the aggregate PPP for auisition of apital goods is 2.9 in 1990, but the aggregate PPP for apital input is only 1.6, refleting lower annualization fators in Japan. iv. Purhasing Power Parities for Labor Inputs We assume that hours worked an be used to approximate one dollar s worth of labor input for eah ategory of labor at the elementary level. Thus, L i and Lˆ i in Euation (4) in the Methodologial Appendix represent labor input and hours worked, respetively. Also, L P, i and ˆ in Euation (5) are L P, i the prie indexes for labor input and average hourly labor ompensation, respetively. Labor uality L Q, ij is given by base-year average hourly labor ompensation. To define PPPs for labor inputs, we follow Nomura and Samuels (2003), revising the estimates to onform to our ommon U.S.-Japan industry lassifiation. 11 For both U.S. and Japanese data sets the labor inputs are ross-lassified by sex, age, eduation, lass of worker, and industry. The U.S.-Japan ommon labor lassifiation system allows us to ompare wages of similar workers. After lassifying the workers by sex, we split the workers by the other ategories industry, age, lass of worker, and eduation. The U.S. data set has eight age lassifiations for workers and Japan has eleven. We hoose a ommon lassifiation of six age groups under 24 years old, 25-34, 35-44, 45-54, 55-64, and over 65 years of age. In both eonomies workers are lassified as employed or self-employed and unpaid family workers. However, labor ompensation for the self-employed and unpaid family workers is estimated differently in the U.S. and Japan. In the U.S. data the hourly wage of self-employed and unpaid family workers is set eual to the hourly wage of the employed. Hours worked times this wage yields an estimate of labor ompensation for self-employed and unpaid family workers. In the Japanese data wage rates of self-employed workers are available for seleted industries that yield an estimate of labor ompensation 10 In measuring apital input in Japan, apital onsumption allowanes, inome allowanes and reserves, speial depreiation, orporate inome tax, business inome tax, property taxes, auisition taxes, debt/euity finaning, and personal taxes are taken into aount. 8

10 for the self-employed group. The wage differential between self-employed and unpaid family workers is estimated from the differential between full-time and part-time employees. Beause of these differenes, we onsider only employed workers when measuring the PPPs for labor input. In the Japanese data there are different levels of detail for the eduational attainment of male and female workers. Male workers are divided among four eduational groups, while female workers are split into three. In the ommon data set we use the most detailed lassifiation available in the Japanese data. After ross-lassifying the data by all the demographi harateristis, we have 1596 groups in total, 912 groups of male employees and 684 groups of female employees. We alulate the industry-level PPP for labor inputs as the translog index of the elementary-level PPPs. III. Empirial Results i. Purhasing Power Parities We summarize our estimates of PPPs in terms of gross domesti produt (GDP) and apital, labor, and intermediate inputs. For eah industry the PPP for value added is defined by the double deflation method, using industry-level PPPs for gross output and intermediate inputs. We define the PPP for GDP as a translog index of the industry-level PPPs for value added, taking the weight hanges over periods into aount. Similarly, the PPPs for the apital, labor, energy, and materials (KLEM) inputs are defined as translog indexes of industry-level PPPs for these inputs. Table 1 presents our estimates of PPPs for Japan. For 2004 our estimate for GDP is yen/dollar; this implies that a unit of GDP osting one dollar in the U.S. was valued at yen in Japan. Sine one U.S. dollar exhanged for yen, Japan s prie for GDP was 23.8 perent higher than the U.S. prie. Our PPP estimates are based on outputs, while the Eurostat-OECD PPPs presented in Table 1 are based on expenditures. 12 Aording to the Eurostat-OECD estimate, the Japanese prie of GDP in 2004 was 23.6 perent higher than the U.S. prie. Although the two PPP estimates are nearly idential in 2004, our output-based estimates are higher than the expenditure-based estimates for 1960 and 1973 and lower for the period Due to data onstraints, we establish a ommon industry lassifiation for 38 U.S. and Japanese industries to alulate PPPs for labor inputs. 12 Organisation for Eonomi Co-operation and Development (2006), GDP PPPs and Derived Indies for all OECD Countries, Paris, OECD, Deember. 9

