Busix

Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=ciai20

Industry and Innovation

ISSN: 1366-2716 (Print) 1469-8390 (Online) Journal homepage: https://www.tandfonline.com/loi/ciai20

Driving business performance: innovation complementarities and persistence patterns

Eleonora Bartoloni & Maurizio Baussola

To cite this article: Eleonora Bartoloni & Maurizio Baussola (2018) Driving business performance: innovation complementarities and persistence patterns, Industry and Innovation, 25:5, 505-525, DOI: 10.1080/13662716.2017.1327843

To link to this article: https://doi.org/10.1080/13662716.2017.1327843

Published online: 05 Jun 2017.

Submit your article to this journal

Article views: 311

View Crossmark data

Citing articles: 1 View citing articles

INDUSTRY AND INNOVATION, 2018 VOL. 25, NO. 5, 505–525 https://doi.org/10.1080/13662716.2017.1327843

Driving business performance: innovation complementarities and persistence patterns

Eleonora Bartolonia,b and Maurizio Baussolac

aISTAT, Italian National Institute of Statistics, Milano, Italy; bDipartimento di Scienze Economiche e Aziendali, Università di Parma, Parma, Italy; cDipartimento di Scienze Economiche e Sociali, UCSC, Università Cattolica del Sacro Cuore, Piacenza, Italy

ABSTRACT

Complementarities between technological and non-technological innovation are crucial determinants of firm performance. Although innovation complementarity has been extensively tested in the empirical literature, it has not been analysed in conjunction with innovation persistence. This fact is mainly due to the lack of data sets able to provide adequate longitudinal information. The capacities to developmarket-orientedbehaviourandintroduceneworganisational innovations, together with technological innovation, are the drivers of a firm’s productivity and profitability. We find that these activities complement technological innovation and that their impact is greater when they persist over time, thus introducing a more general concept of innovation persistence. We present an empirical model based on a large new panel of Italian manufacturing firms covering the period 2000–2012 which enables us to determine the precise impacts of a firm’s innovative attitude, in a broad definition that incorporates non-technological innovation and persistence, on its productivity and profitability.

KEYWORDS Technological and non-technological innovation; complementarities; European community innovation survey; profitability; productivity; unbalanced panel data

JEL CLASSIFICATIONS L25; 030; 032; 033

1. Introduction

The relationship between innovation and firms’ performance has long been debated within the economic and managerial literature. The former has focused on both macro- and microeconomic implications underlining, on the one hand, the role of innovation inputs (e.g.R&Dactivity)indetermininglong-runeconomicgrowth.Thisapproachcharacterised the early R&D endogenous growth models (Romer 1990; Aghion and Howitt 1992; Jones 1995).

On the other hand, the microeconomic approach has focused particularly on the empirical estimation of the impact of innovation on firms’ productivity (Geroski 1989; Crépon, Duguet, and Mairessec 1998; Lööf and Heshmati 2002), thus emphasising the methodological issues underlying such empirical investigations.

On the managerial side, particular emphasis has been devoted to the impact of a firm’s attitude of being an innovator (product and/or process) and, simultaneously, to its ability to be market-oriented (Narver and Slater 1990). This approach embraces a more comprehensive definition of an innovative attitude, which typically brings about other

CONTACT Eleonora Bartoloni [email protected] © 2017 Informa UK Limited, trading as Taylor & Francis Group

506 E. BARTOLONI AND M. BAUSSOLA

forms of non-technological innovations, i.e. organisational and marketing innovations. Indeed, these forms of innovation play a crucial role in affecting firms’ performance in terms of productivity and even profitability, in that the innovation process affects the internal allocation and use of resources, thus enabling innovating firms to be more responsive to changing market conditions (Geroski, Machin, and Van Reenen 1993).

All of these issues imply that for innovation to be effective, it should be persistent, thus enabling those continuously innovating firms to gain a premium with respect to peers that do not act accordingly. This view is also supported on theoretical grounds by theories addressing (i) the existence of sunk costs in innovation activities (e.g. R&D expenditures) (Stiglitz 1987; Mañez et al. 2009); (ii) the positive correlation with past successful innovations (success-breeds-success), which implies a positive impact on firms’ profitability and thus on their future ability to finance more innovative activities (Carpenter and Petersen 2002; Le Bas and Latham 2006); and (iii) the dynamic accumulation of knowledge or, in other words, the dynamic process of innovation that enables a firm to learn and adapt its innovation strategy (David 1992; Geroski, Machin, and Van Reenen 1993; Geroski, Van Reenen, and Walters 1997).

Innovation persistence provides a firm with the ability to exploit competitive advantages withrespecttocompetitorsandthustoearnprofitsthataresystematicallyhigherthanthose gained by non-innovating or at least only occasional innovating firms (Mueller 1992; Cefis 2003; Cefis and Ciccarelli 2005; Bartoloni and Baussola 2009).

However, the role of non-technological innovation has not been completely consid- ered in this framework. Indeed, non-technological innovation is crucially associated with technological innovation (e.g. product or process innovation) and generates technological activities related to new organisational and marketing activities, which affect the success of such new technological practices. In particular, process innovation and organisation innovation may be closely linked to one another, whereas product innovation may be more effectively related (although not exclusively) to marketing innovation.

Percival and Cozzarin (2008) and Evangelista and Vezzani (2010) show how com- plementarities between organisational and technological innovation affect firms’ perfor- mance; in a more recent study Bartoloni and Baussola (2016) underline how emphasis on technological innovation alone is misleading, and that the ability to adopt marketing innovation positively affects firms’ profits.

We propose an empirical investigation in which we explicitly consider the role of persistenttechnologicalandnon-technologicalinnovationsinaffectingfirms’performance in terms of productivity and profitability. We use a panel of Italian manufacturing firms over the period 1998–2012 derived from the Community Innovation Survey and matched with administrative data that enabled us to obtain information on firms’ balance sheets.

The paper is therefore structured as follows. In Section 2, we provide the interpretative framework used to develop the empirical analysis. In Section 3, we describe the charac- teristics of the data-set, we present the empirical model in Section 4, and the results are discussed in Section 5. Section 6 concludes the paper.

INDUSTRY AND INNOVATION 507

2. The interpretative framework

The debate on the persistence of innovation has typically analysed the role of persistent activities as measured by R&D (input) or patents (output) and, to a lesser extent, by technology adoption without considering the role of non-technological innovation.

Cefis (2003) and Cefis and Ciccarelli (2005) analyse the impact of a pattern of persis- tent innovation on firms’ profitability by using patent data, and suggest that persistent behaviour brings about higher profits compared with those achieved by companies that are non-persistent innovators. Johansson and Lööf (2010) adopt an input measure of persistent innovation, referring to the impact on sales, productivity and exports of firms’ long-term R&D strategy. Persistent innovation affects economic performance, and this impact is significantly higher compared with a strategy of occasional innovation.

Raymond et al. (2010) attempt to test whether innovation output, endogenously determined by the decision to undertake R&D, is affected by previous values, thus verifying whether persistence may be partially spurious. In addition, as their data-set is a balanced panel of Dutch firms covering three consecutive Community Innovation Surveys during the 1990s, they are able to take into account initial conditions and adopt a dynamic specification of the model. Using this approach, they disentangle the persistence effect and verify whether a spurious persistent effect does exist. Their results suggest that this is indeed the case, in that persistent innovation activity (either process and/or product innovation) exists only when initial conditions are taken as exogenous. Once endogenised and unobserved individual effects have been taken into account, the persistence hypothesis is rejected.

However, a milder persistence effect is observed concerning innovation output, as the past share of innovative sales does indeed affect the current share. It is worth noting, however, that when they use an innovation input measure, i.e. R&D and the share of R&D expenditures as a ratio of total sales, persistence is observed, thus confirming other evidence for manufacturing and service companies in other economies (Peters 2009).

The impact of innovation on firms’ performance may be analysed with respect to both the input and output of the innovation process. Typically, the former is considered by using R&D expenditure as a proxy for knowledge capital, which therefore contributes, akin to other production inputs, to output growth. Innovation input is also considered, focusing on the adoption of new process technology, which implies the use of new and more efficient capital goods.

