Intangible Capital and Labor Productivity Growth: Panel Evidence for the EU from 1998–2005

Abstract

Using new international comparable data on intangible capital investment by business within a panel analysis between 1998–2005 in an EU country sample, a positive and significant relationship between intangible capital investment and labor productivity growth is detected. This relationship proves to be robust to a range of alterations. The empirical analysis confirms previous findings that the inclusion of business intangible capital investment in the asset boundary of the national accounting framework increases the rate of change of output per hour worked more rapidly. In addition, intangible capital is able to explain a significant portion of the unexplained international variance in labor productivity growth and becomes a dominant source of growth.

Originally published in: Felix Roth and Anna-Elisabeth Thum. Intangible Capital and Labor Productivity Growth: Panel evidence for the EU from 1998–2005. Review of Income and Wealth, Vol. 59, No. 3, 2013, pp. 486–508.

Felix Roth wishes to thank the participants of the INNODRIVE project during workshops in Brussels (2008), Vaasa (2008), Prague (2009), Rome (2009), Ljubljana (2010), and Berlin (2010), as well as the participants at the final INNODRIVE conference in Brussels (February 2011), the SEEK conference at the ZEW in Mannheim (March 2011), the CEGE research colloquium at the University of Göttingen (June 2011), and the EWEPA in Verona (June 2011) for valuable comments. In addition, he wants to thank Bart van Ark, Mary O’Mahony, Jonathan Haskel, Marcel Timmer, Felicitas Nowak Lehmann, Marianne Saam, Thomas Niebel, and two anonymous referees for valuable comments. He is grateful for a grant from the European Commission under the Seventh Framework Programme (FP SSH 2007 1) for the INNODRIVE project (Intangible Capital and Innovations: Drivers of Growth and Location in the EU, contract number 214576). He would also like to thank Massimiliano Iommi, Cecilia Jona-Lasinio, and Stefano Manzocchi for their contribution of the variables on intangible capital, and Raf van Gestel for valuable research assistance. Preliminary versions of this paper were published as a CEPS Working Document 335 in September 2010 and INNODRIVE Working Paper 3 in September 2010.

1 Introduction

As highly developed economies transform more and more into knowledge economies, the input of intangible capital becomes vital for the future competitiveness of their economies (Corrado et al., 2005; World Bank, 2006), as well as the competitiveness of their firms (Teece, 1998, p. 76; Eustace, 2000, p. 6; Lev & Radhakrishnan, 2003, 2005). Although further refinement of the concept of intangible capital is still clearly needed, the overall measurement of the different dimensions of intangible capital has largely improved, and past assessments, which have called into question the possibility of measuring certain dimensions of intangible capital, seem to have been too pessimistic.¹ Nevertheless, it remains an open question as to which range of intangible capital indicators should be incorporated into the asset boundary (Hill, 2009; Stiglitz et al., 2009) and which dimensions should be included in a definition of intangible capital (World Bank, 2006).

This contribution focuses on intangible capital investment by businesses. Using international comparable data on business intangible capital investment generated within the INNODRIVE project² (INNODRIVE, 2011; Jona-Lasinio et al., 2011), this contribution aims to present new econometric evidence about the impact of investments in business intangible capital on labor productivity growth in the business sector. As envisaged in the INNODRIVE framework (Jona-Lasinio et al., 2011), the dimensions of business intangible capital were generated along the framework originally proposed by Corrado et al. (2005, 2009). However, as the author wholly shares the view of the World Bank (2006), namely that the dimensions of human and social capital should also be classified as intangible capital, the dimensions of human and social capital have been included in the Total Factor Productivity (TFP) term of the utilized model specification.

2 Theoretical Links between Business Intangible Capital and Labor Productivity Growth

2.1 Theoretical Relationship between Intangible Capital and Labor Productivity Growth

The importance of Business Enterprise Research & Development (BERD) and innovation was explicitly recognized in the “Lisbon process” and has been adopted by the European 2020 strategy for smart, sustainable, and inclusive growth (European Commission, 2010). Although the importance of business investment in Research & Development has already been widely acknowledged—by policy-makers and in economic theory—our knowledge of the contribution of business intangibles to labor productivity growth is still incomplete. Generating a wider concept of innovation and focusing on the issue of a possible revision of the national accounting framework, Corrado et al. (2005) have grouped various items that constitute the knowledge of the firm into three basic categories: 1) computerized information, 2) innovative property, and 3) economic competencies.

Whereas computerized information includes knowledge that is contained in computer programs and computerized databases, innovative property includes the scientific knowledge embedded in patents, licenses, and general know-how, as well as “the innovative and artistic content in commercial copyrights, licences and designs” (Corrado et al., 2005, pp. 23–26). Corrado et al. (2005, p. 28) define the economic competencies category of intangibles as “the value of brand names and other knowledge embedded in firm-specific human and structural resources”. It includes expenditures on advertising, market research, firm-specific human capital, and organizational capital. These measures indicate that the potential of intangible capital for stimulating productivity growth lies in the provision of knowledge, an increase in the selling potential of a product, and the development of processes and a productive environment for the actual physical production of a good, or as van Ark et al. (2009, p. 63) stress, that products and services are becoming more knowledge-intensive.

While the positive relationship between computerized information, here in particular via an interaction effect with organizational capital (Brynjolfsson et al., 2002), and certain dimensions of innovative property (scientific R&D) (Lichtenberg, 1993; Coe & Helpman, 1995; Park, 1995; Guellec & van Pottelsberghe de la Potterie, 2001) on labor productivity growth have already been extensively discussed it seems important to stress once more the theoretical importance of the single dimensions of economic competencies, namely brand names, firm-specific human capital, and organizational capital.

