Essays on total factor productivity, innovation, education and training: The role of size in Spanish manufacturing firms

Abstract

A) OBJECTIVES AND HYPOTHESES:<br/><br/>The main purpose of this PhD thesis is analysing the behaviour of different factors that contribute to increase total factor productivity (TFP) in the context of Spanish manufacturing firms: mainly innovation, formal education and training. The general underlying hypothesis is that small and large firms take different strategic decisions in relation to their investment in these factors and they may obtain different productivity.<br/><br/>Departing from the general hypothesis that small and large firms play different roles in our economics, obtain different economic results and take different decisions, the first question analysed in this thesis is that large firms may achieve higher productivity levels for two reasons: first, because they innovate more and employ more qualified employees than small firms, and second, because they may be able to obtain higher returns from this effort. The emphasis placed on the analysis of the effect of returns in explaining the productivity differentials between small and large firms is one of the main contributions of this study.<br/><br/>As commented above, human capital is generally considered to have a direct positive effect on productivity. Firms can incorporate more human capital by hiring educated workers, but they can also provide training as a way to increase the skills of their employees. Concretely, training permits adapting workers' skills to the permanent evolution of job requirements and enhances the competitive position of workers and their employers. The main purpose of continuous training is to provide knowledge and adequate skills to occupied employees so that they could adapt to the changing requirements of firms at any moment. Thus, firms' may be interested in providing training to their workers as a way of increasing their productivity.<br/><br/>A general finding is that large firms usually provide more training, while small firms face more difficulties in providing training. This can be seen as an additional limitation for small firms to achieve higher productivity levels. The literature suggests different reasons why large firms provide more training: scale economies, training as a way to reduce monitoring costs, access to cheaper capital or higher pay-offs from their investments, among others. According with the reasoning in the previous paragraphs that firm size is a source of heterogeneity, we depart from the idea that small and large firms behave differently in relation to their strategic decisions. We argue that this behaviour involves firms' decisions on training provision as well as other decisions that may determine such provision of training. Thus, we argue that large firms provide more training because they have certain characteristics that allow them to dedicate more efforts to training workers or that require more training. For example, large firms usually have a more qualified labour force and less temporary workers, which permit obtaining higher returns from their investment in training. Also, large firms innovate more and use more advanced technologies or operate in more competitive markets, which requires additional skills of their employees that can not be found in the labour market. These characteristics are considered relevant determinants of training by different studies in the literature. Thus, we argue that small and large firms may differ in these characteristics, which could explain in part that large firms decide to provide more training.<br/><br/>Spanish workers receive less training than in other countries and the Spanish firms are smaller than in other economies. Thus, if small firms provide less training, the difficulties of workers in small firms in accessing training can be considered as a limitation for the economy as a whole. The second question we analyse in this thesis is the relative contribution of training determinants in explaining the gap between small and large firms in their decisions to provide training.<br/><br/>In addition to the firm characteristics that determine training, those firms that receive subsidies are also expected to provide more training. Actually, there is a system of subsidies in Spain which intends to stimulate firms' provision of training. In this system, firms or groups of firms can obtain public aid to partially finance their training actions so that they make a more intense effort in training their employees. During 2001 and 2002, the eligibility conditions for a subsidy were considerably open to most firms, although the decision on whether to award the subsidy and the amount of it depended upon the evaluation of different authorities and was unknown by the firm until training actions finished.<br/><br/>As an extension of the second question in the thesis, we analyse the impact of subsidies on training. Following the firm-size perspective in the previous questions, we deepen in the analysis of the impact of subsidies in small and large firms. Actually, the subsidies regulation gives special importance to stimulating the provision of training in small firms.<br/><br/>Finally, we would like to clarify that this thesis started as an analysis of the impact of innovation and educative human capital on productivity. However, the later interest in questions related to human capital and the availability of data on training at firm level, led to the analysis of this component of human capital that is generated within the firm. Given that data on training was only available for some years, we decided not to include it as a determinant of productivity in Part II.<br/><br/>B) METHODOLOGY<br/><br/>To address the objectives in the thesis we follow an empirical approach based on a substantive theoretical framework, appropriate quantitative techniques and a comprehensive dataset.