INTANGIBLE INVESTMENT AND CHANGING SOURCES OF GROWTH IN KOREA

Transcription

1 bs_bs_banner The Japanese Economic Review Vol. 67, No. 1, March 2016 The Japanese Economic Review The Journal of the Japanese Economic Association INTANGIBLE INVESTMENT AND CHANGING SOURCES OF GROWTH IN KOREA By HYUNBAE CHUN and M. ISHAQ NADIRI Sogang University New York University doi: /jere We examine how intangible investments change the sources of growth in the Korean economy. After constructing a novel industry-level data set on intangibles, we estimate the contribution of intangible-intensive industries and other industries to aggregate productivity growth in The contribution of intangible-intensive industries to aggregate labour productivity growth has significantly increased, whereas that of other industries has substantially decreased. The increased contribution of intangible-intensive industries is mainly associated with total factor productivity growth rather than with input growth. This suggests that innovations related to intangible investments in these industries might become a new key source of productivity growth in Korea. JEL Classification Numbers: O47, E22, O14, O Introduction Firms in the knowledge economy invest in a wide range of intangible assets rather than a single type of intangible asset such as R&D and information technology (IT). However, systematic attempts to broaden the range of intangible assets in research on the sources of growth are still quite recent. Corrado et al. (2005, 2006) provide a more or less complete list of intangible assets and quantify the amount of intangible investment in the US economy that is also integrated into the national accounts. Based on this widely adopted definition in Corrado et al. (2005, 2006), many studies find that investment in intangible assets has become a new source of economic growth in developed countries. 1 While most of these studies have been conducted for developed countries at the aggregate level, studies on developing countries, in particular at the industry level, are relatively scant. In this paper, we investigate the role of intangible assets in sources of growth for the Korean economy. In particular, we focus on the interrelatedness of intangible investments and industrial transformation in order to explain changes in the sources of growth in Korea. During the past few decades, Korea has transitioned toward a knowledge-based economy by investing massively in intangibles such as industrial R&D and IT. 2 The developing Korean economy also experienced a rapid transformation of its industrial structure by shifting from agriculture to manufacturing and later to services. This transitional experience of Korea to a knowledge economy highlights a unique feature 1 The studies include Jalava et al. (2007) for Finland; Van Rooijen-Horsten et al. (2008) for the Netherlands; Fukao et al. (2009) for Japan; Marrano et al. (2009) for the UK; Edquist (2011) for Sweden; and Baldwin et al. (2012) for Canada. 2 R&D spending as a percentage of GDP reached approximately 3.7% in 2010, which ranked third among OECD countries (OECD, 2012b). Business internet use and business with a broadband connection also ranked in the top five among OECD countries (OECD, 2012a). 50

2 The Japanese Economic Review associated with changing sources of growth, which contrasts with transition in developed countries. 3 Accordingly, we employ a definition of intangible assets suggested by Corrado et al. (2005, 2006), which encompasses a broad range of intangibles, including industrial R&D, computer-related information, and various firm-specific human and organizational resources. 4 The measurement of this broad range of intangible assets, particularly at the industry level, is crucial for our study because of the industry-specific nature of intangible investment, such as industrial R&D in manufacturing-intensive countries and IT in serviceintensive countries. After constructing this novel industry-level data set, we examine two distinctive channels of intangible assets that contribute to aggregate labour productivity (ALP) growth: the total factor productivity (TFP) growth effect and the intangible capital deepening effect. 5 The former is related to technological innovations stemming from knowledge accumulation through intangible investments, whereas the latter is related to the input accumulation effect associated with intangible capital deepening. We employ the methods of Stiroh (2002) and Jorgenson et al. (2007) to link industrylevel productivity growth to aggregate growth. Then, we decompose industries contributions to ALP growth into their direct productivity effects and resource reallocation effects. In particular, we compare the direct effects of two industry groups: intangible-intensive industries and other industries. We define the top one-third of 27 industries with high intangible investment rates as intangible-intensive industries and the others as intangibleless-intensive (hereafter, non-intangible) industries. 6 The contribution of intangibleintensive industries to ALP growth steadily increased during and accounted for more than 60% of ALP growth in Moreover, the contribution of intangibleintensive industries to ALP growth is mainly attributed to their TFP growth rather than to input growth. In contrast, the contribution of non-intangible industries to ALP growth has decreased, which is associated with both sluggish physical investment and lowered TFP growth. Consistent with our finding on TFP growth as the main source of the increased contribution of intangible-intensive industries, reallocation of hours worked to highproductivity industries is relatively small, and its contribution to ALP growth decreases over the sample period. Overall, our findings suggest that the main source of ALP growth in the Korean economy has shifted from the accumulation of tangible capital in nonintangible industries toward TFP growth in intangible-intensive industries. 3 Some studies indicate the diffusion of IT as a key source of recent structural changes, such as high productivity growth in some service industries, in advanced countries (Jorgenson and Timmer, 2011). 4 The positive relationship between these intangibles and firm performance is found in a number of studies, such as Stiroh (2002) and Brynjolfsson and Hitt (2003) for IT; Griliches (1986) and Griffith et al. (2004) for R&D; and Ichniowski et al. (1997) and Bloom and van Reenen (2007) for firm-specific human and organizational resources. 5 This approach is similar to that of Oliner and Sichel (2000), who examine the contribution of IT to the USA. ALP grows through two channels: TFP growth in IT-producing industries and IT capital deepening in IT-using industries. Furthermore, Stiroh (2002) estimates a direct link between TFP growth and IT use at the industry level. Unlike IT hardware, most Corrado et al. (2005, 2006) intangibles are produced and used in the same industry. Therefore, we do not distinguish between intangible-using and intangibleproducing industries. 6 To select intangible-intensive industries, we use the average intangible investment rate over the whole sample period. However, the selection of 9 intangible-intensive industries based on either the whole sample period or a specific period of every decade gives us similar sets of industries because the ranking of intangible intensity is stable over the sample period. 51

