Public preferences for controlling an invasive species in public and private spaces

Land Use Policy 41 (2014) 1–10

Contents lists available at ScienceDirect

Land Use Policy journal homepage: www.elsevier.com/locate/landusepol

Public preferences for controlling an invasive species in public and private spaces John Rolfe ∗ , Jill Windle 1 CQUniversity, School of Business and Law, Bruce Highway, Rockhampton, Queensland 4702, Australia

a r t i c l e

i n f o

Article history: Received 14 November 2013 Received in revised form 13 March 2014 Accepted 18 April 2014 Keywords: Invasive species Red imported fire ants Discrete choice experiments Non-market valuation

a b s t r a c t Discrete choice experiments have been used in this case study to assess community benefits for the control of red imported fire ants, an aggressive ant species that were introduced by accident in 2001 to Brisbane, Australia. This invasive species could have substantial impacts on agricultural production, biodiversity, ecosystem services, infrastructure and communities. Values for avoiding impacts on three particular land uses have been assessed in this study with discrete choice experiments. The results indicated that on a per hectare basis, the value estimates to avoid infestation in public areas (schools and parks), were much higher than for private areas (housing) or natural bushland areas (protected native vegetation). There were high levels of support for eradication rather than containment strategies, despite the additional costs involved. The use of both random parameters logit and latent class models demonstrates that there is a significant heterogeneity in preferences and values for controlling or eradicating the invasive species, indicating that it may be challenging to gain and maintain political support for management options, particularly if these involve large costs or inconvenience to households. © 2014 Elsevier Ltd. All rights reserved.

Introduction Identifying and evaluating appropriate management responses to control invasive weeds and animal pests can be challenging for policy makers. Not only do they have to consider the different types of control programme to implement, and the associated likelihood of success, but they also face budgetary limitations and have to consider the relative costs and benefits of different management measures. Invasive species can have substantial impacts on land uses, through effects on agricultural production, biodiversity, ecosystem services, infrastructure and communities (Pimentel et al., 2005; Lovell et al., 2006). For example Pimentel et al. (2005) estimated that invasive species in the United States cost over US$138 billion per year in damages and control, while McLeod (2004) reported that the impact of invasive animals in Australia generates costs of more than AU$700 million annually. Control efforts can be categorised into three broad strategies: prevention, eradication and containment (Born et al., 2005). Prevention efforts such as quarantine protocols are aimed at preventing the entry and establishment of a new and potentially invasive species. Once an unwanted species has entered the country

∗ Corresponding author. Tel.: +61 74923 2132. E-mail addresses: [email protected] (J. Rolfe), [email protected] (J. Windle). 1 Tel.: +61 74930 9089. http://dx.doi.org/10.1016/j.landusepol.2014.04.013 0264-8377/© 2014 Elsevier Ltd. All rights reserved.

the initial control effort is likely to be focused on eradication measures, and if this fails then control measures are aimed at containing the spread of the species. In all three cases, increasing investment in the particular strategy will increase the likelihood of success. The justification for each of these measures against doing nothing, and the distribution of effort between measures should be assessed by comparing the potential costs of the control against the benefits that may be generated, such as through the application of cost benefit analysis (Born et al., 2005; Burnett et al., 2008). For example, if the containment strategy also proves unsuccessful and the damage caused by the unwanted species is not too great (benefits of control are low), it might not be cost effective to implement further control efforts. While there has been some attention paid to identifying the costs of invasive species (e.g. McLeod, 2004; Pimentel et al., 2005; Lovell et al., 2006), little information is available about the values of benefits associated with avoiding or controlling weeds or animal pests. Assessing these benefits is challenging because most involve direct use, indirect use and non-use components, especially those involving reduced impacts on human health and the protection of environmental assets and ecological processes (Born et al., 2005; Lovell et al., 2006). Some studies (e.g. Sinden and Griffith, 2007) have used treatment costs, replacement costs or defensive expenditures as proxies for values, but these are unlikely to represent underlying community preferences for control, particularly where large non-use values are involved (Shogren et al., 2006).