This approach has been particularly developed within the endogenous growth theoret- ical setting (Romer 1990), in which an R&D sector interacts with a manufacturing sector producing new capital goods and final output. The model implies an equilibrium growth path crucially depending on the resources allocated to R&D.

Innovation output is considered the key variable for increasing productivity in the seminal study by Crépon, Duguet, and Mairessec (1998). By adopting a Cobb–Douglas production function framework, the authors derive a simultaneous equation model that links productivity, innovation output and R&D spending among a cross section of indus- trial firms. They find that innovation output, as proxied either by the share of innovative sales or by patent counts, is positively and strongly affected by R&D. This model, which has inspired an increasing number of studies based on the same methodological approach, focuses on the empirical tools required to overcome the bias related to information being available only for innovative firms when using innovation surveys.

508 E. BARTOLONI AND M. BAUSSOLA

Lööf and Heshmati (2002) use such an approach to develop an empirical analysis of knowledge capital and productivity at the firm level for a sample of Swedish firms par- ticipating in the national Community Innovation Survey. They emphasise how intangible assets are crucial in affecting the results, thus underlining the implicit relevance of their measurement issue.

Another branch of the literature has focused, instead, on panel data investigations to address causality issues (Rouvinen 2002; Frantzen 2003; Battisti, Mourani, and Stoneman 2010), finding support for a causal link running from R&D to productivity.

In our empirical specification, we focus on the relationship between productivity – as measured by real value added per worker – and production inputs while also accounting for the effect of persistently adopted technological and non-technological innovations.

The inclusion of non-technological innovation draws on the Schumpeterian view that new management methods represent another form of innovation (Schumpeter 1934). This dimension has been explored in the current empirical debate dealing with innovation complementarities. In particular, it has been argued that marketing and organisation activities may crucially affect firms’ performance, with the link between technological and non-technological (i.e. managerial and organisational) innovation being analysed by using micro-data typically derived from innovation surveys (e.g. the CIS) (Bartoloni and Baussola 2016; Battisti and Stoneman 2010; Schubert 2010).

Hollenstein (2003) achieves results that are consistent with these findings although he refers to a large sample of Swiss service companies and uses a methodological approach based on cluster analysis. Labour productivity or sales growth are crucially affected by dif- ferent innovation modes, implying the different structural and organisational properties of firms. However, such performance indicators are additionally and significantly influenced by human capital and knowledge capital, thus underlining the need to provide a better understanding of the accumulation of knowledge within and between firms.1

This stream of investigation falls within the debate on innovation complementarities originated by the seminal paper by Milgrom and Roberts (1990). In particular, Mohnen and Röller (2005) set-up an empirical framework in which different complementarity hypotheses are tested for. Following this line of investigation, many authors proposed different tests for complementarities, in particular concerning organisational strategies and technological innovation.

Cozzarin and Percival (2006) tested the impact of various organisational strategies on labour productivity and profitability, also controlling for industry effects and firm size. The results suggest that engaging in organisational strategies involving hiring skilled people and promoting the firm and product reputation, or stimulating R&D and focusing on market or reputation, are pairwise complements and thus have a significant and positive impact either on profits or on productivity.

Also, they find that focusing on hiring highly qualified personnel together with at- tempting to produce a world-first innovation may reduce profits. This fact may depend on the simultaneous combination of the relatively high hiring and innovation costs. However, complementarities vary significantly by firm size and sector once the analysis is concentrated specifically on a subsample of firms by industry and firm size (Percival and Cozzarin 2008).

1This issue, which is beyond the scope of our investigation, is however relevant for identifying possible sources of economic growth that are not accounted for (Arrighetti, Landini, and Lasagni 2015; Montresor and Vezzani 2016).

INDUSTRY AND INNOVATION 509

Product, process and organisational innovation are jointly investigated by Evangelista and Vezzani (2010) using the fourth wave of the Italian Community Innovation Survey (CIS). Complementarities in product, process and organisation innovation enable firms to gain a competitive advantage as measured by the impact on the turnover growth rate, also suggesting that the impact is stronger in manufacturing compared with services. However, the purely cross-sectional nature of the data-set does not allow for testing the robustness of this finding over time.

Complementarities between human resource management and innovation activities may have an impact on firms’ innovativeness and productivity, as highlighted by studies in diverse contexts.

Laursen and Foss (2003) and Antonioli, Mazzanti, and Pini (2010) test this hypothesis for a large sample of Danish and Italian firms, and suggest that firms’ innovativeness is positively affected by human resource management in which personnel training is a key element of a firm’s innovative success and, in the empirical test proposed by Antonioli, Mazzanti, and Pini (2010), productivity.

Profitability is the other aspect of performance. Its relationship with innovation has received less attention compared with the analysis of the determinants of productivity, particularly in recent years. The traditional approach to analysing firms’ profitability is based on the structure–conduct–performance (SCP) paradigm (Bain 1956), in that a firm’s performance is determined by structural characteristics of the industry. In contrast to this approach, the so-called firm efficiency view (Demsetz 1973; Peltzman 1977) emphasises the role of firms’ characteristics in determining their profits. However, empirical studies have generated controversial results, which crucially depend on the characteristics of the data- set used to implement such tests. Slade (2004), Allen (1983), and Delorme et al. (2002) find support for the SCP approach, whereas Roberts (1999, 2001) and Hawawini, Subramanian, and Verdin (2003) recognise the role of managerial capabilities in determining profitability.

Bartoloni and Baussola (2009) emphasise that the traditional SCP effect, although it was verified in a large panel of Italian manufacturing firms in the 1990s, had a very mild effect on profitability and its persistence, whereas firms’ innovative behaviour was more relevant in this respect.

The impact of innovation on profitability has also been analysed in the framework of technology adoption. Geroski, Machin, and Van Reenen (1993) emphasise not only the role of adoption per se but also that such a decision implies a full process that involves other choices and actions within a firm (e.g. organisational changes) that determine different internal allocations of resources.

Mueller and Cubbin (2005) emphasise how technological adoption provides a com- petitive advantage to innovating firms, thus enabling them to increase their profitabil- ity. Technology adoption and profitability are considered in a dynamic perspective by Stoneman and Kwon (1996). They emphasise that multiple adoption may occur, and firms may thus introduce new technologies at different points in time. Profitability – as in the case of technological adoption – should be considered along the diffusion path together with the distinction between older and more recent innovations, as the former are more exposed to greater competition, thus affecting profitability.

Within this interpretative framework, our aim is therefore to conduct an empirical analysis in which the main factors discussed are considered as determining a firm’s performance, and then to test whether: (a) persistent technological and non-technological

510 E. BARTOLONI AND M. BAUSSOLA

innovation enables firms to experience a significant increase in productivity compared with firms that do not innovate persistently; (b) joint occasional technological and non- technological innovation enables firms to experience an increase in productivity which is, however, lower than that achieved by persistent joint innovators. These hypotheses also implytestingforcomplementaritybetweentechnologicalandnon-technologicalbehaviour in both the persistent and occasional modes.

3. Panel data description

Our main data source is the Micro-Manu dataset,2 an unbalanced panel of Italian manu- facturing firms linking consecutive waves of the Italian Community Innovation Survey – which forms part of the EU science and technology statistics and is conducted every two years – with the ASIA archive (Statistical Register of Active Businesses)3 and an administrative data source providing balance sheets and income statements for those firms included in the CIS samples of respondents. The richness of this data-set allows one to enlarge the set of economic indicators typically explored in the innovation survey micro-data and to derive a set of financial and efficiency ratios that are not included in the CIS questionnaire. In accordance with international standards (OECD-Eurostat 2005), firms are classified by their type of innovation activity (technological and non- technological). Information on non-technological aspects of innovation (new marketing and/or organisational methods) allows one to consider comprehensive innovative activities by focusing on the reciprocal interactions between different aspects of innovation.

To analyse firms’ innovative pattern in a longitudinal context, we select an unbalanced panel of firms from the original data-set responding to at least two consecutive non- overlapped4 CIS waves (CIS1, years 1998–2000; CIS2, years 2002–2004; CIS4 years 2006– 2008; and CIS6, years 2010–2012). We have more than 3000 firms, corresponding to nearly 8000 observations over the whole period 1998–2012.