3 H. Chun and M.I. Nadiri: Intangible Investment and Sources of Growth Next, we examine the input accumulation effect of intangibles on ALP growth. To do so, we employ the method proposed by Corrado et al. (2005, 2006) and treat expenditures on intangibles as investments rather than as expenses. Given that the intangible capital deepening effect is included as an additional source of labour productivity growth, other sources of growth, such as tangible capital deepening, labour quality growth and TFP growth are reduced. However, the magnitude of the reductions is relatively small. This confirms that the contribution of intangible-intensive industries to ALP growth through TFP growth is not an artifact of measurement errors due to a missing input. We also find that the slowdown in tangible investment is partially mitigated by an increase in intangible investment in intangible-intensive industries, but such mitigation does not occur in non-intangible industries. From the 1990s to the 2000s, the increased contribution of intangible capital deepening in intangible-intensive industries offset the decreased contribution of tangible capital deepening, whereas this shift was not observed in non-intangible industries. Our study is distinguished from previous works in several ways such that our findings provide meaningful implications for understanding the sources of growth in Korea as well as the role of intangibles in economic growth. We have made the first attempt to explain the changing sources of growth in Korea by examining not only industrial transformation in the development process but also the recent rapid technological progress characterized by massive intangible investments. Our approach provides better explanatory power than the conventional theory on structural transformation to explain the changing sources of growth in the Korean economy: this is because we rely on industrial shifts, first from agriculture to manufacturing, and later from manufacturing to services, during the development stages (Kuznets, 1966; Imbs and Wacziarg, 2003). In conventional theories, a sectoral shift to manufacturing raises aggregate productivity, while a shift to services limits aggregate productivity growth, because productivity in manufacturing is higher than that in both agriculture and services (Baumol, 1967; Restuccia et al., 2008; Duarte and Restuccia, 2010). Rather, the recent technological progress associated with the IT revolution may alter sectoral productivity levels themselves, which is true not only for developed countries but also for developing countries (Inklaar et al., 2005). Contrary to the stylized facts observed in developed countries, the transition from manufacturing to services has evidently not been observed in Korea. In fact, the manufacturing share has been stable since reaching its highest level (almost 30% of the gross domestic product (GDP)) in the late 1980s. 7 Instead, the expansion of intangibleintensive industries, not only in manufacturing but also in services, accounts for recent shifts in the sources of both output and productivity growths in Korea. This suggests that the development process of the Korean economy may not be the same as those of existing developed countries. In contrast to most previous studies on the Corrado et al. intangibles performed at the aggregate level (e.g. Corrado et al., 2006; Fukao et al., 2009; Marrano et al., 2009) or at the sectoral level (Barnes, 2010; Dal Borgo et al., 2013), 8 our study examines the role of 7 Despite the stable manufacturing share since the late 1980s, the service sector has steadily expanded as the agriculture and mining sectors have shrunk. In 2012, the agriculture and mining sectors accounted for only 3% of GDP, and, thus, the future expansion of the service is unsustainable without a decline in the manufacturing sector. 8 See Barnes (2010) for Australia and Dal Borgo et al. (2013) for the UK at the one-digit industry level. 52

