Special Topics

in Business
Economics

Griva
A Model for Technical Inefficiency Effects in a
Anastasia,
Mavropoulos Stochastic Frontier Production Function for
Antonios,
Panteli
Antiopi,
Panel Data
Vasiliki
Spanou, Published by: G.E. Battese, T.J. Coelli
Tsichlia
Afroditi

Introduction
Griva Anastasia, Mavropoulos Antonios, Panteli Antiopi,
Inefficiency
Frontier Vasiliki Spanou, Tsichlia Afroditi
Model for
Panel Data

Empirical University of Patras

Application Department of Economics
Conclusions

Friday, November 13th 2015

1/12
 Theory of the Model

Special Topics
in Business
Economics

Griva
Anastasia,
Mavropoulos
Antonios, Abstract
Panteli
Antiopi,
Vasiliki A stochastic frontier production function is defined for
Spanou,
Tsichlia panel data on firms, in which the non-negative technical
Afroditi
inefficiency effects are assumed to be a function of
Introduction firm-specific variables and time.(Aigner DJ, Lovell CAK,
Inefficiency Schmidt P (1977) Formulation and estimation of
Frontier
Model for stochastic frontier production function models. Journal of
Panel Data Econometrics 6:21-37).
Empirical The empirical model is obtained by using a ten years data
Application
on paddy farmers from an Indian village.
Conclusions

2/12
 Introduction

Special Topics
in Business
Economics
The stochastic frontier production function postulates the
Griva existence of technical inefficiencies of production of firms
Anastasia,
Mavropoulos involved in producing a particular output.
Antonios,
Panteli These papers adopt a two-stage approach, in which:
Antiopi,
Vasiliki
Spanou,
Tsichlia The first stage involves the specification and estimation of
Afroditi
the stochastic frontier production function and the
Introduction prediction of the technical inefficiency effects, under the
Inefficiency
Frontier
assumption that these inefficiency effects are identically
Model for
Panel Data
distributed.
Empirical The second stage involves the specification of a regression
Application
model for the predicted technical inefficiency effects,
Conclusions
which contradicts the assumption of identically distributed
inefficiency effects in the stochastic frontier.
3/12
 Inefficiency Frontier Model for Panel Data

Special Topics
in Business
Economics lnYit = xit + Vit Uit (1)
Griva
Anastasia, This equation specifies the stochastic frontier production
Mavropoulos
Antonios,
function in terms of the original production values.
Panteli
Antiopi,
Vasiliki
Yit denotes the production at the t-th observation
Spanou,
Tsichlia
(t = 1, 2, ..., T ) for the i-th firm (i = 1, 2, ..., N)
Afroditi
xit is a (1 x k) vector of values of known functions of
Introduction inputs of production and other explanatory variables
Inefficiency associated with the i-th firm at the t-th observation
Frontier
Model for
Panel Data is a (k x 1) vector of unknown parameters to be
Empirical estimated
Application

Conclusions
Vit s are assumed to be iid N(0, 2 ) random errors,
independently distributed of the Uit s

4/12
 The technical inefficiency effect

Special Topics
in Business
Economics
The explanatory variables in the inefficiency model may include
Griva
Anastasia, some input variables in the stochastic frontier, provided the
Mavropoulos
Antonios, inefficiency effects are stochastic. If the first z-variable has
Panteli
Antiopi, value one and the coefficients of all other z-variables are zero,
Vasiliki
Spanou, then this case represents the model specified in Stevenson
Tsichlia
Afroditi (1980) and Battese and Coelli (1988, 1992).
Introduction

Inefficiency
Whereas If all elements of the f-vector are equal to zero, then
Frontier
Model for
the technical inefficiency effects are not related to the
Panel Data z-variables and so the half-normal distribution originally
Empirical
Application
specified is obtained. If interactions between firm-specific
Conclusions
variables and input variables are included as zvariables, then a
non-neutral stochastic frontier, is obtained.

5/12
 The technical inefficiency effect, Ui

Special Topics
in Business
Economics

Griva
Anastasia,
Mavropoulos
Antonios,
Panteli
Antiopi,
Ui = zit + Wit
Vasiliki
Spanou,
where the random variable, Wit , is defined by the truncation of
Tsichlia
Afroditi
the normal distribution with zero mean and variance, 2 , such
that the point of truncation is zit , i.e.,Wit >= zit . These
Introduction
assumptions are consistent with U, being a non-negative
Inefficiency
Frontier truncation of the N(zit , 2 )-distribution.
Model for
Panel Data