A strictly technological innovating firm is defined as one that has implemented an innovation only in the technological domain (i.e. a product and/or process innovation, with the exclusion of other non-technological forms of innovation) during the observed period. A complementary innovating firm is defined as one that has innovated in all the technological and non-technological domains (product and process and organisation and marketing). We distinguish between persistent and occasional innovative profiles in both the technological and complementary domains by defining (i) a persistent innovator as one that has innovated in at least two consecutive CIS periods (pers_tech and pers_tech_ntech) and (ii) an occasional innovator as one that has innovated at least once during the entire time span but never in two consecutive periods (tech and tech_ntech).

It is worth noting that the specific nature of the CIS’s sampling design gives rise to potential selection bias when using a longitudinal framework. Indeed, whereas large firms with more than 250 employees are selected on a census basis, small firms are randomly selected, and this sampling mechanism may negatively affect the probability of a firm

2The Micro-Manu dataset is a result of collaboration between the Italian National Institute of Statistics (ISTAT, Regional office for Lombardy) and the Catholic University of the Sacred Hearth.

3This archive is the most relevant administrative register used by ISTAT as the basis for many sample surveys and even census investigations.

4A characteristic that merits attention is that the measurement of the degree of innovation persistence may be over- estimated when two consecutive waves are partially overlapped.

INDUSTRY AND INNOVATION 511

Table 1. Unbalanced panel of manufacturing firms with non-missing accounting information (CIS1, 1998–2000; CIS2, 2002–2004; CIS4, 2006–2008; CIS6, 2010–2012).

Patterns of presence Obs. No. of firms (average) Size (no. of employees, median)

0011 725 363 115 0110 577 289 74 0111 574 191 281 1011 287 96 450 1100 3331 1666 37 1101 633 211 73 1110 747 249 116 1111 1049 262 365 Total 7923 3326 79

Firms by innovative behaviour (sample proportion)

Patterns of presence tech_ntech pers_tech_ntech tech pers_tech non_inn

0011 0.16 0.34 0.10 0.03 0.26 0110 0.15 0.23 0.12 0.04 0.33 0111 0.12 0.44 0.08 0.07 0.21 1011 0.19 0.48 0.07 0.05 0.11 1100 0.16 0.16 0.09 0.02 0.33 1101 0.20 0.24 0.09 0.02 0.27 1110 0.13 0.36 0.07 0.03 0.22 1111 0.06 0.62 0.03 0.11 0.11 Total 0.14 0.30 0.08 0.04 0.26

Notes: The patterns of inclusion indicate absence (0) or presence (1), during the four consecutive innovation surveys. Innovative behaviour: tech – the firm has innovated occasionally only in the technological domain; pers_tech – the firm has innovated persistently only in the technological domain; tech_ntech – the firm has innovated occasionally in both the technological and non-technological domains; pers_tech_ntech – the firm has innovated persistently in both the technological and non-technological domains; non_inn – the firm has never innovated during the observed time span.

being selected in consecutive surveys. Table 1 reports descriptive statistics for each ‘feasible’ pattern of inclusion5 relative to the relevant outcomes of a firm’s innovative activity. Hence, we can observe, for example, that the mean size of firms that are present only in the first two waves is 37 employees, but the size increases to 365 employees when the balanced sample of firms present in all four waves is considered. If we decided to retain this restricted group, we could define a persistent innovator in a more stringent way (i.e. as one that has continuously innovated during a four-period time span). However, by following this approach, we would probably confine our analysis to those firms with higher innovative propensity, with possible bias as a result. On the basis of this consideration, we decided to base our empirical investigation on the full set firms appearing in the unbalanced panel.

It is worth emphasising that balance sheet information for the period 1998–2012 is provided on a yearly basis, whereas the qualitative variables derived from the CIS survey are defined on a three-year basis. To address the problem of different information timing, we averaged accounting information over a three-year period; thus, the economic and financial indexes are provided as average values over the reference CIS time span. One should note that the full samples of firms from the CIS surveys also include small individual firms for which balance sheet information is not available from the Italian public register; thus, our analysis excludes these firms. We have compared the final sample of firms for which there is complete accounting information to the initial CIS samples in the ‘feasible’

5According to the methodology proposed by Raymond et al. (2009), a pattern is ‘feasible’ when the dynamics of innovation are potentially observable. This implies that a firm must be present in at least two consecutive CIS waves.

512 E. BARTOLONI AND M. BAUSSOLA

panel and then concluded that the loss of sampling units due to the use of out-of-sample information is negligible.6 The variables used in the empirical model are described in greater detail below.

Economic performance. We use a measure of operating profitability, return on sales (ros), that is appropriate for investigating the profitability generated by the core business of a manufacturing firm and a measure of labour productivity (Y), which is given by the value added per employee ratio and may be considered an intermediate measure of a firm’s innovation success.7

Financial efficiency indexes. Financial efficiency can be considered by using a measure of a firm’s exposure to external financing sources (lev), which is given by the ratio of shareholders’ funds to total debt, thus reflecting the extent to which a firm uses internal resources instead of borrowing to finance its activity.

Capital deepening. The role of physical capital is captured by considering the capital-to- labour ratio (K, tangible fixed assets per employee). It measures the extent of capital deepening in fostering productivity. Typically, the impact of this variable on labour productivity may be derived from growth accounting exercises, together with the impact that may be exerted by Total Factor Productivity (TFP). Instead, we test its impact by using an econometric approach, which enables us to consider other possible determinants related, in particular, to a firm’s innovative effort. One should note that capital deepening may also incorporate process innovation; this latter determinant typically implies the acquisition of new machinery.8

Innovation input. As noted above, together with physical capital, a firm’s innovative effort should be considered when describing the core determinants of labour productivity. The proxy that we use, R&D activities, may also be considered a proxy for knowledge capital, which can contribute directly to labour productivity growth and exert a positive influence through TFP growth. Because we refer to the entire sample of innovative and non-innovative firms, the aforementioned information is not available for this latter group of firms, given the characteristics of the CIS survey. Therefore, we use a dummy variable indicating whether a firm has undertaken R&D activity in at least two consecutive periods (pers_R&D).9 Thus, the impact of R&D may be considered a shifting parameter in the adopted specification (see the following Section 4).10

Innovation output. The aim of our investigation is to explore the complementary role of technological and non-technological aspects of innovation in determining a firm’s performance relative to innovation that is strictly technological.

We consider marketing and organisational innovation jointly, as these two innova- tive behaviours interact almost simultaneously. As suggested by the market orientation

6Considering the entire period, the manufacturing firms included in the selected CIS waves with balance sheet information are on average almost 80% of the total number of respondents. The Micro-Manu dataset includes more than 90% of the total number of manufacturing limited companies.

7We are aware that the relationship between innovation and productivity produces diverse empirical results. However, followingMohnenandHall(2013), innovationleadstoanincreaseinproductivity,althoughit isnotpossibletodisentangle the price and output effects on growth, given the characteristics of the available data sets.

8This argument is also considered in Hall, Lotti, and Mairesse (2009), who estimate a productivity equation that depends on product and process innovation together with fixed investment.

9Otherwise, a different modelling strategy would have been applied, i.e. focusing only on innovative firms or using a Tobit model with a selection equation. This approach, however, is beyond the scope of our investigation, the aim of which is to specify the different behaviour and performance of innovative and non-innovative firms.

10As it is clarified in the next Section, the inclusion of a persistent R&D dummy variable, which excludes occasional R&D, also depends on the adopted empirical specification.

INDUSTRY AND INNOVATION 513

literature (Slater and Narver 1995), this implies that the creation of superior customer value entails an organisational commitment to learning, information gathering and coordination of consumers’ needs. In other words, market orientation involves a redefinition and easing oftheadministrativeprocesswithinacompany,andthusultimatelyinvolvesorganisational change.11

We aim to reveal the presence of possible performance gains that may be earned by firms developing innovation continuously over time compared with occasional innovators. Thus, we consider the four different proxies for a firm’s attitude towards innovation that are described above. As in the case of R&D, these variables enter the productivity equation as factors that shift the production function (shifting parameters).