2

J. Rolfe, J. Windle / Land Use Policy 41 (2014) 1–10

Most efforts to value the benefits of controlling invasive species have focused on the extractive direct-use values, such as those associated with agricultural losses (Pimentel et al., 2005; Lovell et al., 2006; Olson, 2006). There have also been some studies estimating non-extractive use values for recreation that are affected by invasive species (e.g. Lovell et al., 2006). While direct and indirect use values are important for some types of invasive species, they rarely represent the total economic value, and may not be very relevant for other invasive species, for example, those that impact on biodiversity. A more comprehensive approach to valuing benefits of controlling invasive species requires stated preference techniques such as the contingent valuation method or discrete choice experiments to capture the non-use components together with use and indirect use values. There has been limited application of stated preference techniques to issues involving invasive species (Born et al., 2005; Lovell et al., 2006). Although the number of studies has increased in recent years, the focus has concentrated on invasive plant species, particularly in aquatic environments. Turpie et al. (2003) report estimates for the loss of existence values in the Cape Floristic Region in South Africa from invasive plant species using the contingent valuation method, while Adams et al. (2011) used discrete choice experiments to estimate public preferences to reduce the impact of invasive plant species in State parks in Florida in the United States of America (USA). Nunes and Van Den Bergh (2004) report the use of a joint travel cost study (to estimate recreation benefits) and a contingent valuation study (to estimate non-use benefits) associated with the removal of harmful algal-bloom species along the coast of The Netherlands. Other studies have used discrete choice experiments to estimate the benefits from controlling algal blooms in Lake Tenkiller, Oklahoma (Roberts et al., 2008); the Black Sea Coast of Bulgaria (Taylor and Longo, 2010), and in a region of the South Island in New Zealand (Beville et al., 2012). Both Horsch and Lewis (2009) and Zhang and Boyle (2010) have used hedonic pricing models to estimate the impact of a common aquatic invasive species (milfoil) on lakefront property values in Wisconsin and Vermont, respectively. Provencher et al. (2012) also focused on the problem of milfoil in northern USA and Canada. They applied the contingent valuation method to estimate the impacts of a lake invasion in terms of property welfare losses and specifically incorporated respondents’ subjective opinions about outcome uncertainty and the potential effectiveness of a control programme. Champ et al. (2005) applied contingent valuation to assess the benefits of a weed control programme in the USA, while Carlsson and Kataria (2008) used discrete choice experiments to assess the benefits from weed-control programmes in both Sweden and the USA. Other studies have transferred benefit estimates from recreation, property, health or environmental valuation studies to infer the benefits of controlling an invasive aquatic pest species (Lovell et al., 2006). The focus of this paper is on valuing the non-market benefits of controlling an invasive insect pest species using discrete choice experiments. The case study of interest involved red imported fire ants in Brisbane, Australia, where there was potential for impacts on different land uses, including private residential areas, public use areas and bushland. The choice experiments were framed to address the need for marginal benefit value estimates to align with the cost parameters of other spread and bio-economic models. The relative benefits associated with eradication or containment control measures were also assessed. The paper is structured as follows: In the next section a description of the case study and methodological application of the experiment is provided. Experiment results are described in section three, followed by discussion and conclusions in the final section.

The case study: red imported fire ants in Brisbane, Queensland Red imported fire ants are an aggressive ant species that are native to South America and are viewed as a major risk to Australia for several reasons. They have detrimental impacts on human health, human lifestyle patterns, livestock and wildlife through their aggressive stings, and may also have impacts on crops by feeding on sap, and on electrical equipment and infrastructure through nest building (Jetter et al., 2002; Morrison et al., 2004). The species now inhabits large areas in the southern United States, where it causes extensive damage, mostly through impacts on residential households (Scanlan and Vanderwoude, 2006). There is potential for the species to colonise large areas of non-arid Australia if it is not controlled (Morrison et al., 2004; Scanlan and Vanderwoude, 2006). Red imported fire ants were introduced by accident to Australia, with infestations found in the port areas and south-western suburbs of Brisbane in February 2001 (Moloney and Vanderwoude, 2002). Follow-up surveillance identified scattered infestations in more than 300 km2 of the region (Scanlan and Vanderwoude, 2006). Modelling suggested that the pest could invade half of Australia within 35 years if it was not controlled (Kompas and Che, 2001; Scanlan and Vanderwoude, 2006). The Queensland Government has led a vigorous eradication policy funded by the Australian and Queensland governments to identify and control outbreaks, including regular inspections and control efforts, and restrictions over the movement of soil and garden waste in areas at risk (Fig. 1). While control efforts are reducing the rate of new discoveries, the pest had still not been eradicated in 2009, when the valuation survey was conducted, or at the time of writing in 2013. Ongoing control of the red imported fire ants has involved large and continuing investments of public funds, but no information about benefit values is currently available to assess the efficiency of that funding. In the case study reported in this paper, discrete choice experiments have been applied to assess these values. While discrete choice experiments have been widely used to assess public values for environmental protection (e.g. Bennett and Blamey, 2001), there has been very limited application of these and other non-market valuation techniques to invasive species issues to date, so many practical issues about how tradeoffs are framed and the behavioural responses that are triggered remain to be tested. There were several advantages in using red imported fire ants as a case study for estimating values for controlling invasive species. First, the benefits of controlling red imported fire ants are largely non-market benefits in terms of avoiding health impacts, maintaining lifestyle and amenity values, and avoiding environmental impacts (Moloney and Vanderwoude, 2002). This made it more suitable for assessment with a stated preference technique compared to other pest incursions where the costs may largely relate to agricultural production losses. Second, it represented an ongoing pest incursion, where public and private effort continued to be invested. This meant there was a rationale for the study to be done, helping to frame it as an important issue for respondents to consider in an ex-ante setting. Third, there was high levels of awareness and knowledge of the pest in the case study area, with few other competing pest issues in the region, making it easier to present different control options to households. Application of discrete choice experiments to the case study setting involved identifying the frame of the tradeoffs to be offered, the key attributes, levels and labels that were used to define the tradeoffs, and the way in which the payment vehicle was used to define the opportunity costs involved. To ensure that value estimates would be useful for policy analysis, the frame of the survey matched closely with the infestation and spread scenarios reported in Kompas and Che (2001), Scanlan and Vanderwoude (2006) and

J. Rolfe, J. Windle / Land Use Policy 41 (2014) 1–10

3

Fig. 1. Red imported fire ant restricted area in Brisbane in May 2009. Source: Adapted from Biosecurity Queensland Control Centre (2009) Fire Ant Restricted Area V40, 2009-07-AUG, Department of Employment, Economic Development and Innovation, Queensland Primary Industries & Fisheries, Biosecurity Branch Queensland.

