Since the late 1990s, India has liberalised its economy to an extent unthinkable even 20 years ago. Licensing requirements for production and for access to key inputs have been removed and today India looks more like a capitalist economy than a socialist one. However, a large part of the economy still consists of public sector undertakings of various kinds. Such companies tend to be far less productive than their private-sector counterparts as incentives tend to operate poorly in public firms. Unable to compete with the private sector and seemingly incapable of improvement on their own, are public undertakings doomed?
Public undertakings are often backed by the state and thus can survive even if they are inefficient. They draw resources that could have gone to more efficient firms, which has a negative impact on income and growth. For example, Hsieh and Klenow (2009) use plant-level data to show that productivity in India and China is much more dispersed than in the US and argue that rationalisation of production that would bring this dispersion down to US levels could raise output by 30%-50% in China and 40%-60% in India. Large, inefficient public firms are a significant driver of these results.
There is evidence that private firms are capable of substantial productivity improvements under a threat of competition. Recently, Galdon-Sanchez and Schmitz (2005) showed that when competitive pressures mounted in the market for iron ore due to the collapse in the market for steel in the early 1980s, countries with mines that were close to becoming non-competitive increased efficiency, while others did not. In developing countries, the room for improvement in productivity may be especially considerable. Bloom et al. (2011) show that when textile firms in India were given simple management consulting services, their productivity improved by about 11%. But what about public-sector undertakings? Are they capable of such improvements? Can the threat of competition act as a prod for them to improve productivity? This is the question we study in a recent paper (Das et al. 2010) using a unique floor-level dataset from the Bhilai Rail and Structural Mill, part of the Bhilai Steel Plant, a public-sector undertaking in India.
The Bhilai Steel Plant covers about 17 square kilometres and is a small city in itself. The jobs of regular workers are secure, with excellent fringe benefits including schooling, healthcare and housing, as well as travel benefits and ample leave. As a result, these jobs were, and by all accounts remain, highly valued.
The Rail and Structural Mill is an integral part of the Bhilai Steel Plant. It was commissioned in 1960 with enough capacity to satisfy domestic demand at that time. Since then, it has been the sole supplier of rails for Indian Railways. The plant has had problems keeping up with orders from the Railways, although the stated objective of management was output maximisation. In 1998 and 1999, after a string of train accidents, which were caused by high hydrogen content in the rail steel, Indian Railways suspended orders from the Bhilai Steel Plant. At this time there was considerable questioning of the ability of the plant to adequately provide the rails needed by the Railways. Questions were raised in parliament, and the government was contemplating allowing private players into rail production. In this setting, the Rail and Structural Mill faced not just competition, but a threat to its very existence. A productivity surge ensued. How was this surge obtained?
We use information on daily operations at the Rail and Structural Mill to help answer this question. The mill operates continuously with three production shifts per day and produces two major types of output: rails and heavy structurals (girders of various shapes that are used in construction). We obtain data on the number of steel blooms rolled into the final product in each shift, a list of all workers present during the shift (with their designations), all delay episodes along with their duration, and a description of the cause of each delay. We combine these data with administrative data on worker characteristics and all types of training. Our data set is far more detailed and extensive than that in any previous work in the area, which allows us to look at what occurred on the shop floor and suggests where the apparent productivity improvements came from.
The data in some ways speak for themselves. As is evident from Figure 1, the productivity surge was driven by an increase in the output of rails. After following similar paths, the output per shift in rails and structurals diverged after 2001. This occurred simultaneously with an episode of training directed towards increasing the output of rails. Is this merely a coincidence or is there reason to think that the training caused the surge in productivity?
Figure 1. Output per shift, rails and heavy structurals
We take two approaches to answering this question. First, we use reduced-form regressions and show that after controlling for the type of product and workers, this training episode was associated with a significant increase in the output of rails, but not of structurals. The estimates suggest that the training episode, which was low in terms of cost, raised the output of rails by 20%.
In our second approach, we model the production process in a more structural way. We posit that adverse events occur and these take time to resolve. This creates downtime which reduces output. These events could be triggered by worker errors or by outside forces. We estimate our structural model of the production process using detailed data on delays, their causes and durations. This allows us to see how worker characteristics, including training, affect delays and thereby output. Our results show that the training episode we focus on significantly increased output mainly through decreases in the occurrence of worker errors, while other inputs and training had little or no effect. The effect of the productivity training episode is, however, half the size of that in our first approach. This suggests that the reduced form estimates pick up some spurious correlation between output and the training received by the workers. Nevertheless, the effect of the productivity training remains large.
Although workers and managers in our study experienced no changes in marginal incentives for performance, perceived competition brought about attempts to increase output, which included the highly successful training episode we focus in our paper. Our work suggests that competitive pressures may operate quite effectively with public ownership and the case for privatisation in emerging economies that is so often made may need to be qualified. In fact, competition in these settings may play a much more important role than in industries dominated by private firms precisely because employee compensation in public firms is often above market wages and there is a gap between the actual and potential productivity of workers. In our example, workers were suddenly aware that their plum jobs were at risk. If competition forced the mill to close, they could all be legally fired!
Bloom, Nicholas, Benn Eifert, Aprajit Mahajan, David McKenzie, and John Roberts (2011), “Does Management Matter? Evidence from India”, NBER Working Paper 16658.
Das, Sanghamitra, Kala Krishna, Sergey Lychagin, and Rohini Somanathan (2010), “Back on the Rails: Competition and Productivity in State-owned Industry”, NBER Working Paper 15976.
Galdon-Sanchez, Jose E and James A Schmitz, Jr. (2005), “Competitive Pressure and Labour Productivity: World Iron-Ore Markets in the 1980’s”, American Economic Review, 92(4):1222-1235.
Hsieh, Chang-Tai, and Peter Klenow (2009), “Misallocation and Manufacturing TFP in China and India”, Quarterly Journal of Economics, 124(4):1404-1448.