11 Table 1: PPP and Relative Pries for GDP and KLEM PPP GDP-output based Capital Labor Energy Material (ref)gdp-expenditure based Exhange Rate Relative Pries GDP-output based Capital Labor Energy Material (ref)gdp-expenditure based Note: The PPP for GDP-output based is defined as a translog index of industry-level PPP for value added alulated by the double deflation method. The PPP for GDP-expenditure based is the estimates by the Eurostat-OECD. The PPP and exhange rate are defined by Japanese yen/ US dollar. Figure 1 represents the long-term trends of PPPs for GDP and the KLEM inputs. The yen-dollar exhange rate is drawn as a shadow figure. If the PPP is higher than the shadow figure, the Japanese prie is higher than the U.S. prie. Until the mid-1970s, the Japanese prie for GDP was lower than the U.S. prie and Japanese input pries were lower than the U.S. pries, exept for energy. Lower input pries provided a soure of international prie ompetitiveness for Japanese produts in the 1960s and 1970s. The Plaza Aord of 1985 was a ruial turning point for the Japanese eonomy. In 1985 the yen was undervalued by about 17 perent, based on a omparison between our output-based PPP for GDP and the yen-dollar exhange rate. In the late 1980s the rapid strengthening of the yen relative to the U.S. dollar reversed this relationship, leading to an overvaluation of the yen by 25 perent in The revaluation of the yen ontinued through 1995, leading to an overvaluation of 78 perent! This was followed by a gradual devaluation through 2004, leaving the yen overvalued by 24 perent, very lose to the 1990 level. 10

12 (Japanese Yen/US Dollar) Average Exhange Rate PPP for GDP-output based PPP for Capital Input PPP for Labor Input PPP for Energy Input PPP for Material Input (ref) PPP for GDP-expenditure based (OECD) Figure 1: PPP for GDP and KLEM during The Japanese eonomy spent more than a deade overoming the huge overvaluation of the yen that followed the Plaza Aord. This was aomplished mainly by domesti deflation, rather than devaluation of the yen. The prie of GDP in Japan, relative to the U.S., delined by four perent annually through 2004 from the peak attained in The deline in the PPP for GDP at 2.5 perent per year was the result of modest inflation in the US of 1.4 perent and deflation in Japan of 1.1 perent. In addition, the yen-dollar exhange rate fell by 1.5 perent per year. Figure 2 presents industry-level relative pries for value added and industry origins of the gap in the PPP of yen per dollar of the output-based GDP in The Japanese prie for value added was lower than the U.S. prie in sixteen industries, while the Japanese prie for gross output was lower in only ten industries. Even if the relative prie for gross output is greater than unity, the relative prie for value added an be less than unity as a onseuene of high pries of energy and materials inputs in Japan. The Household setor produes servies of owner-oupied dwellings and onsumer durables and made the greatest ontribution to the PPP for GDP in This reflets high pries of buildings and land in Japan and the large share of value added by the Household setor. Eletriity, whih is very expensive in Japan, pushed up the Japanese PPP by 2.2 perent. 11

13 Relative Prie for Value Added Industry Contribution -4% -2% 0% 2% 4% 6% 8% Household 40.Other Servies 36.Real Estate 4.Constrution 32.Eletriity 5.Foods 1.Agriulture, Forestry, Fishery 34.Wholesale and Retail 11.Printing and Publishing 3.Other Mining 37.Eduation 13.Petroleum Refining 41.Publi Administration 2.Coal Mining 9.Furniture and Fixture 17.Metal Produts 33.Gas Supply 26.Mis Manufaturing 30.Other Trans and Storage 31.Communiations 12.Chemial Produts 16.Primary Metal 6.Textile 7.Apparel 14.Leather Produts 20.Communiations Euipment 38.Researh 25.Preision Instruments 8.Woods and Related Produts 10.Paper and Pulp 35.Finane and Insurane 29.Air Transportation 15.Stone, Clay, Glass 27.Railroad Transportation 28.Water Transportation 24.Other Transportation Euipment 22.Other Eletrial Mahinery 19.Computers 21.Eletroni Components 18.Mahinery 39.Medial Care 23.Motor Vehiles Figure 2: Industry Origins of PPP-for-GDP Gap in

