11. Waste management in the Netherlands Elbert Dijkgraaf and Raymond Gradus

Transcription

1 11. Waste management in the Netherlands Elbert Dijkgraaf and Raymond Gradus INTRODUCTION The Netherlands recycles 32 percent and composts 27 percent of its municipal waste and most of the remainder is incinerated to generate electricity. Lack of space and a growing environmental awareness forced Dutch governments to take measures early on in the 1980s to reduce the landfilling of unsorted waste and to stimulate recycling. Later on, these measures were intensified (see Dijkgraaf 2004). Compared with EU countries with an average of 40 percent of municipal waste recycled and composted, the Dutch percentage is high (see Eurostat 2010). Dutch municipalities are responsible for waste collection and separation (Wet Milieubeheer, art ). By law, Dutch municipalities are obliged to collect two types of waste at the curbside: compostable waste such as vegetable, food and garden waste and unsorted waste. The frequency is such that every municipality collects unsorted and compostable waste in general every week or every two weeks. Since January 1994, compostable waste has been collected at the curbside; this was an important measure in increasing the amount of compostable waste. Municipalities are also obliged to collect separately paper, glass, textiles, plastic packaging, small chemical waste and bulky household waste (such as waste resulting from renovation of houses and gardens). Nevertheless, municipalities may choose how these materials are collected. Therefore, often they are not collected at the curbside, but citizens can deliver them to collection points at central locations. The way citizens pay for waste collection differs by municipality. Most Dutch municipalities still use a fixed fee per year. However, in several municipalities, the introduction of unit- based pricing of (unsorted) waste has been an important measure in increasing the collection of compostable waste and recyclables and in reducing the amount of environmentally unfriendly unsorted waste as well. The main reason is that it introduces a marginal price per unit of waste collected. While marginal costs are zero for citizens with a yearly flat fee, with unit- based pricing citizens have an incentive to reduce waste as there is now a marginal price. As prices are generally much higher for unsorted waste, citizens with unit- based pricing also have an incentive to invest more in sorting waste. In the last ten years, more and 287 KINNAMAN PRINT pt2-3.indd /02/ :55

2 288 Handbook on waste management more Dutch municipalities have implemented unit- based user fees. By 2010, 36 percent of all Dutch municipalities had implemented such a system, this number having risen substantially from 15 percent in Unit- based pricing systems differ in the Netherlands with respect to the basis of pricing. In general, there are systems based on weight, bags, frequency and volume. The weight system is the most refined as each kilogram of waste results in a higher bill. As bags in the Netherlands are much smaller than bins, the bag system is more refined than the frequency system. With the frequency system the bill depends on the number of times the bin is presented at the curbside. Finally, the volume system is the least refined as it allows choice between a small or a large bin, which is the only thing that affects marginal prices. Over time, it seems that the number of municipalities with the more refined weight- and bag- based systems has become stable and the number of municipalities with a frequency system has increased. Based on municipal data, Dijkgraaf and Gradus (2004) find sizeable and significant effects from the different unit- based pricing systems. Based on data until and corrected for environmental activism or municipal fixed effects, the effects of unit- based pricing systems still remain large (see Dijkgraaf and Gradus 2009, Allers and Hoeben 2010). In this contribution, we estimate the effect of different unit- based systems on the quantity of different waste streams. Our main result in comparison with the literature is the stability of effects with respect to timing and to differences between municipalities. Based on a pooled cross- section of municipal data for the Netherlands for , we can correct for municipal fixed effects and test for time- varying unit- based dummies as we have a large data set. It will be shown that there is some deterioration over time in the effectiveness of the weight- based system in reducing the amount of unsorted waste. Time effects might be due to learning or awareness erosion or to an environmental activism effect (see also Dijkgraaf and Gradus 2009). It could be expected that when citizens have more experience with better sorting and environmentally friendly buying behavior, the effectiveness of unit- based pricing might increase over time. On the other hand, it is possible that the introductory incentives erode over time as people get used to paying for waste collection. This is the case if citizens behavior depends not only on prices but also on price changes. Differences between municipalities might also be important. It could be the case that first- movers are more environmentally orientated than municipalities that introduce unit- based pricing in later years. In other words, for municipalities that introduce unit- based pricing later, the effect that can be internalized is larger. Dijkgraaf and Gradus (2009) show that the environmental activism effect indeed erodes over time. KINNAMAN PRINT pt2-3.indd /02/ :55