In theory, brand names should positively affect labor productivity growth since an important aspect of today’s products is the “image” attached to them. Cañibano et al. (2000) argue that the ownership of a brand that is appealing to customers permits a seller to acquire a higher margin for goods or services that are like those offered by competitors. As the consumer’s choice among the products of competing firms is often motivated by a perception of reliability and trustworthiness, the development of this image or brand has to be considered pivotal in the yield of future benefits. Expenditure on market research comprises, next to expenditure on advertising, an important part of the investment in brand equity.

Firm-specific human capital is another important asset of a firm. Cañibano et al. (2000) stress that a firm with more competent employees is likely to earn higher profits than competitors whose workers are less skilled. In this regard, Abowd et al. (2005) argue that the value of companies will increase if the quality of their human resources increases.

In addition to the “image” projected by a firm or a product and the quality of the training of workers, the management of a production process involving highly technological physical capital has also become important. As goods become more and more sophisticated, production processes become more complex and the organizational capital of a firm becomes crucial. Lev and Radhakrishnan (2005, p. 75) define organizational capital as “an agglomeration of technologies – business practices, processes and designs and incentive and compensation systems – that together enable some firms to consistently and efficiently extract from a given level of physical and human resources a higher value of product than other firms find possible to attain.” Organizational capital is seen by them (Lev & Radhakrishnan, 2003, 2005) as the only competitive asset truly owned by a firm, while the others are tradable and thus available to any firm that wants to invest in them.

2.2 The Treatment of Intangible Expenditures

Although, as argued above, the existing literature widely recognizes the importance of the various dimensions of business intangible capital for the enhancement of growth, in contemporary accounting practice, intangibles are treated as intermediate expenditures and are not classified as investments in Gross Fixed Capital Formation (GFCF). This situation has improved with the inclusion of software, mineral exploration, and entertainment, literary, and artistic originals within the asset boundary of the national accounts. Moreover, for innovative properties, such as scientific R&D investment, national accounts have started to set up satellite accounts.

In the economic literature, this situation has gradually improved as a result of Corrado et al.’s (2005) approach to capitalize the above-mentioned intangibles. Utilizing standard intertemporal capital theory, they define investments as “any use of resources that reduces current consumption in order to increase it in the future” (Corrado et al., 2005, pp. 17–19) and treat intangibles, in contrast to the national accounting framework, as investments rather than as intermediate goods; thus, including it in the asset boundary rather than netting it out. Corrado et al. (2009, pp. 663–66) model the impact of capitalizing intangible assets on the sources-of-growth model as follows:

$$ {g}_Q(t)={s}_L(t){g}_L(t)+{s}_K(t){g}_K(t)+{s}_R(t){g}_R(t)+{g}_A(t) $$

(5.1)

where Q is GDP expanded by the flow of new intangibles, g_X(t) denotes the rate of growth of the respective variables, and s_X(t) represents the input shares. L is labor, K is the tangible capital stock, R is the intangible capital stock, and A is the TFP term.

3 Estimates of Intangible Capital Investment

Following Corrado et al. (2005, 2009), the INNODRIVE macro approach (INNODRIVE, 2011; Jona-Lasinio et al., 2011) has classified business intangible capital investment into three groups: 1) computerized information, 2) innovative property, and 3) economic competencies. Moreover, it differentiates two “old” intangible capital variables³—1) software and 2) mineral exploration and copyright and license costs⁴—in to eight new intangible capital variables: 3) scientific R&D, 4) new product development in the financial services industry, 5) new architectural and engineering designs, 6) advertising, 7) market research, 8) firm-specific human capital, 9) own-account, and 10) purchased component of organizational capital.

The first group, computerized information, contains: 1) computer software. Computer software was measured by using data from the EUKLEMS⁵ project, as well as official national account data and the use table from the supply and use framework (Jona-Lasinio et al., 2011, pp. 35–36).

The second group, innovative property, contains the following variables: 2) scientific R&D, 3) new product development in the financial services industry, 4) new architectural and engineering designs, and 5) mineral exploration and copyright and license costs.

To measure investment in scientific R&D, data on expenditure on R&D by businesses were retrieved from Eurostat. To avoid double-counting of software investment and investment in the development of new products within the financial services industry, data for the subsector K72 (computer and related activities) and sector J (financial intermediation) were subtracted from the R&D expenditure. In accordance with Corrado et al. (2005), expenditure in scientific R&D was considered a 100% investment in intangible capital (Jona-Lasinio et al., 2011, pp. 37–39).

Mineral exploration and copyright and license costs were measured with the help of data from the national accounts and the use tables from the supply and use frameworks (Jona-Lasinio et al., 2011, p. 39).

Investment in new architectural and engineering designs has been measured using data from the national accounts (Jona-Lasinio et al., 2011, pp. 40–41).

Investment in new product development in the financial services industry was measured, according to Corrado et al. (2005), on the basis of 20% of total intermediate spending for intermediate inputs by the financial intermediation industry, which is defined as excluding insurance and pension funding (NACE J65) (Jona-Lasinio et al., 2011, pp. 41–42).

The third group, economic competencies, contains the following variables: 6) advertising expenditure, 7) expenditure on market research, 8) firm-specific human capital, 9) own-account development of organizational structure, and 10) purchased organizational structure.

To measure investment in advertising, a private data source (Zenith Optimedia) was used.⁶ Zenith Optimedia data report the expenditure on advertising in newspapers and other media which should capture the purchased and own-account expenditure. Following Corrado et al. (2005), who followed Landes and Rosenfield (1994), only 60% of the actual expenditure was considered investment (Jona-Lasinio et al., 2011, pp. 42–44). In order to measure the investment in market research, data on the turnover of subsector K7413 (Market Research and Public Opinion Polling) from Eurostat’s Structural Business Statistics on Business Services were taken. Following the approach of Corrado et al. (2005), the prevalence of own-account market and consumer research was estimated by doubling the estimate of the data on market research (Jona-Lasinio et al., 2011, pp. 44–46).

