1 The Norwegian KLEMS database Gang Liu Statistics

1 The Norwegian KLEMS database Gang Liu Statistics Norway Presentation at the Fifth World KLEMS Conference Harvard University, June 4 -5, 2018 1

Presentation outline 1. Background and motivation 2. Database coverage 3. Database compilation 4. Aggregation and decomposition 5. Structural changes 6. Use of knowledge-based inputs 7. Hypothesis test 8. Concluding remarks 2

1. Background and motivation § EU-KLEMS (see Timmer et al. 2010); World KLEMS initiative (http: //www. worldklems. net); International standards (e. g. Schreyer, 2001, 2009). § Current practices at Statistics Norway: labor input, index formula. § Distinct character of recent economic growth: intensification of knowledge-based inputs, such as Skilled labor, ICT capital, and R&D assets. § Recent productivity slowdown in Norway. § The KLEMS database can be used for empirical and theoretical research in many areas, e. g. in productivity analysis, skill creation, capital development, technological progress and R&D activities, as well as economic growth more generally. § The construction and maintenance of the Norwegian KLEMS database will also facilitate the systematic production of high quality statistics in general, and of national accounts, growth and productivity statistics in particular at Statistics Norway. 3

2. Database coverage § Time span of the current database: 1997 -2014, determined primarily by availability of data, esp. labor input data. § Market economy at mainland Norway, i. e. market economy excluding offshore oil and gas extracting and transportation industries (KNR 2306, KNR 2348, KNR 2349), owner-occupied housing services (KNR 2368), and private renting (KNR 2369). § Industry coverage: 57 industries (KNR 23 xx) in total, also used by the Norwegian Quarterly National Accounts. § Further allocation of the 57 industries into 6 main sectors (in accordance with EU-KLEMS classification): ICT production (ELECOM: 5 industries), Manufacturing (MEXELEC: 20 industries), Other goods production (OTHERG: 10 industries), Distribution (DISTR: 6 industries), Finance and business services (FINBU: 8 industries), Personal services (PERS: 8 industries). § At the current stage, focus on growth rather than development accounting. 4

3. Database compilation - Multifactor productivity (MFP) • 5

3. Database compilation - Output and intermediate input • 6

3. Database compilation - Output and intermediate input (cont. ) § In Norwegian National Accounts system, there around 950 products defined by following EU’s main product Standard CPA (Classification of Products by Activities). § The Supply and Use Tables (SUTs) in both current and constant prices provide a consistent set of inter-industry transaction accounts (see Simpson and Todsen, 2012). Information drawn from the SUTs allows the calculation of the output from industry in basic prices and the inputs used by industry in purchaser’s prices. § Energy inputs (E) are defined as all energy mining products (code 050000 to code 060058), oil refining products (code 191000 to code 192420) and electricity and gas products (code 351107 to code 353000). § All services (products from code 33 xxxx and above) are included in S, as well as some of the (technically) aggregated products (code 000016, code 000026 to code 000050, code 000150 to code 000379). As a result, all the remaining products are classified as materials (M). § Trade and transportation product is treated as a separate product (service), and the trade and transportation margins on all other products are reallocated to this separate product. 7

3. Database compilation - Labor input • 8

3. Database compilation - Labor input (cont. ) Table 5. 1 Classification of labor force for each industry Dimension Age Education Employment class Gender Number of categories 3 4 2 2 Categories Young (15/16 -29), Middle (30 -49), Elder (50 and above) Low, Intermediate, High S (short), High L (Long) Employees, Self-employed Male, Female § Low = Primary and lower secondary education + Unknown education; § Intermediate = Upper secondary education, general programs + Upper secondary education, vocational programs; § High S (Short) = Tertiary education, lower degree; § High L (Long) = Tertiary education, higher degree. § Given different data quality for two sub-periods (1997 -2007, and 2008 -2014), a number of assumptions were made to derive data for each industry on actual hours worked, and the corresponding labor compensation, cross-classified by age, education, employment class and gender. 9