Other firm-specific characteristics. Firms’ age (years, log values)12 may positively affect their growth if older companies experience better access to external financing, higher capitalisation and more qualified workforce. Haltiwanger, Lane, and Spletzer (1999) find that age is positively associated with a firm’s productivity level, thus exerting an indirect effect on profitability. However, empirical results are controversial, as suggested by Coad, Segarra, and Teruel (2013), in that this clear-cut relationship is not observed within a large longitudinal sample of Spanish manufacturing companies.

Another two variables – available from the CIS survey – reflect a firm’s ownership structure and its propensity to internationalise. Thus, we use two dummy variables: the first indicates whether a firm belongs to a corporate group (gp), and the second indicates whether a firm sells its products in the international market (intern). The first variable may affect a firm’s efficiency, whereas the latter is closely related to the ability to expand internationally and thus increase turnover.

Sectoral structure and localisation. Industry-specific characteristics are accounted for by considering two sectoral dummies that, in line with the Pavitt taxonomy, identify the high and medium-high-technology sectors (pavitt_mh) and the low and medium-low- technology sectors (pavitt_ml). Geographical characteristics are captured by four regional dummies (nwest, neast, centre, south), reflecting a firm’s location in the north-west, north- east, central or southern regions of Italy.

Additionally, we consider the cr5 ratio to capture the SCP mechanism described in Section 2 and the ratio of the sectoral number of technological innovating firms to the total number of firms in that sector (sect_inntech). Descriptive statistics on the full set of variables are reported in Appendix 1.

4. The empirical model

Wemodelproductivityandprofitabilityusinganempiricalspecificationthatcanbederived from an augmented production function and a profit function.

In particular, productivity, which is defined in terms of real value added per employee, may be derived from Equation (1), assuming constant returns to scale.13

11Also, it is worth noting that disentangling product, process, organisation and marketing innovation over four successive CIS surveys may imply the loss of a significant number of observations because companies may not persist in innovation in the same disaggregated way. We therefore prefer to maintain a wider definition, enabling us to preserve an appropriate longitudinal data set, which is however consistent with the interpretative framework we have described.

12This variable is available from the Statistical Register of Active Businesses (ASIA). 13One can specify this equation without imposing constant return to scale. We also estimated such a specification, which provides, however, similar results in terms of capital and shifting factor parameters. A Wald test for constant returns to scale is rejected, but returns to scale are only slightly increasing. Given these issues, we prefer a specification that enables

514 E. BARTOLONI AND M. BAUSSOLA

yit = ait + βkit + uit (1)

where y is the log of per capita real value added of firm i, k is the log of physical capital per employee, and ait is a shifting factor that depends on a firm’s attitude towards technological and non-technological innovation and R&D effort. This latter factor also depends on other firms’ characteristics that may be relevant in shifting productivity. uit is a one-way error component:

uit = μi + �it (2) where:

μi ∼ IID(0, σ 2u ) and �it ∼ IID(0, σ 2� ) (3) are independent of each other and themselves. In addition, the error term �it is assumed to be white noise, that is:

E(�it, �is) = 0 for t �= s (4)

We account for the persistent innovative attitude of a firm by adopting the definition described in the previous section, i.e. a firm is considered a persistent innovator – from both thestricttechnologicalandcomplementaryperspectives(thusincludingnon-technological innovation) – if it has adopted such innovations in at least two consecutive innovation surveys. The persistent R&D effort may be described in the same way, thus defining a persistent R&D firm as one that has undertaken R&D activities over at least two consecutive surveys.Wecanthereforeusetwodifferentdummyvariablestorepresentafirm’spersistent innovative attitude from both an innovation input and output perspective.14

In addition, ait depends on a firm’s specific characteristics, i.e. age, being part of a group, sectoral innovative characteristics and location. Thus, we can define ait as follows:

ait = γ0 + γ1 I it + γ2 Xit (5)

where Iit represents a firm’s innovation attitude and Xit is a vector of firms’ additional characteristics that may affect productivity.

The profitability equation is derived while accounting for both traditional SCP effects and firm efficiency view considerations. Additionally, we account for the role of innovation by considering its effect on productivity and, through the latter, on profitability.

Thus, the empirical specification may be represented as follows:

yit = γ0 + γ1 I it + γ2 Xit + βkit + τTt + uit (6) rosit = α0 + α1yit + α2cr5it + α3levit + α4internit + α5sect_inntechit + vit (7)

where Tt is a time dummy common to every firm and refers to a three-year time span and vit is a one-way error component.

The time variable we consider refers to a three-year time span, i.e. the time interval of the CIS survey, as discussed in Section 3. The estimates therefore refer to contemporaneous relationships over a three-year time span. We are aware of a possible endogeneity issue

us to explicitly consider the capital deepening factor – which may include a firm’s innovative attitude – as a determinant of productivity. Otherwise, we would have had to consider capital and labour separately, thus losing such an interpretation.

14See the variable description in Section 3.

INDUSTRY AND INNOVATION 515

related to the innovative variables; however, given such a time interval, we can also specify a model in which the innovative variables are treated as predetermined, i.e. they may be thought of as independent of current disturbances uit. In other words, we can also introduce a calendar time lag between innovation and balance sheet information, in that the former precedes the latter. Thus, the innovation variables refer to the conventional time t associated with the three-year time span of the CIS Survey, whereas the economic performance variables refer to the time averages covering the three years after the CIS Survey. Given a firm’s innovative behaviour at time t, we can estimate its effect on productivity and profitability at a later calendar time.

In addition, we are aware of possible correlation between the innovation variables and the individual error component, and so we also estimate Equation (2) by using the predicted outcomes of the innovation variables derived from logit models that explain innovation propensities in terms of firm and sectoral characteristics These estimations follow previous studies in which such determinants have been successfully used to derive a firm’s innovative behaviour (Bartoloni 2012), and are reported in the Appendix 2.

From Equations (6) and (7), it appears that the model may be thought of as a recursive system because the matrix of endogenous variables is triangular. Productivity does affect profitability and not vice-versa. In this case, OLS estimates are appropriate, provided that the model is also diagonal recursive, i.e. stochastic disturbances are not correlated.15

Specifically, the productivity equation includes the following explanatory variables:

• a dummy variable reflecting a firm’s attitude towards persistent (occasional) inno- vation (pers_tech, pers_tech_ntech, tech, tech_ntech, depending on the specific case), which is included in the I vector of variables in Equation (7);

• another dummy variable that is also included in the I vector, reflecting whether a firm has persistently undertaken R&D activities (pers_r&d). This variable also reflects a firm’s absorptive capacity, as discussed in Cohen and Levinthal (1990), and its attitude towards sustaining this capability over time;

• physical capital deepening (k); • sectoral innovation characteristics (pavitt_mh and pavitt_ml); • localisation (nwest, neast, centre and south) and other firm-specific characteristics (age and gp).

All variables except for physical capital may be considered as shifting factors for a firm’s production function, as we have previously discussed.16

The explanatory variables in the profitability equation represent, on the one hand, the SCP mechanism (industry concentration) and, on the other, firms’ characteristics related to subjective efficiency (leverage), the ability to sell products on international markets and productivity.Thislattervariablealsoreflectsafirm’sabilitytocompetethroughinnovation, as productivity is crucially affected – as shown in Equation (6) – by a firm’s innovative attitude.17

15We also estimated a SURE model to account for such a correlation. The results are very similar to the OLS estimates, thus suggesting that such a correlation is feeble and that the use of OLS is therefore appropriate.

16Firm size is not considered because – when included – the R&D variable becomes insignificant, as these variables are strictly related in our sample of manufacturing firms. We decided to use the pers_r&d dummy variable because it enters our empirical specification as a shifting factor of the productivity equation and it reflects a firm’s long-term commitment to invest in innovation activities.