Antony et al. (2009), where no further control of fire ants in Brisbane has been modelled to increase levels of spread and infestation over the next 35 years. In the experiments, respondents were asked if they were willing to pay for ongoing control of red imported fire ants over the next ten years to 2020 in order to avoid the impacts of future infestations. Attributes used in the choice experiment matched closely with the outputs of the biological spread models reported in Kompas and Che (2001) and Antony et al. (2009). These focused on the potential levels of infestation in residential areas (private homes), public areas (schools, recreational and sporting areas) and environmental areas (bushland). The attributes were designed to capture both use and non-use values associated with potential health, lifestyle and environmental impacts of red imported fire ants, using levels predicted from the spread models reported in Antony et al. (2009). Other potential impacts on agriculture and infrastructure were not included in the survey to avoid complexity and factors that may not directly relate to respondents. Respondents were informed that these impacts would be assessed separately. The background information provided to respondents is shown in Fig. 2. The cost mechanism chosen was an annual payment over ten years (to 2020). It was described to respondents in the following way: Costs can include different combinations of private control costs, and rates, fees and taxes paid to government to cover public costs of control Defining the payment vehicle in a general way helped to encompass the different types of public and private costs that might be incurred, and was also likely to minimise protests against any specific type of cost mechanism. It also may have helped to make the costs relevant to different groups of respondents, including both home owners and renters. It was realistic to assume that the costs of policy change will be funded through multiple instruments and

pathways. Using only a single payment vehicle has issues of relevance for many people (e.g. not everyone pays rent or taxes) and can attract serious response bias. In addition, particularly in Australia, specific reference to a tax of some kind may not be politically acceptable. Specific reference to the both private and public control efforts was included in description of the payment vehicle to acknowledge that control of an invasive species is not solely the responsibility of the public sector and that some private costs may be incurred to implement control measures on private property. This meant that the elicited value estimates were not solely for public control programmes, but also included a portion of private control costs (in private areas). The experiment was applied in a series of choice tasks, each with three alternatives where there was a common first option. This first option described a situation where there would be no control programme and no cost to individual taxpayers, but where there would be a large level of pest invasion expected across the three attribute areas by 2020. The other two options involved much lower levels of impact, but had an associated payment to reflect the opportunity costs of additional control. The experiment was designed to identify if respondents were willing to pay (WTP) for protection benefits, whether the WTP varied by different attributes, contextual issues and respondent characteristics, and the extent of heterogeneity in support for protection benefits. Other options for framing the experiment trade-offs were considered. One major alternative would have been to describe the current control options to respondents as the ‘status quo’ option, and to then ask about WTP for changes to the control effort. Disadvantages of this approach were that it was difficult to define the levels of infestation risk from this ‘status quo’, there would have been political difficulties involved in presenting different control efforts and infestation levels from the current policy settings, and it would have been necessary to ask respondents to if they were

4

J. Rolfe, J. Windle / Land Use Policy 41 (2014) 1–10

Information about Red Imported Fire Ants The Red Imported Fire Ant is a highly invasive and destructive pest which originates from South America. It was first detected in Brisbane in February 2001, with infestations found around the port facilities at Fisherman Islands and in the south-western suburbs around Richlands. Red Imported Fire Ants are regarded as the worst of the five ant species that are rated in the top ‘100’ of invasive species in the world. The key risks to urban areas are: • Human Health – the ants produce painful stings, which can induce allergic reactions. People have to avoid outdoors areas where the ants might live. • Recreation impacts – the ants interfere with outdoor recreation activities, and affect the use of: o Backyards o Parks o Golf courses and other sporting facilities o School playground and sporting fields • Impacts on native bushland and wildlife – Fire ants are very aggressive and outcompete native species. Heavy infestations are expected to reduce native ant species by up to 90%, as well as impacting on insects, spiders, lizards, frogs, birds and some small mammals. Red imported fire ants can also have some impacts on agriculture and infrastructure (the latter by nesting in electrical equipment). These impacts are being assessed separately, and are not included in this survey. The Queensland and Australian Governments have devoted significant resources to controlling Red Imported Fire Ants since they were found in Brisbane. While they have been containing outbreaks, they have not yet been able to eradicate the pest in all areas. More information about fire ants can be found on the Queensland Government website: http://www.dpi.qld.gov.au/cps/rde/dpi/hs.xsl/4790_4538_ENA_HTML.htm Fig. 2. Background information provided to respondents.

WTP for increased control effort or willing to accept (WTA) compensation for lower control effort. Defining the first option as a ‘no-control’ option had advantages because the scenarios were easier to define (no-control versus control), the opportunity costs could be presented in WTP formats, and it avoided describing exactly the efforts, outcomes and residual risks of current control measures. However, there were risks that some respondents may have found it difficult to distinguish between the current situation with some control (but an unknown outcome) and the hypothetical baseline scenario outlined in the survey with no control programme for the next 15 years. The hypothetical nature of the first option may have made the valuation scenario less realistic with subsequent effects on the choices made by some respondents. Another important issue to consider was whether government should pursue more expensive eradication strategies or containment strategies with lower levels of control at reduced cost. This was an important policy question, but no information existed about whether the community had values for the type of strategies that governments could employ. While there is emerging evidence that the choice of policy mechanisms may be important in influencing value estimation (Johnston and Duke, 2007; Czajkowski and Hanley, 2009; Rolfe and Windle, 2013), this has not been tested in relation to invasive species. To explore this issue an additional split sample labelled experiment was included in the study. The labelled version of the survey (with the first version referred to as the ‘unlabelled’ version) identified whether an Eradication or Containment control strategy would be pursued by including them as labels for the two improvement alternatives. For this version, the attribute levels used in the experimental design were adjusted to show that the eradication policy tended to have higher cost levels