4 The Japanese Economic Review intangibles at the detailed industry level. 9 Because of the limitation of the aggregate data analysis, Corrado et al. (2005, 2006) and most previous studies examine only the input accumulation effect of intangibles on productivity growth and not the TFP effect associated with knowledge accumulation. The growth accounting analysis using industry-level intangible data enables us to identify the two channels of intangibles on ALP growth. In this study, we focus on the TFP effect of intangible-intensive industries on ALP growth. Although the growth accounting analysis cannot be used to identify the underlying mechanism of this TFP effect, 10 our findings shed light on the importance of innovations stemming from intangible investments and invite future research. The remainder of this paper is organized as follows. Section 2 presents our data on industry-level intangible investments following the definition of Corrado et al. (2005, 2006) and other variables for growth accounting analyses. Section 3 presents our methodology. Section 4 reports the results for the industry origins of both aggregate output and productivity growth. Section 5 concludes with a discussion of the results and implications for future research. 2. Data 2.1 Measuring intangibles As discussed in Hulten (1979), any use of resources that reduces current consumption in order to increase consumption in the future qualifies as an investment. This is true for spending on tangibles, including equipment and plants, as well as on intangibles, including software and R&D. However, in measuring GDP, spending on intangibles such as R&D is treated as an intermediate input, such as electricity. Except for mineral exploration and software, the System of National Accounts (SNA) currently treats intangible spending as intermediate consumption (i.e. expenses). The 2008 revision of the SNA treats R&D spending as investments. 11 Compared to the definition of intangibles by Corrado et al. (2006), R&D is only a type of intangible asset. Spending on advertising, training and other intangibles is not regarded by the SNA as investing in intangible assets. In contrast to the narrow definition used in the SNA, a broader definition of intangible investment is developed by Corrado et al. (2005, 2006, 2009). In the present paper, we follow this broader definition and construct three types of intangible investments: computerized information (CI), innovative property (IP) and economic competencies (EC). In measuring each type of intangible investment, we also use other complementary methods suggested by researchers, including Fukao et al. (2009), Corrado et al. (2012) and Dal Borgo et al. (2013). Computerized information consists mainly of software investment. Innovative property includes both scientific and non-scientific R&D. Non-scientific R&D includes costs of: new financial product development; copyright and license; product, machine and architectural design; and mineral exploration. Economic competencies include advertising, training costs and investments in the organizational structure. 9 An exception is Miyagawa and Hisa (2013) for Japan at the two-digit industry level. 10 For example, technology spillovers among firms within intangible-intensive industries may increase the TFP of the firms (Bernstein and Nadiri, 1989). 11 Along with the reclassification of R&D as investments, GDP will be revised in 2013 for the USA and in 2014 for Korea and Europe. 53

5 H. Chun and M.I. Nadiri: Intangible Investment and Sources of Growth First, computerized information at the aggregate level is obtained from aggregate software investment in the National Accounts published by the Bank of Korea (BOK). We allocate the aggregate software investment into 27 industries 12 by using the industry share of total software investment calculated from the Fixed Capital Formation Table, a supplementary table of the Input Output (IO) Table, which provides the industry-by-asset matrix of fixed capital formation. For the second category of intangible investments, we construct two types of investments on innovative property: scientific R&D and non-scientific R&D. We obtain industry-level scientific R&D expenditures from the Survey of Research and Development published by the Ministry of Education, Science and Technology. Scientific R&D of an industry is R&D funded by that industry. Government-funded R&D is not included. The aggregate level of scientific R&D is the sum of industry R&D expenditures. Non-scientific R&D includes: the costs of developing new products in the financial service and insurance industries; copyrights and license costs; product, machine and architectural design costs; and the cost of mineral exploration in the mining industry. Financial product development costs in Corrado et al. (2005, 2006) are assumed to be 20% of the intermediate inputs used by the financial industry. In this paper, we follow the labour-based production cost method for measuring costs of developing new financial products adopted in intangible database projects for the EU and other advanced countries, such as COINVEST and INNODRIVE (Jona-Lasinio et al., 2011). 13 This follows the approach of Hunt (2010), who argues that most of the cost of innovation in the financial service sector consists of researcher compensation. Because data on the detailed occupation structure are usually not available, we use 8% of the compensation of high skilled workers (i.e. college graduate workers) in the financial industry as new product development costs, which is a good approximation and is also adopted in the two projects listed previously. Financial product development costs solely belong to the financial industry. Using IO tables, industry-level copyright and license costs are constructed as intermediate inputs from the newspaper, publishing, broadcasting, movie, music, art and theater industries. Industry-level costs of product, machine and architectural design are similarly estimated using IO tables. These costs include external expenditures on design but not own-account expenditures. 14 The aggregate-level costs for copyrights and license and design are defined as the sum of industry-level costs. Investment in mineral exploration, which is included solely for the mining industry, is obtained from the National Accounts. The third category of intangible investments relates to economic competencies, which include both brand equity (including expenditures on advertising) and firm-specific human 12 Based on the Korea Standard Industry Classification (KSIC) rev. 9 in 2008, the National Accounts publish industry-level GDP for 27 industries, which include 11 manufacturing and 16 non-manufacturing industries. Following this 27-industry classification, we construct the industry-level intangible investment. Detailed descriptions for the 27-industry classification are provided in Table A1 of the Appendix. 13 More detailed information for the COINVEST and INNODRIVE projects is available at and respectively. Details about another project related to these two projects are available at (Corrado et al., 2012). 14 Because data on wage bills of design workers is not available at the industry level, we could not estimate the in-house design costs. Omission of own-account costs may underestimate the total design costs. 54