Data on firm-specific human capital were taken from Eurostat’s Continued Vocational Training Survey. This variable is a measure of the training expenditure by enterprises and is computed as the cost of continued vocational training courses as a percentage of total labor costs multiplied by employee compensation. This training expenditure was considered a 100% investment in intangible capital. The estimation method is applied at the industry level to guarantee that the compositional changes of industries are taken into account. The measures are then aggregated to obtain data on the national level (Jona-Lasinio et al., 2011, pp. 46–49).

Organizational capital is measured by the own-account and purchased investment in the organizational structure of a firm. Data on the own-account component of organizational capital are taken from the EU Structure of Earnings Survey (in 2002) and the EU Labor Force Survey. Own-account organizational capital is represented by the compensation of the management. Manager compensation is computed as the manager compensation share multiplied by the compensation of all employees. The manager compensation share is the share of gross earnings of managers over the gross earnings of all employees. Following Corrado et al. (2005), it was assumed that 20% of manager compensation is spent on investment in the organizational structure of a firm. Data on the purchased component of organizational capital are taken from Eurostat’s Structural Business Statistics on Business Services and the FEACO⁷ Survey of the European Management Consultancy Market. Purchased organizational capital is represented by management consultancy fees. In order to compute the purchased component of organizational capital, the nominal gross output or turnover of NACE 7414 (business and management consultancy activities) was used. It was assumed that 80% of business sector expenditure is considered an investment (Jona-Lasinio et al., 2011, pp. 49–54, 61–62).

As can be inferred by the INNODRIVE macro dataset (INNODRIVE, 2011), all investment rates were constructed for the non-farm business sector and thus for NACE sectors c–k + o.⁸

4 Previous Empirical Results

Several empirical studies have tried to estimate the importance of business intangible assets for labor productivity growth. Up to now, all existing studies have utilized a growth accounting methodology.⁹ There is an extensive body of literature studying intangible capital investment both at the micro (firm) level (e.g., Brynjolfsson & Yang, 1999; Webster, 2000; Brynjolfsson et al., 2002; Cummins, 2005; Lev & Radhakrishnan, 2005) and at the macro (national) level. A detailed summary of the microeconomic studies, as mentioned above, is not undertaken here as this analysis focuses solely on the macroeconomic level.

Recent literature on the macroeconomic level has highlighted three important lines of results once intangibles are capitalized: 1) the share of intangible capital investments as a percentage of GDP or market sector gross value added (MGVA), 2) the contributions of intangible capital on output growth within an accounting framework, and 3) the growth acceleration. Table 5.1 provides an overview of these three dimensions in the literature most recently published in this field. Corrado et al. (2005) find for the United States that the investment in business intangibles represented 10%–12% of existing GDP between 1998 and 2000 and approximately 13% of non-farm business output in 2003 (see Fig. 2 of Corrado et al., 2009, p. 673). In line with Corrado et al., Nakamura (2010, p. S138), analyzing a timeframe from 1959 to 2007, finds that investments in intangibles become as important as investment in tangibles in the US around 2000. Marrano et al. (2009) show that in the United Kingdom the private sector spent a sum equivalent to 13% of adjusted MGVA on business investment in intangibles in 2004. A working paper by Jalava et al. (2007) finds that the Finnish investment in non-financial business intangibles was 9.1% of unrevised GDP in 2005. Fukao et al. (2009) estimate 11.1% of GDP was invested in intangible capital in Japan in 2000–2005. According to Hao et al. (2009), Germany and France invested 7.1% and 8.8%, respectively, and Italy and Spain invested 5.2% in intangibles in the market sector over GDP in 2004. A working paper by Van Rooijen-Horsten et al. (2008) finds an investment share for intangibles of 8.3% of GDP when general government industry is excluded for the Netherlands in 2001–2004. Van Ark et al. (2009) find intangible investment shares in the market sector of 6.5% in Austria, 6.5% in the Czech Republic, and 7.9% of GDP in Denmark in 2006.¹⁰ Investments for the UK, Germany, France, Italy, Spain, and the US are similar to those in the other papers. Edquist (2011) finds that in Sweden, total business investment in intangibles was equivalent to 10% of GDP, or approximately 16% of the business sector gross value added (GVA) in 2006.

Table 5.1 Summary of existing empirical studies

Second, when looking at the contribution of intangible capital to output growth, Corrado et al. (2009) find for the US that 27% of labor productivity growth in 1995–2003 is explained by intangible capital. Marrano et al. (2009) find that 20% of labor productivity growth is accounted for by intangible capital deepening in 1995–2003 in the United Kingdom. Jalava et al. (2007) find that intangible capital accounts for 16% of labor productivity growth in 1995–2000 and for 30% in 2000–2005 in the Finnish case. Fukao et al. (2009) show that intangible capital explains 27% of the Japanese growth rate in 1995–2000 and 16% in 2000–2005. Hao et al. (2009) find that intangibles account for 31% of labor productivity growth in Germany, 37% in France, 59% in Italy, and 64% in Spain. Van Ark et al. (2009) find that in Germany, France, Italy, Spain, Austria, the Czech Republic, and Denmark, intangible capital accounts for, respectively, 21%, 24%, 41%, 26%, 23%, 15%, and 34% of labor productivity growth. Finally, Edquist (2011) finds for Sweden that the contribution of intangible capital drops from 41% in the period 1995–2000 to 24% from 2000 to 2006.

Third, overall the capitalization of intangible capital accelerates productivity growth. Detailed results of the growth acceleration values will be discussed in comparison with the INNODRIVE data in Sect. 6.

