Productivity and firm size in India

Though it is more usual to estimate productivity at the plant level, firm-level productivity is more appropriate in our context as size-restriction is imposed at the firm-level and not the plant-level. Admittedly, they are not identical as a firm may have several plants, and as Winter (1999) shows firm financials may affect plant productivity. Using balance sheet data is a more modern trend, partly because of availability of such databases. One example is Khandelwal and Topalova (2010), who use the same database to estimate TFP.

The unorganized sector accounts for about 45 % of the employment in the manufacturing sector and contributes to about 44 % of the GDP as estimated by the Central Statistical Organisation, India. The data we use does include firms which are not publicly traded. A detail discussion of the informal manufacturing sector in India is beyond the scope of this paper.

Source: Development Commission (SSI), Third Census, Government of India.

Morris and Basant (2006) present a detailed discussion of the financial constraints faced by small firms in India.

We use one more source of data for robustness checks. There is some evidence that large firms pay higher wages (Idson and Oi [1999], Brown and Medoff [1989]). If this is systematically the case, then our calculation of labor is biased, as we use an industry-wide deflator. This will underestimate the TFP of larger firms affecting our regression results. Since firm-level wage data is missing for a large majority of firms, we cannot deal with this directly as it will lead to a drastic loss of sample size. Therefore, we deal with this in a couple of indirect ways. First, we see if, given the limited data, there is any evidence that large firms systematically pay more. We do not find such evidence. We next look at a roughly comparable database on the Indian manufacturing sector, the Annual Survey of Industries, and estimate the average wage premium between small and large firms (Chamarbagwala and Sharma [2012]). The average wage rate of firms in the fifth quintile is approximately three times that in the first quintile. That is, large firms pay almost three times the wage an average small firm pays. We recalibrate our wages to account for this wage premium and then recalculate labor and TFP. On estimating our various specifications, we find that the small firms still have significantly higher TFP than the larger firms, though the coefficients become smaller. The details of these exercises are available upon request.

Please refer to Van Beveren (2012) for a detailed discussion on the various measures of TFP.

For brevity, we exclude the details of TFP estimations. The authors will be happy to make them available upon request.

For brevity, we forego a fuller discussion of the size-R&D literature here. Please see Henderson (1993) for an early exposition and Shefer and Frenkel (2005) and Kim et al. (2009) for a summary of empirical research.

We also use a more popular definition of size, as discussed above—size by quintiles of employment. As mentioned in the data section, we do not have data on employment and impute the same by dividing the wage bill by average wage rate. The imputed employment is then divided into quintiles by year and by industry. We once again find a positive and significant relation between being small by this definition and being productive. However, as this imputation of labor is an imprecise method of measuring employment, we relegate it to the “Appendix”.

We obtain similar results when we define size in terms of quintiles of market share instead of assets. This is expected, as asset-based measure of size and sales-based measure of size are likely to be very similar. These results are available upon request from the authors.