17We have not included an innovative dummy reflecting a firm’s innovative attitude in the adopted profitability specification, as it was not significant in regressions in which it was included. Indeed, the productivity variable does

516 E. BARTOLONI AND M. BAUSSOLA

We also include a sectoral variable to reflect the possible effects on profitability related to the number of innovative firms in each industry. This is a proxy for new technological opportunities brought about by the increase in an industry’s technological knowledge. In this framework, two different mechanisms are operational. On the one hand, we can have a positive effect as an increasing number of sectoral innovators increases a firm’s probability of introducing an innovation (epidemic effect) (Mansfield 1968). This fact may have a positive effect on profitability. On the other hand, this information effect may be offset by a competitive mechanism that implies that the number of competitors in an industry increases, thus squeezing the profits of firms operating in the same market (stock effect) (Karshenas and Stoneman 1993). Thus, the explanatory variables entering the profitability equation are the following:

• market structure (cr5); • financial efficiency (lev); • ability to sell products on international markets (intern); • productivity (y); • technological spill-over (sect_inntech).

5. Results

Table 2 presents the estimates over the entire period, taking previous considerations into account; thus model (1) refers to the base specification, model (2) refers to the specification in which the innovative variables are treated as predetermined, and model (3) specifies these variables as endogenous and so predicted outcomes are endogenised.

The estimates are derived by applying random effect (RE) estimation techniques to the system of Equations (6) and (7).18

With reference to the results, the productivity equation shows that persistent technolog- ical and non-technological innovation increases productivity with an impact that ranges from 13.4% (model 2) to 6.2 % (model 3) compared with non-innovative firms, which form the reference group. Firms that use only persistent technological innovation do not experience a significant increase in productivity in models (1) and (3).

Joint but non-persistent innovation has a positive and significant impact on productiv- ity, although milder compared with persistent innovation. This evidence holds in models (1) and (2) but not in (3), where the impact is not significant. This result depends crucially on the fact that model (3) uses predicted outcomes derived from a logit regression which is less satisfactory in modelling the occasional innovative behaviour. However, the impact of persistent technological and non-technological innovation is significantly higher with respect to occasional innovation, as shown by coefficient values and Wald tests (Table 3).

We also test for complementarity, considering both persistent and non-persistent innovation, in Table 3. These tests suggest that complementarity is confirmed when considering persistent innovation; in other words, persistent technological and non tech-

incorporate a firm’s innovative attitude, which therefore determines the non-significant effect of such an innovative dummy variable.

18The choice of the RE specification depends on the need to control for the effect of time-invariant variables such as regional localisation and industrial sector, and also the persistent innovative variable. In addition, when the target population is large, as in our case, and the selected sample may not be fully representative regarding all the characteristics under investigation, it may be preferable to adopt a random effect model as this permits generalisation of the inferences beyond the sample used in the model.

INDUSTRY AND INNOVATION 517

Table 2. Firms’ economic performance – period 2000–2012.

Productivity RE

Variables 1 – Base model 2 – Predetermined innovation 3 – Endogenous innovation Profitability RE

pers_tech_ntech 0.122*** 0.134*** 0.0617*** [0.0207] [0.0266] [0.0185]

pers_tech 0.0222 0.0781** 0.0253 [0.0373] [0.0392] [0.0186]

tech_ntech 0.0362*** 0.0594*** −0.0053 [0.0114] [0.0176] [0.0159]

tech 0.0524*** 0.0572*** −0.0012 [0.0144] [0.0216] [0.0307]

y 0.0964*** [0.00336]

cr5 0.000366*** [7.83e−05]

sect_inntech −0.00112*** [6.57e−05]

lev 0.0109*** [0.00145]

intern −0.0123*** [0.00191]

k 0.178*** 0.162*** 0.176*** [0.00750] [0.00727] [0.00755]

pavitt_ma 0.118*** 0.107*** 0.122*** [0.0144] [0.0163] [0.0152]

age 0.0485*** 0.0445*** 0.0478*** [0.0104] [0.0126] [0.00937]

pers_r&d 0.0512*** 0.0407* 0.0282*** [0.0195] [0.0237] [0.0201]

gp 0.0914*** 0.116*** 0.0908*** [0.0109] [0.0137] [0.0104]

nwest 0.244*** 0.204*** 0.252*** [0.0235] [0.0265] [0.0210]

neast 0.207*** 0.179*** 0.212*** [0.0230] [0.0252] [0.0219]

centre 0.197*** 0.151*** 0.203*** [0.0274] [0.0303] [0.0248]

d2000 −0.0107 0.136*** −0.0028 [0.0109] [0.0159] [0.0121]

d2004 −0.0220** 0.0968*** −0.0181* [0.00987] [0.0147] [0.01000]

d2012 −0.0355*** 0.116*** −0.0370*** [0.0113] [0.0150] [0.0103]

Constant 8.593*** 8.655*** 8.653*** −0.894*** [0.0847] [0.0869] [0.0840] [0.0351]

Observations 7923 7923 7923 7923 R2 0.323 0.238 0.319 0.295 within 0.040 0.031 0.039 0.416 between 0.364 0.284 0.361 0.273 ρ 0.682 0.564 0.682 0.665 σμ 0.348 0.377 0.347 0.053

Notes: The variables y, k and age are in log values. In order to perform complementarity tests, two additional dummy variables indicating whether a firm has innovated occasionally or persistently in the non-technological domain have been added in the productivity regressions. In models 1 and 2 for productivity and in the profitability model robust standard errors are reported in brackets. In model 3 for productivity we use predicted events for pers_tech_ntech, pers_tech, tech_ntech and tech derived from logistic regressions as shown in Appendix 2. Bootstrapped standard errors in brackets. ***p < 0.01, **p < 0.05, *p < 0.1. ρ is an estimation of the contribution of unobserved heterogeneity to the total unexplained variance. σμ is the estimated standard error of the random effect component μi .

518 E. BARTOLONI AND M. BAUSSOLA

Table 3. Wald tests for innovation complementarity and equality between coefficients.

Complementarity tests: C11 ≥ C10 + C01 1 – Base model pers_tech_ntech vs. pers_tech 7.95*** > 0 (p = 0.997) tech_ntech vs. tech 2.49a > 0 (p = 0.942) 2 – Predetermined innovation pers_tech_ntech vs. pers_tech 2.29c > 0 (p = 0.934) tech_ntech vs. tech 0.28 > 0 (p = 0.701) 3 – Endogenous innovation pers_tech_ntech vs. pers_tech 8.08*** > 0 (p = 0.997) tech_ntech vs. tech 0.02 –

Test for equality between the coefficients of the innovation variables 1 – Base model 2 – Predetermined innovation 3 – Endogenous innovation

pers_tech_ntech vs. pers_tech 7.26*** 2.09c 20.46*** pers_tech_ntech vs. tech 9.97*** 7.85*** 17.61*** pers_tech_ntech vs. tech_ntech 17.66*** 8.65*** 17.60***

Notes: Following Mohnen and Röller (2005) the complementarity test is based on the following null hypothesis: C11 ≥ C10 + C01 where: C11 indicates a joint technological and non-technological innovation; C10 and C01 indicate the introduction of, respectively, a technological and a non-technological innovation in isolation. A Wald χ2 one-sided test is run in two steps. The first step tests the null hypothesis of equality. If the null is rejected, then the second step tests the null of submodularity vs. supermodularity (i.e. complementarity). Thus, a significant Wald χ2 test in the second step reveals the existence of complementarity since the test indicates that introducing only technological innovation has a lower effect on a firm’s productivity than introducing jointly technological and non-technological innovation. Since we are testing one linear restriction at a time, the χ2 distribution has one degree of freedom. ***p < 0.01, **p < 0.05, *p < 0.1. ap = 0.11; bp = 0.13; cp = 0.14.

nological innovation is more effective compared with a strategy that implies technological adoption alone. This is confirmed in all model specifications, although in model (2) the significance level is 0.13. When considering non-persistent innovation, the results of the complementarity test are not clear-cut. Weak complementarity is observed only in model (1).

A positive effect of a firm’s persistent innovating attitude is provided by the impact of the R&D variable, which implies that a firm has invested in R&D in two consecutive surveys. The premium in terms of the productivity gain is between 5.1 and 4.0% in specifications (1) and (2), whereas in (3) the impact is milder (2.8%).

