Persistence and ability in the innovation decisions^* by José M. Labeaga^** **** Ester Martínez-Ros DOCUMENTO DE TRABAJO 2005-16

June 2005

We want to acknowledge to Isabel Gutiérrez Calderón and Samuel Gil-Martín for useful comments to a preliminary version of this paper. We also thank to the participants at the III EURAM Conference hold in Milan and the Workshop Empirical Economics of Innovation and Patenting in Mannhein and specially the discussant Günter Ebling. The paper has also benefited from the suggestions made at seminars at Universidad de Vigo and Universidad Pública de Navarra. The authors also thank financial support from projects BEC2002-04294-C02-C02 and SEC2003-03797, respectively. All errors remain our own.

** FEDEA and UNED, Madrid

**** Universidad Carlos III de Madrid. Correspondence to: Universidad Carlos III de Madrid. Dept. Economía de la Empresa. C/ Madrid, 126. 28903 Getafe (SPAIN). Email: emros@emp.uc3m.es

Los Documentos de trabajo se distribuyen gratuitamente a las Universidades e Instituciones de Investigación que lo solicitan. No obstante están disponibles en texto completo a través de Internet: http://www.fedea.es/.

These Working Documents are distributed free of charge to University Department and other Research Centres. They are also available through Internet: http://www.fedea.es/.

Abstract

This paper explores the effect of persistence and manager ability in the decision to conduct product and process innovations. Managers make strategic decisions about the implementation of a better innovation activity in order to improve their firm’s performance in the market. Many empirical studies have analysed the determinants of the innovation process, but few have considered the effect of experience (the firm’s capacities and routines of organization) and the manager’s ability (skills and capability) as relevant elements. Our aim is to demonstrate the importance of these elements using typical discrete-choice specifications and binary choice models with heterogeneity. We do so in an extensive database that provides information about Spanish manufacturing firms. Our message is that persistence, however measured, is the main determinant of any innovation activity. The experience effect is important in both product and process innovation decisions but the results differ in degree: The experience gained from engaging in process innovations appears to increase the probability of process innovation success, whereas the experience gained from product innovation activity, although important, leaves place to other factors determining product innovation success. Once a firm has a commitment to innovation activity, develops some innovation routines, and learns about the innovation process, it reaches a turning point at which other factors play a less prominent role.

Key words: product innovation, process innovation, panel data, discrete choice. JEL Class.: C33, M20, O32

INTRODUCTION

The topic of innovation has garnered interest since the seminal work by Schumpeter (1942), essentially because it constitutes the main source of economic growth. However, until the advent of panel data sets, there was little empirical evidence to link the innovative stance to firm performance. A recent work by Baldwin (1998) using firm panel databases has demonstrated the importance of innovation to a firm’s growth, which in turn translates into economic growth. Our interest is in the factors that determine if a firm adopts an innovation policy; we leave the relationship between growth and innovation for a future study. Not all firms successfully innovate, despite the intrinsic benefits attached to innovation activity. The advent of innovation surveys that collect data on a variety of firm characteristics provides us with the opportunity to study the differences between firms that innovate and those that do not and to establish the main determinants of innovation activities (see, for instance, Sterlacchini, 1994, in the case of Italy; and Brouwer and Kleinknecht, 1996, in the case of Holland).

Our article has two main objectives. First, we try to determine if a manager’s ability based on firm heterogeneity (Barney, 1991) affects the decision to innovate. We define ‘ability’ as the manager’s capability and experience in developing innovation activities. Additionally, the corporate governance literature recognizes that the motivations of different ownership constituencies and the facilitation of board members affects corporate innovation strategies (Hoskisson, Hitt, Johnson and Grossman 2002). We believe that managers have the ability to control corporate innovation strategy decisions. This phenomenon is difficult to demonstrate, however, because the data do not provide us with observable information on a manager’s skills and abilities in handling innovation decisions. Thus, we assume that a manager’s ability is a fixed effect which can be controlled for, but that could affect the decision to conduct innovation activities, either directly or indirectly, through correlation with other determinants.

Second, we investigate how the organization’s experience (capacities and routines) affects the development of innovations. We assume that a company that is continuously engaged in some type of innovation activities will be encouraged to carry on with innovation, either for investment reasons or to enhance its market image. Continuous innovation over time confers additional experience upon the organization and we label this cumulative experience ‘persistence’. Because technological innovation is one of the driving forces behind a firm’s growth and because there are differences among the levels of innovation activity in various firms, it is of particular interest to question if data on innovative activities at the firm level show persistence at the aggregate economic level (Cefis, 2003). Separating a firm’s experience from a manager’s ability is a difficult task because there are no suitable data available. Nevertheless we assess the existence of some possible associated factors, then attempt to capture the firm’s experience using previous innovation indicators.

The information gathered in our study corresponds to data from 2926 Spanish manufacturing firms observed during the period 1990-1999. The database was provided by the Spanish Ministry of Science and Technology and involves firms belonging to the manufacturing sector. The sample consists of an unbalanced panel that allows us to maintain representativeness and to fully exploit the dynamic nature of the model. Before estimating the model, we conduct an extensive descriptive analysis in an attempt to account for the role of past decisions on current decisions; we compute frequencies conditional on past frequencies and unconditional on other characteristics.

There are several ways to estimate such a model. Previous empirical evidence in this area is based primarily on the estimation of univariate probit analysis, count data, or two-part models (Blundell, Griffith and Van Reenen, 1995 Martínez-Ros, 2000 and Martínez-Ros and Labeaga, 2002). In this paper we use several discrete choice models in an attempt to test all the proposed hypotheses individually and jointly, and in doing so try to control for unobserved firm effects (Blundell, Griffith and Van Reenen, 1995). In this sense, we account for the possibility of the manager’s experience or ability (firm effects) being correlated with some conditioning variables (past innovation and the firm’s experience) using estimation proposals that fully account for unobserved firm effects (see, for instance, Chamberlain, 1984).

The results reveal that past innovations are the main determinants of current innovation decisions (i.e., persistence). Once a firm is engaged in a commitment to innovation activity, there is a learning process that facilitates some routines. Hence, the firm achieves a turning point at which some of the additional factors have no impact and others decrease in importance. It does not imply that elements like ability, size, competition have no effect in achieving such a turning point. We confirm our main hypotheses: large firms that have greater technological opportunities in the market devote large investments in physical capital and perform better in carrying out new process innovations as opposed to new product innovations. That is, some of the internal resources of a firm continue to be important, even when controlling for persistence and ability in innovating.

The paper is organized as follows. In the second section we establish our main hypotheses. The third section presents a description of the methodology and the fourth presents the empirical results. In the fifth section we discuss our findings and the implication of the results. We summarize the main findings and provide some conclusions in the sixth section.

THEORY AND HYPOTHESES

Innovation as a heterogeneous activity

The definition of innovation is broad because it encompasses the creation and introduction of a new product or service as well as improvements or changes in the production process, the materials and intermediate inputs, and the management methods (Schumpeter, 1942). Some of these activities, such as the introduction of new products and the use of new designs (innovation in products) may be related to the improvement of procedures used to manufacture products (innovation in process). We are interested in the factors that could explain why the companies engage in product innovation and/or process innovation. As Milgrom and Roberts (1990) have pointed out, product and process innovation are complements because they mutually reinforce each other, for an increase in the level of any process leads to increases of the marginal profitability of innovations, and vice versa. However, except for a brief description, this issue is beyond the scope of this investigation.

The empirical observation of firm behaviour leads to the conclusion that firms choose strategically between the two alternatives of innovation, usually avoiding a complete specialization in one. Pine, Victor and Boyton (1993) demonstrate that firms structure their organization in order to allow them to engage in both types of innovation. The literature provides scarce evidence on the relationships between product and process innovations. Lunn (1986) and Kraft (1990) have introduced the possibility that innovation could be divided into different types depending on its final purpose. More recent articles (Flaig and Stadler, 1998; Fritsch and Meschede, 2001) found that both activities are related. Implementation of a product innovation can make corresponding process innovation necessary, whereas process innovation may enable a firm to improve the quality of its products considerably or to produce completely new products. Bonano and Haworth (1998) considered a vertically differentiated industry and analysed the firm’s choice of either a product or a process innovation under Cournot and Bertrand competition. Rosenkranz (1996, 2003) analyses simultaneous product and process innovations when demand is characterised by a preference for product variety. She investigates the strategic decisions of two identical duopolists who choose production technology and attempt to achieve product differentiation through research and development (R & D) investment. The firm’s investments are driven to product innovation if there is a high consumer willingness to pay. Those researchers tried to fill the existing gap in the literature relating to the firm’s decisions to direct R & D towards product innovation or towards process innovation, accounting for the degree of market competition in which the firm operates.

