Valuation of agricultural impacts on rivers and streams using choice modelling: A New Zealand case study

Abstract

Increasing substitution of dry land pastoral and arable farming for water-intensive practices is placing pressure on water resources in Canterbury. Although there is a large body of scientific data documenting environmental change, there is a general lack of information on the economic values of agricultural impacts on rivers and streams. This paper applies an economic non-market valuation method to help address this issue. Three impacts are considered: health risks of pathogens from animal waste; ecological effects of excess nutrients; and low-flow impacts of irrigation. This study provides a valuation of outcomes for public agri-environmental policy implemented in Canterbury such as The Dairy and Clean Streams Accord, Living Streams and The Restorative Programme for Lowland Streams. Modelling results indicate that the 5-year economic value for improvements to rivers and streams in Canterbury provided by agri-environmental policy is estimated to be about $186 million.

Keywords:

• non-market valuation
• agricultural environmental externalities
• choice experiment
• agri-environmental policy analysis
• Canterbury

Introduction

Water resources in Canterbury are experiencing escalating demands to satisfy often conflicting environmental and economic objectives. Increasing substitution of dry land pastoral and arable farming for water-intensive dairy farming is a significant current trend in Canterbury that, coupled with increased intensification of production, is a major contributor to growing pressures on water resources. Changes to river and stream environments can have significant economic and social impacts on Canterbury residents. To form an accurate assessment of agri-environmental policy, resource managers require not only scientific descriptions of change, but also information about the economic values of changes to river and stream goods and services.

Choice experiments are a stated preference method that is widely used to estimate the economic value of environmental changes (Bennett and Blamey 2001). In a choice experiment, respondents are presented with a series of questions (choice sets) in which the outcomes of alternative (hypothetical) policy scenarios are given. These outcomes are described by different levels of attributes that are relevant to policy objectives. Respondents choose the alternative within each choice set that they prefer. When respondents choose between alternatives, it is assumed that they are making tradeoffs between the different attributes and levels. By including a monetary attribute (typically a cost to the respondent's household) in each alternative, we can estimate how much the average individual is willing to pay (WTP) for a change in each of the other environmental attributes by dividing the estimated coefficient for the non-marketed environmental attribute by the estimated coefficient for the cost attribute (Bennett and Blamey 2001).

Several choice experiment studies conducted in Canterbury estimate economic values of environmental change associated with agriculture. Takatsuka et al. (2009) estimated values for changes in nitrate leaching to waterways on arable land for Canterbury. Their results indicate that respondents were, on average, WTP $74.95 and $87.73 per year to reduce nitrate leaching by 20% and 50% respectively. Baskaran et al. (2009a) focused on pastoral farming in Canterbury and found that respondents were WTP $11.83 and $38.55 per year for 10% and 30% reduction in nitrate leaching respectively. Households were WTP $8.33 for a 10% reduction in water use for irrigation and $8.90 for a 30% reduction. Baskaran et al. (2009b) employed the same set of attributes as in the previous pastoral study, but disaggregated the pastoral industry further by estimating values for the dairy sector only. They estimated that Canterbury households were WTP $22.67 and $31.82 for 10% and 30% reductions in nitrate leaching respectively, and $20.54 and $26.93 for 10% and 30% reductions in irrigation respectively.

This paper improves on the existing literature in several ways. Firstly, the majority of remaining sources of agricultural environmental impacts are considered to be non-point in nature. This means that it is difficult to separate out the impacts on waterways from various agricultural sector activities as a specific source may not be identifiable. Therefore, to conduct valuation within this context, the agricultural sector must be evaluated as a whole rather than decomposed into various industry activities. This study follows this argument and therefore frames impacts on waterways as being attributable to agriculture collectively. Secondly, this study focuses on the impact on waterways rather than level of agricultural activity. This approach recognises that it is not a change in the level of pollutant per se that people dislike, but rather it is the values that are impinged on by the presence of a pollutant that have meaning to respondents. For example, the quantity of nitrate in waterways has meaning to scientists but it is the description of excess weed growth and other ecological effects that have meaning to residents. Similarly, it is not changes in the amount of water used for irrigation that people value, but rather the impacts on river flows.

The aim of this study is to provide policy makers with estimates of the benefits to mitigating agricultural impacts on rivers and streams in Canterbury. These values can be incorporated into policy formation processes to aid in determining an appropriate level of policy implementation targeted at different environmental outcomes. Furthermore, this study contributes to the development of choice experiment application literature in New Zealand by providing a case study valuing agricultural externalities in Canterbury rivers and streams.

