Discussion Paper
No. 2016-34 | July 06, 2016
David Córcoles, Ángela Triguero and María Carmen Cuerva
Comparing Persistence of Product and Process Innovation: A Discrete-Time Duration Analysis of Innovation Spells
(Published in Recent Developments in Applied Economics)

Abstract

The main purpose of this paper is to analyze the influence of previous experience and learning capabilities on survival in product and process innovation for Spanish manufacturing firms in the period 1990–2010. The authors find past and path dependence and confirm the important effect of R&D effort in both types of innovation. Nevertheless, for product innovation, the level of appropriability and the fact of operating in a high-tech sector are crucial for persistence in comparison with process innovation.

JEL Classification:

O31, O32, L22, L60

Links

Cite As

[Please cite the corresponding journal article] David Córcoles, Ángela Triguero, and María Carmen Cuerva (2016). Comparing Persistence of Product and Process Innovation: A Discrete-Time Duration Analysis of Innovation Spells. Economics Discussion Papers, No 2016-34, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2016-34


Comments and Questions



Anonymous - Referee Report 1
July 06, 2016 - 10:52
his paper studies the persistence of product and process innovation using an extensive panel of Spanish manufacturing firms covering the period 1990 to 2010. The authors use a discrete time methodology for accounting for the persistence of innovation activities, arguing that a discrete time approach is more appropriate to study the persistence of innovation in a panel contest.The findings of the paper show that there is persistence in innovation output at the firm level for both product and process innovation and that it depends strongly on formal R&D activities. Persistent R&D activities have a high impact on the persistence of innovation for both product and process innovation. The authors also find that the probability of persistence in product innovation is higher over time. The authors argue that this finding has policy implications as the innovation strategy choice may affect the persistence of innovation activities. From my point of view, one of the main contributions of the paper is the use of discrete-time duration models to measure the degree of persistence in innovation. The definition of spells (the number of consecutive years during which the firm has an innovative output) to measure persistence in innovation is not very often applied in the empirical literature. As authors point out, this methodology control for unobserved heterogeneity and other narrow assumptions (proportional hazards). Furthermore, they distinguish the differences among the persistence of process and product innovations using a long panel of manufacturing firms (21 years and nearly 22,000 observations). In this regard, I consider that the research has the potential to contribute to the literature on innovation persistence because the different persistence patterns of process and product innovators are worth investigating. Although the former research questions are interesting and relevant, however, some questions after reading the paper could be suggested. For example, the variables related to product and process innovation are reported in the paper, much more information is needed (Do these innovation variables come from CIS-type self-reported data?). Authors do not provide any reason nor argument on the different effect of size on that relationship. Although in their empirical analysis they include size as control variable, they could clarify this influence (unless empirically to check the model robustness).

David Córcoles - Authors' reply
September 19, 2016 - 09:26
Firstly, we would like to thank the referee for reading the article so carefully, and also for the comments made in relation to the article. All of them are going to take into consideration in order to led to an improved version of the paper. The referee’s report starts by noting that the variables product and process innovation are not enough explained. In this sense we are going to add more relevant information about that. Hence, we are going to add the following paragraph: Data, including product and process innovation, come from the Survey of Business Strategies (ESEE), a Spanish dataset compiled by the Ministry of Science and Technology that carries out a panel survey of manufacturing firms located in Spain. This dataset differs from the PITEC (the Spanish CIS). The ESEE is more complete that CIS because is oriented towards capturing information about firms’ strategies, that is to say, about the decisions firms take regarding their competition variables. These cover from the most flexible variables, which experience frequent short-term changes (e.g., prices), to those whose changes require longer periods (e.g., innovation, R&D expenditures). As these decisions are taken in close interaction with the firms’ environment, the survey collects much information about the environment itself (firm’s markets). The information is completed with accounting data that allow measuring the firms’ results. The referee concludes stating that authors do not provide any reason nor argument on the different effects of the size variable on the exit probability of the spell. In this sense, it is true that we have only limited to comment that we have control for the size effect. So, we are going to clarify its influence. The size of the firm is significant for the innovation persistence, that is, large firms have a lower probability of exit from innovation in comparison with small firms (parameter below one and significant). This result is in line with previous literature. But the results of our estimation highlight that the effect of the size is clearly evident when we consider the type of innovation and the size together (the parameter of the interaction is significant). We are including the parameters of the size in the table of the results of estimations (new table 2).