Our paper extents the idea that the decisions and determinants of a firm carrying out some types of innovation are the traditional factors, accounting for the manager’s ability and the firm’s experience. We focus on the decisions to introduce a new product (product innovation) or to introduce new production processes to improve efficiency (process innovation). Both decisions should be related to Porter’s (1980) work, in which he identified two types of generic strategies. Product innovation relates to the generation, introduction, and diffusion of a new product (production process ceteris paribus); whereas process innovation relates to the generation, introduction, and diffusion of a new production process (product ceteris paribus). A product innovation leads to a perception of a new product if the consumer perceives that any attribute of this product, like its service, design, packaging, or quality, has changed. In this case, we are assuming that the firm is conducting a product differentiation strategy. When a firm changes or improves the process of transforming inputs into outputs, it is developing an efficiency strategy, because the impact consists of reducing the cost of production either by being more flexible or by increasing the intensity of capital. Both decisions can be independent, but we test, in a descriptive way, the possibility that they may also happen simultaneously. Companies may acquire new technology by purchasing new capital equipment in which the new technology is embodied. Thus, the capital that embodies the technology is a product innovation, but the buyer is acquiring a process innovation.

Hypotheses

A firm seeks to maximise its value in order to sustain competitive advantages. The competitive advantages have two primarily sources: product differentiation and lower production costs (Porter, 1980). These sources lead to different strategic decisions. Sharing the recognition of firm resources as the vehicle to achieve advantages and better returns than competitors, implies, as Barney (1991) asserted, that resources are distributed heterogeneously across firms and cannot be costless when transferred among firms.

Our focus is on the role of research and development activities as a means of strengthening the knowledge stock of a firm and improving the probability of its developing future innovation (Reinganum, 1989; Blundell, Griffith and Van Reenen, 1995). The relationship between the innovation strategy and the innovation decision constitutes a production function of innovations in which the success of a particular innovation activity depends on the innovation effort already undertaken by the firm in the past, for any innovation strategy achieves returns in the long run (Piergionanni, Santarelli and Vivarelli, 1997). As Geroski, Van Reenen and Walters (1997) pointed out, the above issue is related to the hypothesis of dynamic economies of scale whereby the more innovations a firm produces, the more likely it is to continue innovating. They found that there exists with some probability a threshold of innovation activity associated with the level of previous innovative activity necessary to generate an innovation at time t. So, the two types of innovation decisions are based on internal factors such as the knowledge accumulated by the firm in the past, the experience of the manager, and firm-specific characteristics, and market factors like the technological opportunities offered by the market.

Manager’s ability and firm’s experience

Wernefelt (1984) suggested that a firm’s resources can be a source of competitive advantage in markets when it is difficult for the rivals to obtain like resources. Scarce resources create entry barriers for firms that are running short of resources. Later, Barney (1991) provided the primary baseline definitions of this resource-based view, defining the organizational resources that a firm possesses, such as its assets, capabilities, attributes, and knowledge, which enable it to develop and implement strategies to improve its performance. He assumed that resources are distributed heterogeneously across firms and that these productive resources cannot be easily transferred among firms. In these assumptions, he stated that resources are rare and valuable because they are not freely available in the market and could contribute to efficiency gains. As Dietrickx and Cool (1989) recognised, those resources can produce competitive advantages because they are enduring. But their rarity and value are not sufficient conditions for competitive advantage; they also require nonimitability, non-substitutability, and non-transferability. As recent papers have shown, firms can achieve a competitive advantage by using information technology (Mata, Fuerst and Barney, 1995; Powell, 1997), strategic planning (Michalisin, Smith and Kline, 1997), organization alignment (Powell, 1992), human resource management (Flood, Smith and Derfus, 1996), trust (Barney and Hansen, 1994), organization culture (Oliver, 1997), administrative skills (Powell, 1993), or top management skills (Castanias and Helfat, 1991).

These characteristics offer a framework that supports our aim, because most of them are unobservable to the analyst but are shared by many companies. Those companies have two important, and at least to extent unobservable resources: firm expertise and the manager’s ability. These factors may confer better performance on the firm that is able to control for them properly. Firms with a great deal of experience in innovating develop routines, synergies, and capabilities among departments and employees that increase the probability of success and foster innovation. By analysing these resources, we hope to capture the effect of learning by doing.

The presence or absence of persistence in innovative activities is a major issue of the innovative process and an important feature of the patterns of technological change – one that has significant policy implications. Knowledge stock is a measure that captures the firm’s accumulated investment in R & D. Technological knowledge stock captures the firms’ previous R & D effort, accounting for a depreciation rate. We follow Griliches and Mairesse (1984) or Hall (1990) in that we believe that the search for innovations contributes to the innovation stock by generating a constant stream of incremental innovations. We expect that the technological capital will have a positive impact on the firm’s innovation activities because the quest for innovations, which determines technological capital, is intended precisely to be able to improve products and processes.

H1: Firms with more experience in developing the same type of innovation activity will encourage further innovation.

However, managers make decisions about the types of innovation their firms will undertake. As agency and managerial theories suggest, corporate managers seek to maximise a utility function in which status, power, security, and income are the central components. Those factors create a preference for inefficient strategies through product diversification or reductions in R & D spending (Hoskisson and Hitt, 1988; Hoskisson and Turk, 1990). Status and power are intangible components achieved through prestige, reputation, and image dimensions and will effect managers’ incentives to introduce innovations. Fama and Jensen (1983) argued that the most favourable situation occurs when board proceedings are dominated by independent outside directors rather than inside managers. Baysinger and Hoskisson (1990) or (Lorsch and MacIver, 1989) reached the opposite conclusion because of their difficulty accessing strong data on the strategic direction of firms .

Recent literature on corporate governance (Malatesta and Walkling, 1988; Hansen and Hill, 1991; Wright, Ferris, Sarin and Awasthi, 1996; Li and Simerly, 1998) reveals a positive relationship between strategic innovation decisions and managers’ incentives. Moreover, Hoskisson, Hitt, Johnson and Grossman (2002) found that inside directors tend to carry out more internal corporate innovation as measured by such factors as the intensity of corporate R & D and the frequency of new product announcements because they have greater knowledge and less uncertainty about these issues. Therefore, firms prefer to acquire innovation from internal rather than from external sources

Because managers acquire different rewards from the innovation strategy, we believe that their decision-making abilities could affect the performance of the organization and hence its own experience. We assume that our construct of a manager’s ability captures these effects and that this variable is time invariant, at least in a short term. In these circumstances, variables controlling the experience (either lagged indicators or the knowledge stock) and ability in conducting innovation activities are endogenous to the manager’s decision process, so we account for this problem in the estimation of the models.

H2: The ability of managers to carry out innovation activities enhances the innovation strategy of the organization.

Internal Resources

In the Schumpeterian tradition, the firm’s size has been used as a main element to test the internal resources. Previous empirical research has tested the effect of size on innovation (Acs and Audretsch, 1987; Kleinknecht, 1989; Piergiovanni, Santarelli and Vivarelli, 1997; Martínez-Ros and Labeaga, 2002) with mixed results. In many cases, innovation activity was measured through inputs rather than outputs. Pavitt, Robson and Townsend (1987) found in the UK that innovation intensity was greater for large and small firms, and smaller for medium-sized firms. In contrast, Soete (1979) suggested that R & D intensity increased with firm size in a number of sectors in the USA. Using the innovation counts, Blundell, Griffith and Van Reenen (1995) found that firms in competitive industries and firms with higher market shares have a greater probability of innovating. The apparent inability to obtain consensus on the effect of firm size on innovation activity could be attributed, in some studies, to a lack of control for characteristics of the firm, characteristics of the market, or the functional form of the mathematical model used to determine the effect of size on innovation, despite the tested importance of such effects (Martínez-Ros and Labeaga, 2002). The size distribution of firms varies across industries, in part because of differences in the degree of scale economies in production and distribution. Thus, there is a good reason to believe that fixed industry effects are correlated with firm size and that the omission of such effects will bias estimates of the effects of size on innovation. Moreover, some researchers neglect the possibility of non-linear effects of size on innovation.

Similarly, firm characteristics such as diversification, financial capability, returns on R & D in larger markets, or greater innovation experience within the structure of the organization confirm the positive correlation with large firms (Cooper, 1994; Hitt, Hoskisson and Ireland, 1990; Graves and Langowitz; 1993; Galende and Suárez, 1999). So, in order to isolate the size effect on innovation for a given knowledge stock, it is important to control for the market’s competitive structure and for other firm attributes. For a given stock of technological capital and opportunities, the size of the firm may influence the output of innovations, due, for example, to differences in other physical, human, and financial resources across firms of different sizes. In general, a positive effect of size on innovation output is expected because larger firms tend to be less financially constrained. However, it could also happen that managers in larger firms view their company as being less threatened by competition and decrease the rate of innovation accordingly, so not to erode the profitability of current products and processes. Besides, if the firm has monopoly profits, the incremental profits of innovation will tend to be relatively low in highly competitive markets. Moreover, large firms may also be subject to more bureaucratic controls and dysfunctions, which may negatively affect their capacity to translate capital stock into innovations (Collier, 1983; Williamson, 1985; Hitt, Hoskisson and Ireland, 1990; Cooper, 1994).