but would result in lower levels of infestation compared to a policy of containment. However, the eradication management strategy presented did not necessarily result in complete eradication of the pest, to allow for the possibility that a complete eradication policy might not be achievable within the ten year time frame of the valuation scenario. The background information provided to respondents for both versions of the survey confirmed that both the Australian and Queensland Governments had devoted substantial resources to controlling the fire ants, with control achieved but not eradication in all areas. Example choice sets are provided in Fig. 3 and details of the attribute levels for each of the two experiments in Table 1. The combination of attribute levels for the different options was determined using a D efficient design with nine choice sets, created in the ©Ngene software programme. A separate experimental design was applied for the labelled version because different levels were allocated across alternatives (Table 1). The survey design was tested and refined with four focus groups held with Brisbane residents in 2009. Focus group discussions also allayed some prior concern about potential fatigue effects associated with completing nine choice tasks (with an additional instructional choice set not included in the data analysis). Recent tests and reviews in the literature have also identified little cause for concern (Hess et al., 2011), and suggest that learning effects may dominate over fatigue effects. For example, Caussade et al. (2005) found learning effects dominated over fatigue effects during the first nine tasks. The survey was conducted in an online format with an internet panel of Brisbane residents supplied by a private service provider. Two hundred and twelve survey responses were collected from the Brisbane metropolitan area in August 2009, with 104 responses for the unlabelled split-sample and 108 responses for the labelled

J. Rolfe, J. Windle / Land Use Policy 41 (2014) 1–10

5

Fig. 3. Sample choice set from the unlabelled and labelled versions. Note: Each choice set contained links to further information on fire ants and to background survey information.

Table 1 Attribute and alternative label descriptions and levels. Private areasa

Public areasa

Bushlandb

Cost

Options

Description

No of homes affected by 2020

Recreational, sporting and school areas affected by 2020

Protected areas affected by 2020

How much you pay each year to 2020

All Versions Option 1

No control

500,000 homes (30%)

7500 ha (30%)

73,000 ha (30%)

$0

Version 1 Option 2 Option 3

Unlabelled Option A Option B

17,000; 83,000; 167,000 homes 1%, 5%, 10%

250; 1250; 2500 ha 1%, 5%, 10%

12,000; 24,000; 36,500 ha 5%, 10%, 15%

$20, $50, $200

Version 2 Option 2

Management Labels Containment Strategy (smaller control effort)

500; 1250; 2500 ha 2%, 5%, 10%

Eradication Strategy (Larger control effort)

12,000; 24,000; 36,500 ha 5%, 10%, 15% 1000; 5000; 12,000 ha 0.5%, 2%, 5%

$20, $50, $100

Option 3

33,000; 83,000; 167,000 homes 2%, 5%, 10% 8000; 33,000; 83,000 homes 0.5%, 2% 5%,

125; 500; 1250 ha 0.5%, 2% 5%,

Note: Attribute levels were always described in terms of both an absolute amount and a percentage. a Information sourced from Antony et al. (2009). b Included 10 Local Government Areas.

$50, $100, $200

6

J. Rolfe, J. Windle / Land Use Policy 41 (2014) 1–10

Table 2 Respondent socio-demographic characteristics.

Table 4 Model variable descriptions. Sample

Gender Age Education Income (gross)

b

Female Average age Post school qualification Tertiary degree Less than $499 per week $500–$799 per week $800–$1199 per week $1200–$1999 per week $2000 or more per week

51% 42 yearst 70%* b 43%* b 18%b 16%b 27%b 25%b 14%* b

Populationa

Variables

Description

51% 43 years 56% 24% 17% 18% 21% 24% 21%

Cost Private areas Public areas

Amount households pay each year to 2020 % and no of homes (private areas) affected by RIFA % and ha of public areas (recreational, sporting and school areas) affected by RIFA % and ha of bushland (protected areas affected by RIFA) 1 = children living in the household; 0 = no children Age of respondents in years Data collected in categorical format from 1 = less than $33,000 per year to 5 = more than $104,000 per year The mid point of each category was used for analysis with an additional 25% added to the last category Dummy coded 1 if respondent indicated >66% likelihood that eradiation will be successful 1 = living in a Red Imported Fire Ant restricted area; 0 = not Dummy coded 1 if respondent has a garden or yard that they use Dummy coded 1 if respondent intends to continue living in Brisbane for > 5 years Dummy coded 1 if respondent has lived in Brisbane for > 5 years Dummy coded 1 if respondent owns their own home Alternative Specific Constant associated with Improvement Options 2 and 3 in the unlabelled version (unlabelled experiment only) Alternative Specific Constant associated with Containment Strategy label Alternative Specific Constant associated with Eradication Strategy label

Bushland Children Age Income

a

Australian Bureau of Statistics 2006 Census. * Indicates a statistical difference between the sample and the population when applying: b = the normal approximation to the bionomial test or t = independent samples T-test.