6 The Japanese Economic Review and organizational resources. 15 To obtain advertising spending (including marketing research spending) at the aggregate level, we use the gross output of the advertising industry. Following Corrado et al. (2005, 2006), we assume that only 60% of the total advertising spending has a long-term effect. To estimate the industry-level advertising expenditures, we allocate aggregate advertising expenditures to 27 industries by using the ratio of advertising spending to the total intermediate inputs for each industry, which are sourced from the IO tables. Employer-provided training costs are obtained from the Report on Labor Cost of Enterprise Survey (RLCES) published by the Ministry of Labor. This survey includes training costs only for establishments with 30 or more employees. 16 To estimate the training cost for small firms, we use the Survey on Actual State of SME (SASS) published by the Small and Medium Business Administration. We estimate the ratio of training costs of firms with 30 or fewer employees and firms with 30 or more employees as approximately Then, we apply this number to estimate training costs for firms with 30 or fewer employees in the RLCES. Total training costs are defined as the sum of the direct costs of training and the opportunity costs of training. Because data for the opportunity costs of training are not available, we assume that the direct costs of training are equal to the opportunity costs. The consulting costs are considered as firm-specific investments in organizational resources. To obtain expenditures on consulting at the aggregate level, we the use gross output of the consulting industry. The industry-level consulting costs are also estimated using the IO tables. The construction methods and sources for each type of intangible are summarized in Table A2 in the Appendix Intangible investment in Korea Intangible investment at the aggregate level Figure 1 compares intangible investment with two types of tangible investments: equipment and non-residential structures. Equipment investment is defined as the sum of machinery and transportation equipment, but it does not include software and mineral exploration because these investments are considered to be intangibles. As pointed out in the growth literature on Korea and other fast-growing Asian countries (Krugman, 1994; Young, 1995), Figure 1 confirms a high ratio of tangible investment to GDP. Throughout the period , the average ratios for equipment and non-residential structures were 13.5 and 14.9%, respectively. However, the total tangible investment has significantly declined since the late 1990s. In contrast, intangible investment has rapidly increased from 2.3% in 1981 to 8.3% in 2008, which is equivalent to approximately 72% of equipment 15 Knowledge accumulation from business practices in professional services, such as law and education, are not included in the Corrado et al. (2005, 2006) intangibles. 16 In particular, while the RLCES included firms with 30 or more employees before 1998, it has extended the coverage to firms with 10 or more employees since Our estimate is not much different from that of Corrado et al. (2005, 2006), showing that 69% of small establishments (i.e. with fewer than 50 employees) in the USA provided formal training in 1995 compared with medium-sized ( employees) and large (250 or more employees) establishments. 18 Our intangible data set based on the Corrado et al. (2005, 2006) definition is constructed by revising the industry-level data set reported in Chun et al. (2012). 55

7 H. Chun and M.I. Nadiri: Intangible Investment and Sources of Growth (%) Intangibles Equipment excluding intangibles Non-residential structures Year FIGURE 1. Intangible and tangible investments (as percentages of GDP) in Korea Notes: Intangibles include investments in computerized information, innovative property and economic competencies based on the work by Corrado et al. (2006). Equipment excludes intangibles such as software and mineral exploration. investment. Declines in the intangible investment are observed only for the 2 years of the Asian financial crisis. Combining both tangible and intangible investments shows no evidence of a significant decline in the total investment in the Korean economy during the past two decades. 19 This result suggests that the slowdown in physical investment after the crisis might be partly mitigated by the increase in intangible investment. 20 Both Figure 2 and Table 1 show the share of the five intangible components as a percentage of GDP. Although the intangible share has increased more than three times since 1981, the contribution of each intangible component to growth in the total intangible investment varied substantially. Investments in both computerized information and innovative property (scientific and non-scientific R&D) have increased rapidly, whereas investments in economic competencies (brand equity and firm-specific resources) have increased slowly. In particular, the share of economic competencies declined from the 1990s to the 19 Chun et al. (2008) document the sluggish investment in Korea after the Asian financial crisis in Kinkyo (2007) finds that Korea s lowered investment after the Asian financial crisis was caused by changes to underlying fundamentals, such as deterioration in terms of trade. 20 This may also suggest that there has been substitution away from tangible investment toward intangible investment in Korea. 56

8 The Japanese Economic Review 10 8 Computerized information Scientific R&D Non-scientific R&D Brand equity Firm-specific resources 6 (%) Year FIGURE 2. Intangible investment (as a percentage of GDP) in Korea Notes: The total intangible investment as a percentage of GDP in the figure consists of the sum of the five types of intangibles. TABLE 1 Intangible investment (as a percentage of value-added) in Korea, (1) (2) (3) (4) (4)/(2) Computerized information Innovative property Scientific R&D Non-scientific R&D Economic competencies Brand equity Firm-specific resources Total s, contrasting with the recent experience in the USA, where firm-specific resources, non-scientific R&D RLCES and computer software exhibited relatively faster growth, as reported in Corrado et al. (2005, 2006). As previously mentioned, this difference in the relative importance of the intangible categories might stem from the difference in industry structure. For example, the manufacturing share of GDP is significantly higher in Korea than in the USA, reiterating the importance of industry-level data for a cross-country intangibles analysis. 57