Cohen and Klepper (1996a, b) developed a model, the main hypothesis of which was that the return on an innovation is positively related to the size of the business unit and that this relationship was stronger for process innovation than for product innovation. Fritsch and Meschede (2001) tested the same hypothesis using a different type of R & D expenditure, but their findings are not conclusive. Following the same arguments we hypothesize:

H3: Large firms have a greater proclivity to invest in process innovations than in product innovations.

The characteristics of the production technology may also affect the decision to introduce innovations for a given stock of technological capital; one variable used to differentiate production technologies is the intensity of physical capital. Firms with more capital-intensive technologies will tend to innovate more if, as expected, the rents of innovation are less threatened as, to exploit the innovation, high investment in physical capital is required. It may also happen that more capital-intensive processes provide less room for innovation because they are more automated and rigid. The final effect of capital intensity on innovation activity is uncertain. Kraft (1990) found a positive effect by including only capital intensity in the product equation, but it is somewhat of an empirical issue. We are concerned that although physical capital is more strongly related to production processes, it could also affect the innovation in new products.

H4: The physical capital is more important in the development of process innovations than it is in product innovations.

Technological opportunities in the market

Not only is monopoly power sufficient to explain the technological activities in which firms engage, but, as stressed by Cohen and Levin (1989) the role of other environmental factors have been emphasized. Industries with more technological opportunities are expected to encourage innovation, as the accumulated knowledge of the market, usually shared by many of the firms due to spillovers or other effects, reduces the cost of translating knowledge into new products and processes. At the same time, however, innovation may be hampered if the innovating firms consider the innovation to be susceptible to imitation. This phenomenon is observed particularly in product innovation (Lunn, 1986), which as Crepon and Duguet (1997) have pointed out, captures an externality of R & D capital. Piergiovanni, Santarelli and Vivarelli (1997) found that spillovers from university research are a relatively more important source of innovation in small firms, whereas spillovers from industrial research are more important in producing innovation in large firms. Rosenkranz (2003) found that product innovations exhibit positive externalities as long as consumers have preferences for product variety. As D’Aspremont and Jackquemin (1988) have argued, the effect of this positive competitive spillover on the product market can outweigh negative externalities from process innovation. Our proposal enlists this argument as a reason why technological spillovers may serve as opportunities to develop new products or to improve existing ones.

H5: Higher technological opportunities in the market act as barriers to imitation, leading to increased product innovation.

Industrial factors

Industrial factors are involved in the usually Schumpeterian hypotheses about the extent to which firm size and the degree of competition in the industry environment stimulate innovation. It is sometimes claimed that innovation is fostered by a climate in which firms have market power or in industries in which there is little competition. Although Arrow (1962) made a claim contrary to Schumpeter’s (1942), there is mixed evidence that either factor has an effect (Scherer, 1992). Because of its importance, this issue continues to receive attention, however (Cohen and Klepper, 1996a, 1996b using US data or Martínez-Ros and Labeaga, 2002 using Spanish data). We also tackle the effects of a firm’s ownership on innovation. Several authors have stressed the special role of the multinational firm in transferring special innovation skills among countries.

Market structure is viewed as a key determinant of the sustainable performance of firms. Typically, studies in industrial organizations have approached the degree of competition by assessing market concentration (see Cohen and Levin, 1989 for a complete overview of the relationship between R & D and market concentration and for an extensive discussion about the ambiguous predictions obtained in empirical studies). In general, the empirical evidence supports Schumpeter’s arguments that firms in concentrated markets can more easily appropriate the returns from inventive activity. Other investigations have found evidence that market concentration does not promote R & D because the expected incremental innovating rents are larger the higher the degree of competitiveness (Arrow, 1962; Bozeman and Link, 1983; Delbono and Denicolo, 1998; Yi, 1999). A positive effect would support Schumpeter’s outlook, whereas a negative one would be in accordance with Arrow’s view. This effect could also differ depending on the different types of innovation. The effects found in previous empirical studies, such as those of Lunn (1986) and Kraft (1990), motivates our work: in the former study, concentration is precisely estimated only in the process equation; in the latter, concentration affects only the product equation.

Acs and Audretsch (1987) found that large firms are more innovative in concentrated industries that have high barriers to entry, whereas smaller firms are more innovative in less concentrated and more mature industries. Blundell, Griffith and Van Reenen (1995) obtained a negative relationship between innovation and market concentration. Therefore, in the long run increases in market share may have a net negative effect on innovation if they also incorporate increases in market concentration.

H6: Higher market competition will produce a greater incentive to introduce product innovations than will process innovations.

METHODS

Sample

We used information from manufacturing firms during the sample period 1990-1999. This information was obtained through a survey called ESEE conducted by the Spanish Ministry of Science and Technology.¹ Because some firms stopped providing information during the sample period for several reasons, including mergers, changes to non-industrial activity, or production process shut down, we have an unbalanced panel . New companies enter the survey each year to maintain sample representativeness for the overall sector. We assume (and test by means of descriptive measures) that there is no random attrition innovation decisions. The ESEE survey respondents were CEOs. Their data were collected using direct interviewers supported by a questionnaire. The data constitute a mixed set in which a random sample is drawn for small companies (those with less than 200 employees) keeping the sample representative of the industrial stratification. For large firms (more than 200 employees) the sample is exhaustive.² The definition of firm size in this survey is the number of employees at the year end. A common problem associated with the firm size variable is lack of control for workers’ movements. The solution adopted in the ESEE survey was to consider the firm’s employment average by ruling out possible workforce adjustments during the year.

¹ See the Data Appendix and Martínez-Ros and Labeaga (2002) for additional details about this survey.

Measures

Endogenous variables:

Product innovation is a dummy variable that takes the value 1 when the firm is involved in the creation of a new product and 0 when it is not. Process innovation is a dummy variable that takes the value “1: when the firm introduces some new process innovation and ”0” when it does not. Both variables are provided directly by the person responsible for completing the questionnaire or responding in writing to the interviewer. The information used to construct those innovation types comes from the following questions: “Did your firm introduce product innovation in year ’ t’?” and “Did your firm introduce process innovation in year

Explanatory variables:

Firm’s experience. We used two types of measures to capture this variable: knowledge stock and previous innovation activity. First, the knowledge stock (G) is constructed by using the specification of Griliches and Mairesse (1984) or Hall (1990) and implemented by other authors such as Blundell, Griffith and Van Reenen (1995) or Martínez-Ros and Labeaga (2002).

\[\boldsymbol {G} _ {i t} = \boldsymbol {S} _ {i t} + (1 - \delta) \boldsymbol {G} _ {i t - 1}\tag{[1]}\]

where is the R & D expenditure of firm i in period t and δ is the depreciation rate.⁴ This process of searching for innovations implies that the decision about innovating evolves according to the indicator function [1].⁵

² This aggregation and the threshold used for size are suitable for the typical Spanish structure.

Specifically, CEOs say whether the employment alterations are due to changes in the fixed-worker or the temporary-worker category. For this last category, a question about length of time since hiring is also included.

⁴ We use a depreciation rate of 20 percent. Small changes in this rate do not significantly affect the results presented below.

Second, in previous Innovation, we captured firm experience by controlling for the firm’s previous innovation activities. We accounted for this phenomenon in two ways. First, we included past innovation indicators as determinants of current innovation decisions. Second, we estimated the model on sub-samples defined as conditional on firms having done some innovation in the past.

Managerial ability. We control time-invariant firm effects in the models that are estimated using the panel nature of the data. These unobserved effects are recovering managerial ability (manager’s experience), firm experience in R & D activities, or ability at internal organization, which may in turn affect the production of innovations when all these potential determinants of the innovation strategy of the firm do not show variation along the time span of our sample. In our view, controlling these unobserved time-invariant factors is more important than identifying what they are recovering.

Internal factors. We used several variables to approach internal characteristics. First, the size of firms is measured using the logarithm of the number of employees at the end of each year (SIZE). Because the effects of size on innovation activity could be both positive and negative, we accounted for it by assuming a non-linear relationship between size and innovation and introduced the number of employees squared (SIZE ) among the explanatory variables of the model (Pavitt, Robson and Townsend, 1987) allowing us to identify effects of different sizes in firms of different-sizes. We used a relative measure of size (SHARE) that takes a value of 1 when a firm achieves an improvement in its market share.