14 ii. Eonomi Growth in the U.S. and Japan Table 2 summarizes our empirial results on eonomi growth and its soures for the U.S. and Japanese eonomies. Despite the long eonomi reession in Japan, beginning in the early 1990s, rapid tehnologial progress in the IT-produing industries diffused uikly through investment in information tehnology (IT) euipment and software by the IT-using industries. Jorgenson and Nomura (2005) 13 have disussed these findings through 2000 in greater detail. Table 2 shows that the trends they identified have ontinued through Japanese eonomi growth slowed further after 2000, but the ontribution of ITapital input ontinued to rise. The annual rate of total fator produtivity (TFP) growth in Japan improved from 0.48 perent per year in to 0.57 perent during The ontribution of labor input was negative in both periods, but rose slightly from to The ontribution of Non-IT-apital input in Japan dipped substantially from to , aounting for the slowdown in Japanese eonomi growth. Despite this slowing trend, the ontribution of investment in IT apital aelerated. The ontribution of IT to the growth of apital input expanded from 13.9 perent in to 31.2 perent during , and reahed 51.6 perent for the period , onsiderably exeeding the U.S. ontribution of 43.1 perent in that period. Expanding labor input was an important soure of U.S. eonomi growth during , but this ended abruptly with the dot-om rash of Labor input delined at the rate of 0.17 perent per year during Another signifiant hange in the U.S. eonomy after 2000 was the substantial inrease in the ontribution of TFP growth and the shift in its industry origins. The annual growth rate of TFP during was 1.27 perent per year, lose to twie the growth rate of the late 1990s and 2.7 times the long-term average growth rate during of 0.46 perent. 13 We have extended the Japanese data used in Jorgenson and Nomura (2005) through In addition, we have revised the data after 1995 to inorporate the Annual Report on National Aounts (Eonomi and Soial Researh Institute, Cabinet Offie) published in May This report is the first to inlude the 2000 benhmark revision. Seond, we have revised the imputed ost for owner-oupied housing. In the Japanese national aounts, the estimation method for the ost of owner-oupied housing was substantially improved in the 2000 benhmark. As a onseuene, the imputed ost was redued from 49.9 trillion yen in 2000, based on the 1995 benhmark, to 42.8 trillion. Third, we have revised the alloation of land for dwellings between the Household and the Real Estate industries, based on estimates by Hideyuki Mizobuhi. Finally, we have introdued new estimates for work-inprogress inventories for ultivated assets as a apital input, onsistent with the reommendation of the 1993 SNA. The values for the inventory stok on ultivated assets estimated by Nomura are 3.3, 8.6, and 3.9 trillion yen at urrent pries as of the end of 1960, 1980, and 2000, respetively. 13

15 Table 2: Soures of Eonomi Growth in the U.S. and Japan United States Value Added Capital Input IT Capital Non-IT Capital Labor Input Total Fator Produtivity Agriulture IT-manufaturing Motor Vehile Other manufaturing Communiations Trade Finane & Insurane Other servies Japan Value Added Capital Input IT Capital Non-IT Capital Labor Input Total Fator Produtivity Agriulture IT-manufaturing Motor Vehile Other manufaturing Communiations Trade Finane & Insurane Other servies Note: All figures are average annual growth rates.value added is aggregated from industry GDPs evaluated at the fator ost. Perhaps surprisingly, the substantial improvement in U.S. TFP growth between and was ahieved by a wide variety of IT-using industries rather than the IT-produing setors as disussed more in detail in Jorgenson, Ho, Samuels, and Stiroh (2007). The TFP ontribution by the ITmanufaturing setor was only 0.04 perent, a preipitous drop from 0.48 perent during By ontrast the TFP ontributions in ommuniations, trade, and finane and insurane improved by 0.11, 0.36, and 0.19 perent points, respetively. Also, the negative ontribution of perent per year in other servies during the late 1990s gave way to a positive 0.15 perent after