Anonymous - Re: Comments
July 07, 2016 - 07:53 | Author's CV, Homepage
I find this paper very insightful especially on issues of innovation and their demarcations; product and process. My only suggestion for improvement can be that you explore an additional discrete-time model like probit and logit. our reason for using Clog-log is weak. Please see notes by Prof. Stephen Jenkins of LSE.

David Córcoles - Authors' reply
September 19, 2016 - 09:31
We would like to thank the reader comments specially for the suggestion about a deeper explanation for cloglog model. Following this suggestion, the comment presented about the election of cloglog is going to be more detailed in the following way: Discrete-time models (probit, logit or complementary logit) allow us to solve inherent problems to traditional continuous-time models. Cloglog present two advantages in comparison with other discrete-time models (as a logit or probit): Firstly, cloglog results are the most comparable to the continuous-time Cox model given that both present the same grouped-duration (Hess and Persson, 2012). Secondly, cloglog uses a more flexible frailty function than probit or logit models (Heckman and Singer, 1984). In empirical terms, cloglog is the discrete-time model implemented in most of software programmes (We have used the Stata command hshaz written by Prof. Stephen Jenkins) and it is most used in recent literature (Görg et al., 2007; Brenton et al., 2010; or Triguero et al., 2014 are some examples). Thus, taking into account these advantages, we estimate a discrete-time duration model based on a random-effects complementary log-log (cloglog) model. In either case, probit results are very similar without significant differences in comparison to cloglog model (Probit results are available upon request from the authors). References: -Brenton, P., Saborowski, Ch. and von Uexkull, E., (2010), ‘What Explains the Low Survival Rate of Developing Country Export Flows?’, The World Bank Economic Review 24 (3), 474- 499. -Görg, H., Kneller, R. and Muraközy, B., (2007), “What makes a suceessful export? CEPR Discussion Paper No. 6614. -Heckman, J.J. and Singer, B., (1984), ‘Econometric Duration Data’, Journal of Econometrics, 24, 63-132. -Hess, W. and Persson, M., (2012), ‘The Duration of Trade Revisited. Continuous-Time vs. Discrete-Time Hazards’, Empirical Economics 43(3), 1083-1107. -Triguero, A.; Córcoles, D. and Cuerva, M.C., (2014), ‘Measuring the persistence in innovation in Spanish Manufacturing Firms: empirical evidence using discrete-time duration models’, Economics of Innovation and New Technology 23(5-6), 447-468.

Anonymous - Comment
September 06, 2016 - 10:44
A duration model is estimated to study the factors that affect the persistence in innovation activities. Innovation persistence is defined as the duration of the spells of continuous innovation activities by firms. Some hypotheses are formulated after reviewing the literature, and are checked with the duration model and the available data. The main problem that I see in this paper, and this literature in general, is the complete lack of a guiding theoretical model that would justify the empirical exercise. Some of the hypotheses being tested look trivial without a solid theory behind them. For instance stating that hight-tech firms invest for longer continuous periods in R&D seems an obvious proposition, otherwise they would not be hight-tech. Without this theoretical foundations, the empirical results in the paper are only descriptive, correlations between different variables, and have no causal implications. The paper also uses some incorrect statistical language. A statistical hypothesis is never accepted, and if a null hypothesis is not rejected, the test should be considered inconclusive.