As a second internal factor, we used a proxy for physical capital, which we labelled (KSA) – the ratio of sales to fixed assets of the firm. It was constructed using the traditional literature about the measurement of capital stock (Blundell, Griffith and Van Reenen 1995).⁶ A higher value of the ratio means that the production process is relatively more capital intensive.

Technological opportunities in the market. We used patents as a form to appropriate the returns of innovation. Patents are a good measure of appropriability so we include them in two ways. We used two dichotomous variables (PATSPAIN) and (PATABROAD), which take the value 1 when a firm registers a patent in Spain or abroad, respectively. We are aware of the potential endogeneity in using these variables, even in lagged periods through unobserved effects for instance, but we controlled this effect using previous lags, even in specifications in which the effects are ruled out. Additionally, we included dummy variables to characterize different industries.

Alternatively, we could assume that the knowledge stock is obtained using the number of patents or innovations, as did Blundell, Griffith and Van Reenen (1995).

⁶ It measures the replacement value of the firm’s machinery capital stock.

Finally, we measured the intensity of market competition (COMPETITION) in an inverse way by calculating the average gross profit margin of the industry in order to determine if market competition encourages innovation activity. With this measure we tried to avoid the possible endogeneity problem. The earlier discussion of the hypothesis suggests that there are some theoretical issues to be tested empirically in order to determine the net effect of market concentration on innovation. The lack of a clear theory also reinforces the importance of using econometric estimation procedures that minimize estimation biases. A discussion about the right direction of the effect of market structure needs to be related to the possible endogeneity of the concentration ratio, which is the measure more commonly used in the empirical analysis. As Levin and Reiss (1984) and Levin, Cohen and Mowery (1985) demonstrated, the endogeneity of concentration produces biases in the estimates of the effect of market competition on innovation activity.

Control variables:

Another feature contributing to the explanation of both innovation activities (product innovation and process innovation) is the growth of demand (Schmookler, 1966). A dummy variable GROWTH took the value 1 when the CEO believed that the market was in recession and 0 otherwise. We expect that a recessive demand discourages the level of both innovation activities.

Innovation activity may also be differentiated in terms of the degree of vertical integration (VERTICAL). As firms internalise more activities, there are more opportunities to innovate, ceteris paribus, and there are probably more incentives to engage in it, particularly if the results of innovation can be spread over several activities. Although little quantitative work has been done in this area, some case studies suggest that there are economies of scopeto R & D in vertically integrated industries. Malerba (1985) studied the life cycle of technology in the semiconductor industry and found that the advantages of vertical integration for innovative activity varied along the cycle. VERTICAL is measured inversely by the ratio of purchases to the total value of production, both variables being defined on a yearly basis.

We also considered possible discipline effects of conducting export activities. We defined a dummy variable (EXPORT), which takes the value 1 whenever the firm exports and 0 otherwise. We expected that exports favour at least product innovation competitiveness in foreign markets, and may require firms to adopt leading-edge technology. But it is also true that firms with more innovation activity may have more incentives to export because they also have more intangible resources to sustain growth. So, no clear direction of the causality can be established (see Entorf and Pohlmeier, 1990; Cassiman and Martínez-Ros, 2003).

Foreign ownership is a dummy variable (OWNER) utilised to capture the amount of the firm’s capital that is foreign capital This is a control variable for which no clear effect can be expected from the theory, although it may serve as a proxy disciplinary effect on competitiveness and, in turn, on innovation activities.

Finally, we control possible macroeconomic and business cycle shocks common to all industries using time dummies, as well as common time invariant industry shocks using industry dummies.

Methodology

We performed two types of analyses – conditional frequencies and overall frequencies. First, we tabulated the frequencies for our innovation measures in an attempt to test for persistence in innovation decisions. We constructed conditional frequencies for quitters and entrants and measured whether or not innovating firms contribute to the overall frequencies or if it is simply the differences among quitters and entrants that explain the overall figures. We performed a second type of analysis because we believe that a non-negligible number of firms make conscious decisions to enter and exit innovation activity and we believe that previous experience does not fully explain all the innovation frequencies. This analysis consisted of deriving a specification for innovation decisions, bearing in mind that we only observed whether or not the decisions were taken. In these circumstances, we believe that discrete choice models for the two indicators are adequate. We also identified the determinants of innovation activities of Spanish manufacturing firms. The specification proposed is:

Probability (Innovate) = f (explanatory variables, control variables, time dummies, industry dummies).

In this specification all variables in f(.) are expressed in t-1. In cases in which we exploit the full nature of the panel, we also include in the previous specification the individual time-invariant effects, which can approximate firm effects associated with manager’s expertise or ability. In order to test the different hypotheses, we estimated three different static models. The first estimation was performed on the whole sample and utilised the pooled data, which means that we did not control for different effects across firms. The second model is merely a probit analysis on the pooled data, but it is estimated on the sub-sample of firms innovating in the recent past (last year or t - 1) or in previous periods (t - s). So the equation is:

Probability (Innovate t / Firm innovating t - s) = f (explanatory variables, control variables, time dummies, industry dummies).

This formulation allowed us to test if persistence in conducting innovation activities has any effect on the other condition variables. The third model allows for firm-specific differences according to a common distribution, i.e., a discrete choice, random effects model. The effects of controlling individual heterogeneity on the two equations (product and process innovation decisions) serves to test H2. We also estimated the same three specifications with the inclusion of the lagged indicator in order to put more confidence on the likelihood of the tests for H1 and H2.

RESULTS

The empirical analysis is, first, devoted to performing an extensive descriptive analysis of the innovation frequencies and, second, to formally testing the hypotheses formulated in the theory and hypothesis section of this paper. As a by-product, we tried to identify the determinants of innovation activities of Spanish manufacturing firms in order to establish the main differences among innovation decisions for product versus process innovations and in order to draw some conclusions about strategies concerning this issue. We organised our main results in five tables and two figures. Tables 1 and 2 report a summary of the descriptives.

Table 1a. Unconditional and conditional innovation frequencies

Notes. 1. Figures in the table are frequencies of product innovation. 2. Column under iprod are unconditional frequencies; iprodt1 reports frequencies in t conditional on firms innovating in t-1; iprodt2, iprodt3 and iprodt4 present frequencies in t conditional on firms innovating in t-2, t-3 and t-4, respectively; iprodt2t1 and iprodt3t1 are respectively frequencies conditional on the firm innovation in t-1 and t-2 and t-1, t-2 and t-3, respectively.

Table 1b. Unconditional and conditional innovation frequencies

Notes. 1. Figures in the table are frequencies of process innovation. 2. Column under iproc are unconditional frequencies; iproct1 reports frequencies in t conditional on firms innovating in t-1; iproct2, iproct3 and iproct4 present frequencies in t conditional on firms innovating in t-2, t-3 and t-4, respectively; iproct2t1 and iproct3t1 are respectively frequencies conditional on the firm innovation in t-1 and t-2 and t-1, t-2 and t-3, respectively.

Table 2. Conditional innovation frequencies, enters and exiters

The upper number in each cell reports the innovation frequency in t, conditional on the firm innovating in t-1. The bottom figure corresponds to the difference among frequencies of enters and exits

Before presenting the results of the empirical specifications, our first task consisted of a descriptive study of the frequencies of both innovation activities, conditional on past events but unconditional with respect to other possible determinants. Frequencies presented in Tables 1 and 2 shed some light on the persistence of these activities at the firm level. We calculate the probability of product or process innovation for each firm in the current period and for periods when they have previously conducted these activities. Table 1a shows the probability of engaging in product innovation. The first column shows the unconditional probabilities for the period 1990-1999. Columns 2, 3, 4, and 5 provide innovation frequencies at t, provided that some product innovation has been undertaken at t-1, t-2, t-3, and t-4, whereas Columns 6 and 7 present the frequencies when the conditioning sets are innovations at t-1 and t-2 and t-1 to t-3, respectively. In calculating all these frequencies for every period, we attempted to show possible business cycle effects.

Figure 1. Product Innovation and Firm Experience Table 1b presents the probabilities of process innovation. Again, we report unconditional and conditional probabilities. The first column presents the unconditional probabilities for nine years of the sample. Columns 2, 3, 4, and 5 provide the process innovation frequencies at t given firms that also engaged in some process innovation at , and , respectively, whereas the conditioning sets employed are engaged in process innovations at and (Column 6), and and (Column 7).

Figure 2. Process Innovation and Firm Experience

In Table 3 we present the naïve estimates corresponding to pooled probit models with the two different samples mentioned. Comparisons of the results in this table allow us to test if recent past decisions condition current ones (H1), conditional on common unobserved factors among firms.

