In search of contextual variables in a stochastic DEA framework: effect of regulation on efficiency of Indian cement industry

Abstract

The paper assesses the difference in the nature of relationship between efficiency and its determinants of Indian cement producing firms in the presence and absence of environmental regulation. It provides an intuitive construct to identify the determinants of efficiency, which is followed by an empirical verification using a two-stage stochastic data envelopment analysis. Important quantitative determinants of efficiency include factor intensities of outputs, levels of pollution abatements, gross value added, whereas size, age, ownership type, development indicators of the place of location of a firm are among the qualitative determinants. An initial phase and a matured phase of regulation are considered in the paper. The main results suggest that capital intensity of output plays a positive role in increasing efficiency under regulation in the initial phase, material intensity in both phases and labour intensity, a negative role, in the subsequent phase. Higher levels of pollution abatements cause higher efficiency under regulation, while in the initial phase private firms report higher efficiencies under regulation; it is just the reverse in the subsequent phase. Size, age and development indicator of the state in which a firm is located can play a positive role to achieve higher efficiencies in the matured phase of regulation.

Keywords:

• non-parametric methods
• efficiency
• regulation

Notes

1 Lovell (1993).

2 Analysis on the effect of technological flexibility on technical efficiency.

3 However, we find some studies touching upon the issue of an influence of a reform in an economy on technical efficiencies, a few among them being Movshuk (2004) and Ray and Ping (2006) for the impact state-owned enterprises reforms in China, respectively, on technical efficiencies of iron and steel and the manufacturing industries, Fan (1999) analysing the impact of rural reforms based on decentralization on technical efficiencies of Chinese agriculture farms; Cotfas et al (2000) on the effect of reforms on technical efficiency (TEs) of Romanian Cement Industry; Neogi and Ghosh (1994) on the effect of liberalization policies on technical efficiencies of Indian manufacturing industries.

4 However, out of 68 firms taken for the analysis for 1999–2000, only one firm belongs to 26941 and two firms belong to 26943, the remaining 65 firms belong to 26942 whereas for 2003–2004 out of 243 firms, nine firms belong to 26941, six to 26943 and the remaining 228 to 26942. Therefore the analysis is mainly confined to portland cement.

5 Also, see Bandyopadhyay (2009, pp 167–168).

6 Investment limits in plant and machinery as the criteria determining the size as specified by the industrial authorities in India have changed over the period of our study. For comparability purposes we have used the quartiles to determine the size in each year.

7 We finally solve the model setting α=1, as it can be shown that this does not alter the optimal value of the objective function (Färe et al, 1987).

8 See Bandyopadhyay (2009).

9 For cement, the main air pollutant of concern is the PM. The main sources of PM emissions are associated with intermediate and final materials handling and storage (including crushing and grinding of raw materials) and the operation of kiln systems, clinker coolers and mills. Need for particulate collection is driven by both regulation and material recovery. The main pollution prevention and control techniques intend to improve upon handling and storage of materials by better management and design of systems and investments on rotary bag filling machine, electrostatic precipitators, fabric filters, cyclones, etc. A closer look at the existing prevention and control strategies of international standards reveals that they aim at fulfilling multiple objectives of material recovery and recycling with pollution abatement (http://www.cibo.org; http://www.ifc.org).

10 There are problems constructing even an unbalanced panel from ASI data used in our analysis. CSO changes the frame for stratified sampling every year and the firm identification codes are not disclosed.

11 For details see Table 2.

12 For details see Bandyopadhyay (2009).

13 The regressors are chosen in three groups, combining quantitative and qualitative variables (Models 1 and 4), only quantitative variables (Models 2 and 5) and only qualitative variables (Models 3 and 6). Comparisons of Models 2 and 5 are not explained separately as the results of Models 1 and 4 and Models 4 and 5 for the quantitative variables are the same.

14 The description refers to the OLS regressions results as DEA followed by OLS outperforms many parametric methodologies. Apart from Model 1 for the years 2003–2004, there is no difference in the results for OLS and TOBIT in the second stage. The difference in Model 1 does not alter the basic conclusions however.