David Córcoles - Authors' reply
September 19, 2016 - 09:34
1.- The reader notes that a problem in the paper is: “…the complete lack of a guiding theoretical model that would justify the empirical exercise. Some of the hypotheses being tested look trivial without a solid theory behind them”. Effectively, the paper is organised around a set of “hypothesis” that are meant to guide the empirical findings. However, from our point of view, all priori hypotheses come from an explicit theoretical model. The objective of the paper is to explain how the ability to be constant in R&D, appropriability conditions, technological opportunities and previous episodes of innovation are crucial to current innovation behaviour. As it is stated in the paper, “Building upon the dynamic capabilities framework, we present a model that examines the role of learning capabilities in innovation persistence” (page 3). Hence, theoretical explanations are given in the Section 2. Apart from literature review of previous empirical studies about persistence in innovation, we explain the differences between past and path dependence using the purpose made by Antonelli in several contributions (2012a, 2012b). Furthermore, we clarify that the second theoretical perspective “is closely related to the resource-based theory of the firm and dynamic capabilities, where innovation persistence is linked with the internal characteristics and learning capabilities of firms (including previous innovation behavior) and the changing context in which they are localized (path dependence)”(page 4).These are also the reasons that we believe to justify that our hypotheses are not mere expositions of indisputable ideas. For example, in hypothesis 1 we state that persistence in innovation (product and process) is past dependent taking into account Antonelli´s framework. In similar way, hypothesis 2 is justified with prior studies by Flaig and Stadler (1994) and Geroski et al. (1997) around the importance of R&D success and knowledge accumulation; hypothesis 3 regarding the influence of appropriability on innovative behaviour with the purpose by Dosi et al (2006); and/or hypothesis 4 around the role of technological opportunities with seminal contributions by Dosi (1988) or Cohen and Levinthal (1989). However, we agree about that sub-hypotheses (H1a, H2a, H3a and H4a) could be have set after empirical results in previous works. For example, in hypothesis 1a we argue that persistence in product innovation is more likely to be path dependent than process innovation based upon Martinez-Ros and Labeaga (2009) for Spanish firms and Antonelli et al. (2012) for Italian companies but not a based upon specific theoretical framework. In this regard, we will do an effort to rewrite several parts of the “Introduction” and “Theoretical framework” trying to be more specific about the theoretical perspective and the objectives pursued based upon previous empirical literature. 2.- Also this reader notes that:“The paper also uses some incorrect statistical language. A statistical hypothesis is never accepted, and if a null hypothesis is not rejected, the test should be considered inconclusive..” We agree with the reviewer that it is more appropriate to use the expression “not rejected” for the null hypotheses instead of “accepted”. Therefore, we have corrected the paper in the four occasions in which we use “accepted” to conclude if our hypothesis are fulfilled. References-Antonelli C., Crespi F. and Scellato G., (2012b), ‘Internal And External Factors in Innovation Persistence’, Economics of Innovation and New Technology 22(3), 256-280 -Dosi, G., (1988), ‘Sources, procedures, and microeconomic effects of innovation’. Journal of Economic Literature 26(3), 1120-1171. -Dosi, G., Marengo, L. and Pasquali, C.,(2006), ‘How much should society fuel the greed of innovators? On the relations between appropriability, opportunities and rates of innovation´, Research Policy 35, 1110–1121. -Flaig, G. and Stadler, M., (1994), ‘Success breeds success. The dynamics of the innovation process’, Empirical Economics 19, 55–68. -Geroski, P. A., Reenen, J. and Walters, C.F., (1997), ‘How Persistently do Firms Innovate?’, Research Policy 26, 33-48. -Martínez-Ros, E. and Labeaga, J.M., (2009), ‘Product and process innovation: persistence and complementarities’, European Management Review 6(1), 64–75

Anonymous - Referee Report 2
September 08, 2016 - 08:17
The overall objective of this paper is to investigate the effects of experience and learning capabilities on a firm’s capacity to manufacture a product or process innovation. Its topic is interesting and novel. The study falls within the body of work that deals with the effects of innovation persistence on the capacity of the firms to innovate during the year in progress. The standpoint taken by the paper is original in that only a few studies measure the persistence of innovation by means of the periods of time during which the firm innovates year after year without gaps in its activity. The paper applies a survival model to explain the effect of persistence on innovation at firm level. Another of the paper’s strong points is the source of the data. The data set used is a sample of Spanish firms from the Survey of Business Strategies (ESEE), which despite being an unbalanced panel with some two thousand firms (a lesser number than the panel data from the Spanish CIS), provides data for the period 1990-2010. Overall, the different sections of the paper are well structured, the working hypotheses are correct, the methodology is appropriate to the nature of the data and the interpretation of the data is correct. In my opinion, the paper needs minor revisions before being suitable for publication in a journal such as Economics. With this in mind, some suggestions are offered in the hope they will help to improve this paper or future versions of it. Firstly, the paper would be significantly improved by dealing simultaneously with the two types of innovation. Authors such as Le Bas and Poussing (“Are complex innovators more persistent than single innovators?”, International Journal of Innovation Management, 2014, vol. 18, issue 1) use a heterodox (or innovative) distinction when distinguishing between complex and single innovators. Complex innovators undertake product and process innovation within the same time period, while single innovators undertake only one of the two types. By adopting this distinction, tests can be carried out to discover if the differences in innovation strategies across firms are important in respect of the strategies employed by innovative firms. We would have to expect that complex innovators have a high innovative orientation and single innovators a low innovation orientation. Secondly, especially with regard to hypothesis 3, a better development of the argument and more empirical research are required. We can interpret the patterns as a level of appropriability directly related to levels of persistence in innovation, but there are other proxies that could be included in the analysis. Furthermore, there is a problem that needs to be addressed, namely the direct causal link between the registration of patents and innovation performance.