RIFA area

Table 3 Level of confidence that red imported fire ants can be eradicated. Q. In your opinion how likely is that Red Imported Fire Ants will be eradicated in Brisbane and south-east Queensland in the next five years? Full eradication is very likely (more than 90% probability) Full eradication is likely (67–90% probability) Full eradication is medium likely (34–66% probability) Full eradication is unlikely (10–33% probability) Full eradication is very unlikely (less than 10% probability Don’t know Total

Frequency

Percent

Eradication possible

Garden Cumulative percent

Future in Bne Past in Bne Home owner

8

4%

4%

34

16%

20%

56

26%

46%

ASC Containment

58

27%

74%

ASC Eradication

38

18%

92%

18 212

8% 100%

100%

ASC Control

split-sample. Fifty-six percent of the respondents were living within a Fire Ant Restricted Area (as defined by the Queensland Government dated 15 May 09) and 44% were living outside the restricted area. The socio-demographic characteristics of survey respondents were well aligned with those of the population in terms of age and gender. The education of the respondent group was significantly higher than that of the general Brisbane population and there were significantly less respondents in the highest income category (Table 2). The majority of respondents were long term residents of Brisbane, with 85% having lived there for five years or more and 75% intending to stay in the city for at least the next five years. Most respondents lived in houses, with more than 59% owning their homes. Nearly 90% of respondents had access to a yard or garden, with 70% of them using the space often or very often.

Values for reducing the spread of red imported fire ants Responses to the survey indicated that respondents were generally aware of red imported fire ants in Brisbane (88%), with two-thirds (66%) concerned about the impact on themselves and their community. There was limited confidence that the pest species could be eradicated under current policy settings, with 20% having a confidence level of higher than 66% that the species could be eradicated and only 4% fully confident that eradication was possible (Table 3). These results indicate that the choice scenarios with differing levels of future infestation dependent on the control strategies undertaken were likely to be viewed as realistic. Following Provencher et al. (2012), these subjective perceptions about outcome uncertainty and the likelihood the species could

be eradicated were directly accounted for in the choice models to identify if they influenced values for containment and eradication. In this study, both random parameters logit (RPL) and latent class (LC) models were developed using Nlogit5 (Econometric Software 2012) to examine respondents’ preferences. First, RPL models were developed, for both the unlabeled and labelled versions of the survey, to account for both the panel nature of the data and for unobserved heterogeneity between respondents. Subsequently, LC models were developed to specify identification of multiple ‘classes’ of respondents who have heterogeneous preferences towards the attributes. The three infestation level attributes were presented in the choice sets in both absolute (e.g. hectares of infestation) and percentage levels (see Table 1), but the results were analysed absolute terms as this provided value estimates that could be applied subsequently in benefit transfer if required. A comparison of preferences for relative changes (percentage levels) is provided in the LC models reported below. In each of the RPL models, the non-cost attributes were randomised (with a normal distribution) to capture heterogeneity in the choice processes. However, in the unlabelled experiment the Bushland attribute was not significant as a random variable and was subsequently included as a non random variable. The socio-demographic and other explanatory variables were modelled to explain the choice of the status quo alternative over the two protection alternatives. Descriptions of the variables used in the models are presented in Table 4. The fitted models for both experiments (Table 5) were strong predictors of choices (significant models with high rho-square values). The negative coefficients for the attributes show that respondents preferred options with lower levels of infestation and cost. In the unlabelled version, the ASC (for both the two control options) was not significant indicating the variables included in the model specification explain preferences towards the control options. Significant levels of heterogeneity were present for the two randomised infestation attributes (Private and Public areas),

J. Rolfe, J. Windle / Land Use Policy 41 (2014) 1–10

7

Table 5 Random parameters logit model for unlabelled and labelled experiments. Unlabelled version Coefficient Random parameters in utility functions −0.0152* Private areas (per 1000 homes) −0.0243*** Public areas (per 100 ha) Bushland (per 100 ha) Derived standard deviations of parameter distributions Private areas 0.0139*** 0.0004*** Public areas Bushland Non random parameters in utility functions Bushland (per 100 ha) −0.0011* Cost −0.0142*** Non random parameters in utility function of the No Control (Status Quo) option Children 0.0392 −0.1333*** Age 16E-05 Income 43,160*** Eradication possible 10,917** RIFA −0.5745 Garden −26,894** Future in Brisbane Past in Brisbane −0.4520 16,951* Home owner Non random parameters in utility function of the Control options −79,710 ASC Controla 75E-05* Private areas2 ASC Containment ASC Eradication Model statistics 936 (n = 104) No of observations Log likelihood −643 1410 AIC 0.375 McFadden R-squared Chi squared 771 * ** *** a

Management labels St Error

Coefficient

St Error

0.0086 0.0094

−0.0043** −0.0003*** 0.0019*

0.0018 0.0001 0.0011

0.0017 0.0001

0.0064*** 0.0002 0.0083***

0.0011 0.0002 0.0012

0.0006 0.0015

−0.0144***

0.0022

10,104 0.0407 15E-05 11,145 10,130 14,006 11,222 13,326 11,019

11,427 0.0286 −132E-05 −17,782 12,733 −23,755** −29,589*** 21,921** 15,960*

10,715 0.0386 13E-05 14,461 11,037 11,849 0.9495 0.9881 0.8665

22,507 29,571*

17,401 17,749

48,735 43E-05

972 (n = 108) −661 1397 0.381 814

Significant at the 10% level. Significant at the 5% level. Significant at the 1% level. The same constant applied to both the two control options (Fig. 3).