16 United States -20% -15% -10% -5% 0% 5% Japan -5% 0% 5% 10% Air Transportation Preision Instruments Communiations Wholesale and Retail Other Servies Railroad Transportation Finane and Insurane Other Trans and Storage Other Mining Mahinery Motor Vehiles Medial Care Real Estate Agriulture, Forestry, Fishery Other Transportation Euipment Chemial Produts Mis Manufaturing Woods and Related Produts Eduation 42.Household 41.Publi Administration Textile Stone, Clay, Glass Furniture and Fixture Paper and Pulp Printing and Publishing Apparel Researh Constrution Other Eletrial Mahinery Foods Metal Produts Coal Mining Water Transportation Petroleum Refining Communiations Euipment 16.Primary Metal Eletriity Gas Supply Leather Produts Computers 21.Eletroni Components Figure 3: Changes in Growth of Industry TFP: less

17 United States Japan Other Servies 34.Wholesale and Retail 35.Finane and Insurane 31.Communiations 39.Medial Care 29.Air Transportation 36.Real Estate 25.Preision Instruments 23.Motor Vehiles 30.Other Trans and Storage 1.Agriulture, Forestry, Fishery 18.Mahinery 3.Other Mining 12.Chemial Produts 24.Other Transportation Euipment 27.Railroad Transportation 26.Mis Manufaturing 9.Furniture and Fixture 8.Woods and Related Produts 38.Researh 15.Stone, Clay, Glass 42.Household 41.Publi Administration 37.Eduation 6.Textile 2.Coal Mining 14.Leather Produts 10.Paper and Pulp 28.Water Transportation 7.Apparel 11.Printing and Publishing 22.Other Eletrial Mahinery 20.Communiations Euipment 33.Gas Supply 17.Metal Produts 16.Primary Metal 5.Foods 13.Petroleum Refining 32.Eletriity 4.Constrution 19.Computers 21.Eletroni Components Figure 4: Changes in Contribution of Industry TFP to Eonomi Growth: less

18 Figures 3 and 4 ompare industry-level TFP growth and ontributions to eonomi growth between the periods and In the key IT-produing setors, Eletroni Components and Computers, TFP growth in the U.S. slowed by more than ten perentage points. The improvement of TFP growth in the IT-using setors seems modest by omparison with improvements of 1.9 perentage points in Communiations, 1.7 in Wholesale and Retail Trade, 1.6 points in Other Servies, and 1.2 points in Finane and Insurane. However, as shown in Figure 4, these industries made substantially inreased ontributions to the U.S. eonomi growth, refleting their large Domar weights, as presented in Euation (24) of the Methodologial Appendix. The resurgene of U.S. eonomi growth during was powered by an aeleration of TFP growth in the IT-produing industries and a resulting surge of investment in IT apital by the ITusing setors. After 2000 produtivity growth plummeted in the IT-produing setors, but greatly aelerated in the IT-using industries. The end of the investment boom of the late 1990s with the dot-om rash of 2000 resulted in a substantial fall in the ontributions of investment in IT- and non-it-apital inputs after The growth of labor input plunged by two full perentage points! There was no resurgene of Japanese eonomi growth during ; however, the ontribution of TFP revived during this period and rose further during The ontribution of TFP in IT-manufaturing in Japan rose from 0.09 per year in to 0.42 perent during , before reeding to 0.35 perent after 2000, well above the U.S. ontribution of 0.04 perent. The ontribution of investment in IT-apital input rose steadily from 0.27 perent per year in to 0.32 perent in and 0.37 perent after iii. Level Comparisons of Output, Input, and Produtivity Table 3 summarizes the produtivity gap between the U.S. and Japan. This table ompares relative levels of output, output per apita, input per apita, and total fator produtivity between the two eonomies over the period Differenes in output per apita an be deomposed into differenes in input per apita and TFP, the ratio of output to input. For example, Japanese GDP was 30.2 perent of the U.S. level in GDP per apita in Japan was 69.5 perent of U.S. GDP per apita, while Japanese input per apita was 87.6 perent and Japanese TFP was 79.5 perent of the orresponding U.S. levels. 17