Table 3. Innovation Decisions^{1, 2, 3}

Notes. 1. Sample sizes are 13225 observations in the unconditional models and 3464 and 4420 observations in the conditional product and process innovation equations. 2. In all specifications we include additional controls as time and industry dummies, the knowledge stock and spillovers indicator. 3. T-statistics (in absolute value) are in parenthesis: ♣p<0.10; *p<0.05; **p<0.01. 4. LR is the likelihood ratio test (degrees of freedom in parenthesis).

Table 4 provides coefficients of unconditional and conditional random effects probit models. Comparisons across this table will allow us to test H2 only, and to test H2 once we impose H1. When comparing Column 1 of Tables 3 and 4 (i.e., the decision to engage in product innovations), we obtain significance with most of the variables in this unconditional sample, whether or not the model displays heterogeneity. The only remarkable difference is that the magnitude of the effects for all determinants of the innovation frequencies is smaller when the manager’s ability is controlled. The qualitative results for the process innovation equations (Column 2 of Tables 3 and 4) are not significantly different. A comparison of the results reported in Column 3 of Tables 3 and 4 allows us to test ability while controlling for persistence. Of course, this is the naïve way of controlling for persistence, although we have some confidence in it given the results of the descriptive analysis previously reported. Using the form in which we control ability, we get coefficients more significant in the random effects probit analysis than in the heterogeneous-free, non-random effects probit analysis.

Table 4. Innovation decisions^{1, 2, 3}

Notes. 1. Sample sizes are 13,225 observations in the unconditional models and 3464 and 4420 observations in the conditional product and process innovation equations. 2. In all specifications we include additional controls as time and industry dummies, the knowledge stock and spillovers indicator. 3. T-statistics (in absolute value) are in parenthesis: ♣p<0.10; *p<0.05; **p<0.01. 4. LR is the likelihood ratio test (degrees of freedom in parenthesis).

In order to end up with results on persistence and ability, we present in Table 5 the coefficients of dynamic random effects probit models and of conditional fixed effects logit models. As we have evidence on the importance of the recent past in explaining current decisions, we estimate random effects probits, including as an additional regressor the first lag of the innovation indicator both for product and process decisions. As an alternative to random effects that could present correlation with the regressors, we also provide results obtained using conditional logit fixed effects models. One of the shortcomings of these estimates is that a possible bias as a result of the lagged indicator contains the random effect (which is time invariant) and it is correlated with the composite error. Because the conditional approach of Chamberlain (1994) allows us to rule out the unobserved heterogeneity, we provide these results for comparison. However, we cannot include the lagged indicator in these specifications because the conditional procedure requires strict exogeneity. These results constitute an alternative way of drawing inferences about the determinants of innovation decisions (i.e., the fulfilment of H3 to H6) once we control for ability and persistence.

Table 5. Innovation decisions^{1, 2, 3}

Notes. 1. Sample sizes are 13225 observations in all models. 2. In all specifications we include additional controls as time and industry dummies, the knowledge stock and spillovers indicator. 3. T-statistics (in absolute value) are in parenthesis: ♣p<0.10; *p<0.05; **p<0.01. 4. LR is the likelihood ratio test (degrees of freedom in parenthesis).

DISCUSSION AND CONCLUSIONS

Persistence

We present unconditional and conditional innovation frequencies in Table 1. The jumps from unconditional to conditional probabilities range from 117 to 168 percent. In other words, while 70 percent of the firms in 1991 were noninnovating only 40 percent were classified as non-innovating from those already innovating in 1990. The frequencies reveal that experience in innovation activity in the recent past is a good predictor of current innovation activity . We summarize this information in Figure 1.⁷ The increases in innovation frequencies are not as spectacular as they are before additional past events are taken into account. For instance, the product innovation frequency in 1992 for firms innovating in 1990 and 1991 is 71.5 percent, which must be compared with the conditional 1991 frequency of 64.4 percent. On the other hand, cycle effects are much less pronounced for these last figures. It shows that once a firm has incurred some sunk costs (such as development of an R & D unit, acquisition of capital), the continuation of these activities is less costly. Another implication we can draw is that innovation is an experience-intensive activity and, once a firm embarks upon it, there is a significant reduction in the probabilities of ceasing this activity.

The innovation frequency when extending the conditioning set from t-1 to t-2 is more than 10 percent higher than when the conditioning set is at t-1. The unconditional probability of engaging in product innovation increases by more than 25 percent when comparing the figures to those for the case in which the conditioning set is at t-1 to t-2 and t-1.⁸ The contribution to innovation frequencies coming from firms with or without experience that are continuously making entry and exit decisions regarding innovation, ranges from a maximum of 35 percent of the innovating firms to a minimum of 4 percent. On the other hand, entrant firms exhibit only a 13 percent probability of product innovation on average for the period. This probability reduces to less than 5 percent when we do the analysis accounting for the entrants conditional on not being innovators in previous periods. This simple exercise creates some confidence about the fulfilment of H1 in product innovation decisions. Finally, figures in Table 2 show that the net contribution of entrants (after subtracting the rate of exiters) is small and that some of the net contributions have negative signs, meaning that the percentage of exiters is higher than the percent of enterers. This is true for each of the years studied, except for 1990. Fortunately, we use an econometric specification that does not use the year 1990, as it seems that these figures are measured with some errors. As we do not know the amount of time a firm requires in order to consolidate its innovative activity, the non-diagonal elements in Table 2 report, first, the probability of engaging in product innovation in 1995, conditional on firms always innovating between 1990 and 1994; and, second, the probability of engaging in product innovation in 1999, conditional upon firms always innovating between 1995 and 1998. Independent of the spell of product innovations, these two figures confirm previous evidence.

When we extend the conditioning set to include expenses on R & D or presence of process innovations during past years, the conditional frequencies do not increase substantially.

Although we do not report additional results in this table, we observe that the probability of engaging in product innovation for a firm that conducted this activity in all previous periods actually rises to 94 percent.

The second block in Table 1 reports the same information as does the first block, but in this case for process innovation. The unconditional probabilities are affected more by the business cycle than are those relating to product innovation. The recession began at the end of 1991 and there is a bigger decrease in the frequency of developing new processes than in conducting product innovations. The level of innovation exhibited by firms during the early 1990s recovers after 1996 when the economy began a new boom. Increases in the conditional probabilities are not as large as in the case of product innovations, because the departure point is different. However, the implications from these figures are again that the experience of firms in performing process innovations in the recent past correctly predict current innovation frequencies. We present the unconditional and conditional probabilities In Figure 2.

When extending the conditioning set to previous events, we obtain a similar picture. The innovating frequency in 1992 for firms innovating both in 1990 and 1991 is 68 percent, which must be compared with the conditional frequency of only doing innovation in 1991 of 64 percent. The preliminary implications from all these figures are that recent previous experience strongly conditions current performance. Again, cycle effects are a lesser factor in the change in the decisions of firms already engaged in an innovation process. On the other hand, entrant firms display only an 18 percent process innovation probability, on average, during the period of study. This figure is reduced considerably when we account for the entrants conditional on not being innovators in additional previous periods. In this case the reduction plummets to approximately 10 percent. The net contribution of entrants is small, with most of the signs reflecting a negative relationship to the probability of innovating, according to the business cycle (see Table 2). Although we do not include further conditionings in the analysis, these descriptive statistics allows us to confirm H1 in the case of process innovation decisions. Again, we calculate frequencies aggregated by intervals of several years in an attempt to determine if there is any effect as a function of time required by a firm to consolidate its innovative activity matters. The non-diagonal elements in Table 2 report these figures for two periods: for 1995 conditional on firms that have innovated between 1990 and 1994 and, for 1999 conditional on firms continuously innovating between 1995 and 1998.

9 The conditional frequencies do not change substantially when we extend the conditioning set to include expenses on R & D or presence of product innovations during past years.

¹⁰ The probability of doing process innovation for firms engaged in this activity in all previous periods is 92 percent.

Alternative tests for H1 can be seen when comparing coefficients for unconditional and conditional discrete choice models (see Tables 3 and 4). Once we use the sample on recent past innovators, most of the conditionings lose their significance. But we must be cautious, both because we mis-observed and failed to observe differences amongst firms, and because the results in these specifications are conditional on some assumptions already mentioned. We postpone this discussion on presenting the results of our preferred model.