as evidenced by the significance of the standard deviations of the random parameters. The parameter for Bushland attribute was only weakly significant at the 10% level, and a significant quadratic term was identified for the Private areas (houses) attribute (included only in the improvement alternatives). Older people; and people who thought they would be living in Brisbane for more than five years were more likely to select a control option. People who thought eradication was possible; people who lived in a RIFA area, and people who owned their own home were all more likely to select the status quo option. In the model for the labelled experiment (Table 5), both the ASCs for the management labels were positive (but not significant for the Containment option) indicating they were preferred over the status quo (no control option). The higher coefficient value for the Eradication option indicated this option was preferred over the Containment option Separate analysis confirmed this preference with 56% of respondents selecting this option compared to 33% who selected the Containment option (and 12% the status quo option). All three impact attributes were significant, with Bushland only weakly significant as in the unlabelled model. There was significant heterogeneity in the preferences for control in Private areas and Bushland areas, but not for Public areas (indicting there was more uniform agreement that public areas should be protected). People who owned their own home and those who had lived in Brisbane for more than five years were more likely to select the status quo option while people who had a garden and who intended to live in Brisbane in the future were both more likely to select a control option. The annual willingness-to-pay (WTP) estimates for marginal changes were calculated for the different attributes with a

part-worth formula (WTP = −ˇattribute /ˇcost ), with the Krinskey and Robb (1986) procedure used to generate confidence intervals. For the Private attribute in the unlabelled experiment, the net WTP value was estimated by also taking account of the squared attribute coefficient. The results are reported in Table 6 and represent annual household WTP for a ten year period to 2020. In the unlabelled experiment, respondents were WTP $0.13 to reduce the impact of fire ants per 1000 private homes. The WTP to protect public areas such as recreational sporting and school areas ($172 per 100 ha) was much higher than values for control in protected areas of natural bushland ($0.08 per 100 ha). This is likely to be because of both differences in priorities and scope (there was a much larger area of bushland presented in the choice sets). On an area basis, the WTP to avoid infestation in public areas is much higher than for private housing or natural bushland areas (1000 homes each on 1000 square metre blocks is equivalent to 100 ha). This may be because respondents were concerned about the potential risks to them and their families in public areas (schools and sporting grounds) or they were concerned about the wider risks to the community. Potential reasons why values were lower for private housing were that respondents viewed this as a lower risk, were more confident in their ability to manage risks for their own housing, or did not view control as a public responsibility. In the management labelled experiment, there was a significant coefficient for the Eradication option, but not for the Containment option, confirming that Eradication was a preferred strategy despite the higher cost. Households were WTP at least $206 per household per annum for ten years to pursue a successful eradication strategy (Table 6). Offering the management labels had a substantial impact on choice behaviour, with the proportion of respondents selecting

8

J. Rolfe, J. Windle / Land Use Policy 41 (2014) 1–10

Table 6 Household WTPa and confidence intervals (95%) to reduce the spread of red imported fire ants. Unlabelled version

Private areas (per 1000 homes) Public areas (per 100 ha) Bushland (per 100 ha) Eradication policy option a b

Management labels

Mean WTP

Low CI

High CI

$0.13b $172 $0.08

−$0.17 $0.52 −$0.01

$0.37 $250 $0.13

Mean WTP

Low CI

$0.30 $193 $0.13 $20,563

High CI

$0.07 $108 −$0.02 −$2797

$0.42 $239 $0.22 $63,667

Annual payments for ten years. WTP for private areas calculated with the following formula −(ˇPrivate + 2ˇPrivate 2 *XPrivate )/ˇcost .

Table 7 Latent class models for each split sample and a pooled version. Unlabelled version Coefficient

Management labels St Err

WTP

Coefficient

−$223 −$409 −$168

44%*** −0.016*** −0.065** −0.067*** −0.003

0.002 0.027 0.022 0.019

0.003 0.068 0.050 0.040

45%*** 0.001 −0.130** −0.144*** −0.143***

0.024 0.115 0.204 0.174

11%*** −0.026** 0.069 −0.102 0.060

Class 1: Sample % Cost Private areas (%) Public areas (%) Bushland (%)

67%*** −0.017*** −0.037*** −0.068*** −0.028*

0.001 0.014 0.016 0.015

Class 2: Sample % Cost Private areas (%) Public areas (%) Bushland (%)

18%*** −0.005 −0.316*** −0.160*** −0.132***

Class 3: Sample % Cost Private areas (%) Public areas (%) Bushland (%)

14%*** −0.032 −0.014 0.302 −0.226

Model statistics Observations Log Likelihood AIC McFadden R squared Chi Squared

936 (n = 104) −641 14 0.376 774

* ** ***

Pooled sample St Err

WTP

Coefficient

St Err

WTP

55%*** −0.017*** −0.054*** −0.069*** −0.025**

0.001 0.013 0.012 0.011

−$317 −$404 −$148

0.004 0.051 0.039 0.026

30%*** −0.002 −0.233*** −0.163*** −0.133***

0.002 0.038 0.032 0.021

0.011 0.067 0.089 0.058

15%*** −0.019*** 0.064* −0.030 0.019

0.005 0.036 0.045 0.038

−$402 −$409

972 (n = 108) −665 1397 0.377 806

−$342

1908 (n = 212) −1322 14 0.37 1549

Significant at the 10% level. Significant at the 5% level. Significant at the 1% level.