3 Waste management in the Netherlands 289 In many Dutch municipalities separation of waste is an important topic and policies to separate waste have been developed. There has been an increase in facilities at refuse center for collecting different waste streams such as drinks cartons, flat glass and construction and demolition waste. In this contribution, we describe these policies and try to investigate the relation between the effectiveness of different unit- based pricing systems and the number of separately collected waste streams. In many papers, cost functions for waste collection are estimated (for a recent overview, see Simões and Marques 2012). Most attention has been paid to the influence of the institutional form of waste collection on costs. For the Netherlands, cost functions have been evaluated by Dijkgraaf and Gradus (2003, 2007, 2008 and 2013) and Felsö and de Groot (2011). In this contribution, we extend this literature in two directions by including the unit- based pricing system and the number of separately collected waste streams. It will be shown that these variables are more important from a cost- minimizing point of view than the institutional mode of waste collection. Based on our estimation, it can be concluded that increasing the number of separately collected waste streams will lead to higher costs. However, future research should focus on environmental aspects as well. This contribution is organized as follows. Section 1 discusses the data and method used. Section 2 presents the estimation results for the effect of unit- based pricing systems on quantity of waste. Section 3 discusses the municipal policies on recycling and separating different waste streams. Section 4 estimates a cost function and Section 5 contains some conclusions. 1. DATA AND METHOD In this section, we discuss the data and method used for estimating the effects of unit- based pricing on the quantity of waste, as presented in Section 2 (when we extend this analysis in other sections, the additional data and methods are discussed there). The methodology in Section 2 is characterized by estimating an equation with, on the left- hand side, the quantity of waste collected and, on the right- hand side, dummies for municipalities with a unit- based pricing system and variables correcting for socioeconomic differences between municipalities (see, for example, Kinnaman and Fullerton 1997). 1 As municipal data are available for , we estimate a pooled model using both the cross- section and time- related variation. For each residential waste stream, we estimate Waste w,i,t 5 a s UBP s,i,t 1 gse 1 c 0 1 c i 1 d t 1 e i,t (1) KINNAMAN PRINT pt2-3.indd /02/ :55

4 290 Handbook on waste management where Waste w,i,t is the quantity of waste stream w in municipality i in year t, UBP s,i,t are dummies with the value 1 if municipality i has a unit- based pricing system of type s in year t, SE is a vector of socioeconomic characteristics, c 0 is the general constant, c i are time- invariant municipal fixed effects and d t are year fixed effects. 2 We use the quantity of residential waste collected (Waste w,i,t ) at the municipal level (in kilograms) as the dependent variable. Similar to Dijkgraaf and Gradus (2004, 2009), we distinguish between different waste streams as we have data for compostable waste (such as vegetable, food and garden waste), recyclable waste (such as glass, paper and textiles) and unsorted waste. In previous studies, we used data from 1998 until 2005; for this study, we extend the period to Furthermore, we have added data for other waste streams as well, which are available from 2001 until We include the sum of the other waste streams as a fourth dependent variable. We discuss these streams in more detail in Section 3. Data on the dependent variables the quantities collected of unsorted, compostable, paper/glass/textiles and other waste in kilograms per inhabitant come from studies by the Dutch Waste Management Council (AOO). The AOO uses an annual inquiry from the CBS (the Dutch Central Bureau for Statistics), which is sent to the waste collection units of all Dutch municipalities. These units have reliable figures for the quantity of waste collected, as the bill they have to pay is based on the quantity of waste supplied to waste treatment firms. The response rate of the inquiry is 94 percent on average and is above 90 percent in each year. Thus, our data set comprises nearly all Dutch municipalities. The actual number of municipalities included differs for each dependent variable due to data availability. Table 11.1a presents a summary and availability statistics for dependent and some other variables. Dutch municipalities are free to choose the financing mechanism for collection of unsorted and compostable waste. Most municipalities charge a flat rate for waste collection. However, in order to promote waste prevention and recycling, an increasing number of municipalities have introduced a unit- based pricing system. Starting from 15 percent in 1998, the percentage of Dutch municipalities using such a system was 36 percent in Dutch municipalities have introduced different types of unit- based pricing (UBP) systems. Four different UBP systems can be distinguished: volume-, frequency-, bag- and weight- based. Table 11.1b gives an overview of the pricing systems used in the period The percentage of municipalities using a flat rate decreases from 85 percent in 1998 to 64 percent in The volume- based program allows households to choose between different volumes of collection bin; 4 27 municipalities in the Netherlands used KINNAMAN PRINT pt2-3.indd /02/ :55

5 Table 11.1a Descriptive statistics 1 Waste management in the Netherlands 291 Mean Max. Min. Std dev. Obs. 2 Cross-sections Total waste Unsorted waste Compostable waste Paper, glass, textiles Other waste Households Family size Population density Volume Frequency Bag Weight Notes: 1. Waste is measured in kilograms per inhabitant. Family size is the number of inhabitants per household. Population density is the number of inhabitants per square kilometer. The mean figures for volume, frequency, bag and weight are the proportions of municipalities that have the respective systems. 2. As data are not available for all years for all municipalities, the number of observations is not exactly equal to the number of years multiplied by the number of cross- sections. Table 11.1b Use of UBP by year: proportion of municipalities Volume Frequency Bag Weight Flat rate KINNAMAN PRINT pt2-3.indd /02/ :55