9 H. Chun and M.I. Nadiri: Intangible Investment and Sources of Growth TABLE 2 Intangible investments: Cross-country comparison Korea USA UK Japan Computerized information Innovative property Scientific R&D Non-scientific R&D Economic competencies Brand equity Firm-specific resources Total Notes: All numbers are percentage of value-added output averaged over US and UK data are obtained from the INTAN-Invest database ( Corrado et al., 2012). Japanese data is sourced from the Japan Industrial Productivity intangible database ( Miyagawa and Hisa, 2013). Table 2 presents the investment share of intangibles for Korea, the USA, the UK and Japan from 2001 to The intangible share as a percentage of output is the highest in the USA, whereas that in Korea is approximately 35% below the US level. Shares of computerized information and scientific R&D in Korea approximate those in the other three developed countries, but the share of economic competencies is much lower than those in the USA and the UK. Compared to the USA, the sources of low intangibles vary across countries. Shares of scientific R&D and brand equity in the UK as well as the share of economic competencies in Japan are lower than those in the USA Intangible investment at the industry level Table 3 presents industry-level intangible intensity, defined as the ratio of total intangible investment to industry value-added. 21,22 In Figure 3, industries are sorted by their intangible intensities. Both Table 3 and Figure 3 indicate a substantial variation in intangible intensity across industries. Among manufacturing industries, electrical and electronic equipment (10), transport equipment (12) and machinery equipment (9) show a high intangible intensity that is mainly due to high investment in innovative property (in particular, scientific R&D). Among non-manufacturing industries, information and communication (21), business services (22) and financial intermediation (19) show a high intangible intensity that is associated with high investments in computerized information and innovative property. In general, investment in innovative property is higher in the manufacturing industries, whereas 21 Omission of the own-account costs of software, design, and copyrights and license may affect the ranking of industry-level intangible intensity. However, there is little change in the set of intangible-intensive industries with the inclusion of own-account costs because most software-intensive industries are already included as intangible-intensive industries, and both design and copyrights and license costs account for a relatively small portion of the total intangible investment. We are grateful for an anonymous referee pointing this out. 22 This paper analyses the whole economy, including the private and public sectors, whereas most previous studies based on Corrado et al. (2005, 2006) intangibles examine the business sector. In our 27-industry classification, some industries, such as education (24) and health and social work (25), include both the private and the public sectors. However, our data set is not detailed enough to separate intangible investments in the private sector from those in the public sector for the two industries. 58

10 The Japanese Economic Review TABLE 3 Intangible investment by industry, Industry Total CI IP EC 1 Agriculture, forestry and fishing Mining and quarrying Food, beverages and tobacco Textiles and leather Wood, paper and printing Petroleum, coal and chemicals Non-metallic mineral products Metal and fabricated metal products Machinery equipment Electrical and electronic equipment Precision instruments Transport equipment Furniture and other manufacturing Electricity, gas and water supply Construction Wholesale and retail trade Restaurants and hotels Transport and storage Financial intermediation Real estate and renting Information and communication Business services Public administration and defense Education Health and social work Culture and entertainment Other service activities Intangible-intensive Non-intangible Manufacturing Non-manufacturing Notes: Numbers denote intangible investment as a percentage of industry value-added averaged over The top one-third of industries (9 out of 27 industries) based on the total intangible are in boldface. CI, computerized information; EC, economic competencies; IP, innovative property. that in computerized information is higher in the non-manufacturing industries, further confirming the industry-specific use of intangibles. We choose the top 9 industries that are above the average intangible intensity (approximately 5.5%) of the 27 industries. In addition, a substantial intensity gap exists between financial intermediation (the 9th industry) and furniture and other manufacturing (10th industry). We thus define these 9 industries as intangible-intensive industries, whereas we define the other 18 industries as non-intangible industries. Intangibleintensive industries are in boldface in the first column of Table 3 and are not characterized using a single type of intangible. For example, computerized information is intensively used in business services (22); innovative property (scientific R&D) in electric and electronic equipment (10); and economic competences (brand equity) in food, beverages and tobacco (3). Among the 9 intangible-intensive industries, 6 belong to the manufacturing sector, while 3 belong to the non-manufacturing sector. This implies that intangible-intensive industries belong to neither a particular sector nor consist of a particular type of intangible. 59

11 H. Chun and M.I. Nadiri: Intangible Investment and Sources of Growth Electrical and electronic equipment Transport equipment Machinery equipment Information and communication Precision instruments Food, beverages and tobacco Business services Petroleum, coal and chemicals Financial intermediation Furniture and other manufacturing Culture and entertainment Other service activities Textiles and leather Electricity, gas and water supply Public administration and defense Wood, paper and printing Real estate and renting Education Wholesale and retail trade Metal, Fabricated metal products Restaurants and hotels Non-metallic mineral products Construction Mining and quarrying Transport and storage CI IP EC Health and social work Agriculture, forestry and fishing FIGURE 3. Intangible investment by industry Notes: Industries are sorted based on their intangible investment as a percentage of industry value-added averaged over