the No Control option falling from 20% in the unlabelled experiment to 12% in the labelled experiment. This confirms the findings from other studies that including information about different policy options has an important influence on value estimates (Johnston and Duke, 2007; Czajkowski and Hanley, 2009). The influence of the management labels on WTP by attributes was mixed, with WTP to protect Public areas and Bushland areas higher compared with the unlabelled value estimate, while the values to protect private areas were lower. These results suggest that respondents are more focused on the public benefits of eradication than reducing the risks to their private homes. A further comparison of the heterogeneity in respondents’ preferences was made with LC models. In this case the attribute levels were modelled in relative terms (percentage levels) so that preferences for marginal changes across attributes were more easily compared. Three models were developed; one for each split sample and one that pooled both samples together. In all cases, attribute only models were developed with three classes as the maximum number of classes with a minimum membership of 10 percent. The results are presented in Table 7 and highlight the difference in preferences within each sample. All models were significant with high McFadden rho square and low AIC values. In the unlabelled version, two thirds (67%) of the sample in Class 1 had significant preferences to reduce the area of impact. Preferences were strongest (high coefficient value) and WTP estimates the highest to reduce impacts in Public areas ($409 per household for a one percent reduction) compared to Private

areas ($223) and were the lowest for Bushland areas ($168). In Class 2, 18 per cent of respondents had significant preferences to reduce the impacts in all areas at any cost (i.e. cost was not a significant consideration in their preference selection and is possibly an indication of strategic behaviour). In Class 3 (14 percent of the sample) there were no significant preferences for any of the attributes. In the labelled version class composition was quite different. In Class 1, 44 percent of respondents had significant and similar preferences to reduce impacts in both public and private areas, but not in Bushland areas Forty-five percent (Class 2) had significant preferences for reductions in all areas, but were not considering cost and 11 percent (Class 3) were predominantly cost adverse. When all responses were pooled 70% of respondents (Class 1 and 3) had significant preferences to reduce impacts in private areas (valued at approximately $317 for a one percent improvement) and 55% of respondents (Class 1) had significant values to reduce the impacts in public areas ($404 for a one percent improvement) and Bushland areas ($148 per one percent improvement). Discussion and conclusions The research reported in this paper shows the potential for stated preference techniques such as discrete choice experiments to value the benefits of controlling an invasive pest species. The case study application to red imported fire ants in Brisbane, Australia, has addressed several of the information gaps identified by Born et al. (2005). A key contribution is that the research

J. Rolfe, J. Windle / Land Use Policy 41 (2014) 1–10

demonstrates significant non-market values exist to address future risks of infestation. These values are likely to encompass both use and non-use factors relating to protection of private homes, public spaces (schools and sporting areas) and bushland in the regional area. However, the results also highlight the heterogeneity in preferences to reduce the impacts of an invasive species. The results of the RPL models contribute to our knowledge by providing mean value estimates across the population samples, and highlight the relative importance of other explanatory variables. The LC models highlight the different groups of respondents and their relative preferences to reduce impacts across the three asset classes of private homes, public spaces and natural bushland. The results of the study allow three important conclusions to be made. First, there are significant non-market values to control red imported fire ants across a range of different asset classes, confirming the recommendations of Born et al. (2005), Lovell et al. (2006) and Shogren et al. (2006) that assessment of non-market values is important in evaluating biosecurity policy options. Second, there are significant values for eradication options when respondents are given a choice about management approaches. While the relationship between management values and protection options has been identified in other contexts (Johnston and Duke, 2007; Czajkowski and Hanley, 2009), this study provides key evidence that there are public benefits in having eradication policies for invasive species. Third, there is significant heterogeneity in preferences and values for control and eradication, as demonstrated in both the RPL and LC models. This suggests that it may be challenging to gain and maintain political support for management options for the invasive species, particularly if these involve large costs or inconvenience to households. We note that in this study the majority of respondents thought there was a relatively low probability that full eradication was likely or that it could be achieved as early as 2020. The study also provides estimates of values to avoid infestation for the three key asset classes of private homes, public spaces and natural bushland, which may be suitable for benefit transfer into other applications. Respondents are unlikely to be focused on individual pest species, but rather on the type of damages that might be averted with control or eradication. The values estimated in this study may be suitable for benefit transfer to different species and issues that have similar potential impacts on the three attributes used in this study. Estimating values for biosecurity programmes is challenging because of inherent problems of outcome uncertainty and the difficulty in associating a particular outcome with a specified costs. In light of this, several caveats to the results should be noted. First, there may have been some issues in framing the base option in the experiment as a No Control option, given that government were making vigorous attempts to eradicate the pest. This may have biased responses away from the No Control base Second, the payment vehicle included reference to both private and public costs, and it is unclear if or how respondents considered their own potential private costs. It is possible that respondents did not include their personal efforts and private costs as part of the payment tradeoffs, which is possibly why reducing the impacts of invasions in private areas did not appear to be as important as in other public areas. Third, the study focused on only one invasive species. It is possible that respondents consider that eradication and control measures will also control invasions of other pests, leading to some form of mis-specification bias. Fourth, issues of diminishing marginal returns have not been well captured, although the higher values for eradication options suggest these may be important. These are important issues for future research.

9

Acknowledgements This study was been funded through the Commonwealth Environmental Research Fund for the Environmental Economics Research Hub. The contribution of Tom Kompas and Jeff Bennett in the design of the experiments, and Gail Tucker in the analysis of results is gratefully acknowledged.