Background

Canterbury is New Zealand's largest region, covering an area of 45 346 km² and with a population of approximately 500 000 (SNZ 2007). Environment Canterbury, the regional council for Canterbury, is responsible for a wide variety of functions including environmental monitoring and investigations, regional policy and planning, water permits and discharge permits.

The Canterbury region has a 160-year history of agricultural production and is currently experiencing a significant trend in water-intensive dairy farming replacing traditional dry land pastoral and arable farming. Dairy stock unit numbers have increased rapidly and continue to do so. The environmental implications of these land use changes and intensification of production have been extensively researched, with a growing body of scientific literature outlining the impending consequences if inadequate action is taken. For example, studies of trends in water quality and contrasting land cover indicate a positive relationship between dairy stock numbers and decreasing water quality (Larned et al. 2004). Increases in water-borne pathogens such as Campylobacter have been reported (Ross and Donnison 2003). The rate of fertiliser and pesticide applications has increased dramatically over the past decade and this is forecast to continue increasing (PCE 2004). Coinciding with this trend are reports of increases in nitrogen and dissolved reactive phosphorous in waterways (Cameron and Di 2004). There are risks of irreversible damage in some instances, as long-term consequences such as those from land application of animal effluent cannot not be predicted accurately (Wang and Magesan 2004). There has been a significant increase in groundwater abstraction associated with land use intensification contributing to a decline in groundwater levels and reduced flows in rivers and lowland streams. Environment Canterbury records show a 260% increase in the amount of irrigated land from 1985 to 2005, and some 70% of consumptive use of water in the region is for pastoral purposes.

In the application of agri-environmental water quality policy, some progress has been made in reducing point sources of pollution, but non-point sources remain difficult to manage. Recent water quality planning has spurred the development of policies such as the Dairying and Clean Streams Accord (which targets farming practices on dairy farms), the Restorative Programme for Lowland Streams (which aims to return water to dry streams and ensure minimum environmental flows) (Environment Canterbury 2008) and the Living Streams project (encourages sustainable land use and riparian management practices).

Application details

Survey development

The development of a set of water attributes to be valued consisted of two main procedures. The first was a survey of relevant literature, policy documents and expert opinion, followed by focus groups and cognitive interviews (Dillman 2007) of Canterbury residents. To elicit expert opinion on which impacts were the most significant from a policy maker perspective, dialogue was initiated with Environment Canterbury. Several meetings were conducted and a survey was sent to relevant Environment Canterbury staff. This survey revealed that the variables most relevant to the policy process are scientific and technical in nature. The top four concerns identified were concentrations of E. coli (measured in mpn/100 ml), nitrate (mg/l), phosphate (mg/l) and pesticides (mg/l).

The challenge was to take these scientific measures and match them to descriptions of impacts salient to Canterbury residents. A starting point is to recognise that it is not, for example, the pollutant per se that has disutility for Canterbury residents but the attributes of rivers and streams valued by residents that are impinged on by the presence of pollutant. For example, a quantity of nitrate measured in micrograms per litre has meaning to scientists but it is excess weed growth and other ecological effects that have real meaning to residents. To this end, two focus groups were conducted with Canterbury residents randomly selected from phone listings. One, held in central Christchurch, aimed at gaining an urban perspective; the other was conducted in Lincoln and recruited a rural sample of participants. Cognitive interviews were also conducted in central Christchurch and Lincoln, with 10 interviews conducted in each location.

This search for policy-relevant, resident-salient, measurable water attributes resulted in the identification of three river and stream quality attributes to be valued in the choice experiment (Table 1). The first attribute is the risk of becoming ill from micro-organisms in animal waste that end up in waterways. The risk considered here is from recreational contact, and is measured as the number of people out of one thousand that would become sick annually. Level definition was guided by risk valuation literature (e.g. Adamowicz et al. 2007) and guidance from Environment Canterbury (2007b) The magnitude of changes in levels was guided by the works of Ball (2006) and McBride et al. (2002).

Table 1  Attributes and levels used in choice sets.