12 The Japanese Economic Review TABLE 4 Intangible investment (as a percentage of value-added): Intangible-intensive versus non-intangible industries (1) (2) (3) (4) (4)/(2) Intangible-intensive industries Computerized information Innovative property Economic competencies Total Non-intangible industries Computerized information Innovative property Economic competencies Total Table 4 shows intangible intensity by period for two industry groups: intangibleintensive and non-intangible industries. 23 On average, the intensity of intangible-intensive industries is more than three times higher than that of their counterparts. Among the three components, the difference in innovative properties is the highest, whereas that in computerized information is the lowest. Despite this large gap in intangibles between the two groups, the gap has not widened over the sample period. In other words, no convergence exists among the two groups. This persistent gap again validates our industry grouping. 2.3 Output and input variables Output, labour and capital stock We obtain data for industry-level value-added, labour and physical capital from the Korea Industrial Productivity (KIP) database. The KIP database contains gross output, valueadded output, and five inputs of capital, labour, energy, materials and purchased services (KLEMS) following the guideline of the EU KLEMS project (Timmer et al., 2007) from 1970 to 2008 for 72 industries. 24 Following the industry classification in the National Accounts, we reclassify 72 industries into 27 industries. Based on the EU KLEMS manual, the KIP database has 18 categories of labour: two genders (male and female), three age groups (below 30, 30 49, and 50 or above), and three education levels (middle school or below, high school, and college or above). Labour services are calculated from a Tornqvist aggregation of these 18 types of labour hours with their compensation weights. The growth rate of labour composition (quality) is defined as the growth rate of labour services minus the growth rate of hours. Capital in the KIP database consists of 11 types of assets: (i) residential structures; (ii) non-residential structures; (iii) infrastructure; (iv) transport equipment; (v) computing 23 Value-added used in Table 4 does not include intangibles. However, the use of value-added, including intangibles, generates qualitatively similar results. 24 The KIP database is available at The EU KLEMS database compiled from the raw data set of the KIP database is available at

13 H. Chun and M.I. Nadiri: Intangible Investment and Sources of Growth equipment; (vi) communications equipment; (vii) other machinery and equipment; (viii) products of agriculture and forestry; (ix) other products; (x) software; and (xi) other intangibles. Capital services are calculated using a Tornqvist aggregation of these 11 types of assets. To estimate rental costs of the capital stock, we use depreciation rates from Pyo et al. (2007) and Timmer et al. (2007) and capital gains calculated from the 5-year moving averages of asset deflators. Through standard growth accounting, we use the KIP capital stock that includes intangibles included in the National Accounts, such as software and mineral exploration. To perform a growth accounting analysis with intangible assets, we define tangible and intangible capital stock. Tangible capital stock is KIP capital stock minus the two intangibles of software and mineral exploration. The intangible capital stock consists of computerized information, innovative property and economic competencies, as described in the previous section Intangible capital stock To capitalize intangible investments, it is necessary to measure both the deflator and the depreciation rate for each type of intangible. The software deflator is obtained from the software investment deflator published by the BOK. The initial capital stock of software is set to zero for the year Following Corrado et al. (2005, 2006), the depreciation rate of the software stock is set to The software capital stock is constructed using the perpetual inventory method. The GDP deflator is used for scientific R&D. The initial capital stock of R&D is set to zero for 1969 because the ratio of business sector R&D to GDP (0.07%) is very close to zero in The private R&D investment ratio to GDP was very low in the 1970s (approximately 0.2%) and has rapidly increased since the early 1980s. Among non-scientific R&D, we use the gross output deflator of the newspaper, publishing, movie, music, broadcasting, art and theater industries for copyright and licensing costs, and that of other businesses, including the design industry, for product, machine and architectural design. We also use the gross fixed capital formation deflator for mineral exploration. Following Corrado et al. (2005, 2006), the depreciation rates of both scientific and non-scientific R&D stock are set to We use the gross output deflators of advertising, education and consulting industries for the deflators of brand equity, firm-specific training and organizational structure, respectively. The depreciation rates are 0.60 for brand equity and 0.4 for firm-specific human and organizational resources. The source and method for the deflator and the depreciation rate for each type of intangible asset are summarized in Table A2 in the Appendix Adjustment for output and inputs When we treat intangibles as investments, we must adjust value-added and labour and redefine the tangible capital stock. The National Accounts treats most Corrado et al. (2005, 2006) intangible spending as expenses, except for mineral exploration and 25 The growth accounting results in Section 4 are qualitatively similar to those for alternative depreciation rates. 62