Given these findings concerning the persistent innovation premium, we can discuss the other results in detail (Table 4). The capital-to-labour ratio (k) implies an elasticity of almost 0.18 in model (1) and (3) and 0.16 in model (2). This estimate is consistent with estimates presented in other empirical studies (Mairesse and Sassenou 1991; Crépon, Duguet, and Mairessec 1998).

We have not estimated the return on knowledge capital, as our choice has been to estimate an equation in which we show the impact on productivity of a persistent techno- logical and non-technological attitude, on the one hand, and of positive and persistent R&D expenditures, on the other hand, conditional on a set of firm-specific control variables and the capital-to-labour ratio. However, these estimates provide an indirect measure of the impact of knowledge capital, which implies, on the whole, a significant and non-negligible productivity premium comparable with the impact of the capital deepening variable (k).

Another significant impact reflecting technological opportunities available at the indus- try level is captured by the dummy variable representing an industry’s technological level

INDUSTRY AND INNOVATION 519

Table 4. Marginal effects on performance (selected variables).

Effects on profitability

y (+10%) +0.9 p. p. lev (+10 p. point) +0.1 p. p.

Effects on productivity 1 – Base model (%) 2 – Predetermined innovation (%) 3 – Endogenous innovation

pers_tech_ntech (=1) +12.2 +13.4 +6.2% tech_ntech (=1) +3.6 +5.9 n.s. pers_r&d (=1) +5.1 +4.0 +2.8% k (+1%) +0.18 +0.16 +0.18% pavitt_mh (=1) +11.8 +10.7 +12.2% gp (=1) +9.1 +11.6 +9.1% age (+1) +4.8 +4.4 +4.8% nwest (=1) +24.4 +20.4 +25.2% neast (=1) +20.7 +17.9 +21.2% centre (=1) +19.7 +15.1 +20.3%

Notes: Recall that profitability (ros) is a ratio, whereas productivity (y) is expressed in log values and thus impacts are calculated accordingly.

(pavitt_mh). Its impact is significant and relevant because it implies a productivity gain of about 12 % for those firms operating in medium-high-tech sectors according to the Pavitt taxonomy.

The age and group dummy variables show a positive and significant effect, suggesting that older firms have a productivity premium of approximately 5% and that those firms which belong to a group experience a positive impact on their productivity of more than 9% (model 1 and 3) and 12% (model 2).

Regional differentials are significant and reflect the disadvantage of the South, in that North and Centre Italy exhibit a gain in productivity that is, on average, more than 20%.

Regarding profitability, we can argue that the effect of the variable reflecting the SCP mechanism (cr5) – although significant – is mild, whereas the other variables reflecting firms’ efficiency condition are significant and show non-negligible impacts.

The leverage variable (lev) is significant and positive. A 10% increase brings about a 0.1 p.p. increase in profitability, thus signalling that internal resources are crucial in affecting a firm’s ability to finance its activity and then earn profits. In other words, as the cost of borrowing increases – in particular because of an increasing economy-wide risk caused by the financial crisis – internal resources play a significant role in affecting firms’ investment decisions, as suggested by the pecking order theory (Myers and Majluf 1984).

A negative sign, i.e. a condition in which highly indebted firms earn higher profits, is plausible but prevailing in financial market conditions in which risk is relatively low and a firm’s external debt may amplify the potential gain from investment.

The intern dummy variable represents a proxy for a firm’s internationalisation propen- sity.Itsimpactisnegativeandsignificantbutverylimited(0.01p.p.).Thisevidencesuggests that firms that sell products on international markets earn profits slightly lower than those earned by firms that do not internationalise. This observation may be controversial, as one would expect the opposite result, i.e. a positive sign on the coefficient of this dummy variable. However, one can argue that operating on international markets implies additional costs that may be not fully compensated by the potential increase in revenues that the internationalisation process generates.

520 E. BARTOLONI AND M. BAUSSOLA

The sect_inntech variable shows a very mild and negative impact on profitability, thus signalling that the previously mentioned technological competitive mechanism may prevail, although its effect is feeble.

Productivity, which reflects both a firm’s efficiency characteristics and a technological attitude, enters the profitability equation positively. Highly productive firms receive a profit premium corresponding to 0.9 p.p. when productivity increases by 10%.19

In the adopted specification we have not included, a dummy variable reflecting the persistent attitude of firms in introducing technological and non-technological innovation, as this variable is not significant when included. It does significantly affect productivity, and through this route it indirectly affects profitability.

6. Conclusions

We have presented an empirical model of the determinants of a firm’s productivity and profitability which has enabled us to ascertain the role of factors related to technological and non-technological innovations. In addition, we have underlined how such activities, if undertaken persistently, provide a significant additional increase in a firm’s productivity and profitability. Formal tests suggest that in this framework non-technological innovation is complementary to technological innovation.

Occasional technological innovation either combined with non-technological innova- tion or alone, does have a significant effect on firms’ performance in model specifications in which technology adoption enters the productivity equations as an exogenous or predetermined variable.

We find support to our initial hypotheses thus emphasising the relevance of the innovation process, in that learning, organisational adjustments and market orientation – together with technological innovation – determine a firm’s superior performance.

In addition, we also use an input measure of innovative knowledge, related to a firm’s R&D effort. The underlying productivity premium is significant, with an impact that ranges from 5.1 to 2.8%.

Capital deepening, i.e. the capital–labour ratio, exhibits a positive and significant impact that implies an elasticity of almost 0.18. This finding underlines the role of physical capital accumulation, although a direct comparison with the impact of knowledge capital (R&D activities) cannot be derived, as we proxy for this effect by using a dummy variable.

Additional firm characteristics are taken into account, suggesting that older firms experience a significant and non-negligible productivity premium, which is also acquired by those firms that are part of a group.

Sectoral characteristics related to innovative criteria (Pavitt taxonomy) suggest an increasing relationship between productivity and technological levels.

We also analysed firms’ profitability by estimating a profit function that summarises different mechanisms affecting profits. Thus, we have considered the traditional SCP and efficiency view mechanisms, together with the role played by a firm’s innovative attitude.

The effect of the SCP mechanism (proxied by a concentration index) is negligible, although positive and significant, whereas other firm-level efficiency variables (leverage and the ability to sell products on international markets) show a negative mild impact.

19At sample mean the difference between profits of persistent joint innovators and persistent technological innovators is on average about 2 p.p. over the entire period.

INDUSTRY AND INNOVATION 521

This latter effect in particular – although negative – is feeble, suggesting that possible gains from internationalisations may be offset by the increased fixed costs associated with it, particularly for small- and medium-sized enterprises.

According to the specified empirical model, productivity reflects a firm’s efficiency variable that also incorporates the impact of innovative advances – considered in their extensive definition – on profitability. Its impact on profitability is much larger than that represented by the traditional SCP mechanism, thus underlining the relevance of a firm’s innovative attitude in driving its profitability.

Acknowledgements

The authors would like to thank seminar participants at the EARIE 2016 Conference, Lisbon (Portugal), the Institute of Economics, Scuola Superiore Sant’Anna, Pisa and the Dipartimento di Scienze Economiche e Aziendali, Università di Parma, Italy. Comments received from two anonymous referees on an earlier version of the manuscript have significantly contributed to improving the paper. Needless to say, the usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the authors.

References

Aghion, P., and P. Howitt. 1992. “A Model of Growth through Creative Destruction.” Econometrica 60 (2): 323–351.

Allen, R. F. 1983. “Efficiency, Market Power, and Profitability in American Manufacturing.” Southern Economic Journal 49: 933–940.

Antonioli, D., M. Mazzanti, and P. Pini. 2010. “Productivity, Innovation Strategies and Industrial Relations in SMEs. Empirical Evidence for a Local Production System in Northern Italy.” International Review of Applied Economics 24 (4): 453–482.

Arrighetti, A., F. Landini, and A. Lasagni. 2015. “Intangible Asset Dynamics and Firm Behaviour.” Industry and Innovation 22 (5): 402–422.

Bain, J. S. 1956. Barriers to New Competition: Their Character and Consequences in Manufacturing Industries, vol. 3. Cambridge, MA: Harvard University Press.

