Do competition and ownership matter? Evidence from local public transport in Europe

Abstract

This article investigates how the ownership and the selection procedure of firms operating in the Local Public Transport (LPT) sector affect their productivity. In order to compare different institutional regimes, we carry out a comparative analysis of 77 companies operating in large European cities over the period 1997 to 2006. This allows us to consider firms selected either through competitive tendering or negotiated procedures. Retrieving the residuals we obtain a measure of Total Factor Productivity (TFP), which we regress on firm and city characteristics. We find that totally or partially public firms display lower productivity than privately owned firms. Moreover, firms selected through competitive tendering display higher TFP.

Keywords:

• local public transport
• public ownership
• translog production function

JEL Classification::

• C33
• K23
• L25
• L33
• L91

Acknowledgements

We wish to thank seminar participants at 7th IIOC (Boston, USA), XI SIET (Trieste, Italy), 4th Kuhmo-Nectar Conference (Copenhagen, Denmark), 36th EARIE (Ljubljana, Slovenia) and 50th SIE (Rome, Italy) for their helpful comments on earlier drafts.

Notes

¹ The UK is the sole European country where ‘competition in the market’ has been experienced in urban transports. In Italy some competitive tendering (for the market) took place after 1998. However, large cities were not affected by the tendering process, but for one-fifth of the bus services in Rome, since 2001 (see Boitani and Cambini, 2006, and the references therein).

² Official data on population are sourced from Eurostat.

³ Unfortunately, the Amadeus database does not provide information on sales for firms located in the UK. This forces us to exclude this country's firms as sales are a necessary ingredient of our analysis (as will be explained below). However, the prevalence of competition ‘in the market’ in UK cities (except London) may well have introduced a strong country bias in the empirical analysis.

⁴ For a complete list of the cities included in the sample, as well as further descriptive statistics on the firms analysed, refer to Boitani et al. (2010). Table A1 in the Appendix summarizes the definitions and sources of the variables used in the empirical investigation.

⁵ To the best of our knowledge, this information does not exist. Despite their local monopoly position, firms are quite reluctant to collect these data in a coordinated and fully comparable way, and even more to communicate these data. Even the International Association of Public Transport (UITP), the international association of local public firms in the sector, does not have systematic data of this kind. As highlighted in the previous section, some databases with specific information on the type of contract and output measures are available. Nonetheless, these databases are specific to a single country, or even a single region.

⁶ As Bartelsman and Doms (2000) point out: ‘The choice of output is often dictated by the available data. Where possible, physical output with unchanging quality is the best measure. […] In general, researchers rely on deflating nominal variables at the sectoral level. […] Using deflated production to measure productivity has one drawback, which is the same whether applied at the micro level or at the sectoral or aggregate level: Any quality improvement in output that is not reflected in the deflator will result in a downward bias in productivity’. Notice, however, that the issue of quality measurement is also problematic when direct measures of output are available.

⁷ Although Law decree 422/1997 has introduced competitive tendering in Italy, the Italian firms included in the database were all operating under negotiated procedures during the time period considered in the analysis.

⁸ As will be evident in the next section, this specification is also supported by our data.

⁹ The error term in the production function can be decomposed into two terms: ϵ _it  = ω _it  + η _it , where ω _it represents unobservables that are unknown to the econometrician, but are observed (or predictable) by firms when choosing inputs, and η _it represents unobservables that are not observed by the firm before input decision. For example, ω _it could represent managerial ability, or expected down-time due to breakdowns, while η _it could represent deviations from expected breakdowns. Since a firm has knowledge of its ω _it when making input choices, these choices will be correlated with ω _it , thus incurring endogeneity. Fixed effects estimator assumes that unobserved productivity ω _it is constant over time, allowing to consistently estimate the production function. Given the short-time period considered, constancy of ω _it is not a strong assumption.

¹⁰ This amounts to defining TFP as the unexplained residual of the production function.

¹¹ Wang Chiang and Chen (2005) and Ottoz et al. (2009) introduce nonneutral technical change in a translog cost function on a sample of Taiwanese and Italian LPT companies, respectively.

¹² Take, for example, the role of selection mechanisms: no economic a priori suggests that the amount of output should be statistically different between firms selected by means of a public tendering or negotiated procedures. Indeed, the correlation between the output variable and the procedure variable is 0.06 and not statistically significant. However, economic theory suggests that firms selected through a competitive tendering should be more productive, and this is confirmed by a correlation of 0.11, significant at 5% level, between the output variable and the dummy for competitive tendering.

¹³ We classify as ‘metro companies’ those firms which offer underground transportation services. These firms may be offering exclusively underground transportation services, or both ground and underground transportation services.

¹⁴ Technical progress is labour augmenting in the full sample, but this result is not robust when considering the subsample of ground transportation companies only.

¹⁵ Evidence on large companies in LPT is scant, however the result of diseconomies of scale confirms previous findings in the literature (Bhattacharyya et al., 1995; Matas and Raymond, 1998; Jha and Singh, 2001). Constant and increasing returns to scale are often obtained on samples of small- and medium-sized companies. Cambini et al. (2007) provide a comprehensive review of previous empirical evidence on scale and density economies in LPT. Additionally, notice that the econometric literature acknowledges a downward bias in the estimates of input elasticities in a fixed effects framework in presence of measurement error (see Griliches and Hausman, 1986 for a discussion).

¹⁶ See Section ‘Robustness’ for a discussion on the role of ownership adopting alternative definitions.