Attributes	Base level	Improvement levels
Health risk	60	10 and 30 people sick from contact recreation per 1000 per year
Ecology	Poor	Fair and good
Flow conditions	5	1 and 3 months of low-flow conditions per year
Cost	$0	$15, $30, $45, $60, $75, $90 per household per year

The second attribute allows analysis of the impact of excess nutrients on the ecological quality of rivers and streams. The descriptions of ecological levels were guided by policy outcomes for water quality as defined by Environment Canterbury (2007a) and representing contemporary policy. Elements of these defined outcomes were used to construct the levels. This also involved taking elements of the semi-quantitative macroinvertebrate index used by Environment Canterbury in defining outcomes, using the reports of Environment Canterbury (2003), Stark (1998) and Stark and Maxted (2007a, 2007b). Table 2 shows the descriptions used.

Table 2  Ecology attribute level definitions.

Quality	Definition
Poor	• Weeds are the only aquatic plants present and cover most of the stream channel
	• Stream-bed covered mostly by thick green algae mats
	• Only pollution-tolerant insect populations are present
	• No fish species are present.
Fair	• About 50% of stream channel covered by plants
	• Few types of aquatic plants, insects or fish
	• Algae covering about 20% of stream bed
	• Population densities reduced
Good	• Less than 50% of stream channel covered by plants
	• Algae cover less than 20% of stream-bed
	• Diverse and abundant range of aquatic plants, fish and insects
	• Insect communities are dominated by favourable species, with pollution-sensitive populations present

The third environmental attribute allows for valuation of the impact of low-flow conditions. The description of the impact of low-flow conditions on rivers and streams was guided by Ministry for the Environment (2008a, 2008b). The range in levels was guided by flow rate data from the Environment Canterbury website (http://www.ecan.govt.nz) and Environment Canterbury (2001). The cost attribute is defined as an annual household payment via rates or rent necessary to fund management to achieve improved water quality outcomes. These payments were framed as ongoing, as this view was strongly endorsed in focus groups and interviews during questionnaire development recognising agricultural development trends.

The different levels of attributes were combined to form choice sets according to a linear D-efficient main effects fractional factorial experimental design with dominant and implausible alternatives removed. It has been argued that main effects typically account for 70–90% of explained variance (Louviere et al. 2000). This design consisted of 18 treatments that were randomly blocked into three blocks of six choice sets. Figure 1 shows an example of the choice sets presented to respondents.

Figure 1  Example choice set.

The constant base alternative was assumed to be a worsening condition of rivers and streams if no change in management occurs. There would be no annual tax payment by the household, but the risk of becoming ill would be at its greatest, ecological quality would be poor and the number of low-flow months would be at its highest. The survey consisted of three sections with the first section measuring respondents’ attitudes towards agri-environmental policy and how rivers and streams are important to them. The second section consisted of a description of the impacts of agriculture on Canterbury rivers and streams being considered in the survey and policies that could be funded to mitigate these impacts, followed by choice sets. The third section concluded with household socio-demographic questions. The first and third sections were intended to capture preference heterogeneity that was not captured by the attributes in the choice sets.

Data collection

During the months of July and August 2008, 1500 surveys were mailed to Canterbury residents using random sampling stratified by Territorial Local Authority (TLA) to achieve a geographically representative sample. The mail-out procedure yielded 360 usable responses with an effective response rate of 25%. In order to assess if the sample was representative of the Canterbury population, chi-square tests were conducted. If the null hypothesis was rejected, it could be concluded that the census 2006 population data were statistically significantly different from the sample data. It was apparent that the null hypotheses were rejected for income, education and age. This means that the sample respondents had higher income, higher education and were older. This may indicate sample selection bias toward affluent and educated groups and thus, caution should to be taken when using WTP estimates to represent population values. However, employing a random parameter logit (RPL) model should be able to ameliorate this bias in terms of individual heterogeneity within income groups among respondents when valuing the attributes. To consider the geographical representation of the sample, a chi-square test was conducted for the distribution of respondents according to the regions ten TLAs. This test found that the distributions were not statistically significantly different. A relevant concern when conducting a choice experiment in which the experimental design is blocked is whether a sample contains sufficient representation of the choice sets. The distributions of the three blocks of the experimental design used in this survey were 32% 33% and 34%; the returned surveys therefore represented the choice sets adequately.