14 The Japanese Economic Review software. 26 The value-added in the KIP obtained from the National Accounts includes only software, mineral exploration, and both labour and capital costs in scientific R&D. Thus, we add all other intangibles to the value-added. We also adjust for the double counting of the labour cost of scientific R&D in the KIP labour data. Tangible capital is defined as the KIP capital stock exclusive of software, mineral exploration and capital costs of scientific R&D. 3. Methodology In this section, we estimate the contributions of two groups of intangible-intensive industries and non-intangibles industries to aggregate output and productivity growth. To do this, we employ the methodologies used in Stiroh (2002) and Jorgenson et al. (2007) and then develop them to determine the industry origin for aggregate output and productivity growth. 27 To investigate the industry origin of the aggregate output growth, we employ the production possibility frontier approach used in Jorgenson et al. (2007). The key feature of this approach is that the aggregate value-added does not assume that the price of output is the same across industries. 28 Thus, the aggregate real value-added growth rate (ΔlnY) is defined as ΔlnY = wiδln Yi, (1) i where ΔlnY i is the real value-added growth of industry i. Time subscripts are suppressed for convenience. w i is the 2-year averaged share of industry nominal value-added in the PYi, Yi aggregate as wi = and w i = 05. ( wi, t 1 + wi, t). P Y,i is the value-added deflator for PYi, Yi i industry i, and Δ denotes the difference between 2 adjacent years. Aggregate value-added growth is decomposed into contributions from the two groups of intangible-intensive and non-intangible industries as ΔlnY = wiδlnyi + wiδln Yi, (2) i I where I and N denote the intangible-intensive and non-intangible industries, respectively. The first and second terms on the right-hand side of Equation (2) are the contributions of the intangible-intensive and non-intangible industries to the aggregate real value-added growth, respectively. i N 26 The 1993 System of National Accounts (SNA) treats R&D spending as an expense, but the new 2008 SNA treats it as investment. 27 These methods are also applied by Eicher and Roehn (2007) and Oliner et al. (2007). 28 The production possibility frontier approach allows heterogeneous types of labour and capital inputs, but the price (i.e. rental price and wage rate) of each type of capital and labour is assumed to be identical for all industries. Thus, the aggregate capital and labour service growth rates do not require industry-level data and are calculated as the weighted averages of the heterogeneous types of capital assets and workers in the whole economy. 63

15 H. Chun and M.I. Nadiri: Intangible Investment and Sources of Growth Next, we consider the industries contributions to ALP growth. ALP growth is defined as Δln( Y H)= ΔlnY Δln H, (3) where H is hours worked. The industry-level labour productivity growth is similarly defined as Δln( Y H )= ΔlnY Δln H. (4) i i i i After aggregating industry-level labour productivity over industries and combining Equations (3) and (4), we have Δln( Y H)= wiδln( Yi Hi)+ wiδ Hi Δ H ln ln i = wδ ln( Y H )+ RH, i i i i i (5) where the first term on the right-hand side is a pure productivity effect equal to the weighted average of industry labour productivity growth rates, and the second term (R H ) is the reallocation of hours worked to high-productivity industries (Stiroh, 2002). 29 The aggregate hours growth approximately weights industries by their share of aggregate hours, and so, aggregate productivity rises if industries with value-added shares above their hours shares (i.e. industries with relatively high productivity levels) experience growth in hours (Oliner et al., 2007). Therefore, Equation (5) decomposes ALP growth into two sources: the direct contribution from industry-level productivity growth and the indirect contribution from resource reallocation toward high-productivity industries. Industries direct contributions to ALP growth are further decomposed into contributions of intangible-intensive and non-intangible industries, thus rearranging Equation (5) as follows: Δln( Y H)= wiδln( Yi Hi)+ wiδln ( Yi Hi)+ RH, (6) i I where I and N denote the intangible-intensive and non-intangible industries, respectively. The first and second terms in Equation (6) are the contributions of the intangible-intensive and non-intangible industries to ALP growth, respectively. Finally, we further decompose each industry group s direct productivity effect into three components as: (i) capital deepening; (ii) labour quality growth; and (iii) TFP growth. The first and second terms in Equation (6) are decomposed as follows: wiδln( Yi Hi)= wiνk, iδln( Ki Hi)+ wiνl, iδlnqi + wiδlntfp i i N i I i I i I i I (7a) 29 In this direct aggregation over industries, the price of each type of input is no longer assumed to be the same across industries, so that mobility of inputs is allowed across industries. 64

