R&D IN INDIAN MANUFACTURING FIRMS: EFFECTS OF AGE, SIZE & TECHNOLOGY TRANSFER

1. Introduction

Innovation by firms is recognized as a vital force behind economic growth of a nation (Romer, 1986, 1990; Barro and Sala-i-Martin, 1995). The endogenous growth framework suggests that investing in Research & Development (R&D) helps firms to enhance their productive efficiency, often leading to sustainable competitive advantage (Romer, 1990). Many studies in the past have confirmed positive effects of R&D investment on the performance of a firm (Tsai and Wang, 2005; Hall et al., 2013; Khanna and Sharma, 2018a; Sharma, 2019). It is also well known that in any industry, not all firms invest in R&D. Firms, instead self-select into undertaking R&D investment based on the prevalent market structure or expected net gains from R&D (Feng and Ke, 2016). While most studies have focused on the outcomes of innovation and R&D some of them have confirmed the bidirectional relationship between R&D and performance, where R&D expenditures are also determined by performance and other firm characteristics (Cohen and Levinthal, 1989). In this light, this paper aims to examine the role of various firm characteristics, especially age and size in determining its R&D behavior.

The paper contributes to the existing literature in the following ways. First, many studies have assessed the effects of R&D on productivity performance, while the question of what determines R&D behavior in a firm still remains inadequately answered. We not only examine various firm characteristics that influence its inclination to undertake R&D, but also address the rather unfairly ignored question on the inter-play between in-house R&D and foreign technology transfer. Further, our rich dataset allows us to address these questions by stratifying our full sample into sub-samples based on firm size and age. This helps us to shed more light on the subject as compared to the previous studies that impose homogeneity restrictions on for instance, old and young firms (or large and small firms). Secondly, to compute production function elasticities, we employ two-staged least squares (2-SLS) methodology, which helps us to control for transmission bias or simultaneity associated with input choice. Thus, unlike previous studies that rely on traditional estimators, our estimates of total factor productivity (TFP) are less prone to measurement errors. Lastly, by shedding light on the disparities between R&D undertaking and non-R&D undertaking firms, we contribute immensely to the growth literature on India. This is important as the evidence on the determinants of R&D behavior in emerging economies has remained sparse mainly on account of data limitations. This is probably the first attempt to evaluate our research question in the context on Indian manufacturing. Broadly, our findings suggest that R&D propensity of a firm is strongly influenced by its age and size. We find that young firms are more likely to invest in R&D. More interestingly, our results indicate that small firms possibly use in-house R&D as a complementary input with purchased foreign technology. The remainder of the paper is organized as follows. Section 2 highlights the relevant literature review. Section 3 gives the data and TFP estimation, while Section 4 highlights the results and discussion of the study. Lastly, Section 5 concludes the paper with some future research directions.

2. Literature Review

Schumpeterian hypothesis which states that R&D productivity is an increasing function of firm size has been widely discussed in the literature (Lee and Sung, 2005). Tsai and Wang (2005) partially confirm this hypothesis in the context of firms listed on the Taiwan Stock Exchange. Their study reports that there is a “U-shape” relationship between R&D output elasticity and firm size. They also find that the productivity effects of R&D remain high, irrespective of firm size. The result holds even when the sample is divided into traditional firms and high-tech firms.

Cohen and Klepper (1996) suggest that large manufacturing firms command greater R&D advantages associated with economies of scale despite the problem of high bureaucratic control that they face. Kessler et al. (2000) argue that large firms maintain a diverse portfolio of research projects. The accumulated internal and external knowledge helps these firms to accelerate the process innovation and product development. On other contrary, Scherer and Ross (1990) argue that the attention of a large firm’s technologists is often diverted because of loss of marginal control. This also leads to lower R&D productivity. Nooteboom (1994) and Salavou and Avlonitis (2008) argue that innovation type varies with the size of the firm. Bos-Brouwers (2010) argue that small firms face a higher risk of new product failure. Also, radical innovations take more time and resources to materialize (Gassmann and Keupp, 2007). This often deters small firms from innovating (Allocca and Kessler, 2006). Their findings further indicate that small firms are better at improving the existing technologies. Small firms have a focus on project-driven incremental innovation (Pullen et al., 2009). Incremental innovation consists of making small-scale improvements like adding new features. Fombang and Adjasi (2018) found overdraft, trade credit and asset finance as significant drivers of innovation.