19 Table 3: Relative Levels of Output, Input, and Total Fator Produtivity Output Output per Capita Input per Capita Total Fator Produtivity Note: All figures present the relative values (Japan/U.S.) in eah period. In Table 4 we ompare levels of per apita apital and labor inputs in Japan and the U.S. These are the two soures of differenes in input per apita in Table 3. Japanese labor input per apita in 1960 was only slightly below the U.S. level and Japan has led in labor input per apita sine Japan maintained a lead in hours worked per apita throughout the period , despite a gradual deline after 1973 that aelerated sharply during the 1990s. The U.S.-Japan gap in labor uality, defined as labor input per hour worked, was 78.1 perent in This gap gradually losed during the three deades of rapid Japanese growth, ending in Japan overtook the U.S. in labor uality in 1995 and has maintained its lead through the past deade. This reflets the attainment of high average levels of eduation in Japan, as well as gains in the experiene of the Japanese labor fore. Table 4: Relative Quantities and Qualities of Capital and Labor Capital Input per Capita Capital Stok per Capita Capital Quality Labor Input per Capita Hours Worked per Capita Labor Quality Note: All figures present the relative values (Japan/U.S.) in eah period. Japanese apital input, relative to the U.S., presents a striking ontrast to labor input. In 1960 Japanese apital input per apita was only 23.2 perent of the U.S. level, but staggering levels of investment in Japan during the double-digit growth of redued the gap to 50.0 perent by The gap ontinued to lose through 1995, when Japanese apital input per apita reahed 74.0 perent of the U.S. level. The investment slump that followed the ollapse of the bubble eonomy in Japan at the end of the 1980s and the U.S. investment boom of the late 1990s widened the gap to 68.5 perent in 2000 and 66.2 perent in We onlude that defiient investment in Japan, espeially after 1995, aounts for the remaining gap in input per apita. Table 4 provides a deomposition of the U.S.-Japan gap in apital input per apita between apital stok per apita and apital uality. Capital uality is defined as apital input per unit of apital 18

20 stok and reflets the omposition of the apital stok. In 1960 Japanese apital uality, relative to the U.S. was 61.2 perent, but this rose to 95.4 perent by 1973 and the gap had essentially losed by Capital uality in Japan attained perent of the U.S. level in 1995, before the IT investment boom in the U.S., but delined to 95.5 perent of the U.S. level in 2000 and 93.8 perent in After 1973 the U.S.-Japan gap in apital stok per apita largely paralleled the gap in apital input per apita. Table 5 ompares the levels of apital deepening of IT- and Non-IT apital in the U.S. and Japan, where apital deepening is defined as an inrease in apital input per hour worked. Capital deepening is an important soure of growth in labor produtivity. Capital input per hour worked in Japan rose from 19.3 perent of the U.S. level in 1960 to 67.1 perent in 2000 before falling bak to 62.6 perent in 2004 as hours worked delined in the U.S. IT-apital deepening in Japan peaked in 90.2 perent of the U.S. level in 1990 before delining to 87.3 perent in 1995 and falling preipitously to 56.6 perent in The trend in Non-IT-apital deepening was similar to overall apital deepening. 14 Table 5: Relative Levels of Capital Deepening Capital Deepening IT Capital Deepening Non-IT Capital Deepening Note: All figures present the relative values (Japan/U.S.) in eah period. Table 6 gives level omparisons of labor produtivity and apital produtivity, defined as output per hour worked and output per unit of apital stok, respetively. Labor produtivity in Japan was only 21.2 perent of the U.S. level in The labor produtivity gap shrank rapidly until 1995, when Japanese labor produtivity reahed almost 70 perent of the U.S. level. In sharp ontrast Japanese apital produtivity exeeded the U.S. level throughout the period , peaking at perent in The trends in relative labor and apital produtivity reflet relative fator supplies in the two eonomies. Japan has had a substantially higher labor/apital ratio than the U.S. throughout the period. The high labor/apital ratio is onsistent with the low apital/labor PPPs presented in Table The trend of apital deepening in both ountries is onsistent with the long-term hanges in relative pries for labor and apital inputs. The prie for labor input in the U.S. inreased annually at 5.1 perent for the period , by ontrast with 3.1 perent for the prie of apital input. In Japan, the growth rates in input pries for labor and apital were 6.3 and 1.9 perent, respetively. Thus the prie for labor input inreased relative to the prie of apital input by more than 2.0 perent in the U.S. and 4.4 perent in Japan. 19