<br/><br/>In the case of TFP differences across firm size, previous to analysing such differences, we need to measure TFP at the firm level. More concretely, we are interested in a measure of TFP, which summarizes information about the relationship between output and the main inputs involved in the production process (labour, physical capital and materials). The large quantity of theoretical papers that suggest alternative ways of measuring productivity indicates that it is far from easy to suggest a unique measure of productivity. Thus, we compare the most usual TFP indices in the literature on the basis of some a priori-defined mathematical properties and on the basis of the production functions from which they are derived. According with these criteria, we select a TFP index and use it to measure TFP for a sample of Spanish manufacturing firms over the period 1990-2002.<br/><br/>Departing from this measure of TFP, we perform a preliminary descriptive analysis on the relationship between productivity, firm size, innovation and education and training. This analysis intends to characterize the evolution of TFP and our variables of interest for the Spanish manufacturing firms in the period of analysis. Particularly, we investigate the growth pace of productivity for small and large firms, paying special attention to the evolution of the differences in TFP between firms of different size classes. Moreover, we study whether the TFP gap by firm size is homogeneous along the distribution and whether the evolution of this gap differs for firms at different points of the distribution. Furthermore, we also analyse whether small and large firms follow different patterns in the intensity of use of technological and human capital. Finally, we provide preliminary evidence on the relationship between productivity, innovative activity, human capital and firm size. This descriptive provides further insights in the relationship between these variables, and constitutes the basis for the analysis of the differences in TFP between small and large firms.<br/><br/>The TFP differential between small and large firms is evaluated in the mean of the distribution using the Oaxaca-Blinder decomposition. This methodology permits decomposing the TFP differential and obtain the individual contribution of innovation and human capital in determining TFP. The decomposition departs from an auxiliary regression for each of the two groups under comparison. Every variable determining TFP may contribute to explain the TFP differential in two ways: as differences in characteristics, which means that the differences in TFP are due to the fact that one group invests more in technological or human capital, or as differences in returns, which means that the differences in TFP are due to the fact that one group obtains higher returns from these investments. One of the main contributions of this analysis is using the Oaxaca-Blinder decomposition in the field of empirical industrial organization, and concretely, in analysing the TFP differences between small and large firms.<br/><br/>The Oaxaca-Blinder decomposition permits evaluating the TFP differential in the mean of the distribution. However, in the presence of high heterogeneity among firms, the results for the whole distribution may differ from those at the mean of the distribution. Departing from the idea of the Oaxaca-Blinder methodology, we perform a counterfactual distribution analysis. The idea behind the counterfactual distribution analysis is transferring the Oaxaca-Blinder decomposition to the whole distribution, in the sense that we try to separate differences in firms' characteristics and returns at any point of the distribution. The counterfactual distribution analysis compares the density function of the estimated TFP of small firms with the counterfactual (or hypothetical) TFP, obtained by evaluating small firms under the returns of large firms. Thus, the difference between these density functions can be attributed to differences in returns between small and large firms. All in all, this methodology permits assessing the contribution of returns of a given variable at any point of the TFP distribution, which permits identifying the non-homogeneous behaviour of certain groups of firms. As for the analysis of the second question raised in the thesis, the determinants of training, we consider that firms' decision on the provision of training is a double decision process, in which firms first decide whether to provide training or not and then, the quantity of it. Moreover, we argue that in the presence of fixed costs, some potential training providers may not decide to provide training and possible sample selection effects may appear. Thus, we propose estimating firms' decisions on training as a two part model and discuss the possible sample selection problems. On the basis of these estimations, we analyse the effect of the determinants of training on firms' provision decisions: first, we estimate our specification for the subsample of small and large firms and use the results to further analyse the differences in the provision of training by firm size using the Oaxaca-Blinder decomposition; second, on the basis of this specification, we introduce a variable on subsidies on training.