6 292 Handbook on waste management this rather crude UBP system in 1998 and 39 used such a system in A more refined marginal price results from a frequency- based system, in which the household pays for the number of times the bin is presented at the curbside. 6,7 This type of system was used by only 21 municipalities in 1998, but 73 municipalities had such a system in In the bag- based system, households buy a special bag with specific marks. 8 This is a more refined pricing system than the frequency- based one, as the volume of the bags is significantly less than that of the bins. 9 Importantly, the bag system allows households to change volume each week. Although 22 Dutch municipalities used a bag- based system in 1998, this number decreased over the period The relatively low number of such schemes is due to Dutch legislation limiting the number of bags carried per waste collection employee and the incentive for households to put as much waste as possible in each bag, which makes them difficult to handle (see also Fullerton and Kinnaman 1996). The same issues also make it hard to have a bag for compostable waste and therefore, contrary to the other systems, the bag- based system is generally used for unsorted waste only. 10 Maximum price incentives result from a weight- based system, pricing the waste per kilogram. The collection vehicle weighs the bin before emptying and combines this information with the identity of the owner, stored in a chip integrated in the collection bin. 11 A disadvantage of this system is the high administrative cost; therefore only 12 Dutch municipalities had such a system in 1998, while 26 had one in Table 11.1c shows the proportion of unit- based pricing systems weighted by the number of inhabitants for 1998 to In the last column, the weighted proportion of municipalities using a flat rate is given; it decreases Table 11.1c Use of UBP by year: weighted by inhabitants Volume Frequency Bag Weight Flat rate KINNAMAN PRINT pt2-3.indd /02/ :55

7 Waste management in the Netherlands 293 % of municipalities Volume Frequency Bag Weight Years Figure 11.1 Use of UBP by year: number of municipalities from 90 percent in 1998 to 73 percent in It is clear that unit- based pricing systems are used more in small municipalities and flat rates are applied more in large cities. The shares based on inhabitants are much smaller for weight- based systems as this type of system seems especially applicable in municipalities or cities with few blocks of flats. The development of different unit- based pricing systems over time is also illustrated by Figure 11.1, which clearly shows that the use of frequency- based pricing systems has especially increased, more than quadrupling between 1998 and Prevalence of the volume system increased more steadily. However, the use of the more refined weight- and bag- based systems has stabilized after To correct for differences between municipalities, we include the following socioeconomic characteristics in the estimation: the number of households, the number of inhabitants of a municipality per area (population density) and the average family size. Data for the socioeconomic characteristics come from the CBS. See Table 11.1a for descriptive statistics of these variables. 2. RESULTS Table 11.2a presents the estimation results. 12 It only shows the estimations for the dummy variables for different unit- based pricing systems and the socioeconomic variables. Results for municipal and time fixed effects are available upon request. KINNAMAN PRINT pt2-3.indd /02/ :55

8 294 Handbook on waste management Table 11.2a Effectiveness of different UBP systems over time 1 Unsorted Compostable Paper, glass and textiles Other Households 0.98*** 1.01*** 0.98*** 1.13*** Family size Population density *** Constant 5.58*** 4.30*** 4.67*** 6.25*** Volume ( ) *** ( ) *** ( ) * *** *** ** ** ** ** ** *** ** *** ** 0.06 Frequency *** 0.16** 0.01 ( ) *** 0.29*** 0.04* ( ) *** 0.33*** 0.07*** ( ) *** 0.35*** 0.07*** *** 0.37*** 0.07*** *** 0.43*** 0.08*** *** 0.48*** 0.07*** *** 0.51*** 0.06*** *** 0.54*** 0.06*** *** 0.56*** 0.08*** *** 0.52*** 0.07*** *** 0.49*** 0.08*** *** 0.49*** 0.10*** 0.01 Bag *** *** ( ) *** ** ( ) *** ** ( ) *** ** *** * *** *** ** *** KINNAMAN PRINT pt2-3.indd /02/ :55

9 Waste management in the Netherlands 295 Table 11.2a (continued) Unsorted Compostable Paper, glass and textiles Other Bag *** *** *** ** *** * *** * 0.20 Weight *** 0.82*** 0.11** ( ) *** 0.68*** 0.06 ( ) *** 0.68*** 0.07** ( ) *** 0.64*** 0.10*** 0.17* *** 0.66*** 0.10*** *** 0.71*** 0.09*** *** 0.74*** 0.11*** *** 0.81*** 0.08*** *** 0.81*** 0.06** *** 0.80*** 0.08*** *** 0.73*** 0.09*** *** 0.86*** 0.10*** *** 0.75*** 0.10*** 0.07 R 2 (adjusted) F-statistic Note: Coefficients with */**/*** are significant at 10%/5%/1% level. Pricing waste on the basis of weight has a highly negative and significant effect on unsorted waste of 45 percent in 1998 and 34 percent in ,14 For unsorted waste, we find that the effectiveness of unit- based weight pricing is deteriorating over time, especially in the period (see Figure 11.2a). For the bag- based system, we find that the effect of pricing waste is increasing over time since The effect of pricing waste on the basis of a bag increases from 30 percent in 1998 to 41 percent in In terestingly, for the frequency- based system, we find that the effect on unsorted waste of pricing waste more than tripled from 6 percent in 1998 to 22 percent in However, especially in 1999, there was a jump in effectiveness. The effect of introducing a system based only on the volume of the collection is small, at between 3 percent and 6 percent. 16 This result is not surprising since the volume- based system is less refined than the other systems. KINNAMAN PRINT pt2-3.indd /02/ :55