Ability

One can argue that firms continuously making decisions about product and process innovation could have managers and/or workers with some special abilities. It should be true as long as firm’s organization or its managers do not differ across periods. However, we are able to observe some changes (like ownership of the company), whereas others (like internal organization) are completely unobserved. In our sample, we observe that the majority of firms (67%) use the limited liability corporation as a legal form, implying that those corporations have available the possibility to use agency controls to align the managers’ objectives with the owners’ and hence, changes in abilities are relative small. Looking at the data, only 3% of firms change the legal corporation that they employ. This information as proxy of ability we could assess that, ability is a fixed effect along the panel. Even in these circumstances, the identification of the effects of ability on innovation strategies is a difficult task. Another difficulty turns up when it comes to separating unobserved effects that differ across firms not showing time variation from persistence. With all these caveats in mind, comparisons among coefficients on unconditional models across Tables 2 and 3 can provide evidence about the importance of controlling unobserved heterogeneity, without having account of persistence. On the other hand, comparisons of coefficients corresponding to conditional models allow us to test for the importance of firm effects once we account for persistence in the innovation decisions.

In both specifications only the dummy for exporters, the dummy indicating that the firm has registered a patent in Spain, and the dummy collecting the state of the market conditions remain significant in the conditional sub-sample corresponding to the product equation decision. The dummy indicating whether or not the capital of a firm is controlled 50 percent or more by foreign firms and the knowledge stock are marginally significant, although they are non-significant at standard levels. In the decision to innovate in process, ability is less important than persistence, because once persistence is controlled, the results do not change whether controlling ability or not.¹¹ In both cases, only market share and the firm’s registration of a patent in Spain do not lose their significance.¹² This confirms that it is more important to have experience innovating in process for the success of future process innovations than having experience in innovating in product for the success of future product innovations, all remaining characteristics being equal.

Finally, we test for persistence and ability in an alternative context. Table 5 reports coefficients of dynamic random effects probit models and conditional fixed effects logit models. In the probit specifications the lagged indicator is significant, which merely confirms previous evidence. The implied probabilities of innovating for firms engaging in product and process innovation in the past (without having account of other determinants) are greater than 80 percent. In the conditional logit specification, we cannot use the lagged indicator because of endogeneity problems. We must bear in mind that this last procedure relies on changes of regime from innovating to non-innovating and vice versa, meaning that innovators and non-innovators (in all periods) never contribute to the likelihood of the model. Qualitatively, most of the results sustain previous evidence, although some of the coefficients could be affected by bias, because in this model all regressors should be strictly exogenous and we do not have complete confidence in knowledge stock fulfilling this requirement. However, the coefficient of knowledge stock is merely indicating that the probability of changing from innovation regimes is negligible (the rest of conditionings taking mean values). This means that the probability of observing a change from innovating to non-innovating for a firm with a high knowledge stock is statistically zero, whereas the probability of observing a change from noninnovating to innovating for a firm with a low level of knowledge stock is also zero, statistically. Together, these results place confidence in the validity of H1 and, to some extent, H2.

Control of the ability of the manager has some effect on the determinants of innovation, but it is less important than the effect of previous experience. We must note, however, that heterogeneity could be correlated with some of the explanatory variables because the more skills, the greater the propensity to innovate; but in order to continue innovating, firms must commit more resources. These feedback effects induce correlation among skills and input variables. Perhaps when a firm decides to enter into innovatation and remains in this activity, it achieves a threshold of capabilities and routines inside, that makes it difficult to give up innovation. On the other hand, either conditioning on past decisions or including lagged innovation indicators could control not only for experience but ability when firms do not change managers among periods.

11 We should also note that the lag of the innovation decision includes both persistence and ability, because we assume that ability is an unobservable, heterogeneous time invariant effect.

¹² These results are confirmed when we condition on innovation in t-2 instead of innovation in t-1. We do so in order to control for first-order serial autocorrelation in the errors. Moreover, we get less significant coefficients when we extend the conditioning on previous years. These results are available upon request.

In those models including lagged innovation indicators as proxies for experience, the results are similar to those in which we estimate on the subsample. This is plausible because once we condition on past indicators, the lag of the innovation variable became a constant. In this sense, we are estimating similar models in both samples. However, some differences need to be emphasised and clarified. These differences arise for several reasons. First, we lose almost 60 percent of the observations when conditioning on past events. Second, by these cuts in the sample size, the variance of the coefficients is also affected.

Additional determinants of innovation activities

Once we have evidence about persistence and ability, individual or jointly considered, we move on to test the significance of some variables in determining innovation frequencies. In other words, we try to confirm the hypotheses established in the theoretical model above. H1 emphasizes the importance of past innovations and, as a result, once experience is controlled for, either estimating the model in the sub-sample of firms innovating in the recent past or including lagged innovation indicators, the accumulated knowledge stock lacks its significance almost everywhere. Knowledge stock and technological opportunities get the expected estimates confirming our hypotheses. When the firm accumulates knowledge, it serves and encourages itself to innovate. And it is true for both decisions. Spanish patents provide incentives to continue the development of both innovation activities. Because patents create both a barrier and a protection from imitation, such activities will be more product-innovation intensive.

As regards H3, the evidence we find is interesting. First, there is a quadratic effect of size in the decision to conduct product innovation. Both small and large firms innovate more in product than do medium-sized firms, confirming previous results for Spain and other countries (Martínez-Ros and Labeaga, 2002 and Pavitt,Robson and Townsend, 1987, respectively). On the other hand, innovation size seems to play a crucial role in developing process, independently of the controls we include in the specifications. However, once we condition on the existence of past innovations, size becomes irrelevant in explaining current innovation decisions. This confirms the existence of a threshold beyond which companies innovate, no matter what their size.

Physical capital is more important for the development of process than for the development of product innovation, thus confirming H4. As is commonplace in the empirical literature, large firms have more capabilities and resources to continue process innovation. Small firms, however, concentrate on the capture of small parts of the market using the introduction or improvements of products.

Technological opportunities within the market appear, as expected, to be a barrier to new entrants in the competition to introduce or change in the product market. The fact that firms could register patents of new products (or any attributes such as quality, service, and packaging) either in Spain or in foreign countries constitutes a defence to imitators, which confirms H5. However, results also show the importance of registering patents in Spain, when we model the decision to innovate in process, although its effect is lower than in the product equation. Finally, H6 is not confirmed at standard significance levels, probably because the effect of internal factors (firm characteristics, experience and ability) is more important than is the level of competition.

We also find common results with respect to a control variable: the export activity. Export and innovation decisions are highly positively correlated. It seems that competition in foreign markets induces a higher propensity for both innovation activities. These results are highly robust across different specifications, but we should be cautious about their magnitude because of potential endogeneity. This problem could exert an effect, even including the first lag of the export indicator, if innovation is a previous step for exporting rather than the other way around.

In sum, the main aim of the paper is the determination of innovation process considering the experience effect of the firm (capacities, routines as organization) and the manager’s ability (skills, capability) as relevant elements. We show the importance of these elements using several alternatives of discrete choice models for panel data. Preliminary evidence indicates that there are different determinants (or effects) of innovation in the two equations. In fact, we find that experience or persistence in engaging in these activities are relevant, whereas other conditionings remain crucial determinants of the innovation frequencies even after controlling for experience.

Persistence in innovation, whatever form it may take, is crucial to the firm. It implies the existence of “dynamic capabilities of the firm” in which the concentration of knowledge accumulation and the underlying ability of firms is absorbed to produce and use the knowledge (Teece and Pisano, 1994). Moreover, those dynamic capabilities trigger the economies of scale from a threshold level of innovation activity associated by some organizational and internal factors.

The ability of the manager, as a proxy for firm-specific time invariant effects is another factor influencing the firm’s performance, unless persistence is considered. We also test several hypotheses and can conclude that even in an environment of managers with high propensities to innovate and firms developing experience in conducting these activities, some particular characteristics are needed in order be successful in the innovation strategy. Although the history of the innovation activities in the firm (firm experience) and the unobserved heterogeneity (manager’s ability) are two factors that, to the best of our knowledge are being accounted for in this paper for the first time, some of the internal and organization resources continue to be important in developing innovation activities.

Data Appendix

The database is provided by the Spanish Ministry of Science and Technology and involves 18,095 observations over the period 1990-99 on 2926 firms belonging to the manufacturing sector in Spain. These firms provide information on between 1 and 10 reporting periods, as shown in the figures in Table A.1. The descriptive statistics of the main variables for the subsamples of firms innovating either in product or in process are shown in Table A.2.

Table A.1. Structure of the unbalanced panel

Note. 1. Number of observations = 18,095. Number of firms = 2926.

Table A.2. Descriptive statistics of the main variables

1. Sample in each innovation type corresponds to the observations in the period 1990-1999. The percentages over the total number of observations are expressed in brackets. Notes.