16 wiδln( Yi Hi)= wiνk, iδln( Ki Hi)+ wiνl, iδln Qi + wiδlntfp i, (7b) i N i N i N where ν K and ν L are 2-year averages of capital and labour cost shares, respectively. Q represents labour quality, and labour quality growth is defined as Δ ln Q i = Δ ln L i Δ ln H i. Industry-level capital and labour growth rates are defined as the weighted averages of the heterogeneous types of capital assets and workers, respectively. Capital and labour growth PKki,, Kki, are Δln Ki = wk, iδln Kk, i and Δln Li = wl, iδln Ll, i, where wki, = and k l PKki,, Kki, k PLli,, Lli, wli, =. Subscripts k and l denote the type of capital and labour, respectively. P L l Lli,, li, After treating spending on intangibles as investments, we consider a model with two types of capital: tangible and intangible capital. Equations (7a) and (7b) can be rewritten as follows: wδln( Y H )= wv i KT, iδln( KTi Hi)+ wv i KI, iδln( KIi Hi) i i A i i I i I i I + A i L, i ln i + i ln i I i I The Japanese Economic Review wv Δ Q wδ TFP (8a) wiδln( Yi A Hi )= wv i KT, iδln( KTi Hi)+ wv i KI, iδln( KIi Hi) i N i N i N + A wv i L, i ln Qi + wi ln TFPi A, i N i N i A i N Δ Δ (8b) where KT and KI are tangible and intangible capital, respectively. Adding intangibles requires adjustments for the value-added (Y A A i ), labour cost share ( v Li, ) and TFP (TFP A i ) variables Results Following the method used in the previous section, we decompose the aggregate output growth into contributions from intangible-intensive and non-intangible industries. Then, we decompose ALP growth into direct productivity contributions of intangible-intensive and non-intangible industries and the reallocation effect. Furthermore, we reexamine these analyses when expenditures on intangibles are treated as investments. 4.1 Industry origins of aggregate output growth Table 5 reports the average growth rate and share for each industry group by period and its contribution to the aggregate real value-added growth. The contribution is calculated using the average growth rate multiplied by the share of each industry group. In Panel A, the 30 Labour hours and industry share variables also change, but the same variable names are used for convenience. 65

17 TABLE 5 Contributions to aggregate real value-added growth: Intangible-intensive versus non-intangible industries Panel A. Conventional case (1) (2) (3) (4) (3) (2) (4) (3) Growth rates Whole economy Intangible-intensive Non-intangible Value-added share Intangible-intensive Non-intangible Contributions Intangible-intensive Non-intangible Panel B. Including intangibles H. Chun and M.I. Nadiri: Intangible Investment and Sources of Growth (1) (2) (3) (4) (3) (2) (4) (3) Growth rates Whole economy Intangible-intensive Non-intangible Value-added share Intangible-intensive Non-intangible Contributions Intangible-intensive Non-intangible Notes: Aggregate real value-added growth rates are averages of industry-level valued-added growth rates weighted using the industries nominal value-added share within the whole economy, averaged over 2 adjacent years. Contributions are calculated as industries real value-added growth rates multiplied by their value-added shares. Thus, the sum of the two industry groups contributions equals the value-added growth rate of the whole economy. whole economy grew, on average, at an annual rate of 6.64% during the period Approximately each half of the aggregate growth is attributable to intangible-intensive and non-intangible industries. 31 Contrary to the results for the whole sample period, the contribution of the intangible-intensive industries increased rapidly from 36.7% in the 1980s to 65.1% in the 2000s. The contribution of the intangible-intensive industries in the 2000s is almost two times larger than that of the non-intangible industries. This indicates that the industry origin of economic growth in Korea shifted from non-intangible industries toward intangible-intensive industries over the past three decades. In contrast, the contribution of the manufacturing industries to the aggregate output growth showed little change: 35.1% in the 1980s, 37.5% in the 1990s and 37.8% in the 2000s. 32 The contrasting results between the manufacturing and intangible-intensive 31 Among the intangible-intensive industries, the contribution of two IT-producing industries, i.e. electrical and electronic equipment (10) and information and communication (21), has increased more rapidly than that of other intangible-intensive industries, but the magnitude is relatively small because of the small output share. 32 Detailed results are available on request. 66

18 100 Intangible-intensive versus non-intangible The Japanese Economic Review Manufacturing versus non-manufacturing 80 (%) Intangible-intensive (left) manufacturing (right) Non-intangible (left) non-manufacturing (right) FIGURE 4. Contributions to the aggregate real value-added growth: Intangible-intensive versus manufacturing industries Notes: The three bars on the left present the relative contributions of intangible-intensive (light blue) and non-intangible industries to the aggregate real value-added growth. The three bars on the right are similarly defined for manufacturing and non-manufacturing industries. industries are presented in Figure The stable contribution of the manufacturing industries may reflect a divergence in the performance of two industry groups within manufacturing: intangible-intensive manufacturing industries have grown quickly, whereas non-intangible manufacturing industries have grown more slowly. In Panel B of Table 5, we report the results for the contributions of the two industry groups to the aggregate output growth when intangible expenditures are treated as investments. As previously mentioned, this treatment may change both the level and the growth of value-added. The aggregate growth rate increased from 6.64 to 6.78% in the period This treatment increases the aggregate growth rate by adding 0.16, 0.07 and 0.18 percentage points in the 1980s, 1990s and 2000s, respectively. However, the magnitudes were too small to reverse the slowdown in economic growth during the past three decades. The adjustments for the contributions of intangible-intensive and non-intangible industries are not the same because of differences in intangible investments in the two industry groups. As expected, the contribution of intangible-intensive industries to aggregate growth increased from 3.13 to 3.28%, whereas that of non-intangible industries showed little change. Similar to the results for the whole economy, the effects on the two industry groups 33 Consistently, the value-added share of the manufacturing sector in the whole economy hardly changed, from 24.9% in the 1980s to 25.8% in the 2000s, whereas the value-added share of the intangible sector increased significantly, from 24.5% in the 1980s to 37.0% in the 2000s. 67