10 296 Handbook on waste management % of unsorted waste Volume Frequency Bag Weight Years Figure 11.2a Effectiveness of UBP systems over time: unsorted waste % of compostable waste Volume Frequency Bag Weight Years Figure 11.2b Effectiveness of UBP systems over time: compostable waste For compostable waste, we also find significant results for the weightbased system: it will reduce compostable waste by 56 percent in 1998 and 53 percent in So, the effect of weight pricing on composting has been quite stable over time (see Figure 11.2b). The system stimulates recycling KINNAMAN PRINT pt2-3.indd /02/ :55

11 Waste management in the Netherlands 297 of biodegradable waste so that, for example, garden waste is left in the garden and garden and kitchen refuse, such as vegetable and fruit peelings, is home composted. For the bag- based system, the effect on composting is insignificant. The reason for this is that most municipalities with a bagbased system price unsorted waste only and not compostable waste. 17 The effect of pricing compostable waste on the basis of frequency increases from 16 percent in 1998 to 38 percent in This may be because an alternative for curbside collection (home waste composting) is easily available. The effect of the volume- based pricing system on compostable waste is insignificant. For recyclable waste such as paper, glass and textiles, pricing waste on the basis of weight has a positive effect. This effect varies from 5 percent to 11 percent with peaks at the beginning and the end of the estimation period (see also Figure 11.2c). We find that the effect on recycling of pricing bags varies from 5 percent to 14 percent. The effect of pricing waste on the basis of frequency increases from small (or even negative) at the end of the previous century to 10 percent in The effect on recycling of introducing a system based only on the volume of the collection is small and only significant in 2004, 2009 and For other waste streams such as bulky waste, construction and demolition waste and gravel there seems to be no relation with the unit- based pricing system. Only for pricing waste on the basis of weight is there % of compostable waste Volume Frequency Bag Weight Years Figure 11.2c Effectiveness of UBP systems over time: recyclable waste (paper, glass and textiles) KINNAMAN PRINT pt2-3.indd /02/ :55

12 298 Handbook on waste management a significant (only at the 10 percent level) and negative effect, in However, this effect is not present in other years and the conclusion is that the size of these streams depends on other policy variables. This is an important finding as it was suggested that the positive effects of unit- based pricing in the literature might be explained by better sorting of the waste into streams not included in the analyses. 18 Turning to the socioeconomic characteristics in Table 11.2a, we find constant returns to scale for unsorted and compostable waste and paper, glass and textiles, so an increase in the number of households results in the same proportional increase in waste. For other waste streams, we find increasing returns to scale. A possible explanation for this is that large cities have larger refuse centers with more separate waste streams. Furthermore, family size plays no role in the amount of unsorted or compostable waste or paper, glass and textiles. In addition, municipalities with higher population densities have lower amounts of other waste. In Table 11.2a, it was shown that unit- based pricing has a significant effect on the amount of collected waste and this was investigated using a time- varying dummy. Based on our data set, it is also possible to investigate whether there is a difference in effect related to the introduction date and whether time effects are influenced by learning effects, awareness erosion or environmental activism effects. It might be that systems that are currently being introduced are more advanced and more effective than earlier systems. Furthermore, households might need time to change their behavior in an optimal direction (the learning effect). This means that it is possible that effectiveness increases when the system is used longer. However, the opposite might occur. The awareness effect is eroded over time if households get used to waste being priced and go back to their old behavior. This might be the case when behavior depends not only on price levels but also on price changes. 19 Another mechanism is environmental activism (see also Dijkgraaf and Gradus 2009). Differences between municipalities can be important. It could be the case that first- movers are more environmentally friendly than municipalities that introduce a unit- based pricing system in later years. This means that the effect for firstmovers might be smaller as the behavior of citizens before the introduction of unit- based pricing is already more environmentally friendly. In other words, for municipalities that introduce unit- based pricing later, the effect that can be internalized is larger. These mechanisms together will shape the time pattern of the effect of unit- based pricing. From a theoretical point of view, the pattern is undetermined. Therefore, we now include information on the period that the unit- based pricing system is introduced and an age variable, which gives information on years of experience after the introduction of the UBP KINNAMAN PRINT pt2-3.indd /02/ :55