A related stream of literature has investigated the pay-offs of purchased technology from foreign firms viz.-viz. firm’s own R&D efforts (Braga and Willmore, 1991; Katrak, 1997; Majidpour, 2017; Pietrobelli, 2018). Katrak (1997) argues that a firm’s technology imports may be positively linked with its R&D intensities in a competitive market situation. However, the relationship may not hold in a protectionist environment. However, none of these papers has specifically focused on understanding the interplay between R&D and technology purchase for firms differing in age, size and other firm characteristics such as export status or magnitude.

Thus, the previous studies have provided relevant evidence on innovation patterns in firms. With reference to these studies, the present work not only gives a more comprehensive view but also addresses some questions that the past literature has failed to answer adequately.

3. Data and TFP Estimation

3.1. Data

For the purpose of analysis, data is sourced from Enterprise Surveys (ES) published by the World Bank. Enterprise Surveys capture data through interviews with the manufacturing and service sector firms in order to create globally comparable business environment indicators. For the objective suited to this paper, we extract data on the manufacturing sector that covers 7165 firms. Details on sectors are given in Table A1 of the Appendix. We use the year 2014 data, which is the most recent survey on India. Table 1 presents the description of the variables used in the model.

Table 2 summarizes the descriptive statistics for these variables. All variables except for export to sales ratio are expressed in terms of natural logarithm. The export to sales ratio therefore shows a higher variation; however, this is not problematic for our study. The variable lncapacity, on the other hand, shows the minimum variation. The correlation matrix for these variables is given in Tables A2 and A3 of Appendix A.

3.2. TFP estimation

Some scholars suggest that TFP may have an influence on the R&D behavior of the firm (Peters et al., 2017). To measure TFP, we first define a production function that may be used for our cross-sectional dataset. Given our cross-sectional data, we define and estimate the following value-added based production function :

We assume a Cobb–Douglas² production function specification, where we define the TFP index as ${\theta }_i=e^{{\mu }_i}$. The PF can therefore be specified as :

Taking natural logarithm of both sides in Equation (2), we get :

Our dataset provides us information on heterogeneous quality of labor based on their skillset. Exploiting this feature of this rich dataset, we estimate factor shares for skilled and unskilled labor separately. This greatly helps us to reduce any measurement error in our TFP series.

To get our baseline parameters, we estimate Equation (3) using ordinary least squares (OLS) methodology. However, considering that PF estimation is typically plagued by the problem of simultaneity as firms constantly adjust their quantity of labor input based on the unobserved productivity shocks they face (Levinsohn and Petrin, 2003), we apply two-staged least squares approach (2SLS) (see for example, Khanna and Sharma, 2018b). In this methodology, we instrument our labor variable using “average years of education” of production workers, which is highly correlated with number of labor but appears to have no direct impact on firm’s unobserved productivity shocks. This helps us to circumvent the problem of simultaneity in PF estimation. Estimation results using both OLS and 2SLS methodologies are summarized in Table 3.

We find that the production technology is dependent more upon labor as compared to capital. Further, there is not much difference in the estimated factor shares of skilled and unskilled labor. The factor shares of labor and capital are found to be 0.93 and 0.25, respectively using the OLS estimator. As expected, accounting for simultaneity leads to a downward correction in the labor coefficient and slightly raises the size of the coefficient of capital (Levinsohn and Petrin, 2003).

