A Panel Data Stochastic Frontier Model with Autocorrelated Inefficiency Debdas Bandyopadhyay 1 and Arabinda Das 2 Department of Statistics University of Kalyani Kalyani-741 235 West Bengal India 1. Introduction: Evolution of the economic efficiency of a firm over time takes places through a process where the level of efficiency at a particular point of time depends upon its past level of efficiency. Be it the firm’s ability to adopt the technology or to take the “right’ decision under uncertainty regarding the scale of operation and expansion, the effect of the “past experience” on the current level of efficiency can not be denied. This process is, in fact, a part of the general process of ‘growth through learning by doing’. Thus, estimation of the nature and the extent of the time-dependence of a firm’s productive efficiency is important not only for the measurement of a firm’s economic efficiency in a dynamic context but also for understanding its relation to the overall growth process of an economy. In the panel data stochastic frontier literature the evolution of firm’s efficiency over time has been studied by the so-called “stochastic frontier models with the time–varying efficiency” (Cornwell, Schmidt and Sickles,1990, Kumbhakar, 1990, Battese and Coelli1992, Lee and Schmidt 1993 and Kumbhakar and Hjalmarsson 1993).Although these studies differ in individual details, they take a common approach in modelling the temporal dependence of a firm’s economic efficiency, In this approach the temporal variation of economic efficiency is captured by an exogenously specified function of time where the efficiency of a firm at any time point is obtained by multiplying this function with the time –independent efficiency of the firm. Although this approach has the intrinsic merit of explaining time variation of efficiency by the ‘all catch’ variable “time” and offers the flexibility of choosing empirically the time path of the economic efficiency of a firm, it ignores an important aspect of the dynamics of the economic efficiency viz. the dependence of a firm’s economic efficiency ‘today’ on its economic efficiency of ‘yesterday”. In other words, in this approach the evolution of a firm’s efficiency overtime is not viewed as a process and ,consequently , in this approach a firm’s economic efficiency at different points of time are independently but not identically distributed random variables. 1 Address for correspondence: Debdas Bandyopadhyay, Department of Statistics, University of Kalyani, Kalyani-741 235, West Bengal, India. E-Mail: debdas.banerjee@gmail.com 2 Arabinda Das, Department of Statistics, University of Kalyani, Kalyani-741 235, West Bengal, India. E-Mail: ara.das@gmail.com . Research of the second author is supported by a fellowship from Council for Scientific and Industrial Research, Govt. of India.