21 Table 6: Relative Levels of Labor Produtivity and Capital Produtivity Labor Produtivity Capital Produtivity Note: All figures present the relative values (Japan/U.S.) in eah period. The soures of U.S.-Japan labor produtivity gap are shown in Figure 5. The y-axis gives the logarithm of the relative produtivity level. In 1960 lower Non-IT-apital deepening in Japan explained 50.1 perent of the U.S.-Japan labor produtivity gap, while lower Japanese TFP explained 40.4 perent. The lower uality of labor input and lower IT-apital deepening ontributed 8.8 perent and 0.7 perent, respetively. As a onseuene of rapid apital deepening in Japan, the TFP gap has been the major soure of lower labor produtivity sine the mid-1990s. Lower TFP explains 57.0 perent of the labor produtivity gap in 2004, while Non-IT-apital deepening aounts for 37.3 perent. log (relative produtivity level: Japan/US) IT Capital Deepening Labor Quality Labor Produtivity Gap Non-IT Capital Deepening TFP Figure 5: Soures of U.S.-Japan Labor Produtivity Gap iv. Industry Origins of the U.S.-Japan Produtivity Gap Figure 6 presents long-term trends of TFP gaps in manufaturing and non-manufaturing setors. In 1960 both gaps were very large. However, the TFP gap for manufaturing had almost disappeared by 20

22 1990, 15 so that the overall gap refleted lower TFP in non-manufaturing. The expansion of the TFP gap after 1990 was due mainly to deterioration in the relative TFP level in Japanese manufaturing. Finally, we fous on the industry origins of the U.S.-Japan TFP gap. Figures 7 and 8 present industry-level TFP gaps and the ontributions of eah industry to the overall TFP gap for 1990 and 2004, respetively. Industries are ordered by the magnitude of their ontributions to the TFP gap in eah year. The first olumn in eah figure gives the U.S.-Japan TFP gap, defined as the ratio of TFP in Japan to TFP in the U.S. Note that TFP gaps for Publi Administration and Household setors are zero by definition, sine the outputs of these industries onsist entirely of apital inputs. The seond olumn gives the ontribution of eah industry to the aggregate TFP gap, using Domar weights from Euation (24) in the Methodologial Appendix. log (relative produtivity level: Japan/US) IT Manufaturing Motor Vehile TFP Gap for the Whole Eonomy Non-IT Mnf.(ex. Vehile) Non Manufaturing TFP Gap for Manufaturing Figure 6: TFP Gap in Manufaturing and Non-Manufaturing during Cameron (2005) analyzes the produtivity onvergene of Japanese manufaturing setor to the U.S. and estimates the TFP gap in 1989 as Our result indiates a gap of 97.8 in the same year. The main soure the gap may be that his researh used the PPP estimates from one of our earlier studies. Our new estimates are based on a onsiderably revised framework, implementing the various onepts of PPP at the ommodity level. On apital PPPs, our new estimates reflet the improved estimates of the relative annualization fators, based on the ommon asset lassifiation, onsidering land as apital. 21