Empirical
Application

Conclusions

6/12
 Technical Inefficiency Production function

Special Topics
in Business
Economics

Griva
Anastasia, The technical efficiency of production for the i-th firm at the
Mavropoulos
Antonios,
t-th observation is defined by equation:
Panteli
Antiopi,
TEit = exp(Uit ) = exp(zit Wit )
Vasiliki
Spanou,
The assumption that the U,s and the V - ts are independently
Tsichlia
Afroditi
distributed for all t = 1, 2,..., T, and i = 1, 2, ..., N, is
obviously a simplifying, but restrictive, condition. Alternative
Introduction
models are required to account for possible correlated structures
Inefficiency
Frontier of the technical inefficiency effects and the random errors in
Model for
Panel Data the frontier.The method of maximum likelihood is proposed for
Empirical simultaneous estimation of the parameters of the stochastic
Application
frontier and the model for the technical inefficiency effects.
Conclusions

7/12
 Empirical Application

Special Topics
in Business
Economics Information on the age and years of schooling for 14 paddy
Griva
Anastasia,
farmers from Aurepalle are used to explain the differences in
Mavropoulos
Antonios,
the inefficiency effects among the farmers.
Panteli
Antiopi,
A total of 125 observations are involved for a ten-year period
Vasiliki from 1975-76 to 1984-85. The stochastic frontier production
Spanou,
Tsichlia function is estimated to be:
Afroditi

Introduction
ln(Yit ) = 0 + 1 ln(Landit ) + 2 (PILandit ) +
Inefficiency
3 ln(Labourit ) + 4 ln(Bullocksit ) + 5 ln[Max(Costsit , 1
Frontier
Model for
Dit )] + 6 (Yearit ) + Vit Uit
Panel Data

Empirical
where the technical inefficiency effects are assumed to be
Application
defined by:
Conclusions
Uit = 0 + 1 (Ageit ) + 2 (Schoolingit ) + 3 (Yearit ) + Wi

8/12
 Estimated Model

Special Topics
in Business
Economics

Griva
Anastasia, Maximum-likelihood estimates of the parameters of the model:
Mavropoulos
Antonios, Stochastic Frontier:
Panteli
Antiopi,
Vasiliki ln(Y ) =
Spanou,
Tsichlia 2.86 + 0.37 ln(Land) + 0.38 (PILand) + 0.85 ln(Labour )
Afroditi (0.60) (0.12) (0.21) (0.12)
0.33 ln(Bullocks) + 0.071 ln(Costs) + 0.014(Year )
Introduction (0.11) (0.031) (0.60)
Inefficiency
Frontier Inneficiency Model:
Model for
Panel Data
U = 1.5 + 0.035 Age 0.006Schooling 0.57 Year
Empirical (2.8) (0.034) 0.077) (0.60)
Application

Conclusions

9/12
 Empirical Analysis

Special Topics
in Business
Economics Table 1. Tests of hypotheses for parameters of the inefficiency
Griva
Anastasia, frontier model for paddy farmers in Aurepalle, = 0.952
Mavropoulos (0.047)
Antonios,
Panteli Null Hypothesis x 2 -value Test Statistic*
Antiopi,
Vasiliki H0 : = 0 = = 3 = 0 12.59 29.99*
Spanou,
Tsichlia H0 : = 0 7.82 26.97*
Afroditi
H0 : 1 = 2 = 3 = 0 7.82 10.69*
Introduction * An asterisk on the value of the test statistic indicates that it
Inefficiency exceeds the 95th percentile for the corresponding
Frontier
Model for x2-distribution and so the null hypothesis is rejected.
Panel Data

Empirical
The estimate for the variance parameter, , is close to one,
Application which indicates that the inefficiency effects are likely to be
Conclusions highly significant in the analysis of the value of output of the
farmers.

10/12
Special Topics
in Business Generalized likelihood-ratio tests of null hypotheses, that the
Economics

Griva
inefficiency effects are absent or that they have simpler
Anastasia, distributions, are presented in Table 1.
Mavropoulos
Antonios, The first null hypothesis, which specifies that the inefficiency
Panteli
Antiopi, effects are absent from the model, is strongly rejected. The
Vasiliki
Spanou, second null hypothesis, which specifies that the inefficiency
Tsichlia
Afroditi effects are not stochastic, 6 is also strongly rejected. The third
Introduction
null hypothesis, considered in Table 1, specifies that the
Inefficiency
inefficiency effects are not a linear function of the age and
Frontier
Model for
schooling of the farmers and the year of observation. This null
Panel Data hypothesis is also rejected at the 5 o level of significance. This
Empirical
Application
indicates that the joint effects of these three explanatory
Conclusions variables on the inefficiencies of production is significant
although the individual effects of one or more of the variables
may not be statistically significant.
11/12
 Conclusions

Special Topics
in Business
Economics

Griva
Anastasia,
Mavropoulos
Antonios, The results indicate that the model for the technical
Panteli
Antiopi, inefficiency effects, involving a constant term, age and
Vasiliki
Spanou, schooling of farmers and year of observation, is a significant
Tsichlia
Afroditi component in the stochastic frontier production function. The
Introduction
application also illustrates that the model specification permits
Inefficiency
the estimation of both technical change and time-varying
Frontier
Model for
technical inefficiency, given that inefficiency effects are
Panel Data stochastic and have a known distribution.
Empirical
Application

Conclusions

12/12