Age and productivity as determinants of firm survival over the industry life cycle

Abstract

This paper contributes to fill the gap between the literature on the determinants of firm survival and the empirical works on the industry life cycle (ILC). Using a representative sample of Spanish firms with 10 or more employees over the period 1993–2009, the role played by firm age and productivity in firm survival is empirically analysed across three stages of the life cycle of forty-seven 3-digit manufacturing sectors. In the ‘early’ stage of the ILC, firm age is negatively correlated with hazard rates while firm productivity is not. Firm productivity is associated with lower hazard in the ‘mature’ stage of the ILC, when competition is primarily efficiency-driven, while firm age does not play a significant role for firm survival. In the ‘intermediate’ stage, both age and productivity play a role in reducing firms’ hazard rates.

Keywords:

• Firm survival
• industry life cycle
• Spanish manufacturing firms
• discrete time survival methods

JEL classification:

• C41
• L10
• L60

Disclosure statement

No potential conflict of interest was reported by the authors.

Acknowledgements

The authors wish to thank the associate editor Sandro Montresor and three anonymous referees for their useful remarks on previous versions of the paper. Comments by Giulio Bottazzi, Anna Maria Ferragina, Tamara de la Mata, Erol Taymaz, Davide Castellani and seminar participants at the XI Jornadas sobre Integración Económica (Universitat de Valencia, 2014) and the International Workshop on ‘Innovation, Globalisation and Firm Dynamics’ (Università degli Studi di Salerno, 2014) are also greatly acknowledged. The usual disclaimer applies. Silviano Esteve-Pérez and Fabio Pieri acknowledge financial support from the Spanish Ministry of Science and Innovation (Project MINECO/FEDER ECO2015-68057-R) and from the Valencian Regional Government (PROMETEOII/2014/053).

Notes

¹ As a non-exhaustive list: Mata and Portugal (1994), Audretsch and Mahmood (1995) have explored the role of firm initial size; Evans (1987) has explored the role of firm current size; Freeman, Carroll, and Hannan (1983), Mata and Portugal (1994), Mata, Portugal, and Guimaraes (1995) have explored the role of firm age; Agarwal (1997) has explored the role of firm past growth rate; Hannan and Freeman (1977) have explored the role of narrowness/wideness of the niche a firm occupies in the market; Hall (1987), Esteve-Pérez and Máñez (2008) have explored the role of firm R&D spending; Bruderl, Preisendorfen, and Ziegler (1992), Cefis and Marsili (2005, 2006) have explored the role of firm’s innovative strategies; Doms, Dunne, and Roberts (1995) have explored the role of firm technological capabilities.

² See, among others, the seminal paper by Utterback and Abernathy (1975) and the formal model provided by Klepper (1996).

³ The relevant literature has indistinctly made use of the expressions ‘industry life cycle’ and ‘product life cycle’ to refer to the changing competitive setting in terms of entry/exit rates, number of firms, innovative activities and firm boundaries. Naturally, the term ‘industry’ has a broader rendition than that of ‘product’ and the interpretational issue has been also recognised by Malerba and Orsenigo (1996, 64) and Klepper (1997, 148, footnote no. 1). In this work, given that the indicator which identifies the stages of the life cycle has been defined at the 3-digit (NACE rev.2) level (see Table A.3 for the list of industries considered in this work), the expression ‘industry life cycle’ and its ‘ILC’ acronymic have been adopted along the entire text. The reader is cross-referred to Sections 2.1 and 3.2 for the definition of the stages of the ILC.

⁴ See, among others, the articles contained in the special issue of the International Journal of Industrial Organization (Vol. 13, Issue 4) regarding ‘The Post-Entry Performance of Firms’, published in December 1995 (guest editors: David B. Audretsch and José Mata).

⁵ Seminal studies in the ILC tradition aimed to identify and analyse the life cycle of specific products (approximately comparable to 5- or even 6-digit levels of disaggregation in a standard industry classification) in a historical perspective. For instance, Gort and Klepper (1982) developed a framework for analysing 46 new products introduced over the last century, from commercial inception to 1972; Klepper and Graddy (1990) extended the time-series for those products until 1981; Klepper (2002) focused on four products only, i.e. automobiles, tires, televisions and penicillin.

⁶ Agarwal (1997) studied the role of firm size, growth and product diversification across five stages of the ILC, identified on the basis of information on entry rates. Agarwal and Gort (1996, 2002) applied the same procedure to study the role of learning-by-doing and endowments in firm survival across subsequent stages of the ILC. Agarwal and Audretsch (2001) studied the role of firm size in survival across two stages (formative and mature) of the ILC identified by using information on net entry rates. Agarwal, Sarkar, and Echamebad (2002) used the information on entry rates to define two stages (growth and maturity) of industries’ evolution. All these works have made use of the information contained in the Thomas Register of American Manufacturers.