5 Model Specification, Research Design, and Data

5.1 Model Specification

The model specification within this contribution follows an approach by Benhabib and Spiegel (1994), which Temple (1999) coined “cross-country growth accounting” or “growth accounting with externalities” (p. 124). The model by Benhabib and Spiegel (1994) differs from the framework of a traditional single growth accounting methodology as depicted by Eq. (5.1) by two components. First, the output elasticities are estimated, rather than imposed. Second, part of the model can be designed to explain the international variance in TFP growth.

One advantage of the utilization of stock data for tangible and intangible capital, in contrast to other econometric growth estimations (e.g., Mankiw et al., 1992), is that one is able to estimate the production function without the term for initial efficiency and thus without “the complexities of dynamic panel data models, provided that TFP growth is unrelated to initial income” (Temple, 1999, p. 125). Like the approach by Fleisher et al. (2010), the empirical model by Benhabib and Spiegel (1994) is applied in a panel context.

Following the theoretical framework of Corrado et al. (2009) as depicted in Eq. (5.1), Benhabib and Spiegel’s (1994) model specifications are expanded by intangibles. The starting point for the estimation is then an augmented Cobb-Douglas production function,

$$ {Q}_{i,t}={A}_{i,t}{K}_{i,t}^{\alpha}{L}_{i,t}^{\gamma}{R}_{i,t\kern0.5em}^{\beta}{\varepsilon}_{i,t}, $$

(5.2)

where intangible capital R is added to the conventional production function because it is treated as investments rather than intermediate expenses. Q_i,t is GVA (non-farm business sectors c–k + o excluding real estate activities) expanded by the investment flows in intangible capital in country i and period t. Similar to Eq. (5.1), K is the tangible capital stock, L is labor, and A is TFP. Assuming constant returns to scale and rewriting the Cobb-Douglas production function in intensive form, the following equation is obtained: with lower case letters indicating variables in terms of total hours worked. If differences in natural logarithms are taken, the annual growth relationship can be expressed as follows:

$$ {q}_{i,t}={A}_{i,t}{k}_{\mathrm{i},t}^{\alpha}{r}_{i,t\kern0.5em}^{\beta}{\varepsilon}_{i,t} $$

(5.3)

with lower-case letters indicating variables in terms of total hours worked. If differences in natural logarithms are taken, the annual growth relationship can be expressed as follows:

$$ \left(\ln {q}_{i,t}-\ln {q}_{i,t-1}\right)=\left(\ln {A}_{i,t}-\ln {A}_{i,t-1}\right)+\alpha \left(\ln {k}_{i,t}-\ln {k}_{i,t-1}\right)+\beta \left(\ln {r}_{i,t}-\ln {r}_{i,t-1}\right)+{u}_{i,t} $$

(5.4)

where

$$ {u}_{i,t}=\ln {\varepsilon}_{i,t}-\ln {\varepsilon}_{i,t-1}. $$

(5.5)

Unless the TFP growth term is estimated, the estimation of this equation will be biased. (Benhabib & Spiegel, 1994; Temple, 1999). Therefore, using a similar but extended approach to Benhabib and Spiegel (1994), a model for (lnA_i,t −  ln A_i,t − 1) is specified as follows:

$$ \left(\ln {A}_{i,t}-\ln {A}_{i,t-1}\right)=c+g{H}_{i,t}+m{H}_{i,t}\frac{\left({Q}_{\max, t}-{Q}_{i,t}\right)}{Q_{i,t}}+n\left(1-{ur}_{i,t}\right)+p\sum \limits_{j=1}^k{X}_{j,i,t}+{cd}_{i,t=2001} $$

(5.6)

where the constant term c represents exogenous technological progress, the level of human capital (gH_i,t) reflects the capacity of a country to innovate domestically, the term $m{H}_{i,t}\frac{\left({Q}_{\max, t}-{Q}_{i,t}\right)}{Q_{i,t}}$ proxies a catch-up process, the term n(1 − ur_i,t) takes into account the business cycle effect,¹¹ the term p$\sum \limits_{j=1}^k{X}_{j,i,t}$ is a sum of k extra policy variables which could possibly explain TFP growth, and cd_i,t = 2001 is a crisis dummy to control for the economic downturn in 2001 after the IT (information technology) bubble burst in the year 2000 and the 9/11 attack in 2001. Inserting Eq. (5.6) into Eq. (5.4) provides the baseline model to be estimated within the econometric estimation in Sect. 7:

$$ \left(\ln {q}_{i,t}-\ln {q}_{i,t-1}\right)=c+g{H}_{i,t}+m{H}_{i,t}\frac{\left({Q}_{\max, t}-{Q}_{i,t}\right)}{Q_{i,t}}+n\left(1-{ur}_{i,t}\right)+p\sum \limits_{j=1}^k{X}_{j,i,t}+{cd}_{i,t=2001}+\alpha \left(\ln {k}_{i,t}-\ln {k}_{i,t-1}\right)+\beta \left(\ln {r}_{i,t}-\ln {r}_{i,t-1}\right)+{u}_{i,t} $$

(5.7)

5.2 Research Design

The econometric analysis covers 13 EU-27 countries for a time period from 1998 to 2005. The countries included are Germany, France, Italy, the United Kingdom, Spain, Sweden, Denmark, Finland, Austria, Ireland, the Netherlands, Slovenia, and the Czech Republic.¹² Although the INNODRIVE Macro approach (INNODRIVE, 2011) has managed to produce a complete set of intangible capital variables for all 27 EU countries plus Norway, the following econometric analysis had to be restricted to a maximum of 13 EU countries, due to a lack of sectoral tangible capital input data within the EUKLEMS database. With the 13 countries and the given timeframe, this leaves the analysis with 98 overall observations (the Czech Republic and Slovenia, each miss three time observations from 1998 to 2000). Following existing empirical studies (e.g., Bassanini & Scarpetta, 2001), annual growth rates over the time period 1998–2005, rather than long-term growth rates from 1998 to 2005, have been chosen to be able to apply a panel data analysis.¹³ With intangible capital stocks having been calculated for the period 1995–2005, the econometric analysis was restricted to a time-frame of 1998–2005 as the calculation of capital services (as can be depicted in Supplementary Appendix A2) needed intangible capital stock information from 1995 to 1997 to produce intangible capital service growth for the year 1998. The whole research design applies to non-farm business sectors c–k + o. The output measure is GVA for the non-farm business sectors c–k + o excluding real estate activities. Tangible and intangible are non-farm business investments c–k + o. Tangible capital investments excluded residential capital.