References

1. Acs, Z.J. and Audretsch, D.B. (1987), “Innovation, market structure and firm size”. The Review of Economics and Statistics 71, pp.567-574.
2. Arrow, K. (1962). “Economic welfare and the allocation of resources for inventions”. In R.R. Nelson ed. The Rate and Direction of Inventive Activity. Princeton University Press.
3. Baldwin, J. (1998), “Innovation and entry”, in Z. Acs (ed.), Are Small Firms Important? Dordrecht: Kluwer.
4. Baldwin, J. Hanel and D. Sabourin (1999), “Determinants of innovation activity in Canadian manufacturing firms: The role of intellectual property rights”. Paper presented at the TSER Network on Innovation and Economic Change, Delft.
5. Barney, J.B. (1991), “ Firm resources and sustained competitive advantage”. Journal of Management 17, pp. 99-120.
6. Barney, J.B. and Hansen, G.S. (1994), “Trustworthiness as a source of competitive advantage”. Strategic Management Journal 15, pp.175-190.
7. Baysinger, B.D. and Hoskisson, R.E. (1990), “The composition of board directors and strategic control: Effects on corporate strategy”. Academy of Management Review 15, pp.71-87.
8. Blundell, R.W., Griffith, R. and Van Reenen, J. (1995), “Dynamic count data models of technological innovation”. The Economic Journal 105, pp. 333- 344.
9. Bound, J., C. Cummins, Z. Griliches, B.H. Hall and A. Jaffe (1984). “Who does R & D and who patents? In Z. Griliches (ed.) R & D, Patents and Productivity. Chicago, University of Chicago and NBER.
10. Bonano, G. and B. Haworth (1998), “Intensity of competition and the choice between product and process innovation”. International Journal of Industrial Organization 16, pp. 495-510.
11. Bozeman, B. and Link, A.N. (1983), Investments in technology: corporate strategies and public policy alternatives. New York: Praeger.
12. Brouwer, E. and Kleinknecht, A. (1996), “Determinants of innovation. A microeconometric analysis of three alternative innovation output indicators”, in A. Kleinknecht (ed.), Determinants of Innovation: The Message from New Indicators. Amsterdam: Macmillan. pp. 99-124.
13. Bughin, J. and Jacques, J.M. (1994), “Managerial efficiency and the schumpeterian link between size, market structure and innovations revisited”. Research Policy 23, pp. 653-659.
14. Cassiman, B. and Martínez-Ros, E. (2003), “Innovation driving export performance. Evidence from Spanish manufacturing”, mimeo.
15. Castanias, R.P. and Helfat, C.E.(1991), “Managerial resources and rents”. Journal of Management 17, pp. 155-171.
16. Cefis, E. (2003), “Is there persistence in innovative activities?”. International Journal of Industrial Organization 21, pp.489-515.
17. Chamberlain, G. (1984), “Panel data” in Handbook of Econometrics vol. II ed. Z. Griliches and M. Intriligator, pp. 1247-1318. Amsterdam: North-Holland.
18. Cohen, W.M. and Klepper, S. (1996a), “A reprise of size and R . Economic Journal 106, pp. 925-951.
19. Cohen, W.M. and Klepper, S. (1996b), “Firm size and the nature of innovation within industries: the case of process and product R & D”. Review of Economics and Statistics 78, pp. 232-243.
20. Cohen, W.M. and Levin, R.C. (1989), “empirical studies of innovation and market structure” in Handbook of Industrial Organisation, R. Schmalensee and R. Willig (Eds.). Amsterdam. Elsevier.
21. Collier, D. (1983), “Technology in diversified, decentralised companies”. Journal of Business Strategy 3, pp. 91-93.
22. Cooper, A. (1994), “R & D is more efficient in small companies”. Harvard Business Review 42, pp. 75-83.
23. Crepon, B. and Duguet, E. (1997), “Research and development, competition and innovation. Pseudo-maximum likelihood and simulated maximum likelihood methods applied to count data models with heterogeneity”. Journal of Econometrics 79, pp. 355-378.
24. D’Aspremont, C., and Jacquemin, A. (1988), “Cooperative and noncooperative R & D in duopoly with spillovers”. American Economic Review 78, pp. 1133-1137.
25. Delbono, F., and Denicolo, V. (1990), “R & D investment in a symmetric and homogenoeus oligopoly”. International Journal of Industrial Organization 8, pp. 297-313.
26. Dietrickx, I. and Cool, K. (1989), “Asset stock accumulation and sustainability of competitive advantage”. Management Science 35, pp. 1504-1511.
27. Entorf, H. and W. Pohlmeier, (1990), “Employment, innovation and export activity: evidence from firm level data”, in Florens, J.P., M. Ivaldi, J.J. Laffont and F. Laisney (eds.), Microeconometrics. Surveys and Applications, Oxford: Basil Blackwell. pp. 394-415.
28. Fama, E.F. and Jensen, M.C. (1983), “Separation of ownership and control”. Journal of Law and Economics 26, pp. 301-325.
29. Flaig, G. and Stadler, M. (1998), “On the dynamics of product and process innovations”. Jahrbücher für Nationalökonomie und Statistik 217, pp. 401-417.
30. Flood, P.C., Smith, K.A. and Derfus, P. (1996), “Guess editors’ introduction-top management teams: a neglected topic in strategic human resource management”. Ibar 17, pp.1-17.
31. Fritsch, M. and M. Meschede (2001), “Product innovation, process innovation, and size”. Review of Industrial Organization 19, pp. 335-350.
32. Galende, J. and Suárez, I. (1999), “A resource-based analysis of the factors determining a firm’s R & D activities”. Research Policy 28, pp. 891-905.
33. Geroski, P.A. Van Reenen, J. and Walters, C.F. (1997), “How persistently do firms innovate?”. Research Policy 26, pp.33-48.
34. Graves, S. and Langowitz; J. (1993); “Innovative productivity and return to scale in the pharmaceutical industry”. Strategic Management Journal 14, pp. 593-605.
35. Griliches, Z. (1990), “R & D, patent statistics as economic indicators: a survey”. Journal of Economic Literature 28, pp. 1661-1707.
36. Griliches, Z. and Mairesse, J. (1984), “Productivity and R & D at the firm level” in Z. Griliches (ed), R & D, Patents and Productivity. Chicago, University of Chicago Press, pp. 339-74
37. Hall, B.H. (1990), “The impact of corporate restructuring on industrial research and development” Brooking Papers on Economic Activity: Microeconomics, pp.85-124.
38. Hall, B.H., Griliches, Z. and Hausman, J. (1986), “Patents and R & D: is there a lag?”. International Economic Review 27 pp.265-283.
39. Hall, B.H. and Mairesse, J. (1993), “Exploring the productivity of research and development in French manufacturing firms”. National Bureau of Economic Research and Centre de Reserche en Economie et Statistique. WP 9326.
40. Hansen, G.S. and Hill, C.W. (1991), “Are institutional investors myopic? A time series study of four technology-driven industries. Strategic Management Journal 12, 1-16.
41. Hausman, J., Hall, B. H. and Griliches, Z. (1984), “Econometric models for count data and an application to the patents-R & D relationship”. Econometrica, 52, pp. 909-938.
42. Hitt, M., Hoskisson, R. E. and Ireland, R. (1990), "Mergers and acquisitions and managerial commitment to innovation in M-Form firms". Strategic Management Journal, 11 pp. 29-47.
43. Hoskisson, R.E., Hitt, M.A, Johnson, R.A. and Grossman, W. (2002), “Conflicting voices: the effects of ownership heterogeneity and internal governance on corporate strategy”. Academy of Management Journal 45, pp. 697-716.
44. Hoskisson, R.E and Hitt, M.A. (1988), “Strategic control systems and relative R & D investment in large multiproduct firms”. Strategic Management Journal 9, pp. 605-621.
45. Hoskisson, R.E. and Turk, T.A. (1990), “Corporate restructuring: governance and control limits of the internal market”. Academy of Management Review 15, pp. 459-477.
46. Kamien, M.I. and Schwartz, N.L. (1970), “Market structure, elasticity of demand, and incentive to invent”. Journal of Law and Economics 13, pp. 241-252.
47. Kleinknecht, A. (1989), “Firm size and innovation. Observations in Dutch manufacturing industries”. Small Business Economics 1, pp. 215-222.
48. Kraft, K. (1990), “Are product- and process-innovations independent of each other?”. Applied Economics 22, pp. 1029-1038.
49. Levin, R.C. and Reiss, P.C. (1989), “Cost-reducing and demand-creating R & D with spillovers”. Rand Journal of Economics 19, pp. 538-556.
50. Levin, R.C., Cohen, W.M. and Mowery, D.C. (1985), “R & D appropriability, opportunity and market structure: new evidence on some schumpeterian hypotheses”. American Economic Review Papers and Proceedings 75, pp.20-24.
51. Li, M. and Simerly, R.L. (1998), “The moderating effect of environmental dynamism on the ownership and performance relationship”. Strategic Management Journal 19, pp.169-180.
52. Lorsch, J.W. and MacIver, E. (1989), Pawns or potentates: The reality of America’s corporate boards. Boston: Harvard Business School Press.
53. Lunn, J. (1986), “An empirical analysis of process and product patenting: a simultaneous equation framework”. The Journal of Industrial Economics 34, pp. 319-330.
54. Malatesta, P.H. and Walking, R.A. (1988), “Poison pill securities: stockholder wealth, profitability, and ownership structure”. Journal of Financial Economics 20, pp. 347-376.
55. Malerba, F. (1985), The semiconductor business: The economics of rapid growth and decline. Madison: University of Wisconsin Press.
56. Martínez-Ros, E. (2000), “Explaining the decisions to carry out product and process innovations: the Spanish case”. The Journal of High Technology Management Research 10, pp. 223-242.
57. Martínez-Ros, E. and J.M. Labeaga (2002), “The relationship between firm size and innovation activity: a double decision approach”. The Economics of Innovation and New Technology 11, pp. 35-50.
58. Mata, F.J., Fuerst, W.L. and Barney, J.B. (1995), “Information technology and sustained competitive advantage: A resource-based analysis”. MIS Quarterly 19, pp 487-506.
59. Michalisin, M.D., Smith, R.D. and Kline, D.M. (1997), “In research of strategic assets”. International Journal of Organizational Analysis 5, pp. 360-387.
60. Milgrom, P. and Roberts, J. (1990), “The economics of modern manufacturing: technology, strategy and organization”. American Economic Review 80, pp.511-528.
61. Miravete, E.J. and Pernías, J.C. (2000), “Innovation complementarities and scale of production”. Mimeo.
62. Oliver, C. (1997), “Sustainable competitive advantage: Combining institutional and resource-based views. Strategic Management Journal 18, pp. 697- 713.
63. Pavitt, K. Robson, M. and Townsend, J. (1987), “The size distribution of innovating firms in the UK: 1945-1983”. Journal of Industrial Economics 35, pp. 297-316
64. Piergiovanni, R., Santarelli, E. and Vivarelli, M. (1997), “From which source do small firms ferive their innovative inputs? Some evidence from Italian industry”. Review of Industrial Organization 12, pp. 243-258.
65. Pine, B.J. Victor, B. and Boyton, A.C. (1993), “Making mass customisation work”. Harvard Business Review 1, pp. 108-119.
66. Porter, M.E (1980), Competitive Strategy. New York: Free Press.
67. Powell, T.C. (1992), “Strategic planning as competitive advantage”. Strategic Management Journal 13, pp. 551-558.
68. Powell, T.C. (1993), “Administrative skill as competitive advantage-extending Porter’s analytical framework”. Canadian Journal of Administrative Science 10, pp. 141-153.
69. Powell, T.C. (1997), “Information technology as competitive advantage: the role of human, business, and technology resources”. Strategic Management Journal 18, pp. 375-405.
70. Reinganum, J. (1989), “The timing of Innovation”. Handbook of Industrial Organization vol. I Schmalensee and Willig (eds.)
71. Rosenkranz, S. (1996), “Simultaneous choice of process and product innovation”. Discussion Paper nº 1321. CEPR. London.
72. Rosenkranz, S. (2003), “Simultaneous choice of process and product innovation when consumers have a preference for product variety”. Journal of Economic Behavior and Organization 50, pp. 183-201.
73. Scherer, F.M. (1992), “Schumpeter and plausible Capitalism”. Journal of Economic Literature 30 pp. 1416-1434.
74. Schumpeter, J.A. (1942), Capitalism, Socialism and Democracy. New York, Harper.
75. Schmookler, J. (1966), Invention and economic growth. Cambridge Mass: Harvard University Press.
76. Soete, L.G. (1979), “Firm size and inventive activity: the evidence reconsidered” European Economic Review 12, pp. 319-340.
77. Spence, A.M. (1975) “Monopoly, quality, and regulation”. Bell Journal of Economics, 6. pp. 417-429.
78. Sterlacchini, A. (1994), “Technological Opportunities, Intraindustry Spillovers and Firm R & D Intensity”, The Economics of Innovation and New Technology, 3, 123-37.
79. Teece, D. and Pisano, G. (1994), “The dynamic capabilities of firms: An introduction”. Industrial and Corporate Change 3, pp. 537-556.
80. Wernefelt, B. (1984), “A resource-based view of the firm”. Strategic Management Journal, 5 pp. 171-180.
81. Williamson, O. E. (1985), The Economics of Institutions of Capitalism: Firms, Markets, Relational Contracting. Free Press, New York.
82. Wright, P., Ferris, S.P., Sarin, A. and Awasthi, V. (1996), “Impact of corporate insider, blockholder, and institutional equity ownership on firm risk taking”. Academy of Management Journal, 39 pp. 441-463.
83. Yi, S. (1999), “Market Structure and Incentives to Innovate: the case of Cournot Oligopoly”. Economic Letters 65, pp. 379-388.