4. Results and Discussion

4.1. Full sample results

In order to formally estimate the effect of different characteristics of a firm on its propensity to undertake R&D, we estimate the following models in a Probit framework. In this set up, we analyze the effect of firm’s age, size, proportion of total exports to sales, access to foreign technology and a few dummies on the firm’s propensity to undertake in-house R&D investment. The general regression model that we use in this analysis is of the following form :

where RD is a dummy indicating whether firm $i$ undertakes R&D investment, age and size are measured in terms of years and number of employees and technology transfer captures firm’s access to foreign technology. X refers to firm-specific control variables. In particular, we include firm’s exports to total sales ratio, capacity utilization ratio and dummies indicating whether or not firm holds an international quality certification, whether or not firm’s production employees have received formal training and industry dummies, etc. We use the following baseline model for regression analysis :

We test alternate specifications of this model in a binary-Probit framework and present these results in Table 4. Wald test for all models is statistically significant suggesting that the estimated regressions are overall significant. To test for multicollinearity, we report the mean variance inflating factor (VIF) in Table 4 for all the models. The estimated value of the VIF is less than 10 suggesting that multicollinearity may not be a serious problem in our set up (Gujarati, 2009).³ We find that a firm’s unobserved productivity shocks and export ratio positively impact the inclination to undertake R&D. Further, we find that as firms grow in size, they are more likely to invest in R&D. This finding is consistent with the results of previous studies such as Dosi (1988), while it is in contrast with several others such as Shefer and Frenkel (2005), and Poddar and Singh (2020) that report no significant relationship between innovation activity in the firm and its age. One justification behind our result might be that in developing countries such as India, where there are a large number of small firms, innovation is often embedded in technology adoption. Since undertaking in-house R&D involves a huge sunk cost, smaller firms in the manufacturing sector find it more convenient to instead purchase technology from foreign firms. Further, we find that younger firms are more likely to undertake R&D and as they grow in age, their inclination to invest in R&D reduces. This is a rather peculiar finding. One way to justify this result is that young firms especially in the high-tech industries often face tough competition in the market that results in a high incentive for undertaking in-house R&D. The result is in line with Seenaiah and Rath (2018) that reports firm’s innovation activity to be negatively associated with its age. Thus, our results indicate that both age and size are important factors in influencing a firm’s inclination to undertake R&D. The coefficient of lnTFP is positive and statistically significant suggesting that more productive firms have a higher inclination to invest in R&D. This appears consistent with the self-selection hypothesis of R&D, which suggests that given the high sunk cost of establishing R&D labs and R&D equipment, productive firms often self-select into R&D depending upon the prevailing market structure (Sasidharan and Kathuria, 2011).⁴ Besides, we find that firms that hold an international quality certification and those that undertake training of their employees are more likely to undertake in-house R&D. On the other hand, we do not find any relationship between capacity utilization and firm’s inclination for R&D investment.

Our results confirm that firms that are larger in size are more likely to invest in R&D. This may seem reasonable given the high sunk cost of R&D, which makes technology transfer a more feasible alternative for smaller firms. In order to formally test this hypothesis about the possible interplay between purchase of technology and firm size, we introduce an interaction term between firm size and technology transfer dummy (columns 3–5 of Table 4). While all previous results hold, we find the coefficient on the interaction term to be negative and statistically significant. This suggests that small firms that undertake technology purchase have a higher probability of investing in R&D and this is true of the sample firms. Moving forward, our results (columns 2–5 of Table 4) suggest that younger firms show a higher inclination towards investing in R&D. Thus, it would also be interesting to know the inclination of young firms towards transfer of technology. To this end, we introduce an interaction term between firm age and technology transfer (columns 4 and 5 of Table 4). However, we do not find any conclusive result as the estimated coefficient of the interaction term is insignificant across both the models.

