Economies of scale and cost efficiency in the postal services: empirical evidence from Switzerland

Abstract

This paper investigates the scale and cost efficiency of the Swiss Post using an econometric approach. Our empirical analysis considers a sample of postal offices operating in the Italian-speaking part of Switzerland. The paper considers estimation of a log-log cost function employing a data set for the year 2001 with information on 47 small local post offices. The empirical evidence indicates the existence of economies of scale. Further, the outcome of this analysis shows that approximately 50% of the postal offices operate close to the regional standard for efficiency.

Acknowledgement

We would like to thank Giacomo Jurietti and Mauro Andreotti from Swiss Post for their support and also for providing the data. The views expressed in this paper are strictly personal and the responsibility for all remaining errors lies solely with the authors.

Notes

Different approaches can be used to estimate a frontier cost function. A good overview is given by Battese (1992), Fabbri et al. (1996) and Lovell and Kumbhakar (2000).

Variables for environmental characteristics such as customer density, network length or area size, have been introduced in different cost model specifications for network industries. See, for example, Norsworthy et al. (1991) for the post offices and Roberts (1986), Filippini (1998) for the electricity distribution companies.

A complementary approach to address the problem of unobserved heterogeneity across production units is to use cost frontier models for panel data. Unfortunately, for this study only cross-sectional data are available. See Farsi and Filippini (2004) for an example of the efficiency measurement using panel data and Greene (2004) for a discussion on recent developed cost frontier models for panel data.

See Cornes (1992), p. 106.

In the first part of the analysis a homothetic version of the translog functional form was used. However, the econometric results were not satisfactory. All the coefficients of the second term variables were not statistically significant. For this reason it was decided to apply an F-test to test the hypotheses that all second order coefficients are equal to zero (in this case the translog collapses to the Cobb–Douglas functional form). The observed F-value was not significant at 10%. Finally, the authors are aware that using the Cobb–Douglas functional form they are not able to verify the existence of economies of scope. This important topic should be included in future research activities.

It is appreciated, that using this relatively small data set a small sample estimation bias could be encountered. For a discussion on this issue see Jensen (2003).

For the econometric estimation LIMDEP, version 8 was used.

Return to scale and economies of scale are equivalent measures if and only if the production function is homothetic (see Chambers, 1988, p. 72). Here, by using a Cobb–Douglas functional form this assumption is being imposed.

EFF is calculated as the ratio of actual costs to the efficient level of costs. The values of inefficiency scores can be interpreted as follows: an efficiency score of 1.12 means that the firm's cost is 12% higher than the cost of an equivalent firm that is efficient. The values of EFF are calculated using the procedure suggested by Jondrow et al. (1982).