Bartoloni, E. 2012. “The Persistence of Innovation: A Panel Data Investigation on Manufacturing Firms.” International Review of Applied Economics 26 (6): 787–810.

Bartoloni, E., and M. Baussola. 2009. “The Persistence of Profits, Sectoral Heterogeneity and Firms’ Characteristics.” International Journal of the Economics of Business 16 (1): 87–111.

Bartoloni, E., and M. Baussola. 2016. “Does Technological Innovation undertaken Alone have a Real Pivotal Role? Product and Marketing Innovation in Manufacturing Firms.” Economics of Innovation and New Technology 25 (2): 91–113.

Battisti, G., A.-G. Mourani, and P. Stoneman. 2010. “Causality and a Firm-level Innovation Scoreboard.” Economics of Innovation and New Technology 19 (1): 7–26.

Battisti, G., and P. Stoneman. 2010. “How Innovative are UK Firms? Evidence from the Fourth UK Community Innovation Survey on Synergies between Technological and Organizational Innovations.” British Journal of Management 21 (1): 187–206.

Carpenter, R. E., and B. C. Petersen. 2002. “Capital Market Imperfections, High-tech Investment, and New Equity Financing.” The Economic Journal 112 (477): F54–F72.

Cefis, E. 2003. “Persistence in Innovation and Profitability.” Rivista Internazionale di Scienze Sociali 111: 19–37.

Cefis, E., and M. Ciccarelli. 2005. “Profit Differentials and Innovation.” Economics of Innovation and New Technology 14 (1–2): 43–61.

522 E. BARTOLONI AND M. BAUSSOLA

Coad, A., A. Segarra, and M. Teruel. 2013. “Like Milk or Wine: Does Firm Performance Improve with Age?” Structural Change and Economic Dynamics 24: 173–189.

Cohen, W. M., and D. A. Levinthal. 1990. “Absorptive Capacity: A New Perspective on Learning and Innovation.” Administrative Science Quarterly 35: 128–152.

Cozzarin, B. P., and J. C. Percival. 2006. “Complementarities between Organisational Strategies and Innovation.” Economics of Innovation and New Technology 15 (03): 195–217.

Crépon, B., E. Duguet, and J. Mairessec. 1998. “Research, Innovation and Productivity: An Econometric Analysis at the Firm Level.” Economics of Innovation and New Technology 7 (2): 115–158.

David, P. A. 1992. “Knowledge, Property, and the System Dynamics of Technological Change.” The World Bank Economic Review 6 (Suppl 1): 215–248.

Delorme Jr., C. D., D. R. Kamerschen, P. G. Klein, and L. F. Voeks. 2002. “Structure, Conduct and Performance: A Simultaneous Equations Approach.” Applied Economics 34 (17): 2135–2141.

Demsetz, H. 1973. “Industry Structure, Market Rivalry, and Public Policy.” Journal of Law and Economics 16: 1–9.

Evangelista, R., and A. Vezzani. 2010. “The Economic Impact of Technological and Organizational Innovations. A Firm-level Analysis.” Research Policy 39 (10): 1253–1263.

Frantzen, D. 2003. “The Causality between R&D and Productivity in Manufacturing: An International Disaggregate Panel Data Study.” International Review of Applied Economics 17 (2): 125–146.

Geroski, P. A. 1989. “Entry, Innovation and Productivity Growth.” The Review of Economics and Statistics 71: 572–578.

Geroski, P. A., S. Machin, and J. Van Reenen. 1993. “The Profitability of Innovative Firms.” RAND Journal of Economics 24: 198–211.

Geroski, P. A., J. Van Reenen, and C. F. Walters. 1997. “How Persistently Do Firms Innovate?” Research Policy 26 (1): 33–48.

Hall, B. H., F. Lotti, and J. Mairesse. 2009. “Innovation and Productivity in SMEs: Empirical Evidence for Italy.” Small Business Economics 33 (1): 13–33.

Haltiwanger, J. C., J. I. Lane, and J. R. Spletzer. 1999. “Productivity Differences across Employers: The Roles of Employer Size, Age, and Human Capital.” The American Economic Review 89 (2): 94–98.

Hawawini, G., V. Subramanian, and P. Verdin. 2003. “Is Performance Driven by Industry-or Firm- specific Factors? A New Look at the Evidence.” Strategic Management Journal 24 (1): 1–16.

Hollenstein, H. 2003. “Innovation Modes in the Swiss Service Sector: A Cluster Analysis Based on Firm-level Data.” Research Policy 32 (5): 845–863.

Johansson, B., and H. Lööf. 2010. Innovation Strategy and Firm Performance: What is the Long-run Impact of Persistent R&D? Stockholm: CESIS, KTH, Royal Institute of Technology.

Jones, C. I. 1995. “R&D-based Models of Economic Growth.” Journal of Political Economy 103: 759–784.

Karshenas, M., and P. L. Stoneman. 1993. “Rank, Stock, Order, and Epidemic Effects in the Diffusion of New Process Technologies: An Empirical Model.” The RAND Journal of Economics 24 (4): 503– 528.

Laursen, K., and N. J. Foss. 2003. “New Human Resource Management Practices, Complementarities and the Impact on Innovation Performance.” Cambridge Journal of Economics 27 (2): 243–263.

Le Bas, C., and W. Latham. 2006. The Economics of Persistent Innovation. Berlin: Springer. Lööf, H., and A. Heshmati. 2002. “Knowledge Capital and Performance Heterogeneity: A Firm-level Innovation Study.” International Journal of Production Economics 76 (1): 61–85.

Mairesse, J., and M. Sassenou. 1991. R&D Productivity: A Survey of Econometric Studies at the Firm Level. Working Paper 3666, National Bureau of Economic Research.

Mañez, J. A., M. E. Rochina-Barrachina, A. Sanchis, and J. A. Sanchis. 2009. “The Role of Sunk Costs in the Decision to Invest in R&D.” The Journal of Industrial Economics 57 (4): 712–735.

Mansfield, E. 1968. Industrial Research and Technological Innovation: An Economic Analysis. New York: Norton.

INDUSTRY AND INNOVATION 523

Milgrom, P., and J. Roberts. 1990. “The Economics of Modern Manufacturing: Technology, Strategy, and Organization.” The American Economic Review 80: 511–528.

Mohnen, P., and B. H. Hall. 2013. “Innovation and Productivity: An Update.” Eurasian Business Review 3 (1): 47–65.

Mohnen, P., and L.-H. Röller. 2005. “Complementarities in Innovation Policy.” European Economic Review 49 (6): 1431–1450.

Montresor, S., and A. Vezzani. 2016. “Intangible Investments and Innovation Propensity: Evidence from the Innobarometer 2013.” Industry and Innovation 23 (4): 331–352.

Mueller, D. C. 1992. “The Persistence of Profits.” In Profits, Deficits and Instability, edited by D. B. Papadimitriou, 82–102. Heidelberg: Springer.

Mueller, D. C., and J. Cubbin. 2005. The Dynamics of Company Profits. Cambridge: Cambridge University Press.

Myers, S. C., and N. S. Majluf. 1984. “Corporate Financing and Investment Decisions When Firms have Information that Investors Do Not Have.” Journal of financial economics 13 (2): 187–221.

Narver, J. C., and S. F. Slater. 1990. “The Effect of a Market Orientation on Business Profitability.” The Journal of Marketing 54 (4): 20–35.

Peltzman, S. 1977. “The Gains an Losses from Industrial Concentration.” Journal of Law and Economics 20 (2): 229–263.

Percival, J. C., and B. P. Cozzarin. 2008. “Complementarities Affecting the Returns to Innovation.” Industry and Innovation 15 (4): 371–392.

Peters, B. 2009. “Persistence of Innovation: Stylised Facts and Panel Data Evidence.” The Journal of Technology Transfer 34 (2): 226–243.

Raymond, W., P. Mohnen, F. C. Palm, V. der Loeff, and S. Schim. 2009. Innovative Sales, R&D and Total Innovation Expenditures: Panel Evidence on their Dynamics. CESifo Working Paper Series.