3. Database compilation - Capital input • 10

3. Database compilation - Capital input (cont. ) • 11

3. Database compilation - Capital input (cont. ) § In the Norwegian National Accounts (NNA) system, there about 40 detailed asset types that make up broad asset groups classified by the SNA. In particular, three asset types are regarded as ICT capital: Office and computing equipment, Communications equipment, and Software. The first two are Hardware (ITH), and the last is Software (ITS). R&D asset is newly incorporated. § Two important asset types, land inventories, are missing in the current NNA. § In the NNA system, long time series of gross fixed capital formation (GFCF) for different assets exist, dating back to 1970, which enables the estimation of capital stock for each asset by following the PIM. § The PIM and geometric depreciation are applied to all assets for each industry, and the associated service lives vary by industry, institutional sector and asset type. The choice of service lives is based on expert advices, other countries’ estimates, as well as empirical estimates drawn from recent survey questionnaires (see Barth, et al. , 2016). 12

3. Database compilation - Average labor productivity (ALP) • 13

4. Aggregation and decomposition • 14

4. Aggregation and decomposition (cont. ) • 15

5. Structural changes - Output and employment Figure 5. 1. Ratio of services over goods production, 1997 -2014 Ratio (Value added) Ratio (Hours worked) 2. 5 2. 0 1. 5 1. 0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Source: Calculations are based on Norwegian KLEMS database, July 2017. 16

5. Structural changes - Output and employment share by sector Table 5. 1. Share of value added and hours worked by sector (%) Value added Hours worked 1997 2014 Total market economy of mainland Norway 100 100 ICT production (ELECOM) 7. 76 6. 86 6. 06 5. 57 Goods 36. 06 30. 99 38. 54 32. 39 Manufacturing (MEXELEC) 19. 00 12. 89 18. 58 13. 76 Other goods (OTHERG) 17. 06 18. 10 19. 96 18. 62 Services 56. 18 62. 16 55. 40 62. 05 Distribution (DISTR) 24. 78 20. 23 27. 59 25. 46 Finance and business (FINBU) 22. 39 33. 56 16. 56 23. 26 Personal (PERS) 9. 01 8. 37 10. 89 13. 33 Source: Calculations are based on Norwegian KLEMS database, July 2017. 17

5. Structural changes – Average labor productivity (ALP) (cont. ) Figure 5. 2. ALP level by sector (1997=100), value-added based 250 200 ELECOM MEXELEC OTHERG 150 DISTR FINBU PERS 100 50 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Source: Calculations are based on Norwegian KLEMS database, July 2017. 18

5. Structural changes - Average labor productivity (ALP) Table 5. 2. ALP growth by sector (%), value added based Total market economy of mainland Norway 1997 -2014 2. 15 1997 -2006 2. 89 2006 -2014 1. 33 ICT production (ELECOM) 4. 90 5. 21 4. 51 Goods Manufacturing (MEXELEC) 2. 11 3. 28 1. 86 2. 04 2. 41 5. 02 Other goods (OTHERG) 1. 02 1. 66 0. 34 Services Distribution (DISTR) 1. 50 2. 30 2. 82 4. 35 0. 09 -0. 21 Finance and business (FINBU) 1. 60 2. 59 0. 70 Personal (PERS) -0. 92 -0. 52 -1. 41 Notes: Average annual compound growth rates. Source: Calculations are based on Norwegian KLEMS database, July 2017. 19

5. Structural changes – Multifactor productivity (MFP) (cont. ) Figure 5. 3. MFP level by sector (1997=100), value added based 250 200 ELECOM MEXELEC OTHERG 150 DISTR FINBU PERS 100 50 199719981999200020012002200320042005200620072008200920102011201220132014 Source: Calculations are based on Norwegian KLEMS database, July 2017. 20

