Intangible capital and productivity at the firm level: a panel data assessment

Abstract

The econometric literature on measuring returns on intangible capital is vast, but we still know little about the effects on productivity of different types of intellectual capital (R&D and patents) and customer capital (trademarks and advertising). The aim of this paper is to estimate the marginal productivity of different types of intangibles by relying on the theoretical framework of the production function, which we apply to a large panel of Italian companies. To this end, the European accounting system makes it possible to compare the impact on productivity of intangibles measured from expenditures (as usual in Anglo-American studies) with the impact of intangible assets reported by companies in their balance sheets (a measure which is available in the Italian context, for example, but less common in the literature). Our results contribute two main findings to the literature. First, among the intangible components, the highest marginal productivity is that of intellectual capital, customer capital and intangible assets. Second, the use of accounting information on intangible investments is crucial to find high effects of intangible assets on productivity, while intangibles measured from expenses seem to play a more limited role. Preliminary results obtained from sub-samples mimicking the presence of spillovers deliver higher effects of intellectual capital on productivity, suggesting that intangibles’ social value is larger than the part we can estimate with individual firm data.

Keywords:

• capitalised intangibles
• expensed intangibles
• production function
• CES
• panel data
• robustness

JEL classifications:

• C23
• C52
• D24
• M41
• M44
• O39

Acknowledgement

Supplementary material (Appendices) is available at the link http://www2.dse.unibo.it/bontempi/ricerca/BontempiMairesse_EINT_Supplementary%20material.pdf. We are thankful for comments of participants at the 2005 DRUID Summer Conference, the 2005 EIASM Workshop on ‘Visualising, Measuring, and Managing Intangibles and Intellectual Capital’, the 2006 International Conference on Panel Data, the 2010 COINVEST Conference ‘Intangible Investments at Macro and Micro Levels and Their Role in Innovation, Competitiveness and Growth’, Instituto Superior Técnico, Lisbon, as well as seminar participants at Carlos III Universidad de Madrid, the Free University of Bolzano, the Bicocca University of Milan, the EU Commission and OECD. Many thanks are due to Carlo Bianchi and Roberto Golinelli for helpful suggestions and encouragement.

Notes

1. FASB and IFRS accounting standards are due to converge, and the USA will make the switch to IFRS in 2016. For an international comparison of accounting principles concerning intangibles, see Stolowy and Jany-Cazavan (2001).

2. As pointed out by a referee, there are also risks of impure heteroscedasticity, induced by specification problems such as not valid assumption of homogeneous slope parameter and/or omitted variables and/or incorrect functional form.

3. We also estimated using end-of-period capital measures; results were qualitatively the same.

4. We also experimented with different measures of the share of labour cost in the value added, such as industry medians, company medians and the Törnqvist measure . Results are robust.

5. When we use the broader definition of tangibles, estimation results do not change significantly. All non-reported results of the present paper are available upon request.

6. Software, mineral explorations and entertainment and artistic originals are the only components that are considered investment in current national accounts data.

7. Based on article 2424 of the Italian Civil Code, on Legislative Decree no. 127/91 implementing the Fourth European Commission Directive which modified a number of accounting standards, and on principle no. 24 of the Commissione per la Statuizione dei Principi Contabili of the Consiglio Nazionale dei Dottori Commercialisti e Ragionieri.

8. An intangible asset should be recognized at cost if and only if it is identifiable, it is probable that specifically attributable economic benefits will flow from the assets, and its cost can be measured reliably.

9. Intangibles initially recognized as an expense should not be recognized as part of the cost of an intangible asset at a later date.

10. The results are assessed in terms of whether they are robust enough to ‘start-up’ to the inclusion of the start category.

11. The reallocation procedures also take into account the legislative changes introduced in 1992, when the fourth European Commission Directive was implemented by statutory law (Legislative Decree no. 127/91). Bontempi (2011) illustrates the procedures we followed in order to link the reporting rules of Italian GAAP with the accounting information available for our sample of Italian companies, and the empirical variables suitable for productivity analysis.

12. See also the estimation results reported in Table 10 of Bontempi and Mairesse (2008).

13. At the parameter-estimation stage, we used two approaches: focusing on the sub-sample of K never equal to zero and using the full sample and including specific dummy variables indicating observations with null values for intangibles. The results are quite robust, especially, as expected, in the additive and CES specifications, Equations (2) and (3).

14. These facts suggest the use of the first, second and third quartiles in computing MRTS in the multiplicative specification, and the elasticity of output with respect to intangibles in the additive specification.

15. Confirming that median values in the total factor productivity approaches do not bias results.

16. See Table 1 for the summary of the relationships between model parameters, specific capital ratios, elasticity and marginal productivity. Estimates of the other parameters are available in Bontempi and Mairesse (2008), Table 7(a)–(d).

17. Estimates obtained by using other specifications of the production function are available upon request.

18. The IK variable includes IT and telecommunications, engineering and design, R&D-related services, filings for patents and registration of industrial designs for copyright and engaging in production process innovation or organisational and operational innovation or product innovation, while the CK variable includes marketing, advertising, promotions, market research and trademarks.

19. Estimates of the productivity of total intangibles computed by excluding advertising show results qualitatively similar to the ones in Table 7(a)–(d) (results not reported).

20. Note that, besides sample period and composition, in the ‘new’ data set we also modified (due to data availability) the measurement of capital inputs, which here are based on the replacement values (rather than book values) for all capital inputs. However, the use of book values, albeit resulting in a smaller sample, would not qualitatively alter the results from replacement value regressions.

21. The within transformation was chosen because of its close similarity to the results of long differences (also shown in Table 8(a)), while preserving more observations. This parsimony of the within-transformation estimator will be extremely useful for making better inferences (i.e. with more observations) in small sub-samples.

22. Where the provider of the support function is a different firm, i.e. a non-group firm, Italian university and research centre or foreign university and research centre. The reference period here covers the years 2004, 2007, 2008, 2009 and 2010. The alternative is that in which the support function is provided internally, either by the firm itself or by another firm in the group.

23. More specifically, the sub-sample includes firms that receive support functions from firms located in another region of Italy, in an EU country (on 31 December 2003), in other European countries, in China/India, in USA/Canada, or in the rest of the world.

24. Regarding the econometric assumptions, it should be noted that specification and estimation with panel techniques are able to cope with (or, at least, mitigate) a number of potential issues, such as unknown individual and temporal effects and simultaneity and measurement errors.