Results and discussion

Agri-environmental attitudes and importance of river and stream resources

The attitudes of respondents towards various aspects of agriculture and associated agri-environmental policy can influence how they value changes in impacts on waterways. Likewise, the importance that respondents place on different uses for water resources may condition the values that they place on impacts. Table 3 indicates that the vast majority of respondents disagree that agriculture is environmentally safe. The next two statements in Table 3 attempted to gauge the contentious issue of who should pay to clean up and prevent agricultural impacts on water resources. The survey results suggest some conflicting responses, as 64% agree that the burden should fall on rate payers collectively while 84% agree that costs should be borne by farmers. Although more respondents agree that farmers should pay, this means that some respondents agree with both statements. Perhaps this reflects the public nature of the resource and indicates that although rate payers are prepared to contribute, farmers must bear a proportionately greater cost. This argument was supported by the next two statements. Almost all respondents agreed that the Canterbury agricultural landscape is important, and so there are public benefits from the private actions of farmers. However, 75% of respondents agreed that a price should be charged for water for irrigation, reflecting that some of the private benefits enjoyed from the extensive use of a public resource as essentially a free production input should be channelled into the public domain. This result is also a reflection of public desire to see a valuable resource used efficiently and is a refutation of the current allocation mechanism that incentivises a race to the bottom of the well. A third of respondents agreed that organic farming methods should be employed across the sector. This could be considered relatively extensive support for policies in this area.

Table 3  Agri-environmental attitudes and river and stream resource importance.

	Attitude (% of sample)
	Disagree strongly	Disagree		Agree		Agree strongly		Don't know
Agri-environmental attitude
Agricultural production today is environmentally safe	21	48		20		4		7
Canterbury ratepayers as a whole should pay the costs of cleaning up and preventing agricultural impacts on water resources	3	22		35		29		11
Farmers should pay for the costs of cleaning up and preventing agricultural impacts on water resources	3	10		43		41		3
The agricultural landscape is important in Canterbury	2	2		41		54		1
A price should be charged for water for irrigation	5	13		43		32		7
Agriculture should fully convert to organic farming methods	13	43		21		12		11
River and stream resource importance
Resource for future generations	88		—		—		—
Habitat for plants and animals	83		—		—		—
Recreational opportunities	75		—		—		—
Drinking water resource for public	74		—		—		—
I just like knowing that they are there	28		—		—		—
Resource for commercial development	14		—		—		—

Table 3also shows which uses of rivers and stream resources are important to respondents and is an effort to indicate the relative weights given to use and non-use values by respondents. Inter-generational equity is important. This is not really considered a use per se, but instead reflects the sentiment of sustainability: the resource should not be degraded for the short-term gain of present generations at the cost of those to come. Use as habitat for plants and animals was also judged very important, as was recreational opportunity. There is some overlap here as healthy habitats also provide enhanced recreational opportunities, particularly for game hunting and fishing enthusiasts. Also very important is use for drinking water. The next statement could be considered to reflect existence values: just over a quarter of respondents considered these to be important. Finally, use as a resource for commercial development was considered important by only 14% of respondents. At first glance this seems a relatively small endorsement for water as a resource for commercial development. This finding may reveal that few respondents considered this to be a significant source of employment, or could reflect the fact that only a small proportion of respondents would actually require water resources for commercial uses.

Random parameter logit model estimates

In this study a RPL model was fitted to the choice data using NLOGIT 4.0™ statistical software. A significant benefit of employing this model specification is the incorporation of heterogeneity around the parameter mean estimates (Train 2003). To examine non-linear preferences the attributes are effects-coded with the lowest level of quality for each attribute being the base comparator. The attitudes of the respondents towards agri-environmental issues, resource importance questions and demographic variables were re-coded and interacted with the alternative specific constant for inclusion in modelling. The variables included in the final model specification are given in Table 4.

Table 4  Model data summary.

Variable	Description
Attribute variables
Risk 10	10 people per 1000 per year sick from recreational contact
Risk 30	30 people per 1000 per year sick from recreational contact
Ecology good	Ecological quality is good
Ecology fair	Ecological quality is fair
Flow 1	1 month of low-flow conditions per year
Flow 3	3 months of low-flow conditions per year
Cost	$15, $30, $45, $60, $75 and $90 per household per year
Non-attribute variables
ASC	Alternative specific constant: 1 if alternative 2 or 3; 0 if base alternative
Safe	Respondent agrees that agriculture is environmentally safe
Commercial	Respondent indicates that commercial use is important
Farmers pay	Respondent indicates that farmers should pay for agri-environmental water policy
Income	Household gross yearly income
Gender	1 if respondent female; 0 if male

To check if the effects-coded variables for an attribute should be combined into a single linear variable, a Wald test was conducted to observe whether the two parameters (one for each of the two effects-coded attribute levels) were equal. The null hypothesis of inequality was retained for all attributes. Thus, preferences for the two attribute levels were statistically significantly different and we can conclude that preferences were not the same across different attribute levels. Model estimates are presented in Table 5. To incorporate heterogeneity around the parameter mean, the most common distributional functional forms for the random parameters are normal, lognormal, uniform and triangular. In this work, different distributions were tested and finally a bounded triangular distribution was chosen for all attributes. The bounds were set to take into account a degree of heterogeneity whilst obtaining meaningful WTP estimates, with the spread of each random parameter distribution restricted to be equal to the mean. 500 shuffled Halton draws were used in maximising the simulated log-likelihood function. A panel specification that allows for error correlation between choice observations from a given individual was employed (Train 2003).