5.3 Data

The data for the following econometric analysis were taken from various different data sources, as described below.

• Data on the single components of intangible capital were taken from the INNODRIVE macro dataset (INNODRIVE, 2011). The INNODRIVE macro data to a large extent conforms to EUKLEMS data, and GFCF data are provided for all intangible assets and all EU-27 countries plus Norway.

• Data on GVA (non-farm business sectors excluding real estate activities), tangible capital stocks, capital compensation, gross fixed tangible capital investments, tangible investment price indices, labor input (number of hours worked per persons engaged), and depreciation rates for tangible capital were calculated from the EUKLEMS database. Tangible capital included communications equipment, computing equipment, total non-residential investment, other machinery and equipment, transport equipment, and other assets, but excluded residential capital.

• Human capital is measured as the “percentage of population who attained at least upper secondary education”, which is taken as a proxy for the inherent stock of human capital. These data are provided by Eurostat.

• The variable rule of law is taken from the Worldwide Governance Indicators project (Kaufmann et al., 2010). The World Bank (2006) uses this indicator as a proxy for generalized trust, an important indicator of social capital (Roth, 2009).

• The data on openness to trade are retrieved from the Penn World Table Version 6.2 (Heston et al., 2006).

• The data on unemployment rates, the stocks of foreign direct investment (FDI) (as a % of GDP), total government expenditure (as a % of GDP), total expenditure on social protection (as a % of GDP), total general government expenditure on education (as a % of GDP), inflation rates (annual average rate of change in Harmonized Indices of Consumer Prices), taxes on income (as a % of GDP), and the stock market capitalization (as a % of GDP) are taken from Eurostat.

5.4 A Note on the Construction of Intangible Capital Stocks

Intangible capital stocks for the 11 EU-15 countries for the time period 1995–2005 were constructed by applying the perpetual inventory method (PIM) to series of intangible capital investments going back to 1980 and using the depreciation rates, δ_R, suggested by Corrado et al. (2009): 20% for scientific R&D, new architectural and engineering designs, and new product development in the financial services industry; 40% for own and purchased organizational capital and firm-specific human capital; and 60% for advertising and market research. In accordance with EUKLEMS for software, a depreciation rate of 31.5 was used.¹⁴ Intangible capital stocks for the two transition countries, Slovenia and the Czech Republic, were calculated by applying the PIM to investment flows from 1995 to 1999, constructing stocks for the 6-year period 2000–2005. For the calculation of the intangible capital stock R_t the PIM takes the following form:

$$ {R}_t={N}_t+\left(1-{\delta}_R\right){R}_{t-1} $$

(5.8)

which assumes: 1) geometric depreciation, 2) constant depreciation rates over time, and 3) depreciation rates for each type of asset are the same for all countries. The real investment series for N_t use a GVA price deflator which is the same for all intangible assets.

5.5 A Note on the Construction of Intangible and Tangible Capital Services

Data on intangible capital services were generated according to the literature by Oulton and Srinivasan (2003) and Marrano et al. (2009), and are consistent with the EUKLEMS approach (Timmer et al., 2007). This literature argues that rather than using a wealth measure for capital (like the capital stock), it is crucial to derive the flow of services a capital stock can provide to production. An overview of the technical steps in how intangible and tangible capital services were constructed is given in Appendix A2.

6 Descriptive Analysis

Table 5.A1 in Appendix A1 shows the descriptive statistics of the data utilized over the 13 EU countries and over the time period 1998–2005. Annual labor productivity growth increases by 0.10 (from 2.3 to 2.4) percentage points, or by 4.4%, when taking into account the contribution of intangibles in the measure of GVA. This value is smaller than most values reported by the existing literature as depicted in Table 5.1, which find that productivity growth increased by 11% in Germany, 9%–14% in France, 11%–37% in Italy, 31%–40% in Spain, 19% in Austria, 37% in Denmark, and 13% in the UK when adding intangible capital to the asset boundary. The mean value of 4.4% however is larger than the value for Sweden of −2.3%, for Finland of 2.1%, and the Czech Republic with 2.2%. The descriptive results in Table 5.A1 also show that the services of the intangible capital stock (4.1%) grow on average faster than the services of the tangible capital stock (3.3%).

Figure 5.1 shows the business intangible capital investments over GVA for the ten single intangible assets and the three core dimensions as described in Sect. 3. Overall business intangible capital investments differ considerably across the 13 EU countries used in the econometric estimation.¹⁵ Sweden ranks first, with an overall investment of 13.6% of GVA. As can be inferred from Table 5.1, this is within the range of the values pointed out by Edquist (2011), who finds investment rates of 10% over GDP and 16% over business GVA. Sweden is followed by the UK, which has an investment rate of intangible capital of 12.4%. As depicted in Table 5.1, this is quite similar to the value for the UK, as measured by Marrano et al. (2009), of 13%. The UK is followed by the second largest economy, France, with an investment rate of 11.6%. France’s investment rate is in the range of the results of Hao et al. (2009), who report that it invests 8.8% of GDP. The two large Mediterranean economies of Italy and Spain are situated within the two last positions in the distribution. Spain invests 6.3% on intangibles and Italy around 7.0% on intangible capital. This again fits with the reported investment rate by Hao et al. (2009) of 5.2% and 5.2%, and van Ark et al. (2009), with investment rates of 5.5% and 5.0% over GDP. The largest European economy—Germany—is positioned in the middle of the distribution, with an overall investment of 9.6%.

