سرمایه انسانی شرکت ها، تحقیق و توسعه و عملکرد: یک مطالعه بر روی شرکت های فرانسوی و سوئدی

This paper studies the effects of human and technological capital on productivity in a sample of large French and Swedish firms. While the role of technological capital as measured by R&D has been intensively investigated, almost no work has been done on the role of human capital as measured by firm-sponsored training and even less its interaction with technological capital. The level of intangible capital may also have a lasting effect on productivity growth, as emphasised by some endogenous growth models in a macroeconomic setting. The study uses data from two panels of large French and Swedish firms for the same period (1987–1993). It constructs measures of a firm's human capital stock, based on their past and present training expenditures. The results confirm that firm-sponsored training and R&D are significant inputs in the two countries, although to a different extent, and have high returns. However, except for managers and engineers in France, we do not find evidence of positive interactions between these two types of capital. Finally, growth effects at the firm level do not appear.

Why do firms perform differently? Among the least analysed factors, we can point to intangible assets, whose share in total assets is becoming increasingly important. Some estimates for the period 1974–1983 for France mention a rise from 21.2% of total investments to 32.2% (Marion, 1987). Eliasson (1990, p. 81) offers several estimates for Sweden ranging from 14% to 30%. R&D capital and marketing capital are the most frequently cited items, but workers' human capital is also important. The firm is able to augment this capital by hiring educated workers and/or by training its existing workers and conversely, it can reduce it by its separation policy (its attitude towards layoffs, quits, and retirement). It is also responsible for the organisation of the individual workers' human capital and any resulting efficiency. Human capital exercises its effects on the firm's productivity through several mechanisms: (1) an efficiently organised firm with a manager who has substantial human capital will make better decisions than its rivals with lower human capital; (2) innovation will be stimulated by the quality and training of the personnel in the R&D department; (3) learning-by-doing is also higher if workers have high human capital. The neoclassical theoretical literature treats the firm's human capital mainly as an input, which has static effects only. The evolutionary school, based on Schumpeter's hypotheses, is built on the cognitive abilities of workers interacting in organisations and has developed into concepts such as “economic competence” (Eliasson, 1990), or “absorptive capacity” (Cohen and Levinthal, 1990). These concepts can help to explain differences in performance between firms. Several new mechanisms have been proposed. Firms well endowed in human capital have the capacity to imitate other firms and therefore to benefit from relatively cheap spillovers coming from other firms' technological knowledge. These firms also have the ability to undertake the necessary organisational changes. The evolutionary theory suggests also that intangible capital may have dynamic effects. This can be translated into quantitative terms by stating that the human capital stock may generate not only a higher productivity but also a continuous growth of productivity. Thus, differences between firms' human capital stock values may yield differences in performance in terms of future growth, while human capital in the production function should not generate such a difference. The purpose of our study is to present empirical evidence relating to the effects of intangible capital on firms' performance, with an emphasis on firm-sponsored training expenditures. Numerous studies analyse R&D effects on performance, while very few study the effects of training. Our access to data on R&D allows us to deal with the possible interactions between “training” and “technological” capital, an issue which to our knowledge has yet to be analysed at the firm level. Another novelty is to allow training and R&D capital to intervene both as inputs and determinants of total factor productivity growth. Section 2 presents the theoretical scheme and previous work. Section 3 discusses the data, and notably the construction of the training capital measure. Section 4 presents the estimation results. Section 5 offers some conclusions.