13 Waste management in the Netherlands 299 Table 11.2b Descriptive statistics for added variables Mean Max. Min. Std dev. Obs. 1 Crosssections Volume UBPT UBPT Total Frequency UBPT UBPT Total Bag UBPT UBPT Total Weight UBPT UBPT Total Volume AGE Frequency AGE Bag AGE Weight AGE Note: 1. As data are not available for all years for all municipalities, the number of observations is not exactly equal to the number of years multiplied by the number of cross-sections. system. These variables test whether systems differ in effectiveness if they are introduced in different periods and the total effect of time- related developments. For each waste stream (unsorted waste, compostable waste, paper/glass/textiles and other waste streams), we estimate Waste w,i,t 5 a s UBP s,i,t 1 b s UBPT97 s,i 1 d s UBPT05 10 s,i 1 m s AGE s,i,t 1 f s AGE 2 s,i,t 1 gse 1 c 0 1 c i 1 d t 1 e i,t (2) where the UBPT97 dummy has the value 1 if the municipality already has a unit- based pricing system in 1997 and the UBPT05 10 dummy has the value 1 if the UBP system was introduced between 2005 and Table 11.2b gives descriptive statistics for UBPT97 and UBPT05 10 and also for the variable AGE, which measures the number of years the system is used. Coefficient a S gives the time- invariant effect, b S gives the effect of early introduction, d S gives the effect of late introduction and m S gives the effect of using the system longer. As there is a large correlation between the variable AGE and especially UBPT97, 21 we split equation (2) into KINNAMAN PRINT pt2-3.indd /02/ :55

14 300 Handbook on waste management Table 11.2c Estimation effects with age variables 1 Unsorted Compostable Paper, glass and textiles Volume UBP 0.057*** ** AGE * 0.012* AGE * Frequency UBP 0.194*** 0.285*** 0.060*** AGE *** 0 AGE *** 0 Bag UBP 0.354*** ** AGE 0.016** 0.049** 0 AGE *** 0.002* 0 Weight UBP 0.519*** 0.700*** 0.077*** AGE 0.018*** 0 0 AGE R 2 (adjusted) F-statistic Note: 1. All coefficients tested at the 10% level. Insignificant coefficients are presented as zero. Effects for other waste streams are not shown as they are all insignificant. Coefficients with */**/*** are significant at the 10%/5%/1% level. two separate estimations (one including AGE and AGE 2 and the other including the time dummies). Tables 11.2c and 11.2d present the estimation results. For AGE and the time dummies, we only give the significant results. 22 For unsorted waste, we find a significant time- invariant effect for all four systems in both estimations. The effect of introducing a system based only on the volume of the collection is small, at 5 percent to 6 percent. For the frequency- based system, we find that the effect on unsorted waste of pricing waste is 17 percent to 18 percent. The effect is 30 percent to 33 percent for the bag- based system and 33 percent to 40 percent for the most advanced system, the unit- based weight system. For unsorted waste, we find age and time effects. In Table 11.2c, the age effect for weight is significant and positive, leading to less effectiveness when the weight system is used longer. For the bag system, we also have a significant positive age effect. However, the effectiveness reaches a maximum after four years and is declining from then on to zero after eight years as the coefficient on age squared is negative. In addition, we find a significant negative effect for municipalities that introduce the weight system between 2005 and 2010 (see Table 11.2d). This can be explained if KINNAMAN PRINT pt2-3.indd /02/ :55

15 Table 11.2d Estimation effects with period dummies Waste management in the Netherlands 301 Unsorted Compostable Paper, glass and textiles Volume UBP 0.049*** 0 0 UBPT ** 0 UBPT *** Frequency UBP 0.204*** 0.530*** 0.049*** UBPT * UBPT *** 0.065*** Bag UBP 0.402*** ** UBPT * 0 0 UBPT *** 0 0 Weight UBP 0.394*** 0.665*** 0.061*** UBPT UBPT *** 0.438*** 0.124*** R 2 (adjusted) F-statistic Notes: All coefficients tested at the 10% level. Insignificant coefficients are presented as zero. Effects for other waste streams are not shown as they are all insignificant. Coefficients with */**/*** are significant at the 10%/5%/1% level. less environmentally active municipalities introduced such a system later on. Interestingly, we now have a significant positive effect for municipalities that introduce the bag system before 1998 and between 2005 and Thus, the effectiveness of the bag- based system is increasing in this period. So, there are awareness and learning effects. A possible interpretation for the learning effect is that there is an innovation or that the issue of illegal use of bags is solved (see also endnote 15). For compostable waste, we find significant effects for the age variables for the bag and frequency systems (see Table 11.2c). We now have a significant negative age and positive age- squared effect for the frequency- based system, so the amount of compostable waste increases after the introduction of this system. A possible interpretation is that people are learning to avoid the fee on unsorted waste by increasing their amount of compostable waste. In Table 11.2d, we also find a significant negative effect for municipalities that introduce the weight system between 2005 and This can be explained if less environmentally active municipalities introduced such a system later on. For the frequency system, we find a positive significant effect for this dummy. A possible interpretation is that there is an innovation in the frequency system, although this should be a topic of future research. KINNAMAN PRINT pt2-3.indd /02/ :55