Fig. 5.1

Business intangible investment (as a percentage of GVA) in 13 EU countries, 1998–2005 Notes: The left bar chart for each country shows all ten single intangible capital items and the right bar chart indicates the three dimensions: Computerized Information, Innovative Property, and Economic Competencies. All variables in the graph are compared to GVA (non-farm business sector c–k + o excluding real estate activity). Soft = Software; Arch = New architectural and engineering designs; NFP = New product development in the financial services industry; R&D = Scientific research and development; Other NA = Other national account intangibles (mineral exploration and copyright and license costs), OKP = Organizational capital (purchased component); OKO = Organizational capital (own-account component); FSHK = Firm-specific human capital; ADV = Advertising; MKTR = Market research; CompInf = Computerized Information; InnoProp = Innovative Property and EconComp = Economic Competencies.

Source: INNODRIVE data (INNODRIVE, 2011).

This is in accordance with van Ark et al. (2009) and Hao et al. (2009), who find investment rates of 7.2% and 7.1%, respectively. On average the included 13 EU countries invest 9.9% in intangibles over GVA.

The right-hand bar charts in Fig. 5.1 make clear that overall the largest shares of intangibles are in either economic competencies or innovative properties, and only a small part of the investment consists of investment in software. In order to identify the distribution between the three dimensions in each country more clearly, Fig. 5.2 shows a scatterplot between the two larger dimensions, innovative property and economic competencies. These are the countries that can be classified as being highly innovative and investing strongly into economic competencies, and which can be detected in the upper right corner, namely Sweden, Slovenia, and France. In addition, France and Sweden score high on computerized information. On the other hand, these are the economies that invest more in innovative properties than in economic competencies, such as Denmark, Finland, and Germany. The third category includes those countries that score low on innovative property and high on economic competencies: the UK, the Netherlands, and the Czech Republic.¹⁶ The fourth category includes those countries that score low on both dimensions: Austria, Ireland, Italy, and Spain. Overall, Fig. 5.2 clarifies that the sole focus of the Europe 2020 strategy (European Commission, 2010) on R&D investment seems to be too narrow in view of the significant investments in economic competencies.

Fig. 5.2

Scatterplot between innovative property and economic competencies (as a percentage of GVA), 1998–2005

Notes: EU-13 mean value of all 13 countries; es = Spain, it = Italy; fr = France; fi = Finland; de = Germany; dk = Denmark; at = Austria; ie = Ireland; cz = Czech Republic; nl = the Netherlands; si = Slovenia; uk = United Kingdom; ♦ Low Computerized Information; ● High Computerized Information.

Source: INNODRIVE data (INNODRIVE, 2011).

Figure 5.3 shows a comparison between business investments in intangible capital and tangible capital as it is used in the econometric estimation. Interestingly, one detects that when including intangible capital investments, the average investment for the 13 EU countries is over 30% of GVA. This value is significantly higher than when solely considering tangible investments. Values on top of the bar charts depict the ratio of intangible/tangible capital investment. Ratios close to but still less than 1 indicate that intangible capital investment is almost as large as tangible capital. France, Sweden, the UK, the Netherlands, and Finland have reached ratios of larger than 0.6, with France already having reached a value of 0.78. It thus seems sound to conclude that some EU countries have started to converge towards the US, for which Nakamura points out that investment in intangible capital has become as large as investment in tangible capital. In the transition countries Slovenia and the Czech Republic, and the Mediterranean countries Spain and Italy, tangible capital still dominates investments, with ratios below 0.4 (ratios of 0.37, 0.23, 0.29, and 0.32, respectively).

Fig. 5.3

Business tangible and intangible investment (as a percentage of GVA) in 13 EU countries, 1998–2005

Notes: CT = Communications equipment; IT = Computing equipment; OCon = Total nonresidential capital investment; OMach = Other machinery and equipment; Other = Other assets; TraEq = Transport equipment; IC = Intangible capital. Residential capital has been excluded.

Source: INNODRIVE data (INNODRIVE, 2011) and EUKLEMS data.

7 Econometric Analysis

Without a lagged initial income term on the left-hand side, the baseline model in Eq. (5.7) may be estimated without the complexities of a dynamic panel analysis.¹⁷ Thus, when estimating Eq. (5.7), the standard methods of panel estimation are fixed effects or random-effects. The fixed effects are calculated from differences within each country across time; the random-effects estimation, in contrast, incorporates information across individual countries as well as across periods.¹⁸ The major drawback with random-effects is, although being more efficient, they are consistent only if the country-specific effects are uncorrelated with the other explanatory variables. A Hausman specification test can evaluate whether this independent assumption is satisfied (Hausman, 1978; Forbes, 2000, p. 874). The Hausman test applied here indicates that a random-effects model can be used.¹⁹ In addition, to control for potential cross-sectional heteroskedasticity, a robust VCE estimator has been utilized.²⁰ As highlighted in Sect. 5.2, the random-effects estimation uses 13 countries with a total of 98 observations. It is a balanced panel, with two countries (the Czech Republic and Slovenia) missing three time observations from 1998 to 2000. Regression 1 in Table 5.2 shows the estimation results when estimating the traditional production function (without the inclusion of intangible capital and specifically excluding software from the tangible capital investment).²¹ The overall R-square value is 0.40, with a within R-square value of 0.20 and a between R-square value of 0.63. The growth of tangible capital services is positively associated with labor productivity growth and has a coefficient of 0.47, explaining a 65% share of labor productivity growth.²² Regression 2 shows the same model specification when including intangible capital investment. When including intangible capital investment in the asset boundary, the overall R-square value increases by 11% points to 0.51, the within R-square value increases by 16% points to 0.36, and the between R-square value increases by 9% points to 0.72. Growth of intangible capital services is positively related to labor productivity growth, with a coefficient of magnitude 0.29. With this magnitude, intangible capital is able to explain around 50% of labor productivity growth. As can be inferred from Table 5.1, a value of 50% is in close range to the results of the growth accounting results for the relevant country cases presented in this contribution; in particular, the results from Hao et al. (2009), who find 59% for France, 59% for Italy, and 64% for Spain. It is larger than the value of Marrano et al. (2009) of 20%, however. Once including intangible capital, the impact of tangible capital diminishes to 40%. TFP then changes from 35% to 10% and is thus diminished by 25%. As intangible and tangible capital are able to explain 90% of labor productivity, the finding by Corrado et al. (2009) that capital deepening becomes the dominant source is sustained.

