Technology and spillovers: evidence from Indian manufacturing microdata

Abstract

This article finds that technology stocks and spillovers have significantly affected the output of Indian manufacturing firms over the period 1994 to 2006. The technology of a firm is measured, as embodied in its recent stock of plant and machinery, as well as generated through its own R&D. Moreover, investments in both these types of capital by a firm, also generate significant knowledge spillovers, for all other firms in that industry.

Acknowledgements

Helpful comments and suggestions from two anonymous referees of this journal, and from Jakob B. Madsen, Russell Smyth, Mita Bhattacharya and Rebecca Valenzuela, as well as from conference participants at the ‘Second annual Max Planck international conference on innovation, entrepreneurship and economic growth’, Indian Institute of Science, India, and the Department of Economics workshop at Monash University, are gratefully acknowledged.

Notes

¹ If entrepreneurship indeed leads to spillovers, then it is appropriate to consider intra-industry spillovers, as against inter-industry spillovers. This is because if an entrepreneur commercializes an idea, s/he is more likely to have used an idea generated by firms of his own industry. Two existing studies from Indian manufacturing support this claim. Veeramani (2004) has shown that intra-industry trade promotes growth of output of firms within that industry. Lall and Rodrigo (2000) use a cross-section plant-level manufacturing dataset for 1994, to show that locating plants within proximity to other similar firms, improves the chances of knowledge spillovers.

² King and Robson (1993) suggest that an idea embodied in an investment process, spills over immediately as the project is initiated, with no subsequent spillover effects. According to this argument, knowledge spillovers from investments in equipment should be measured by the industry-wide investments in equipment and not by the industry-wide stock of equipment as done here. However, an implicit assumption in this argument is that workers learn immediately when working with new equipment. This may not necessarily be true. A worker can also learn new ideas when working with old capital–an argument very similar in nature to that which suggests R&D affects not only current period productivity but also productivity in future periods. As Shaw (1992) remarks, ‘the productivity of a given firm is assumed to be an increasing function of cumulative aggregative investments for the industry.’ Nonetheless, as a robustness check, the learning spillovers from such knowledge in the present study were also measured using the cumulative industry-wide investments in equipment. Results remained robust to this change, and are therefore, not reported here.

³ Regressions inclusive of a time trend were also tried. Results were found to be of the same sign and significance for all inputs, as those reported here. However, the coefficient on R&D and its spillover were found to be relatively lower when such a trend was included. This gives credence to the assumption, that technology embodied in R&D captures technological progress.

⁴ Nonetheless, Equation 2 was also estimated using the level of value-added as the dependant variable, but the results remained robust to this change, and hence, are not reported here.

⁵ The financial year in India runs from 1 April to 31 March. For expositional convenience, 1993–94 is written as 1994, hereafter, and the practice is followed throughout this article.

⁶ Data on capital deflators with base 1994 is available only up to 2004. Thereafter, the new series with base 2000 has been introduced. For 2005 and 2006 therefore, this study has generated deflators using the growth rates of deflators provided by the new series.

⁷ ASI data for 2005 and 2006 is still not available electronically. For these 2 years, data has been extrapolated based on the mean growth of wages over the period 1994 to 2004.

⁸ The RPM is chosen to comprise of investments in the past 4 years. Different values of 2 and 3 years were also tried, but results remained robust to these changes.

⁹ Most empirical studies decide to use a less structured approach as used here, when measuring embodied technology for the following reason: the production function-based approach dates back to Nelson (1964) and Solow (1960), who suggest that when empirically measuring capital stock, a higher weight should be assigned to new vintage equipment to emphasize improved technology embodied in that equipment. They use a production function of the following form: where Y, A, L, J and α are output, Total Factor Productivity (TFP), labour, an index of quality weighted machines and the elasticity of output with respect to labour, respectively, and where t indexes time. In the above equation, J is derived as follows: where K_vt is capital of vintage v in use at time t and λ is the rate of technological progress over time. In this equation, K is derived using the efficiency units of capital, H, and by assuming a common depreciation rate for all years, δ, as follows: K _vt = H _t + (1 − δ)H _t−1 + ··· + (1 − δ) ^t H ₀ where, H, in turn, is derived as a multiplicative function of investment in new equipment, I_vt , and an index of technical efficiency, Φ, as follows, H _t = I _vt Φ _t . Both Nelson (1964) and Solow (1960) assume the index of technical efficiency, Φ, to equal one. However, this assumption is questionable (Hulten, 1992). Even though, Sakellaris and Wilson (2004) have shown a tractable way of measuring the embodied technological change, λ, the complexity involved in measuring Φ, the index of technical efficiency, and the derived assumptions thereof in measuring it, lead others to use a less structured approach (see, e.g. Bregman et al., 1991; Hasan, 2002). A further advantage of using this approach in the Indian case is that a recent study focused on Indian manufacturing, Hasan (2002) has used this approach, and therefore the results from the present study can be compared with those of Hasan.

¹⁰ Most empirical studies, which consider the role of inter-industry knowledge spillovers, assign weights to represent technological closeness. However, most empirical studies, which measure intra-industry knowledge spillovers, as done in the present study, usually measure intra-industry knowledge spillovers by the unweighted sum of the knowledge stock of all other firms in the industry of the firm under consideration (see, e.g. Raut, 1995; Basant and Fikkert, 1996; Chen and Yang, 2005; Lee, 2005). Nonetheless, a few estimations were tried to capture technological absorption of a firm, as is suggested by a few studies when measuring foreign knowledge spillovers or inter-industry spillovers (see, e.g. Griffith et al., 2003). To measure technological absorption of a firm, the R&D spillover available to a firm was pre-multiplied with that firm's R&D intensity. Results were not found to be robust as those presented here, and hence, are not reported.

¹¹ Estimations based on the pooled OLS, FE, RE and PCSE estimators were also tried, but results were not found to be as robust as those reported here. This suggests that the estimators used here have adequately accounted for the problems cited above.

¹² The reason is twofold: (i) the instruments used in the latter approach are weak when the variance of the time-invariant firm-specific technical efficiency, u_i increases relative to the variance of the autocorrelated error term, v_it (Blundell and Bond, 1998). (ii) Moreover, the difference GMM estimator may be susceptible to finite sample biases when the variables in the data follow a random walk process (Mairesse and Hall, 1996; Blundell and Bond, 2000).

¹³ The one-step estimator is used here instead of the more efficient two-step estimator because, as argued by Blundell and Bond (2000), the asymptotic variance matrix based on the former method is more reliable.

¹⁴ Adamou and Sasidharan (2007) find some evidence that current R&D expenditures have a statistically significant impact on growth of sales of Indian manufacturing firms in some specifications used by them.