<br/><br/>As in the case of the TFP differential, we decompose the training provision differential between small and large firms in the mean of the distribution. In this case, we decompose the differential between small and large firms in the probability of providing training and the differential in the quantity of training per worker. The main objective of the Oaxaca-Blinder decomposition in this case is analysing which are the main variables that contribute to explain the gap in the provision of training between small and large firms. Both differences in characteristics and differences in the impact of these characteristics on training may contribute to explain the training gap. The data used in this thesis are mainly drawn from the "Encuesta sobre Estrategias Empresariales" (Survey on Business Strategies, ESEE), an unbalanced panel which covers the period 1990-2002. This survey has been used in many papers on empirical industrial organization for Spain. The reference population of the ESEE is firms with 10 or more employees in the Spanish manufacturing sector. Small firms are defined as those with 10 to 200 employees and large firms, as those with more than 200. Since 1990, an average sample of 1800 firms has been surveyed yearly, on the basis of a questionnaire with more than 100 questions. According to data drawn from the Observatory of European SMEs, firms with 10 or more employees represent 5.3% of the firms in the total Spanish economy in 2000. Although this percentage is low, these firms employ 53% of the workers and produce 80% of the added value. According with data drawn from the DIRCE,3 firms with 10 or more employees represent 18% of the firms in the Spanish manufacturing sector in 2001, indicating that the manufacturing sector concentrates larger firms than other sectors.<br/><br/>As for the preliminary descriptive of TFP, we obtain results for a sample of more than 13000 observations over the period 1990-2002, which means that we have around 800-1000 observations per year for more than 2000 different firms. Due to some methodological and econometric requirements, the analysis of the differences in TFP between small and large firms is based on three waves of the survey: 1994, 1998 and 2002. For the each period we count on more than 800 observations per year. Finally, the analysis on the determinants of training, firm size and subsidies is based on data for the last two years of the survey, when data on training were available. In this case, we obtain results based on data for more than 1500 firms per year. In addition, to measure the effect of subsidies on training provision, we use data from other sources.<br/><br/>Particularly, data on subsidized training are provided by the Tripartite Foundation for Employment Training and data on the number of workers and worked hours are drawn from General Treasury of the Social Security and the Ministry of Labour and Social Affaires. It should also be mentioned that the software used to obtain the empirical results in the thesis is Gauss v.6.0 and Stata v.9.0, specifieally the commands: "heckman" and "xtprobit" and "xtreg", in the panel data module.<br/><br/>C) STRUCTURE OF THE THESIS<br/><br/>As previously mentioned, this PhD thesis is structured in three main Parts, each one containing two Chapters that share a common empirical framework or perspective. Due to this structure, an important effort has been done so as not to repeat information and specifying where it is possible to find it. Part I of the thesis includes Chapters 1 and 2. Chapter I presents the different methods that permit measuring productivity and, focusing on the method of index numbers, it reviews the main suggestions in the literature. Different index numbers are compared on the basis of the production functions from which they are derived and on the basis of some a priori-defined mathematical properties or axioms. Next, an index that incorporates desirable properties is chosen to develop the remainder of the analysis in Parts I and II. Chapter 2 presents the ESEE and the particularities of the variables involved in the measurement of TFP, as well as those of firm size, the innovative activity and human capital. Next, it offers a descriptive analysis of the evolution of TFP, as well as a descriptive of TFP in relation with the main variables of interest.<br/><br/>Part II of the thesis comprises Chapters 3 and 4 and it investigates the differences in TFP between small and large firms. Chapter 3 starts with a review of theoretical and empirical literature that relates firms' productivity, size and innovation and human capital, in their role of TFP determinants. Next, we explain the empirical specification and the methodology of estimation. This Chapter contains also a descriptive analysis and the results of the estimation. Finally, the size TFP-gap is evaluated in the mean of the TFP distribution using the Oaxaca-Blinder decomposition. Chapter 4 departs from the framework, specification and results in the previous Chapter and evaluates the TFP gap between small and large firms along the distribution using the counterfactual distribution analysis.<br/><br/>The last Part of the thesis contains Chapters 5 and 6 and it studies two different questions related with firms' determinants of training. Chapter 5 analyses the differences between small and large firms in their training provision decisions. This Chapter starts with a revision of the literature on the relationship between firm size and training and determinants of training. This Chapter also discusses the empirical specification and some methodological issues related with its estimation. Next, it provides a descriptive analysis and the results of the estimation. Finally, it explains the decomposition of training decisions using the Oaxaca-Blinder methodology and its results. Chapter 6 departs from the theoretical framework in the Chapter 5 and analyses the impact of subsidies on firms' provision decisions. First it describes the system of subsidies in Spain and the theoretical approach. Then, it describes the data used to measure subsidies. On the basis of the empirical model and specification in the previous Chapter, it provides the results for the estimation the effect of subsidies on training. Such effects are analysed for the total sample, as well as for the small and large firms' subsamples.