⁷ Interestingly enough, McGahan and Silverman (2001, 1156) claim that identifying the stages of the ILC via the inflection points in a long time series of the number of active enterprises (or by means of the analysis of the net entry rate) may generate remarkable difficulties in defining correctly the beginning and the end of the stages. For this reason, also other dimensions of an industry’s evolution and different mechanisms for identifying the stages should be considered.

⁸ Nonetheless, it is acknowledged that the lack of information on both net entry rates and the entire history of specific products may be a limitation for the scope of the present work.

⁹ In the words of Suarez and Utterback 1995, 416: ‘A dominant design is a specific path, along an industry’s design hierarchy, which establishes dominance among competing design paths’.

¹⁰ However, the idea that the evolution of the vertical structure of the firm in an industry depends only on the extent of the market and the attendant division of labour may not hold in all industries. Indeed, Helfat (2015, 808–811) furnishes some explanations of why Stigler’s hypothesis does not apply in all cases: the evolution of the vertical structure of a firm depends on several factors that Helfat points out as ‘contextual (to the industry) factors’. For example, the type of innovation, either ‘systemic’ (i.e. which needs a strong coordination and alignment among the stages of the innovation process) or ‘autonomous’ (i.e. which may be undertaken in its different stages by different agents) plays a role for firms at an industry’s inception in their choice of being, respectively, vertically integrated or not. Thus, the choice made in this work ought to be considered as a simplification (based on Stigler’s hypothesis) that is functional to the identification of the different stages of the ILC.

¹¹ In the ‘passive’ learning model by Jovanovic (1982), firms become more conscious about their unknown type (level of efficiency) as time passes and adjust their growth rates with the updated expectation about their type. In ‘active’ learning models (Ericson and Pakes 1995), a firm’s type can be partially modified through purposive investments in the development of new technologies. In both models, firm age helps in dispelling the uncertainty about the firm type but does not provide any advantage to survive per se. Conversely, if a ‘learning-by-doing’ process á la Arrow (1962) is at work, young firms may be truly disadvantaged with respect to their older counterparts in terms of (efficiency and thus) survival because of less time accumulated for practice and self-perfection strategies.

¹² Efforts have been made to minimise attrition and incorporate each year new firms with by following the same criteria used in the base year. This helps in maintaining the representativeness of the sample over time (see http://funep.es for further details).

¹³ Note that the ESEE is not a mandatory survey.

¹⁴ Therefore, information in 2010 is used to identify those firms exiting in 2009.

¹⁵ For the list of the industries considered in the empirical analysis the reader is cross-referred to Table A.3.

¹⁶ A firm which in a given year has introduced both a product and a process innovation adds 1 observation to the numerator and the denominator of the TINNOV_m,t variable. The percentage of innovative firms (which are about 45% of all firms contained in the ESEE) that simultaneously introduce product and process innovations is about 35%.

¹⁷ Actually, the concept of ‘relevant market’ is narrower than that of ‘industry’ defined at the 3-digit level. Firms surveyed by the ESEE are asked to provide their view on the real number of competitors they face in their relevant market. Thus, a market is the effective locus of competition for the firms, while the industry gathers all firms sharing a common technology and similar production processes.

¹⁸ Table A.1 in the Appendix 1 displays the Pearson correlation coefficients between all pairs of the four dimensions of the ILC. These variables capture different dimensions of the ILC.

¹⁹ That is, from ‘early’ to either ‘intermediate’ or ‘mature’, and from ‘intermediate’ to ‘mature’.

²⁰ A classic example is the major entry made into the market shares of the U.S. leaders by foreign producers of smaller cars (such as Toyota and Volkswagen) during the 1960s.

²¹ Christensen, Verlinden, and Westerman (2002, 972–975) discuss the case of the 2.5-inch disk-drive industry during the 1990s as an example of vertical re-integration performed by the most technological experienced firms, such as IBM, Toshiba, Hitachi and Fujitsu to manage complex and interdependent technologies in order to cope with a large dissatisfaction for notebook computers in terms of disk-drive capacity, weight and power consumption.

²² Following Lieberman (1991), uncertainty is measured as the average of squared residuals of the following regressions.

where y_im,t is (the log of) the real output (i.e. deflated by firm-specific price variations), t = 1, … ,17 is an integer increasing by one in each year and m = 1, … ,47 refers to the industry.

²³ Table 9 in Section 5.3 shows that the main results of the paper do not change significantly when both age and productivity are included as continuous variables.

²⁴ See Jenkins (2005) for an excellent overview of complementary log–log and proportional hazards models.

²⁵ The test regarding whether the variance of the frailty term is statistically different from zero is performed in Section 5: if this variance is not statistically different from zero, a non-frailty model will be the preferred specification. Under the null hypothesis, the statistic is distributed as a chi-squared with one degree of freedom.

²⁶ Interestingly enough, when the lagged growth rate is not included in the analysis, the higher hazard ratio of small firms corresponds to a statistically significant, and higher than 1, coefficient. This may well be the result of the statistical relation between firm size and firm growth.

²⁷ We thank an anonymous referee for having suggested to further explore the interaction effect of firm age and productivity on firm survival.