Table 5.2 Intangibles and labor productivity growth; random-effects estimations

In order to assess which dimensions of intangible capital services are the main drivers of the positive relationship between intangible capital and labor productivity growth, regression 3 includes the three dimensions of 1) computerized information, 2) innovative property, and 3) economic competencies, instead of the overall intangible capital index. Interestingly, the main driver is not innovative property as expected from the guidelines of the Europe 2020 strategy (European Commission, 2010), but rather economic competencies.

When running growth regressions, such as in Eq. (5.7), one must be aware of the possibility that the left-hand side and the right-hand side variables will affect each other. More specifically, the growth of the factor inputs intangible and tangible capital deepening might be endogenous, affected by a common event such as an economic shock (Temple, 1999, p. 125), or stand in a bi-directional relationship with labor productivity; thus, an increase in labor productivity growth might, for example, influence the agent’s decision to invest in tangible and intangible capital. Following Temple (1999, p. 125), as the authors have not been able to retrieve valid external instruments,²³ for example, for intangible capital, lagged levels of intangible and tangible capital as instruments were chosen. Regression 4 shows the estimation results when instrumenting with lagged levels of intangible and tangible capital.²⁴ A Sargan–Hansen test of overidentification confirms the validity of the utilized instruments.²⁵ After controlling for endogeneity, the relationship between intangible capital and labor productivity remains significant (90% level) and the coefficient is only slightly reduced to 0.25. Therefore, it seems valid to conclude that the estimation results from regression 2 were indeed unbiased and not affected by uncontrolled endogeneity. Thus, the following sensitivity analysis will further explore the robustness of the coefficient of intangible capital on labor productivity growth, from regression 2, permitting us to conduct an analysis with a maximum of 98 observations.

7.1 Sensitivity Analysis

Table 5.3 shows a sensitivity analysis of regression 2 in Table 5.2. The first row, under the title Baseline regression, depicts the coefficient of regression 2 in Table 5.2. The second and third row exclude potential influential cases from the country sample.²⁶ Thus, in row 2 of Table 5.3, Ireland and Italy are excluded from the country sample. After the exclusion, the intangible capital coefficient is reduced (0.24) and remains significant at the 95% level. With Sweden being an outlier in opposition to the positive relationship, when excluding Sweden in row 3 the relationship gets significantly larger, with a coefficient of 0.35. When restructuring the data in rows 4 and 5 into the two time periods 1998–2001 and 2002–2005, we detect that the relationship seems to be slightly stronger in the time period 1998–2001 (0.32) than in 2002–2005 (0.30). In addition, the relationship is highly significant in the 1998–2001 time period and only significant at the 90% level in the 2002–2005 time period. Since the EU-13 country sample analyzed is very heterogeneous considering its economic structure, rows 6–10 explore the different regime typologies.²⁷ Excluding the two transition countries, Czech Republic and Slovenia, from the country sample does not alter the coefficient in a significant manner. Whereas the four coordinated (Germany, Austria, the Netherlands, and France) and two Mediterranean cases (Spain and Italy) remain highly significant, the two liberal cases (Ireland and the UK) and the three Scandinavian cases (Sweden, Denmark, and Finland) lose significance. Moreover, the coefficient increases significantly for the Mediterranean case.

Table 5.3 Sensitivity analysis for the baseline random-effects model

Since labor productivity growth might be related to many other determinants of labor productivity growth, in particular, characteristics of the institutional settings within the 13 EU economies, rows 11–19 include a range of policy variables. The magnitude of the coefficient of intangible capital remains robust after their inclusion, and thus none of the included policy variables is able to alter the relationship. Rows 20–22 alter the included Business Cycle (as after a downturn in the economy, unemployment usually starts to rise with a lagged effect), incorporate 8-year dummies, and add an additional crisis dummy for the year 2000. Using a forward lagged business cycle in row 20 does not alter the coefficient. The 8-year dummies or an additional crisis dummy for the year 2000 are only able to alter the significance of the coefficient partially (to the 95% level) but tend to reduce its size to 0.22. When utilizing a jackknife post-estimation command (Stata Corporation, 2007, p. 22) in row 23, the coefficient is also around 0.22. A coefficient of 0.22 would still represent an impact of 39% on labor productivity growth.

8 Conclusion

Using new international comparable panel data on business intangible capital investment within a panel analysis from 1998 to 2005 in an EU country sample, this contribution detects a positive and significant relationship between business investments in intangible capital and labor productivity growth within the business sector. Five findings emerge. First, the empirical analysis confirms the view that intangible capital investment is able to explain a significant portion of the unexplained international variance in labor productivity growth, and becomes the dominant source of growth of labor productivity. Second, this result is robust to a range of alterations and holds when controlling for endogeneity. The result is stronger in Mediterranean and Coordinated countries and within the time period 1998–2001. Third, the empirical analysis confirms the finding that the inclusion of intangible capital investment in the asset boundary of the national accounting framework implies that the rate of change of output per worker increases more rapidly. Fourth, the empirical analysis demonstrates that when incorporating intangibles into the national accounting framework, tangible and intangible capital become the unambiguously dominant source of growth. Fifth, the most important intangible capital dimension seems to be the dimension of economic competencies. Innovative property and software do not seem to have an impact on labor productivity growth within the given research design of this contribution.