16 302 Handbook on waste management For recyclable waste, we do not find significant (at the 5 percent level) effects for the age variable. For the volume, frequency and weight systems, we find significant positive results for municipalities that introduce the system between 2005 and 2010, implying that these systems are introduced later in municipalities that employ less separation of waste. Let us compare our results with previous estimates of Dutch unitbased pricing systems based on cross- sectional analysis of municipalities. In Dijkgraaf and Gradus (2004), we find sizeable and significant effects of the different unit- based pricing systems. For unsorted waste, the weight and bag systems had the largest effects, reducing quantities by 50 percent and 52 percent respectively, followed by the frequency system (27 percent) and the volume system (12 percent). In Dijkgraaf and Gradus (2004, 2009), the effects of different unit- based pricing systems are smaller if corrected for environmental activism. 23 According to our 2004 estimates, the municipalities with a UBP system had a 7 percent lower level of waste before the introduction of unit- based pricing. In Dijkgraaf and Gradus (2009), we show that this environmental activism effect is decreasing over time as municipalities that introduce unit- based pricing later on are probably less environmentally friendly. In Allers and Hoeben (2010), a fixed effect at the municipal level is included and, for unsorted waste, the weight system has the largest effect, reducing quantities by 39 percent, followed by the bag system (28 percent) and the frequency system (21 percent). 24 Comparing this with our results from Tables 11.2c and 11.2d, it is in line with the estimation results of our time- invariant dummy (weight- based system percent, bag- based system percent and frequencybased system percent). For compostable waste, our time- invariant effects are in line with Allers and Hoeben (2010) as well. They find a 51 percent reduction for the weight system and a 43 percent reduction for the frequency system, while we find a percent reduction for the weight system and a percent reduction for the frequency system. It is well known that a unit- based pricing system has a positive effect on recycling of paper, glass and textiles. However, these effects differ between the mentioned studies. In Dijkgraaf and Gradus (2004), it is shown that this effect is between 10 percent and 30 percent (the effect is insignificant for the volume system). Allers and Hoeben (2010) suggest that 18 percent of the reduction in unsorted waste is due to better recycling of paper, glass and textile (PGT). Calculating the PGT/Unsorted- share from Tables 11.2c and 11.2d gives percent (frequency), percent (bag) and percent (weight). In the next section, we show that other waste streams can be important for recycling, although we did not find a significant effect for these waste streams in Table 11.2a. In conclusion: it is important to correct for fixed effects, otherwise the KINNAMAN PRINT pt2-3.indd /02/ :55

17 Waste management in the Netherlands 303 reduction effect of a unit- based pricing system may be overestimated. Moreover, the overall conclusion is that especially the weight and frequency systems generate sizeable reductions in unsorted and compostable waste. The bag system may still have its advantages, given the large reduction in unsorted waste and the fact that municipalities can save a lot of money with this system as its administrative costs are the lowest (see also Section 4). Based on a model incorporating ageing, we can conclude that the reduction effects are quite stable. The estimates show some awareness erosion, learning and introduction (environmental activism) effects. However, these are small and not consistent between different waste streams. Learning effects are only present for compostable waste. The introduction of unit- based pricing systems may, however, have adverse effects. Citizens may take their waste to neighboring municipalities or may dump it illegally. Dijkgraaf and Gradus (2004) and Allers and Hoeben (2010) show that there is no evidence that surrounding municipalities without unit- based pricing systems in fact collect part of the waste produced in municipalities with unit- based pricing systems. Allers and Hoeben (2010) analyze whether illegal dumping may account for part of the reduction in refuse due to pricing of waste. In their opinion, it is not a serious problem in the Netherlands, as one would expect that many municipalities would have abolished user fees if this were the case. This has not happened: only five municipalities have stopped applying user charges for unsorted waste; of these, one reintroduced them after two years, and the others dropped them when they merged with municipalities without user charges MUNICIPAL POLICIES TO SEPARATE WASTE Separation of waste has become an important issue in the Netherlands, as elsewhere. 26 Every Dutch municipality has been obliged to supply a free collection system for paper, glass and textiles since the beginning of the 1990s and for plastic packaging since 1 January In addition, Dutch municipalities can supply a free curbside collection system for paper, glass and textiles. In some municipalities, there is a free curbside collection program for recyclable paper organized by local associations, such as sports clubs and schools. 27 For municipalities without curbside collection of paper, glass and textiles, the number and location of drop- off centers must be such that the collection infrastructure is easily accessible for all citizens. 28 For example, municipalities place collection units at shopping centers and at entrance roads of neighborhoods. For waste such as bulky household waste, most Dutch municipalities have one or more refuse centers to which inhabitants can take their waste. Some municipalities have KINNAMAN PRINT pt2-3.indd /02/ :55