David Córcoles - Authors' reply
September 19, 2016 - 09:38
Firstly, we would like to thank the referee for the attention devoted to the article. We are grateful for the excellent comments that were made. All of them have been taken into consideration, and most of them will be take into account in order to try to improve the paper. The referee’s report starts by highlighting the novelty of the paper to the innovation literature (thank you very much), however, the referee suggests considering complex innovators. Following Le Bas and Poussing (2014), we are going to carry additional estimations to check the robustness of our results and consider this comment. For this, we only have to introduce an additional dummy variable explaining if the firm is a complex innovator (simultaneously reports product and process innovation in the current year) or a single innovator (reports only product or process innovation) in our empirical models. With respect to the hypothesis 3 regarding the influence of appropriability, he/she also recommends a better development of the argument and more empirical research. In this regard, we acknowledge that the link between innovative performance and appropriability at the firm-level must be incorporated in the discussion. Both of them capture the output of the innovation process but they are not comparable. Our dependent variable is defined as the length of time that a firm is continuously innovating following Geroski et al. (1997) (a categorical variable indicating the discrete exit probability of a spell of product or process innovation) while the explanatory variable related to patents is a dummy reflecting whether the firm have o not patents. Since only a minority of Spanish manufacturing firms use patent system mechanism as appropriation mechanism of innovation results, the persistence of innovation must be influenced by firm-specific characteristics associated with knowledge protection technological capabilities. Hence, the persistence of innovation might be positive influenced by this IP protection mechanism or not (there could exist other knowledge protection mechanisms as industrial secret, imitation cost or lead time innovations related to firm appropriability and innovative performance). Unfortunately, our data -the ESEE- does not give information about these IP protection strategies. Since this paper addresses in the influence of this type of learning capabilities on innovation performance (other than through the past dependence, R&D continuity and technological opportunities), we only use this categorical variable related to patents. As this question must be clarified, we are going to incorporate the comment in a new footnote in the following way: Although numerous authors (Harabi, 1995; Brouwer and Kleinknecht, 1999; Gonzalez and Nieto, 2007) have showed that patents are the least used mechanism to appropriate the innovative results, it is needed to consider the influence of appropriability on the innovation persistence. Since our database only gives information about patents, we only consider this particular IP mechanism. Finally, we believe that a paragraph could be incorporated in the conclusions to provide some self-criticism and some limitations of the study. Our findings are subject to several important caveats. Firstly, the limitations related to the causal link between innovation and appropriability used measure based on patents. Namely, other IP mechanism as industrial secret and strategies as the use of imitation cost and time to protect their innovations from their competitors or the technological leadership race must be controlled for. In this sense, it would be desirable to consider the influence of external sources as university-industry networks and continuous innovation as a method of appropriation (Malerba and Torrisi, 1992; Segarra and Arauzo, 2008). Thank you very much for all the suggestions. References:- Brouwer, E., Kleinknecht, A., 1999. Innovative output, and a firm propensity to patent. An exploration of CIS micro data. Research Policy. 28, 615-624. - González, N., & Nieto, M. 2007. Appropriability of innovation results: an empirical study in Spanish manufacturing firms. Technovation, 27(5), 280-295. - Harabi, N., 1995. Appropriability of technical innovations: an empirical analysis. Research Policy 24(2), 981-992. - Le Bas and Poussing, N. 2014. Are complex innovators more persistent than single innovators?”. International Journal of Innovation Management, vol. 18, issue 1. - Malerba, F., Torrisi, S., 1992. Internal capabilities and external networks in innovative activities: evidence from the software industry. Economics of Innovation and New Technology, 2, 49-71. - Segarra, A., & Arauzo, J. M., 2008. Sources of innovation and industry–university interaction: Evidence from Spanish firms. Research Policy, 37(8), 1283-1295.