The effect of innovation on productivity: evidence from Turkish manufacturing firms

Abstract

This paper systematically explores the effects of firms’ innovation activities on their productivity changes for Turkish manufacturing firms, differentiating between different typologies of innovation. We employ endogenous switching methodology, controlling for endogeneity and selection bias issues, as well as analysing counterfactual scenarios. The main finding of the study points to firm heterogeneity in terms of propensity both to innovate and to benefit from innovation activities. Our results indicate that all types of innovation activity have positive effects on the productivity of firms when compared with non-innovating firms. We find robust evidence for the differential impact of innovation on firm productivity across different innovation types. Further, this relationship alters across different phases of the economy with respect to the 2008 financial crisis.

Keywords:

• Innovation
• productivity
• economic crisis

Subject classification codes:

• D22
• L25
• O30

Acknowledgements

We thank TURKSTAT (Turkish State Institute of Statistics) and particularly, TURKSTAT staff Doğan Böncü, Şehmus Şenol Bozdağ, Nusret Kılıç, Nilgün Arıkan, Erdal Yıldırım, Kenan Orhan, Sabit Cengiz Ceylan, Ferhat Irmak and Esra Sazak for providing access to firm level data.

Notes

¹ These datasets are available under a confidential agreement and all the elaborations can only be conducted at the Microdata Research Centre of TURKSTAT in respect of the laws on the confidentiality of statistics and personal data protection.

² Coefficients of labour and capital from the Levinsohn and Petrin (2003) estimations are available upon request.

³ Note that, we do not distinguish between product/service innovation new to the firm and product/service innovation new to the market and use the ‘other innovation activity’ instrument for product/service innovation for both types.

⁴ Firms with number of employees 20–100; 101–249 and 250 + are defined as small, medium and large, respectively.

⁵ See Table A3 in Appendix 1 for the marginal effects estimated from Equation (1).

⁶ The marginal effects estimated from the regarding specification are available upon request.

⁷ The standard errors of all equations as well as of the treatment effects are bootstrapped.

⁸ The likelihood ratio test shows that we can reject the null hypothesis (with p-value 0.000), indicating that the equations measuring each type of innovation indicator and productivity are independent.

⁹ The growth process of the Turkish economy accelerated remarkably during 2002–2007. While real GDP grew on average by 6.8% annually, manufacturing industry experienced a higher growth rate, on average 8.5%.

¹⁰ In 2009, the economy shrunk by 4.7%.