Table 5  Random parameter logit model estimates.

	Parameter estimate (standard error)
Attribute
Risk 10	0.592 (0.066)^3
Risk 30	0.072 (0.058)
Ecology good	0.805 (0.078)^3
Ecology fair	0.356 (0.681)^3
Flow 1	0.346 (0.074)^3
Flow 3	0.032 (0.085)
Cost	−0.016 (0.002)^3
Non-experimental variables
ASC	0.086 (0.337)
ASC^*Safe	−1.229 (0.208)^3
ASC^*Commercial	−1.071 (0.293)^3
ASC^*Farmers pay	−0.945 (0.058)^3
ASC^*Gender	−0.518 (0.204)^2
ASC^*Income	0.179 (0.048)^3
Standard deviation of parameter distributions
Risk 10	0.592 (0.066)^3
Risk 30	0.072 (0.058)
Ecology good	0.805 (0.078)^3
Ecology fair	0.356 (0.681)^3
Flow 1	0.346 (0.074)^3
Flow 3	0.032 (0.085)
Cost	0.016 (0.002)^3
Log-likelihood	−1410
McFadden psuedo-R ^2	0.37
AIC	1.35
BIC	1.39
Pr(chi^2) > z	0.000
Iterations	22
Observations	2160
^1Significant at 10% level
^2Significant at 5% level
^3Significant at 1% level

Considering the attribute variables, all attribute means were significantly different from zero at a 99% level except Flow 3 and Risk 30, which were not significant. All attributes have a priori signs with improvements in the levels of the attributes increasing the probability of that option being chosen. The magnitude of the probability increases as the attribute level improves indicating that respondents are sensitive to the scope of provision. The psuedo-R ²can be considered to be high, indicating a good model fit with the data. This model diagnostic was considered in conjunction with the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) in assessing model performance. Respondents derived the most utility from Ecology good, followed by Risk 10, Ecology fair, Flow 1 and Risk 30. The Cost variable was negative, indicating that alternatives with higher cost were less likely to be chosen. The alternative specific constant (ASC) was not significant, indicating that the sample did not prefer to have the status quo policy settings. The ASC interactions reveal that respondents who:

1.	agree that agriculture is environmentally safe
2.	agree that farmers should pay to clean up and prevent agricultural impacts on rivers and streams
3.	indicate that rivers and streams are an important resource for commercial development
4.	or are male

are less likely to choose an alternative with improvements to water quality. Respondents with higher household income are more likely to choose an alternative with improvements in water quality. The estimated standard deviations for the water quality attributes indicate significant heterogeneity in respondents’ preferences for these attributes, thus supporting the choice of a RPL model.

Welfare estimates

Estimates of WTP are shown in Table 6. These represent the annual payments respondents indicated they are WTP for marginal changes in river and stream improvements. Distributions of WTP were constructed by simulating the population using unconditional parameter estimates (population moments) (Hensher et al. 2005). A concern when dividing two random parameters, remembering that WTP is calculated as the environmental attribute parameter divided by the cost parameter, is that confidence intervals cannot be calculated as expected values for distribution moments may not exist (Daly et al. 2011). A pragmatic approach is then to report quantiles, which is done here by presenting lower and upper quartiles. The lower quartile tells us that 25% of the population is WTP less than a certain amount, while the upper quartile tells us that 25% of the population is WTP more than this amount. This approach was also used in the calculation and presentation of values of combinations of attribute changes as shown below.

Table 6  Estimates of average WTP ($NZ 2008). Lower and upper quartiles of WTP distributions in parentheses.

Attribute	WTP per household per year
Risk 10	$27 (20, 35)
Flow 1	$52 (42, 66)
Ecology good	$84 (62, 105)
Ecology fair	$64 (50, 80)

Different combinations of attribute outcomes were valued collectively as compensating surplus (CS) estimates. These estimates represent households overall willingness to pay for improvements in policy outcomes from the ‘no change’ alternative. The ‘no change’ base and two improvement scenarios are shown in Table 7.

