Collusion in the Dutch waste collection market

Abstract

In this paper we analyse whether collusion exists in the Dutch waste collection market, which shows a high degree of concentration. Although scale effects might be in accordance with this market outcome, the question is whether concentration is in fact a result of fair competition. Using data for (nearly) all Dutch municipalities, we estimate whether collusion exists and what the impact is on tariffs for waste collection. The results indicate that high concentration increases prices and therefore (partly) offsets the advantage of contracting out. The presence of competing public firms might be essential to ensure more and fair competition.

Notes

1 Note that we use a broad interpretation of the term collusion. We do not distinguish between overt or tacit collusion. It is therefore possible that a collusion effect is found while none of the private firms made an explicit deal with one or more other private firms. In that case tacit collusion is present.

2 In Dijkgraaf et al. (2003), where data for 1998 are used, 42 per cent of all municipalities use private firms. Because some small villages have merged, and due to the fact that private collectors are especially active in small villages, the number in 2002 is comparable to the number in 1998.

3 The Herfindahl index measures the size of firms in relationship to the size of the relevant market and is, therefore, an indicator for concentration. We calculate the Herfindahl index as the sum of the squared market shares of private (and, if applicable, public) companies in The Netherlands or a Dutch province.

4 Due to (past) Dutch legislation there is some evidence that the relevant market is the province. The Netherlands had 12 provinces and 496 municipalities in 2002.

5 The C3 ratio, an alternative measure for concentration in the relevant market, is defined by the sum of the market shares of the three largest private (and, if applicable, public) companies in The Netherlands or a province.

6 Price variables for the different inputs are not included as no reason exists ex ante why factor prices would differ between municipalities. Wage bargaining takes place at a national level. Moreover, higher input prices due to inefficiency should show up in the coefficients for the institutional form dummies.

7 There are 11 plants in the Netherlands. Note that in 2002 in the Netherlands none of the municipalities used another form of treatment, like landfill.

8 As we have only provided observations on the costs for municipalities this assumption cannot be tested directly. Nevertheless, we can estimate the total effect as profits are included in the tariffs municipalities pay.

9 We multiply the Herfindahl indices and C3 ratios by the private ownership dummy as we are interested in the effects of concentrated markets on the behaviour of private firms.

10 As a starting point we include only municipalities within a distance of 15 miles. Substituting this distance for 25 miles does not influence the conclusions.

11 The definition of C_P and C_I could suggest that there is an autocorrelation problem as the number of inhabitants and households and density influence these two variables, while they are also included directly in the estimations. However, correlations between these variables are very low (absolute value is less than 0.16).

12 The AOO presents figures for actual tariffs and for the extent to which these tariffs cover total costs. If actual tariffs do not cover total costs, we use the coverage factors to calculate cost covering tariffs.

13 As total collection costs are about 40 per cent of TC (Dijkgraaf, 2004) and VAT is 19 per cent, we reduce TC by 7.6 per cent for O_I and O_P. If input VAT is taken into account the correction might be slightly (1 percentage point) lower, but this will not influence the conclusions. By correcting O_I for VAT, we assume that public companies are involved in a tender. However, this will not be true in all cases, and therefore we might underestimate the costs of public companies. Note that if we do not correct the costs of public companies for VAT, only the coefficient of O_I changes (and becomes insignificant). All other coefficients and their standard errors are exactly the same. Results are available upon request. (See also Wassenaar & Gradus, 2004.)

14 We have no concern about biases due to missing data as we have no indication that exclusion is non-random with respect to regions or institutional form. In fact, all average values for regional presence and institutional forms do not differ between samples and all municipalities (we have data for these variables for all municipalities).

15 Due to data availability the specification was somewhat different in Dijkgraaf and Gradus (2003), while we can use a more extensive dataset in this paper.

16 As the estimations are in logs the effect can be calculated using e^x−1. Note that this effect has to be multiplied by 2.5 as collection costs are on average 40 per cent of TC.

17 Note that this result depends on the correction we made for VAT for public companies. If in fact no VAT is paid, the coefficient for public companies is insignificant. The other results described in this section do not change if no correction for VAT is made. See note 13.

18 Estimations with the C4 ratio show that results are comparable. At the average value the cost advantage is now 19 per cent for private provision, while at monopoly no cost advantage exists.

19 Note that an alternative interpretation could be that private firms have to lower prices if public firms are present in their neighbourhood, if these public firms have lower tariffs, for example resulting from different firm goals or circumstances (such as cost minimalisation or lower labour costs). However, the first estimation shows that in general tariffs are comparable between public and private firms.

20 We present results for Moran's I including all neighbouring municipalities within a distance of 15 or 25 miles. Alternatives with smaller or larger distances lead to comparable results.