References Adams, D.C., Anafrida, N.B., Donna, J.L., Sherry, L.L., Janaki, R.R.A., 2011. Public preferences for controlling upland invasive plants in state parks: application of a choice model. Forest Policy and Economics 13, 465–472. Antony, G., Scanlon, J., Francis, A., Kloessing, K., Ngeyen, J., 2009. Revised benefits and costs of eradicating the Red Imported Fire Ant. In: Paper Presented at the 53rd Annual Conference of the Australian Agricultural and Resource Economics Society, 10–13 February, Cairns. Bennett, J., Blamey, R., 2001. The Choice Modelling Approach to Environmental Valuation Edward Elgar. Cheltenham, UK. Beville, S.T., Kerr, G.N., Hughey, K.F.D., 2012. Valuing impacts of the invasive alga Didymosphenia geminata on recreational angling. Ecol. Econ. 82, 1–10. Born, W., Rauschmayer, F., Brauer, I., 2005. Economic evaluation of biological invasions – a survey. Ecol. Econ. 55, 321–336. Burnett, K.M., D’Evelyn, S., Kaiser, B.A., Namtamanasikarn, P., Roumasset, J.A., 2008. Beyond the lamppost: optimal prevention and control of the Brown Tree Snake in Hawaii. Ecol. Econ. 67, 66–74. Carlsson, F., Kataria, M., 2008. Assessing management options for weed control with demanders and non-demanders in a choice experiment. Land Economics 84, 517–528. Caussade, S., Ortuzar, J., Rizzi, L., Hensher, D., 2005. Assessing the influence of design dimensions on stated choice experiment estimates. Transport. Res. B 39 (7), 621–640. Champ, P.A., Alberini, A., Correas, I., 2005. Using contingent valuation to value a noxious weeds control program: the effects of including an unsure response category. Ecol. Econ. 55, 47–60. Czajkowski, M., Hanley, N., 2009. Using labels to investigate scope effects in stated preference methods. Environ. Resour. Econ. 44 (4), 521–535. Hess, S., Hensher, D.A., Daly, A., 2011. Not bored yet – revisiting respondent fatigue in stated choice experiments. Transport. Res. A 46, 626–644. Horsch, E.J., Lewis, D.J., 2009. The effects of aquatic invasive species on property values: evidence from a quasi experiment. Land Economics 85 (3), 391–409. Jetter, K.M., Hamilton, J., Klotz, J.H., 2002. Red imported fire ants threaten agriculture, wildlife and homes. California Agric. 56, 26–34. Johnston, R.J., Duke, J.M., 2007. Willingness to pay and policy process attributes. Am. J. Agric. Econ. 89 (4), 1098–1115. Kompas, T., Che, N., 2001. An Economic Assessment of the Potential Costs of Red Imported Fire Ants in Australia. Report for the Queensland Department of Primary Industries, Brisbane. Krinskey, I., Robb, A., 1986. On approximating the statistical properties of elasticities. Rev. Econ. Stat. 68, 715–719. Lovell, S.J., Stone, S.F., Fernandez, L., 2006. The economic impacts of aquatic invasive species: a review of the literature. Agric. Resour. Econ. Rev. 35, 195–208. McLeod, R., 2004. Counting the Cost: Impact of Invasive Animals in Australia. Pest Animal Control Cooperative Research Centre, Canberra, Australia. Moloney, S., Vanderwoude, C., 2002. Red imported fire ants: a threat to eastern Australia’s wildlife? Ecol. Manage. Restor. 3, 167–175. Morrison, L.W., Porter, S.D., Daniels, E., Korzukhin, M.D., 2004. Potential global range expansion of the invasive fire ant, Solenopsis invicta. Biol. Invasions 6, 183–191. Nunes, P.A., Van Den Bergh, C.J., 2004. Can people value protection against invasive marine species? Evidence from a Joint TC–CV survey in the Netherlands. Environ. Resour. Econ. 28, 517–532. Olson, L., 2006. The economics of terrestrial invasive species: a review of the literature. Agric. Resour. Econ. Rev. 35, 178–194. Pimentel, D., Zuniga, R., Morrison, D., 2005. Update on the environmental and economic costs associated with alien-invasive species in the United States. Ecol. Econ. 52, 273–288. Provencher, B., Lewis, D.J., Anderson, K., 2012. Disentangling preferences and expectations in stated preference analysis with respondent uncertainty: the case of invasive species prevention. J. Environ. Econ. Manage. 64, 169–182. Roberts, D.C., Boyer, T.A., Lusk, J.L., 2008. Preferences for environmental quality under uncertainty. Ecol. Econ. 66 (2–3), 584–593. Rolfe, J., Windle, J., 2013. Including management policy options in discrete choice experiments: a case study of the Great Barrier Reef. Can. J. Agric. Econ. 61, 197–215. Scanlan, J., Vanderwoude, C., 2006. Modelling the potential spread of Solenopsis invicta Buren (Hymenoptera: Formicidae) (red imported fire ant) in Australia. Aust. J. Entomol. 45, 1–9. Shogren, J.F., Finnoff, D., McIntosh, C., Settle, C., 2006. Integration-valuation nexus in invasive species policy. Agric. Resour. Econ. Rev. 35, 11–20.

10

J. Rolfe, J. Windle / Land Use Policy 41 (2014) 1–10

Sinden, J., Griffith, G., 2007. Combining economic and ecological arguments to value environmental gains from control of 35 weeds in Australia. Ecol. Econ. 61, 396–403. Taylor, T., Longo, A., 2010. Valuing algal bloom in the Black Sea coast of Bulgaria: a choice experiment approach. J. Environ. Manage. 91 (10), 1963–1971.

Turpie, J.K., Heydenrych, B.J., Lamberth, S.J., 2003. Economic value of terrestrial and marine biodiversity in the Cape Floristic Region: implications for defining effective and socially optimal conservation strategies. Biol. Conserv. 112, 233–251. Zhang, C., Boyle, K.J., 2010. The effect of an aquatic invasive species (Eurasian watermilfoil) on lakefront property values. Ecol. Econ. 70 (2), 394–404.