Table 7  ‘No change’ base and two improvement scenarios.

No change	• 60 people per 1000 become sick from recreational contact each year
	• Ecological quality poor
	• 5 months of low-flow conditions per year
Management fair	• 30 people per 1000 become sick from recreational contact each year
	• Ecological quality fair
	• 3 months of low-flow conditions per year
Management good	• 10 people per 1000 get sick from recreational contact each year
	• Ecological quality good
	• 1 month of low-flow conditions per year

Table 8shows that, on average, Canterbury households are WTP $154.00 per annum for the middle levels of water quality improvement and $213.00 per annum for the best outcomes considered. Household level CS is aggregated across Canterbury for a 5-year period, reflecting the definition of the payment vehicle. To aggregate CS across the population, an assumption had to be made about the preferences of those who did not return the survey. The aggregate CS was calculated under two assumptions: non-respondents’ CS is half that of sample respondents and non-respondents have the same CS as sample respondents (Carson and Mitchell 1989). A declining discount rate (DDR) approach (Birol et al. 2010; Gollier et al. 2008) was used to calculate the present value of a 5-year horizon as policy outcomes in this context are considered to be ongoing. DDRs should be used when evaluating projects or policies with long-term impacts. In comparison with the use of a constant discount rate, the use of a DDR increases the weight attached to the welfare of future beneficiaries of water quality improvements (for a review of discounting for public policy see Hepburn et al. (2009). The 5-year discounted CS assuming non-respondents have the same preferences as the sample is about $186 million for the outcomes of the best management scenario and about $134 million for the middle level of policy outcomes. Assuming half the respondent CS for non-respondents, these values fall to about $112 million and $81 million respectively.

Table 8  Estimates of average compensating surplus (NZ$ 2008). Lower and upper quartiles of CS distribution in parentheses.

Scenario	Annual household	5-year Canterbury	Non-response adjusted
Management fair	$154 (125, 187)	$133 872 980	$80 969 948
Management good	$213 (169, 260)	$185 026 964	$111 946 872

Discussion and conclusions

This research used a choice experiment to provide welfare estimates for external effects of agriculture on Canterbury streams and rivers. More specifically, the number of low-flow months caused by irrigation, the impact of excess nutrients on ecological quality and the risk of becoming sick from contact with animal waste were attributes of agri-environmental policy valued by Canterbury residents. Survey and model results show that there is strong support among residents for protection and improvement of rivers and streams in Canterbury from these effects. The 5-year economic value of changes to river and stream quality in Canterbury is considerable and could be as high as $186 million. This finding should afford support for policy makers to continue funding to agri-environmental projects.

A major element of the debate over water quality and quantity centres on the perceived property rights of differing user and non-user groups in a community. The use of willingness to pay as the elicitation method for preferences for environmental improvements implies that the polluter has the property right to the use of the water resource. Conversely, the use of willingness to accept approaches (which estimate the value that someone is willing to accept for deterioration of environmental quality) implies that the property right is with the non-polluter. The allocation of property rights is often unclear. Although members of the general public may feel they have a moral right to water resources, they may not have a legally enforceable property right. Farmers, on the other hand, have their rights to water use more formally constructed since activities such as effluent discharge and water abstraction consents are legally enforceable under the Resource Management Act. The monetary value of preferences for changes in environmental quality can be estimated under either approach, although they may differ as a result of endowment effects, and the implication for who should pay for management of environmental changes is reversed.

Focus groups and interviews revealed that residents have informed awareness of the general issues involved. This is not surprising given the regular media coverage that water rights, water quality and related issues receive. Extractive water use, often accompanied by the subsequent disposal of agricultural waste back into the environment, versus alternative uses for Canterbury water resources by other groups within the region is at the heart of this sensitive and essential debate. A collaborative management strategy that encompasses the values held by diverse interests is crucial to forming policies that are acceptable to the general public. Such a strategy must aim to avoid the adversarial and legalistic approaches used to date. The inclusion of economic values for water quality is an important management component because it reveals both the level and relative importance of public preferences for policy outcomes that can be used to allocate management resources efficiently.

Acknowledgements

Funding for this research was provided by the Agricultural and Marketing Research and Development Trust, (http://www.agmardt.org.nz) and Lincoln University. The authors thank Environment Canterbury for contributing to the survey development.