Three tentative results are derived from this study. First, both types of intangible assets, R&D and training stocks, appear as significant inputs in the production function, but one factor seems to be more influential than the other in each country, and it is not the same one. We have checked this result with several estimators, and the recent GMM-SYS method developed by Blundell and Bond, 1998 and Blundell and Bond, 1999 yields the best results by alleviating the weakness of instruments in the standard GMM. Second, some interaction effects between R&D and training stocks indicate an interesting complementarity, but the results are not very robust. Third, we have found no growth effects. The first result is based on the simultaneous study of training and R&D as inputs and appears to be a new result. It suggests the importance of training as input, especially for France, while controlling for another intangible asset, R&D, often considered as capturing the knowledge capital. It is supported by a recent study for France by Ballot and Fakhfakh (1996) on the same data, but with different specifications. The previous other studies we reviewed Carriou and Jeger, 1997 and Boon and van der Eijken, 1997 do not control for R&D, but find generally a positive effect of training. As we mentioned, Delame and Kramarz (1997) have found a significant effect only for managers, engineers and technicians, and only in the firms which spend more than the legal minimum. Their results do not contradict ours, since our firms are large firms which spend generally more than the legal minimum, and their estimations cover a balanced panel of 495 firms for the period 1982–1987, which precedes our study period. It is widely acknowledged that there is an increasing awareness by French firms over the period of the importance of continuous training as an investment. A study by Kazamaki-Ottersten et al. (1999) on cost functions in Swedish firms also finds positive returns to sponsored training. R&D seems the most important intangible capital in Sweden, while its direct effect seems to disappear in France, when the interaction with the training stock is taken into account. This contrasting position in the two countries is of interest because of the difference in the institutional arrangements for the sponsorship of training. As we have seen, there is a minimum compulsory expenditure in France while the expenditure by Swedish firms is entirely voluntary. One might then suspect that some French firms would not consider training as an investment but as compulsory non-monetary benefit they offer to their employees. The effect on value added could then be insignificant in France. Our study shows that the levy does not seem to have induced the firms' executives to ignore the potential for well designed training programs to have productivity effects. We have computed the rates of return (in terms of value added) implied by the GMM-SYS estimators in the aggregate labour cases. Rates of return in terms of value added can be computed for the training and the R&D stock. For R&D the rates of return at the mean value is 38% for France and 32% for Sweden. These figures fall within the range of estimates on firms data for industrial countries which range from 13% for USA, 20% for Japan, to 78% for France (Mohnen and Mairesse, 1999). For training capital, the rates reach 288% for France and 441% for Sweden. High returns on training then should not be linked to an institutional arrangement, and this appears as an interesting outcome of this comparative study. Moreover, as the estimation results of time-interaction terms reveal, the average rate of return to training changes over time and there may be significant differences between rates of return to training of various groups of employees, that have important implications for policy. While being extremely high, these figures are in the same class of magnitude as those obtained by Boon and van der Eijken (up to 288%). This result can be interpreted as revealing that employers underestimate the returns to training. Otherwise they would spend more since financial constraint has little appeal as an explanation except in some cases.14Ichniowski et al. (1997) in a broader study on the performance effects of Human Resource Management systems also mention underestimation, which comes from a lack of experimentation with some practices, such as in our case high levels of training expenditures. Baron and Kreps (1999, p. 391) in a detailed study of firm-sponsored training as a strategic resource mention the difficulty of measuring the value of human capital as a source of underinvestment. The very high rates of return in terms of productivity may imply high rates of return in terms of profit which would then be contrary to the prevailing view of standard human capital theory according to which firms have no private interest to invest in general training. Some of the training measured in our study is understood, especially in France, to be general training. It then gives some credit to the hypothesis that the firm can sponsor general training even if some firms poach their workers and some workers quit Ballot, 1994 and Ballot and Hammoudi, 1998. The second result is that some positive interactions between R&D and training stocks as inputs appear, but they are not very robust. In Sweden, it appears only when training capital is homogenous. On the other hand, it is very interesting to note that in France R&D interacts with managers/engineers' training capital but not with other employees' training capital, which confirms the higher importance of training for innovation than for adaptation. If the results are not robust, it may be because the dominant type of human capital interacting with R&D is the education and/or the experience of employees rather than the training sponsored by firms, as suggested for Finland (Leiponen, 1995). The quality of the data may also be insufficient or the data set may be too small to capture the effects, if they exist. The main conclusion here is that the topic, which presently induces vigorous theoretical research, deserves further investigations. The third conclusion is that intangible capital has few positive growth effects. Only training capital of low-skilled workers in Sweden has an impact. We find, at the firm level, no evidence for the Nelson–Phelps and “Schumpeterian” mechanism modelled by Aghion and Howitt (1998) at the aggregate level. More detailed data would be required to pursue research on the topic. Despite these reservations, the results justify our distinction between firms' technological and human capital, both from a theoretical point of view and in the production function. To our knowledge, this distinction has not previously been the subject of quantitative analysis. It motivates potential further exploration of the same specifications on larger data sets, and extension to other countries.