In light of these five points, three main policy conclusions can be drawn from our analysis of European economies. First, measuring innovation by solely focusing on R&D, as is currently proposed in the European 2020 agenda, seems to be problematic, and the R&D benchmark measure should be substituted by a wider intangible capital benchmark. Second, incorporating intangible capital into today’s national accounting framework seems to be necessary as developed economies transition into knowledge societies and thus the significant change of investment from tangible to intangible investment is not acknowledged in today’s national accounting framework. The current accounting framework seems to be inaccurate as it incorrectly depicts levels of capital investment within European economies that are too low. In reality, European economies’ levels of capital investment are significantly greater once incorporating investment in intangible capital. Third, incorporating a wider dimension of innovation investments seems to be a first important step in revising today’s national accounting framework, in particular when focusing on the business sector. In addition, a next step seems to involve the wider adaptation of the national accounting framework by environmental, educational, health, and social capital.²⁸ Moreover, wider reform of the national accounting framework should be envisaged to achieve a more accurate signaling of real economic performance, to allow developed and emerging countries to strive for sustainable economic growth.

Appendices

Appendices

1.1 Appendix 1 Descriptive Statistics

Table 5.A1 Descriptive statistics, EU13, 1998–2005*

1.2 Appendix 2 Construction of Intangible and Tangible Capital Services

The consecutive steps required to construct the services are briefly depicted in this section. They broadly follow the literature by Oulton and Srinivasan (2003) and Marrano et al. (2009) and the EUKLEMS approach (Timmer et al., 2007). In Eq. (5.7), the growth of tangible and intangible capital services (lnk_i,t −  ln k_i,t−1) and (lnr_i,t −  ln r_i,t−1), or respectively Δ ln k_i,t and Δ ln r_i,t are specified as:

$$ \Delta \ln\ {k}_{i,t}={\sum}_{i=1}^m{\overline{v}}_{i,t}\Delta \ln {a}_{i,t} $$

(5.A1)

$$ \Delta \ln\ {r}_{i,t}={\sum}_{i=m+1}^n{\overline{v}}_{i,t}\Delta \ln {a}_{i,t} $$

(5.A2)

where small letters for k, r, and a are reflecting variables divided by hours worked in the entire industry and for all assets, and where capital services from each asset i are assumed to be proportional to the capital stocks. More concretely, in the equation, a_{i, t} is the asset-specific capital stock in hours worked obtained by applying the PIM on the asset investments and dividing the stock by the hours worked. Assets 1 to m refer to tangible capital and assets from m + 1 to n refer to intangible capital. The capital stocks are weighted by a two period average weight ${\overline{v}}_{i,t}$ given as:

$$ {\overline{v}}_{i,t}=\frac{1}{2}\left[{v}_{i,t}+{v}_{i,t-1}\right]\ \mathrm{and}\ {v}_{i,t}=\left(\frac{p_{i,t}^A{A}_{i,t}}{\sum_{i=1}^n{p}_{i,t}^A{A}_{i,t}}\right) $$

(5.A3)

The weight consists of the user cost of a specific asset (tangible or intangible) ${p}_{i,t}^A$ and the capital stocks A_{i, t} for each asset i. The user cost equation can be displayed as:

$$ {p}_{i,t}^A={p}_{i,t-1}^I{i}_t+{\delta}_{i,t}{p}_{i,t}^I-\left[{p}_{i,t}^I-{p}_{i,t-1}^I\right] $$

(5.A4)

with the capital gain term $\left[{p}_{i,t}^I-{p}_{i,t-1}^I\right]$, following Niebel and Saam (2011), being expressed as

$$ \left[{p}_{i,t}^I-{p}_{i,t-1}^I\right]=\frac{1}{2}\left(\mathit{\ln}\left({p}_{i,t}\right)-\mathit{\ln}\left({p}_{i,t-2}\right)\right){p}_{i,t-1} $$

(5.A5)

where ${p}_{i,t-1}^I$ is the investment price derived from the price index for gross fixed capital investments, i_t is the overall time-specific rate of return, and δ_{i, t} is the asset-specific, time-variant depreciation rate. As relative contributions of industries to the total business sector (c − k + o) change over time, depreciation rates are allowed to vary over time and are calculated from the EU KLEMS dataset as:

$$ {\delta}_{i,t}=\frac{A_{i,t-1}+{I}_{i,t}-{A}_{i,t}}{A_{i,t-1}} $$

(5.A6)

where A_{i, t} is the asset-specific capital stock and I_{i, t} the asset-specific investment. The rate of return i_t is common to all assets (tangible and intangible) under assumption of profit maximization (Oulton & Srinivasan, 2003) and it is the only unknown element in the user cost equation (Eq. 12). Following Timmer et al. (2007) it can be estimated using an ex post approach where it is assumed that total capital compensation is equal to the sum of all rental payments. It can as such be calculated using the following equation:

$$ {i}_t=\frac{p_t^A{A}_t+{\sum}_i\left[{p}_{i,t}^I-{p}_{i,t-1}^I\right]{A}_{i,t}-{\sum}_i{p}_{i,t}^I{\delta}_{i,t}{A}_{i,t}}{\sum_i{p}_{i,t-1}^I{A}_{i.t}} $$

(5.A7)

Following Timmer et al. (2007), the first term at the right-hand side is the total nominal capital compensation and has been obtained by subtracting the labor compensation from the value added.