5. Structural changes – Multifactor productivity (MFP) Table 5. 3. MFP growth by sector (%), value added based Total market economy of mainland Norway 1997 -2014 1. 35 1997 -2006 1. 55 2006 -2014 1. 13 ICT production (ELECOM) 4. 06 3. 81 4. 38 Goods Manufacturing (MEXELEC) 1. 85 2. 58 1. 10 1. 01 2. 76 4. 78 Other goods (OTHERG) 1. 17 1. 19 1. 16 Services Distribution (DISTR) 0. 72 2. 15 1. 50 3. 54 -0. 12 0. 44 Finance and business (FINBU) 0. 27 0. 76 -0. 18 Personal (PERS) -1. 53 -1. 71 -1. 30 Notes: Average annual compound growth rates. Source: Calculations are based on Norwegian KLEMS database, July 2017. 21

6. Use of knowledge-based inputs - Measure of input intensity • 22

6. Use of knowledge-based inputs - Skilled labor in total market economy Figure 6. 1. Labour services share in value added, 1997 -2014, total market economy 80% 70% 60% 50% Total labor 40% High L High S 30% 20% 10% 0% 199719981999200020012002200320042005200620072008200920102011201220132014 Notes: Labour includes employees and self-employed. Source: Calculations are based on Norwegian KLEMS database, July 2017. 23

6. Use of knowledge-based inputs – Capital in total market economy Figure 6. 2. Knowledge-based capital services share in value added, 1997 -2014, total market economy 4% 3% ITH 2% ITS R&D 1% 0% 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Source: Calculations are based on Norwegian KLEMS database, July 2017. 24

6. Use of knowledge-based inputs – High S labor by sector Figure 6. 3. Compensation of High S share in sector value added, 2008 -2014 30% 25% 20% ELECOM MEXELEC OTHERG 15% DISTR FINBU 10% PERS 5% 0% 2008 2009 2010 2011 2012 2013 2014 Notes: Labour includes employees and self-employed. Source: Calculations are based on Norwegian KLEMS database, July 2017. 25

6. Use of knowledge-based inputs – High L labor by sector Figure 6. 4. Compensation of High L share in sector value added, 2008 -2014 16% 14% 12% ELECOM 10% MEXELEC OTHERG 8% DISTR 6% FINBU PERS 4% 2% 0% 2008 2009 2010 2011 2012 2013 2014 Notes: Labour includes employees and self-employed. Source: Calculations are based on Norwegian KLEMS database, July 2017. 26

6. Use of knowledge-based inputs – Hardware (ITH) by sector Figure 6. 5. Hardware (ITH) share in sector value added, 1997 -2014 20% 18% 16% 14% ELECOM 12% MEXELEC 10% OTHERG DISTR 8% FINBU 6% PERS 4% 2% 0% 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Source: Calculations are based on Norwegian KLEMS database, July 2017. 27

6. Use of knowledge-based inputs – Software (ITS) by sector Figure 6. 6. Software (ITS) share in sector value added, 1997 -2014 6% 5% 4% ELECOM MEXELEC OTHERG 3% DISTR FINBU 2% PERS 1% 0% 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Source: Calculations are based on Norwegian KLEMS database, July 2017. 28

6. Use of knowledge-based inputs – R&D by sector Figure 6. 7. R&D share in sector value added, 1997 -2014 12% 10% 8% ELECOM MEXELEC OTHERG 6% DISTR FINBU 4% PERS 2% 0% 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Source: Calculations are based on Norwegian KLEMS database, July 2017. 29

7. Hypothesis test § Economic growth in last decades has revealed a growing importance of skilled labour and ICT assets in production (e. g. Jorgenson et al. , 2005). § One appealing explanation to this phenomenon is that there exists complementarity between increased use of skilled labour and ICT capital (e. g. O’Mahony et al. , 2008; Timmer et al. , 2010). § This complementarity hypothesis is tested by using the Norwegian KLEMS data. § Is there any complementarity relationship between skilled labor and other knowledgebased capital? § Define knowledge capital as: ITH + ITS + R&D; Intellectual Property Products (IPP) as ITS + R&D. 30

