Investigating the Prevalence of ‘Missing Middle’ in Indian Manufacturing

Several earlier studies have observed that the size structure of Indian manufacturing is characterized by the ‘missing middle’ phenomenon – a relatively low appearance/occurrence of middle-size industrial enterprises among industrial enterprises of different size classes. The authors of the studies have commonly drawn this inference by comparing the employment shares of industrial enterprises belonging to different size classes, a methodological approach whose validity has been seriously questioned by Hsieh and Olken (2014). In this paper, we carry out a fresh investigation into the prevalence of ‘missing middle’ in Indian manufacturing’s size structure by using the unit records of the NSS (National Sample Survey) 73rd round survey, which covered the year 2015-16, along with the unit records of the Annual Survey of Industries (ASI) for 2015-16. The NSS 73rd round survey covered unorganized (unincorporated) manufacturing. By contrast, ASI covers organized manufacturing. A distinguishing feature of the present study is that it follows the empirical strategy suggested and adopted by Tybout (2014), and thus the prevalence of the ‘missing middle’ phenomenon in Indian manufacturing is investigated using an estimated Pareto distribution. Based on this analytical approach, we find some empirical evidence pointing towards the prevalence of the ‘missing middle’ phenomenon in Indian manufacturing. We find other supportive/corroborative empirical evidence by examining manufacturing enterprises’ survival/closure probability in different size classes. However, a fundamental doubt arises about the validity of this deduction regarding the prevalence of ‘missing middle’ in Indian manufacturing because the two datasets on industrial enterprises (NSS 73rd round and ASI) for 2015-16 that have been utilized for the present study do not cover between them a large number of industrial enterprises in India, somewhere in the range of 100 to 150 thousand, probably accounting for about five per cent of the aggregate manufacturing employment. This argument about data under-coverage applies equally well to most earlier studies dealing with the ‘missing middle’ issues in Indian manufacturing