Reforming the local public sector: economics and politics in privatization of water and solid waste

Abstract

Several empirical studies have analyzed the factors that influence local privatization. We examine the influence of transaction costs and political factors on local governments’ choices through new variables. We consider two relevant services accounting for different amounts of transaction costs: water and solid waste delivery. Our results show that privatization is less common for water, which has higher transaction costs. Furthermore, we find that municipalities with a conservative ruling party privatize more often regardless of the ideological orientation of the constituency. Finally, we find that intermunicipal cooperation may be a suitable organizational form for some municipalities.

Keywords:

• privatization
• contracting‐out
• local governments

JEL codes:

• L33
• R51
• H72

Acknowledgement

Our research on public sector reform has received financial support from the Spanish Ministry of Education (SEJ2006‐04985). We have received useful comments in the 2007 conference ‘The role of the State in public service delivery’ (Lee Kuan Yew School of Public Policy, National University of Singapore, September 2007), and the 2007 Fall Conference of the Association of Policy Analysis and Management (APAM, Washington DC, November 2007). We thank Trevor Brown, Harvey Feigenbaum, Leonard Gilroy, Raymond Gradus, Steve Kelman, R.K. Mishra, Steve Savas, David van Slyke, Mildred Warner, Wu Xun, and Anna Na Yi for useful suggestions.

Notes

1. Joint effects of political interests and ideological attitudes have been considered in Bel and Miralles (2006) when analyzing the economics and politics of local public services financing.

2. It is worth noting that whereas intermunicipal cooperation in the US is not compatible with privatization (Warner and Hefetz 2002a, b; Levin and Tadelis 2007), intermunicipal cooperation in Spain – as well as in most European countries – is compatible with private production. Hence, decisions on engaging in cooperation and/or privatizing are not taken simultaneously. Section 3 deals in detail with this issue.

3. Beyond the decision to privatize local services, there is an interesting recent literature that analyzes the politics of the formation of public or private governments (Carr and Feiock 2004). In this view, creating private governments is a solution to collective action problems that emerges because of a political process of crafting a relational contract among members of a community (Baer and Feiock 2005).

4. This double dimension of the politician has been named the citizen‐candidate approach. Osborne and Slivinski (1996) and Besley and Coate (1997) offer theoretical insights; Levitt (1996) and Lee et al. (2004) offer empirical support.

5. Another variable used in several studies is the percentage of public employees over population. On this we must recall that the determination of service delivery choices and the percentage of public employees is done simultaneously. Hence, the variable for the weight of public employees is statistically biased. Indeed, a more intense use of external suppliers implies per se a reduction in the number of public employees. Because of this, we do not consider here results obtained from using the variable percentage of public employees. A complete discussion of this methodological issue can be found in Bel and Fageda (2007).

6. Still for the US, following the proposition that politicians obtain the support of public employees under public production, Lopez‐de‐Silanes et al. (1997) test the hypothesis that political patronage affects service delivery choices. They find that state laws imposing accountability requirements in contracting for personnel encourage privatization. However, later studies by Kodryzcki (1998) and Walls et al. (2005) do not confirm these results.

7. Miralles (2006) obtains mixed results. Christoffersen and Paldam (2003) find no relation between ideology and privatization in their univariate study for Denmark.

8. Bel (2006) contains more detailed information on the sample. The rate of response for the survey was 24.9% for municipalities of over 2000 inhabitants, and 100% for municipalities of over 30,000 inhabitants.

9. Note that including this variable does not pose a problem of multicollinearity, which would only be an issue if the sample size were within a very small population range. Other works (e.g. Warner 2006) find an inverse U‐shape curve for local privatization too. However, in Warner’s case, the curve is related to the geographical continuous metro‐suburbs‐rural, whereas our curve relates to the city/town population.

10. Levin and Tadelis (2007) refer to exceptional cases of this type of city partnership or public agencies partnership in California, and they consider this a special form of public sector contracting. Instead, they are much more frequent in continental Europe, and especially in France, the Netherlands and Spain (Kelly 2007).

11. In this way, Brown and Potoski (2005) measure asset specificity and ease of measurement for 64 local services in the US. They build indicators ranging from 1 (low specificity, or easy measurement) and 5 (high specificity, and difficult measurement). They find asset specificity of 3.94 and ease of measurement 2.44 for water distribution; assets specificity is 3.00 and ease of measurement 2.06 for residential solid waste. For commercial waste, ratings are 3.06 and 1.97. Levin and Tadelis (2007) build indicators on contract difficulty, as perceived by city managers, and find that contract difficulty is over the average for water services, whereas for all services related to waste, contract difficulty is below the average. Bel (2006) provides evidence that contract terms are longer in water than in solid waste, thus reflecting higher relevance of sunk costs in water.

12. Local elections in Spain are based on party lists. In this way, independent parties are those parties that are strictly local. The members of the city council are elected on a proportional basis (corrected with the d’Hondt system). The election of the mayor is not direct: after being elected, the city council members elect the mayor in the first meeting of the council. An absolute majority (50% + 1) of votes from the city council members is required to be elected as mayor in the first round. If no candidate obtains such a majority, then the top member of the list that obtained the largest fraction of votes in the election is automatically elected as mayor.

13. Otherwise, the correlation between the variables industrial_interests and major is low.

14. We exclude from the estimation those observations in which a mixed firm is in charge of delivering the service. Observations excluded are 76, which represents 7% of the sample. Mixed firms are firms whose stakeholders are both the corresponding governments and private investors. Such mixed public/private firms are usual in many European continental countries, but they are a very rare organizational form in countries like the US (Warner and Bel 2008). Whenever a mixed firm is involved, we are not able to make clear standard inferences about the ownership status of the producer. In this way, we do not know with precision the structure of the firm ownership. Including mixed firms in the estimation would distort the analysis of the choice between public and private form of production since mixed firms are a hybrid organizational form.

15. After pooling the data for our two local services, the number of observations excluded in the estimations with the variable ‘mayor’ has been 95. We had initially 111 observations for cities with ‘independent mayor’. However, 16 of them had already been excluded because of mixed firms or unavailability of fiscal data.

16. One could argue that the variable for intermunicipal cooperation implies an endogeneity bias as the decision to cooperate could be made simultaneously with the decision of contracting out. However, as we mention above, intermunicipal cooperation as it is understood in Europe is a decision that can be undertaken under any production form setting. Thus, the possible endogeneity bias is not an issue in our context.

17. In the empirical analysis, we exclude observations in municipalities that first externalized the delivery of the local services considered before local democracy was restored in Spain in 1979. Observations excluded are 78, which represents 7.2% of the sample. Inclusion of service delivery choices made in the authoritarian political context of Franco’s dictatorship might distort the analysis of explanatory factors, particularly concerning the political factors, since these decisions were not taken within a democratic decision context. Many of these concessions, particularly on water, were still in place in the early 2000s. Considering in the estimation cities that first externalized the service before democracy was restored yields similar results to those obtained in our analysis. Political variables show less robust results, even if they keep significance.

18. We exclude from the estimation the following dummy variable that combines political and ideological factors: P‐P: Progressive major – Progressive constituency. We exclude this variable from the estimation to avoid perfect multicollineality with the other dummy variables for political and ideological factors. This variable is set as the reference variable for these other dummy variables.