7. Hypothesis test (cont. ) Table 7. 1. Correlation coefficients between use of skilled labour and of knowledge-based capital Total market economy ICT production (ELECOM) Goods Manufacturing (MEXELEC) Other goods (OTHERG) Services Distribution (DISTR) Finance and business (FINBU) Personal (PERS) Knowledge capital (ITC+R&D) High (S+L) High S High L 0. 34 -0. 69 0. 74 0. 47 -0. 90 0. 23 -0. 72 0. 68 0. 45 -0. 91 0. 41 -0. 63 0. 78 0. 48 -0. 83 -0. 43 -0. 86 -0. 73 0. 27 -0. 84 -0. 87 IPP capital (ITS + R&D) High (S+L) High S High L ITC (ITH+ITS) High (S+L) High S High L -0. 79 -0. 85 0. 66 0. 34 -0. 90 -0. 71 -0. 83 0. 61 0. 33 -0. 86 -0. 83 -0. 80 0. 70 0. 34 -0. 92 -0. 34 -0. 86 -0. 90 -0. 85 High (S+L) 0. 41 -0. 84 R&D High S High L 0. 91 -0. 26 0. 76 0. 32 0. 38 0. 82 -0. 24 0. 70 0. 31 0. 27 0. 95 -0. 26 0. 81 0. 31 0. 57 0. 73 0. 46 0. 60 0. 39 0. 25 0. 62 0. 41 0. 55 0. 37 0. 12 0. 79 0. 46 0. 64 0. 40 0. 50 -0. 04 0. 61 -0. 75 0. 63 0. 79 0. 56 -0. 06 0. 65 -0. 74 0. 68 0. 75 0. 56 Source: Calculations are based on Norwegian KLEMS database, July 2017. 31

8. Concluding remarks § Theoretical methodologies and practical compilation procedures are documented as regards the construction of the Norwegian KLEMS database over the period 1997 -2014, based mainly on official statistics including national accounts data. § The database consists of five accounts: output and intermediate input, labor, capital, and multifactor and labor productivity accounts at disaggregated industry level, all being organized by the modern growth accounting framework. § Further classifications (intermediate inputs into E, M and S; labor inputs into hours worked and changes of labor composition; capital inputs into Hardware (ITH), Software (ITS), R&D, and Others) make it possible for the decomposition of labor productivity growth into various detailed components. § Based on Norwegian data for 1997 -2014, an increasing share was found in output and employment of services at the expense of goods production; services had become the largest sector in terms of output and employment; productivity growth in goods production sector was higher than in services sector over the entire period. § However, when considering the changes between two subperiods (i. e. 1997 -2006, and 2006 -2014), productivity performance in the goods production sector was weaker in the first subperiod, while much stronger in the second, than in the services sector. 32

8. Concluding remarks (cont. ) § A more detailed sector analysis reveals very substantial heterogeneities both within the goods production sectors and among the services sectors, leaving the traditional distinction between goods and services outdated. In particular, the characterization of services as stagnant in terms of productivity growth and input structure is no longer true. § Based on the calculated input intensity measures, an increased share of skilled labour in value added is found for the total market economy of mainland Norway over the entire period, as well as for almost all the sectors, at least for the latter period (2008 -2014). § For the total market economy, the shares in value added of both Software (ITS) and R&D capital increased, while those of Hardware (ITH) decreased, for the whole period 19972014. With a few exceptions, this finding also holds for almost all the sectors, at least for the latter period (2008 -2014). § Finally, the test results show that the complementarity hypothesis between the use of ICT capital and skilled labour is not supported by the Norwegian data. On the other hand, the existence of complementarity between the use of IPP (ITS + R&D) capital and highly skilled labour (High L) seems to be suggestive. Furthermore, the complementarity relationship between R&D and highly skilled labour is strongly suggestive based on the Norwegian data. 33