RELACION DE DOCUMENTOS DE FEDEA

DOCUMENTOS DE TRABAJO

References

1. 2005-16: “Persistence and ability in the innovation decisions”, José M. Labeaga y Ester Martínez-Ros.

References

1. 2004-15: “Measuring Changes in Health Capital”, Néboa Zozaya, Juan Oliva y Rubén Osuna.

References

1. 2005-14: “Discrete choice models of labour Supply, behavioural microsimulation and the Spanish tax reforms”, José M. Labeaga, Xisco Oliver y Amedeo Spadaro.

References

1. 2005-13: “A Closer Look at the Comparative Statics in Competitive Markets”, J. R. Ruiz-Tamarit y Manuel Sánchez-Moreno.

References

1. 2005-12: “Wellbeing and dependency among European elderly: The role of social integration”, Corinne Mette.

References

1. 2005-11: “Demand for life annuities from married couples with a bequest motive”, Carlos Vidal-Meliá y Ana Lejárraga-García.

References

1. 2005-10: “Air Pollution and the Macroeconomy across European Countries”, Francisco Álvarez, Gustavo A. Marrero y Luis A. Puch.

References

1. 2005-09: “The excess burden associated to characteristics of the goods: application to housing demand”, Amelia Bilbao, Celia Bilbao y José M. Labeaga.

References

1. 2005-08: “La situación laboral de los inmigrantes en España: Un análisis descriptivo”, Ana Carolina Ortega Masagué

References

1. 2005-07: “Demographic Uncertainty and Health Care Expenditure in Spain”, Namkee Ahn, Juan Ramón García y José A. Herce.

References

1. 2005-06: “EL NO-MAGREB. Implicaciones económicas para (y más allá de) la región”, José A. Herce y Simón Sosvilla Rivero.

References

1. 2005-05: “The real picture: Industry specific exchange rates for the euro area”, Simón Sosvilla-Rivero y Sonia Pangusión.

References

1. 2005-04: “A Residential Energy Demand System for Spain”, Xavier Labandeira, José M. Labeaga y Miguel Rodríguez.

References

1. 2005-03: “The Evolution of Retirement”, J. Ignacio Conde-Ruiz., Vincenzo Galasso y Paola Profeta.

References

1. 2005-02: “Housing deprivation and health status: Evidence from Spain” Luis Ayala, José M. Labeaga y Carolina Navarro.

References

1. 2005-01: “¿Qué determina el éxito en unas Oposiciones?”, Manuel F. Bagüés.

References

1. 2004-29: “Demografía y empleo de los trabajadores próximos a la jubilación en Cataluña”, J. Ignacio Conde-Ruiz y Emma García

References

1. 2004-28: “The border effect in Spain”, Salvador Gil-Pareja, Rafael Llorca-Vivero, José A. Martínez-Serrano y Josep Oliver-Alonso.

References

1. 2004-27: “Economic Consequences of Widowhood in Europe: Cross-country and Gender Differences” Namkee Ahn.

References

1. 2004-26: “Cross-skill Redistribution and the Tradeoff between Unemployment Benefits and Employment Protection”, Tito Boeri, J. Ignacio Conde-Ruiz y Vincenzo Galasso.

References

1. 2004-25: “La Antigüedad en el Empleo y los Efectos del Ciclo Económico en los Salarios. El Caso Argentino”, Ana Carolina Ortega Masagué.

References

1. 2004-24: “Economic Inequality in Spain: The European Union Household Panel Dataset”, Santiago Budría y Javier Díaz-Giménez.

References

1. 2004-23: “Linkages in international stock markets: Evidence from a classification procedure”, Simon Sosvilla-Rivero y Pedro N. Rodríguez.

TEXTOS EXPRESS

References

1. 2004-02: “¿Cuán diferentes son las economías europea y americana?”, José A. Herce.

References

1. 2004-01: “The Spanish economy through the recent slowdown. Current situation and issues for the immediate future”, José A. Herce y Juan F. Jimeno.