Raymond, W., P. Mohnen, F. Palm, and S. S. Van Der Loeff. 2010. “Persistence of Innovation in Dutch Manufacturing: Is it Spurious?” The Review of Economics and Statistics 92 (3): 495–504.

Roberts, P. W. 1999. “Product Innovation, Product-market Competition and Persistent Profitability in the US Pharmaceutical Industry.” Strategic Management Journal 20 (7): 655–670.

Roberts, P. W. 2001. “Innovation and Firm-level Persistent Profitability: A Schumpeterian Framework.” Managerial and Decision Economics 22 (4–5): 239–250.

Romer, P. M. 1990. “Endogenous Technological Change.” Journal of Political Economy 98 (5): S71– S102.

Rouvinen, P. 2002. “R&D-productivity Dynamics: Causality, Lags, and ‘Dry Holes.” Journal of Applied Economics 5 (1): 123–156.

Schubert, T. 2010. “Marketing and Organisational Innovations in Entrepreneurial Innovation Processes and their Relation to Market Structure and Firm Characteristics.” Review of Industrial Organization 36 (2): 189–212.

Schumpeter, J. A. 1934. The Theory of Economic Development: An Inquiry into Profits, Capital, Credit, Interest, and the Business Cycle, Vol. 55. Transaction Publishers.

Slade, M. E. 2004. “Competing Models of Firm Profitability.” International Journal of Industrial Organization 22 (3): 289–308.

Slater, S. F., and J. C. Narver. 1995. “Market Orientation and the Learning Organization.” The Journal of Marketing 59: 63–74.

Stiglitz, J. E. 1987. “Learning to Learn, Localized Learning and Technological Progress.” In Economic Policy and Technological Performance, edited by P. Dasgupta and P. Stoneman, 125–153. Cambridge: Cambridge University Press.

Stoneman, P., and M. J. Kwon. 1996. “Technology Adoption and Firm Profitability.” The Economic Journal 106 (437): 952–962.

524 E. BARTOLONI AND M. BAUSSOLA

Appendix 1. Descriptive statistics

Period 1998–2000 2002–2004 2006–2008 2010–2012 Tot.

No. of observations 2462 2855 1471 1135 7923

Variable name Type Variable description

tech_ntech 0/1 1 if the firm has occasionally introduced a technological inno- vationinconjunctionwithanon- technological innovation

0.22 0.21 0.28 0.30 0.24

pers_tech_ntech 0/1 1 if the firm has persistently introduced a technological inno- vationinconjunctionwithanon- technological innovation

0.07 0.06 0.12 0.13 0.08

tech 0/1 1 if the firm has occasionally introduced a technological inno- vation

0.21 0.22 0.25 0.25 0.22

pers_tech 0/1 1 if the firm has persistently innovated in the technological domain

0.03 0.03 0.05 0.05 0.04

ros c Return on sales. The ratio between gross operating profits and sales. An index of operating profitability.

0.12 0.11 0.10 0.09 0.11

Y c Value added per employee (thousands euros)

59.4 60.1 71.3 74.2 64.0

K c Tangible fixed assets per em- ployee (thousands euros)

53.1 54.5 69.7 82.8 60.9

lev c The ratio of shareholders’ funding to total debts

0.71 0.76 0.83 1.03 0.80

cr5 c Concentration index (Pavitt sec- tors)

0.33 0.35 0.26 0.29 0.30

sect_inntech c Share of sectoral technological innovators %

50.1 46.8 66.0 66.2 54.2

intern 0/1 1 if the firm sells its products in the international market

0.71 0.67 0.80 0.86 0.73

pavitt_mb 0/1 1 if in the low and medium-low technology sectors

0.66 0.66 0.63 0.61 0.65

pavitt_ma 0/1 1 if in the high and medium-high technology sectors

0.34 0.34 0.37 0.39 0.36

AGE c Firm’s age (years) 23 26 30 33 27 r&d 0/1 1 if the firm has undertaken R&D

investments 0.31 0.37 0.53 0.50 0.40

pers_r&d 0/1 1 if the firm has persistently undertaken R&D investments

0.26 0.26 0.42 0.43 0.32

gp 0/1 1 if the firm belongs to an industrial group

0.30 0.35 0.62 0.76 0.44

nwest 0/1 1 if the firm is localised in the North-West

0.35 0.36 0.39 0.36 0.36

neast 0/1 1 if the firm is localised in the North-East

0.34 0.34 0.35 0.36 0.34

centre 0/1 1 if the firm is localised in the Centre

0.16 0.16 0.13 0.14 0.15

south 0/1 1 if the firm is localised in the South

0.15 0.15 0.13 0.14 0.14

Notes: Y and K have been deflated using sectoral deflators (base year 2010). ‘Persistently’ means in at least two consecutive periods. ‘Occasionally’ means at least one time but never in two consecutive periods.

INDUSTRY AND INNOVATION 525

Appendix 2. RE Logistic regressions for the innovative and R&D binary vari- ables – period 2000–2012

Variables pers_tech_ntech tech_ntech pers_tech tech pers_r&d

size 0.544*** −0.0237 −0.0133 −0.152*** 0.601*** [0.0266] [0.0308] [0.0531] [0.0391] [0.0282]

cr5 −0.00924*** 0.00216 0.00733 0.0121*** 0.000276 [0.00250] [0.00263] [0.00493] [0.00336] [0.00250]

sect_inntech 0.0448*** 0.0240** 0.0161 −0.0304* 0.0374*** [0.0101] [0.0106] [0.0220] [0.0183] [0.0104]

lev −0.0275 −0.0700** 0.0692 0.0475 0.0129 [0.0267] [0.0353] [0.0453] [0.0374] [0.0277]

intern 0.910*** 0.208** 0.360** 0.285*** 1.075*** [0.0850] [0.0849] [0.172] [0.107] [0.0892]

k 0.110*** 0.0453 0.155*** 0.0178 0.101*** [0.0272] [0.0317] [0.0518] [0.0396] [0.0275]

pavitt_ma −0.0912 −0.553** −0.282 0.655** 0.374* [0.190] [0.216] [0.420] [0.322] [0.194]

gp 0.0336 0.0268 0.0306 −0.123 0.111 [0.0692] [0.0861] [0.158] [0.109] [0.0709]

nwest 0.326*** −0.00099 0.795*** 0.261* 0.781*** [0.104] [0.108] [0.234] [0.140] [0.113]

neast 0.598*** −0.0128 0.616*** 0.141 0.814*** [0.103] [0.108] [0.236] [0.142] [0.112]

centre 0.0469 0.129 0.587** 0.154 0.399*** [0.120] [0.121] [0.261] [0.161] [0.130]

d2000 0.542*** 0.943*** −0.207 −1.494*** 0.379** [0.172] [0.187] [0.373] [0.326] [0.178]

d2004 0.727*** 0.0302 −0.0975 −0.481 0.557*** [0.205] [0.223] [0.432] [0.363] [0.212]

d2012 0.0327 0.168 −0.0562 −0.333** −0.0914 [0.0896] [0.113] [0.180] [0.148] [0.0945]

Constant −8.271*** −3.834*** −6.660*** −0.646 −8.816*** [0.679] [0.720] [1.438] [1.200] [0.710]

LR χ2(14) 1769.675 180.51 56.31 151.92 1498.38 pseudo R2 0.18 0.03 0.02 0.03 0.23 Observations 7.923 7.923 7.923 7.923 7.923

Notes: Robust Standard errors in brackets. ***p < 0.01, **p < 0.05, *p < 0.1. Following Bartoloni (2012), it is possible to estimate a firm’s innovation probability using logit models that incorporate explanatory variables causing different firms’ innovativebehaviours.Weusethefollowingexplanatoryvariables:firmsize(size,numberofemployees, logvalues),financial efficiency (lev), physical capital deepening (K), industrial group membership (gp), ability to sell products on international markets (intern), market structure (cr5), technological spill-over (sect_inntech), and regional, sectoral and time dummies. We derive predicted probabilities that can then be used to predict the estimated events (pers_tech, pers_tech_ntech, tech, tech_ntech, and pers_R&D) used in the productivity regression.