While the above analysis is useful, it does not allow us to understand the magnitude of impact of these variables on the firm’s inclination to invest in R&D. For instance, these results suggest that a unit increase in the firm’s size (age), raises (reduces) the z-score or probit index by 0.150–0.179 (0.071–0.078) across various models. Given the complexity of interpretation, we also report the marginal effects for our main variables of interest in Equation (5). These are presented in Table 5. The marginal effects reflect by how many percentage points inclination to undertake R&D varies with changes in the value of each of these explanatory variables. Our results from Model 1, for instance, suggest that with a percentage rise in the size (exports ratio), the probability of undertaking in-house R&D improves by 0.049 (0.002) percentage points. Further, these results point towards a sizeable impact of both age and size on the R&D inclination of a firm.

4.2. Sub-sample results

Our results above suggest that the R&D propensity of a firm is strongly influenced by its characteristics, specifically its size and age and a few other variables such as TFP and exports as a percentage of its total sales, etc. However, what is not very clear above is whether firms of different age and size also exhibit differential R&D behavior with respect to the above explanatory variables specifically, technology transfer. In particular, we want to assess whether firms consider technology transfer as a substitute or a complement for R&D investment and whether the relationship between R&D and technology transfer (and other explanatory variables) vary with firm age and size. In order to formally assess the same, we estimate a simplified version of our benchmark model in Equation (5) for various sub-samples depending upon the age and the size of the firm. Results are summarized in Table 6 below.

As seen in the table, we exclude the control variable on capacity utilization as the variable does not come out to be significant. As firms in a sub-sample are homogeneous in terms of a given criteria, we exclude control variable on age (size) from sub-samples of old or young firms (large or small firms). Our results suggest that positive productivity shocks do not significantly influence the R&D decision of a firm that is more than 15 years in age (column 1). However, this is not true of young firms (column 3), where TFP appears to be a significant factor influencing the decision to undertake R&D. We find the coefficient of technology transfer to be negative and significant for large firms (column 2) possibly suggesting that technology transfer works as a substitute for in-house R&D. This result is in congruence with Sharma (2019) that studies the innovation behavior of Bangladesh-based firms. Next, a rather interesting finding of our study is that for both young and small firms (columns 3 and 4) technology transfer possibly works as a complement to in-house R&D. This suggests that young firms, especially those that employ less than 100 people, exhibit a significantly different R&D behavior as compared to market giants. This seems understandable as young (and small) firms often attract tough competition, in the face of which innovation becomes essential. Our results suggest that instead of choosing between technology transfer and R&D, these firms undertake both, and technology transfer positively impacts the firm’s likelihood to invest in in-house R&D. A plausible explanation behind this finding could be that small and young firms import technology and then undertake R&D in order to adapt or assimilate this technology in the production process (Katrak, 1985; Majidpour, 2017). This is evident from our results based on firms that are both young and small (column 6). Moving forward, we find that irrespective of firm’s age and size, export to sales ratio affects R&D propensity positively (columns 1–6). Similarly, the sign and significance of age, size and training dummy match our previous results based on the full sample.

Further, the marginal effects of our key variables of interest for the sub-sample results are summarized in Table 7. Our results suggest that with a percentage rise in the exports ratio, the probability of undertaking in-house R&D improves by 0.002–0.003 percentage points across various models for all firms irrespective of their age and size. More importantly, these results show that as a firm moves from no technology transfer to technology transfer, its probability for undertaking R&D reduces (improves) by 0.145 (0.073) for large (small) firms. For firms that are both (old and large) young and small, the probability of undertaking R&D reduces (improves) by 0.114 (0.089) as the firm moves from no technology transfer to technology transfer.