18 304 Handbook on waste management a facility to collect, on request, items such as bulky waste at the curbside with or without an extra payment. In addition, in recent years, there has been an increase in facilities at refuse centers for collecting different waste streams such as construction and demolition waste and gravel. Nowadays, almost all Dutch municipalities have facilities for bulky household waste. These centers also have facilities for collecting different waste streams such as chemical waste (batteries, medicine, solvents, paint, etc.), construction and demolition waste, gravel and flat glass. 29 The provision of these facilities has increased in recent years. Batteries can be handed in at many shops and supermarkets as well. Table 11.3a is based on CBS data and shows, for 24 different waste streams, the percentage of Dutch municipalities that collect that waste stream separately. Table 11.3a Percentage municipalities that collect waste streams separately Unsorted waste Vegetable, fruit and garden waste Paper Glass Textile Bulky waste Construction and demolition waste Chemical waste (KCA) Metal packaging Drink- cartons Plastic Kitchen appliances Bulky garden waste Useful house council Floor covering Flat glass Metals Wood Debris Tar roofing Waste with asbestos Car tires Garden ground Other waste Note: for convenience reasons we only present data for 2001, 2004, 2007 and KINNAMAN PRINT pt2-3.indd /02/ :55

19 Waste management in the Netherlands 305 Number of separated waste streams Years Figure 11.3a Average number of separately collected waste streams per municipality Separation of the first five streams listed in the table is obligatory by law, so only a very few municipalities do not collect these streams separately. For bulky household waste, 97 percent of the Dutch municipalities now have such a facility. The percentage of municipalities that separate household articles has increased from 9 percent in 2001 to 86 percent in For construction and demolition waste, chemical waste and flat glass, the number of facilities has also increased since the beginning of this century. In some municipalities, as many as 23 different waste streams are collected. The average number of separate waste streams has increased over time from 14 in 2001 to 16.5 in 2010 (see Figure 11.3a), with an average of 15.5 over the period. Municipalities with a unit- based pricing system have 16 different streams and municipalities without such a system have 15 different streams, on average. So, municipalities with a unit- based pricing system are slightly more eager to separate waste. An empirical estimation can test for such a relation and therefore we now estimate Waste w,i,t 5 a s UBP s,i,t 1 gse 1 bus i,t 1 dus i,t UBP s,i,t 1 c 0 1 c i 1 d t 1 e i,t (3) where US i,t is the number of different waste streams collected by municipality i in year t. So, coefficient b gives the marginal effect of extra waste streams and coefficient d presents the effect of extra waste streams times the UBP system. Table 11.3b presents the estimation results. 30 KINNAMAN PRINT pt2-3.indd /02/ :55

20 306 Handbook on waste management Table 11.3b Effectiveness of UBP related to number of separately collected waste streams 1 Unsorted Compostable Paper, glass and textiles Other Number 0.05** 0.31*** 0.08*** 0.88*** times volume *** times frequency 0.15*** *** times bag 0.42*** times weight 0.35*** R 2 (adjusted) F-statistic Note: 1. Coefficients with */**/*** are significant at 10%/5%/1% level. The estimation of this relationship gives some interesting results. First, the quantity of unsorted waste is negatively related to the number of separately collected waste streams. There is an extra effect if the more sensitive pricing systems (weight, bag or frequency) are chosen. Second, the quantity of compostable waste is positively related to the number of separately collected waste streams. An interpretation could be that those municipalities with more separately collected waste streams place recycling higher on the political agenda and thus provide more facilities for home composting, for example. This is confirmed for some recyclables by the positive relation between the quantity of paper, glass and textiles and the number of separately collected waste streams. Another reason for this relation could be that, in many municipalities, the separate collection points for paper, glass and textiles are combined with the collection points for other waste streams. Figure 11.3b shows that, for all four unit- based pricing systems, there is a positive relationship between the reduction in unsorted waste and the number of separately collected waste streams. For more refined pricing systems, the relationship is stronger, as one should expect. 4. MUNICIPAL POLICIES TO SAVE COSTS OF SOLID WASTE There is a large literature discussing the economic performance of waste services. Recently, Simões and Marques (2012) have shown that, in more than 100 papers, cost or performance functions were estimated in the KINNAMAN PRINT pt2-3.indd /02/ :55

