Measurement of productivity and quality in non‐marketable services. With application to schools
| DOI | https://doi.org/10.1108/09684880610643593 |
| Published date | 01 January 2006 |
| Date | 01 January 2006 |
| Pages | 21-36 |
| Author | R. Färe,S. Grosskopf,F.R. Forsund,K. Hayes,A. Heshmati |
| Subject Matter | Education |
Measurement of productivity and
quality in non-marketable
services
With application to schools
R. Fa
¨re and S. Grosskopf
Department of Economics, Oregon State University, Corvallis, Oregon, USA
F.R. Forsund
Department of Economics, University of Oslo, Oslo, Norway
K. Hayes
Department of Economics, Southern Methodist University, Dallas,
Texas, USA, and
A. Heshmati
TEPP, College of Engineering, Seoul National University, Seoul, South Korea
Abstract
Purpose – This paper seeks to model and compute productivity, including a measure of quality, of a
service which does not have marketable outputs – namely public education at the micro level. This
application is a case study for Sweden public schools.
Design/methodology/approach – A Malmquist productivity index is employed which allows for
multiple outputs or outcomes such as test results and promotions without requiring price data with
which to aggregate these outputs. It also allows one to account for inputs such as teachers and
facilities as well as proxies for quality of the inputs (e.g. experience of teachers) and outputs. This
model generalizes the basic data envelopment analysis (DEA) models – used successfully to measure
performance in many educational applications – to the intertemporal case. A way of computing
quality and quantity components of overall productivity is employed.
Findings – The case study is an application to the Swedish primary and secondary school system
over the 1992 to 1995 period. It was found that quality “matters”, i.e. productivity growth changes
when one accounts for quantity.
Research limitations/implications – Thedata available implied that the specification is restricted
to an intermediate production model, i.e. the output data only account for the intermediate outcomes of
education like grades and promotions, but not the longer term outcomes related to success in the job
market or higher education, which one proposed as a task for future research.
Originality/value – The indices which are computed at the micro level are of value for policy
purposes (does investment in quality matter?) and in an evaluation context.
Keywords Quality, Productivity rate, Publicschools, Sweden
Paper type Research paper
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/0968-4883.htm
This paper was initiated while the first three authors were visiting the Department of Economics
at Gothenburg University.
Measurement of
productivity and
quality
21
Quality Assurance in Education
Vol. 14 No. 1, 2006
pp. 21-36
qEmerald Group Publishing Limited
0968-4883
DOI 10.1108/09684880610643593
1. Introduction
In many countries education of young children and adolescents is a publicly provided
service. This particular government activity has many desirable characteristics such
as improving the human capital of the students and providing (somewhat) uni form
citizenship training. To adequately measure the productivity of publicly provided
education, we not only need to measure the outputs as described above, but also to
include some aspects of the quality of the training. Often the provision of primary and
secondary education is decentralized, i.e. locally provided and financed, with the
consequence that the efficiency with which local government resources are used to
achieve these goals varies widely. Central governments and voters are interes ted in the
relative productiveness of these local governments in their efforts to educate.
The current paper provides a technique for identifying a local government’s
productivity in providing education focusing particularly on the need to measure
productivity when outputs are not marketed, and both inputs and outputs can vary in
quality. Another contribution of this paper is that we derive an explicit quality index as
a component of productivity. The overall productivity index is thereby decomposed
into a quality and a quantity component. From a policy perspective these may be used
to assess existing allocations of resources and policies targeted at improving
performance. We illustrate this model with an application to the Swedish primary and
secondary school system.
Both quality and productivity changes are defined here in terms of distance
functions, consistent with the Malmquist productivity index introduced by Caves et al.
(1982). We follow Fa
¨re et al. (1989, 1995) and compute this index “directly” by applying
linear programming techniques to compute the underlying distance functions. One of
the key advantages of computing the Malmquist index as opposed to say the To
¨rnqvist
index, is that the former does not require data on prices or cost shares. This is
particularly useful for publicly provided, non-priced services such as the public sector
education context addressed here, where there are no obvious prices for sch ool outputs.
The technique we use to compute the indexes also has the advantage of providing
individual measures of productivity and quality and allows identifi cation of
“benchmark” observations, which has proved useful as a management tool. The
individual distance functions which are used to construct the Malmquist index are
themselves performance measures – they are equivalent to the technical efficiency
measures familiar from Farrell (1957) and data envelopment analysis (DEA).
Earlier work using this data set employed econometric techniques to estimate cost
and production functions to determine the effect of quality characteristics on the
production and costs of education. Heshmati (2002) uses both a hedonic price and index
number approach to including quality characteristics in the estimation of producti on
and cost functions for the data set we use here. In contrast to the approach taken in this
paper, Heshmati used quality characteristics to arrive at a quality-adjusted measure of
output. We include quality characteristics in an overall index of productivity, and
isolate a quality change index. Also using this data set for 1993/1994, Heshmati and
Kumbhakar (1997) estimate stochastic frontier production and cost functions to predict
efficiency in provision of school services for each municipality, which they find to be on
the order of 90 percent to 92 percent on average.
Here we take an index number approach rather than an econometric approach. We
follow Fixler and Zieschang (1992) and augment the Malmquist index to include
QAE
14,1
22
To continue reading
Request your trial