5. Conclusion and Directions for Future Research

Our findings suggest that young firms are more likely to invest in R&D as are firms that are large in size. We find that capacity utilization is not a significant factor affecting a firm’s propensity to undertake R&D. Contrary to previous studies that portray technology transfer as a possible substitute of in-house R&D, we find that the result holds only for large firms. We find that small firms exhibit a unique R&D behavior. These firms possibly rely on both R&D and technology transfer, and in fact, the instance of technology purchase seems to be positively associated with investing in R&D. This is possibly indicative of the synergy benefits that young firms draw from the combined use of technology transfer and in-house R&D. Nooteboom (1994) notes that as against large firms that invent new technologies, small firms are better at making incremental improvements on acquired technologies. In this light, our results on the complementary relationship between purchased technology and in-house R&D for young firms seem justifiable. From the policy perspective, this paper throws light on how a firm’s age and size affect its innovation behavior including its inclination for undertaking R&D, and purchasing technology from abroad. Investing in technologically advanced inputs such as R&D will be crucial for attaining the target of the Make in India, 2014 campaign that aims to increase the output contribution of manufacturing to around a quarter of the nation’s GDP by the year 2025. Given the abysmally low penetration of R&D in India, the decade of 2010–2020 was recognized as the “Decade of Innovation”, where the government attempted to raise R&D expenditure to 2% of the nation’s GDP. In this context, the results of our study throw light on significant disparities in the R&D inclination of firms differing in terms of their age and size. Further, our results show that export to sales ratio are positively associated with its inclination to undertake R&D. Thus, for encouraging R&D activity among firms, policy measures must take into consideration the magnitude of the firm’s exports. As a part of its Atmanirbhar Bharat 2020 initiative, the Government of India has devised the Production-Linked Incentive (PLI) scheme that aims to promote manufacturing capabilities and exports in key manufacturing sectors of the country. In this backdrop, the results of this study suggest that enhancing manufacturing exports would impact the sector’s R&D propensity positively, which might result in meaningful spillover benefits for the rest of the economy.⁵

Besides offering valuable insights, the paper also provides interesting insights for future research. For instance, the paper demonstrates a firm’s innovation behavior by looking into whether it invests in R&D. Future studies can focus on various distinct forms of innovation such as product innovation, process innovation, organizational innovation, etc. and their interaction with the firm’s R&D activity. A notable limitation of this study is that it is based on cross-sectional data covering only one time-period. Future research may extend this analysis for different time-periods combining data from several rounds of the survey. Further, circumventing data challenges on cross-country innovation statistics, a future study can also focus on a wider coverage in terms of both countries and firm and industry-level characteristics such as competition, ownership structure, governance quality, etc. that may be significantly identified with its R&D behavior.

Notes

¹In the Enterprise Surveys, the distinction between skilled and unskilled labor is made based on the occupation or tasks performed by the employees and not their qualification. Further, the “skilled” (“unskilled”) employees category in this dataset corresponds to skill levels 2–4 (level 1) of the International Labor Organization (ILO) classification.

² The extant literature on TFP estimation uses three main specification forms for production function: Cobb–Douglas (CD) specification, Constant Elasticity of Substitution (CES) specification and Transcendental Logarithmic (TL) specification. The most widely adopted is the CD specification due to its advantages in terms of parsimony, uniformity and flexibility. Given these advantages, it suffers from a few limitations, for instance, its assumptions such as constant RTS and perfect competition in the factor and the goods markets. However, despite these limitations, it is widely used in the existing literature (see Trivedi, 2004; Mitra et al., 2016). Murthy (2002) highlights that the CD functional specification is better suited to the Indian setting as against other functional forms such as the CES and the Translog forms.

³ Additionally, we also report the pair-wise correlations between the regressors in Table A3 in Appendix A. As per the rule of thumb, multicollinearity is a serious problem if the correlation is higher than around 0.80 (Gujarati, 2009).

⁴ While traditional research has more often focused on productivity as an outcome of R&D (Khanna and Sharma, 2018a), some scholars have also of late attempted to explore the effects of firm-performance on its decision to opt for R&D (Peters et al., 2017). To assess the robustness of our result against a plausible simultaneous effect between R&D and TFP, we estimate all regressions excluding TFP. Our results remain quite consistent in statistical terms.

⁵ Cabinet approves PLI Scheme to 10 key Sectors for Enhancing India’s Manufacturing Capabilities and Enhancing Exports — Atmanirbhar Bharat. URL: https://pib.gov.in/PressReleasePage.aspx?PRID=1671912, accessed on 8 December 2020.