21 Waste management in the Netherlands 307 % of unsorted waste Volume Frequency Bag Weight Number of separately collected waste streams Figure 11.3b Effectiveness of UBP related to number of separately collected waste streams: unsorted waste context of discussing the efficiency of municipal waste services. Most of these studies look at the so- called public private dichotomy of solid waste delivery, as there was some evidence in the early literature that private delivery provides efficient services well adapted to needs and a reduction in the costs to the taxpayer. However, the current evidence for cost savings from private delivery is more mixed. In a recent overview article, Bel et al. (2010a) conduct a meta- regression analysis, dominated by the refuse collection literature, and show that there is no unambiguous evidence for obtaining significant cost savings from private production. This is in line with recent evidence for the Netherlands, where Dijkgraaf and Gradus (2013) show that, in the case of refuse collection, the cost advantage of inter- municipal cooperation is larger than that of privatization. However, less emphasis in this literature has been given to other elements such as the effects of unit- based pricing or recycling policies on costs. Simões and Marques (2012) conclude that there is not much attention given to discussing the relationship between collection, recycling and incineration. 31 Including the different UBP systems as explanatory variables in cost function estimation has an additional advantage. As administrative costs differ significantly between the systems, it is interesting to know which system is preferred from a cost- minimizing perspective. Another topic is whether recycling of different waste streams is economically advisable. 32 This is an empirical question, as more recycling can come KINNAMAN PRINT pt2-3.indd /02/ :55

22 308 Handbook on waste management with higher costs because more equipment and containers are required, but also with higher benefits because of lower total treatment costs for unsorted waste and because of recycling revenues and fees for bringing separate waste streams. 33 In the literature, such a research question is investigated by estimating the following cost function (for the Netherlands, see Dijkgraaf and Gradus 2008, 2013): lntc i 5 f(lnq i, lni i, lns i,o N,O C,O P,O I,UBP s,us i, a i, b t ) 1 e i,t (4) where TC are the (total) waste costs per household. 34 Comparison of total costs between municipalities is only possible when a correction is made for all relevant differences in exogenous factors. Most factors we use follow directly from the literature that estimates cost functions for waste collection. Total costs will change: 35 if the number of stops made by the collection vehicle increases (Q is the number of pick- up points, measured as the number of households); if the time spent at each pick- up stop increases (more bags or bins) (I is the number of inhabitants per pick- up point); if the time to arrive at the different pick- up points increases (S is the area served per pick- up point); if the institutional form in which waste is collected changes (O N is a dummy with value 1 for municipalities where waste is collected by neighboring municipalities, O C is a dummy with value 1 for municipalities that collect waste in cooperation with other (neighboring) municipalities, Oo is a dummy with value 1 for municipalities that use a municipality- owned collection firm, O P is a dummy with value 1 for municipalities that use a private collection firm and O I is a dummy with value 1 for municipalities that collect waste themselves); as the time- invariant municipal fixed effects (a i ) change; therefore we include fixed effects at the municipal level; as the time fixed effects (b t ) change; therefore we include a fixed effect for each year. For the purpose of our analysis, we also include unit- based pricing systems based on volume, frequency, bags and weight (UBP) and the number of separately collected waste streams (US). In addition to the data discussed earlier, we now also need data on costs and institutional mode (see Table 11.4a). We have data for 500 municipalities for the period , with a total of 5878 observations. We have KINNAMAN PRINT pt2-3.indd /02/ :55

23 Waste management in the Netherlands 309 Table 11.4a Descriptive statistics for other variables in the cost function Variable Unit Mean Max. Min. Std dev. Collection costs (TC) Euro per household per year Collection by: neighbor (O N ) Dummy cooperation (O C ) Dummy municipality-owned Dummy company (O O ) private company (O P ) Dummy in-house (O I ) Dummy No. of separately collected streams Number fewer observations for US (2653 instead of 5878) and therefore we estimate two models: one including this variable and one without it. Institutional data on waste collection come from the Dutch Waste Management Council (AOO). TC is calculated for each municipality by multiplying the average cost per household by the number of households, based on figures from Agentschap NL. 36 TC is in real terms as we correct for price developments on the basis of the index of consumer prices. Dutch municipalities have a legal obligation to provide a waste- collection infrastructure for municipal waste, but they are free to choose whether to carry out this task themselves or to contract out waste collection to outside firms (private or municipality- owned). Of all observations, 37 percent represent contracting out waste collection to a private firm and 21 percent to a municipality- owned firm (see Table 11.4a). It should be noted that a municipality- owned firm operates under commercial law, whereas the shares are publicly owned by municipalities. A third group of observations (16 percent) represent collection via a municipal service in cooperation with neighboring municipalities. In the Netherlands, municipal cooperation means maintaining public production. 37 For 3 percent of municipalities, the waste is collected by a neighboring municipality. The remaining observations (23 percent) represent collection by municipalities themselves. 38 Table 11.4b shows the estimation results for the cost function. In both models, the estimate for the number of pick- up points (that is, households) indicates constant returns to scale as the coefficient is not (significantly) different from In both models, household size has a positive and significant effect on total costs, although only at the 10 percent level in the model excluding the number of waste streams. For area per household, KINNAMAN PRINT pt2-3.indd /02/ :55