Estimating the Extent of CITES Noncompliance among Traders and End‐Consumers; Lessons from the Global Orchid Trade

The Convention on the International Trade in Endangered Species of Wild Flora and Fauna (CITES) regulates trade in over 35,000 species, over 70% of which are orchids. To investigate rule‐breaking behavior among traders and buyers in a specific international wildlife trading community, we used direct questions (DQs) and the unmatched count technique (UCT) to survey the orchid growing community about CITES compliance and their knowledge and opinions of the rules. In DQ, 9.9% had smuggled, 4.8% had laundered, and 10.8% had been sent orchids from online purchases without paperwork; UCT estimates did not differ significantly. Growers with greater knowledge of CITES rules were more likely to break them, and there were widespread negative views of CITES among respondents. We recommend targeted enforcement focusing on both online trade and at the point of import, coupled with efforts to encourage traders and end‐consumers to engage with discussions on CITES rule implementation


Introduction
The international wildlife trade is a lucrative market involving thousands of species. Well-managed trade can be sustainable, but monitoring and regulation are essential, as unsustainable and illegal trade can threaten species, affect the livelihoods of communities dependent upon those resources, and strengthen organized criminal networks (Haken 2011;South & Wyatt 2011). National and international policy may prohibit the trade in certain species, set export quotas, or restrict the source of wildlife permitted in trade. The primary international wildlife trade legislation is the Convention on the International Trade in Endangered Species of Wild Flora and Fauna (CITES), which regulates trade in over 35,000 species (CITES 2013). Species are listed by governments when trade poses a direct threat or adds pressure to vulnerable wild populations, or as a "lookalike" of a listed species, to aid enforcement of the Convention (CITES 2013). While CITES and corresponding national legislation can successfully control trade (Doukakis et al. 2012), the continued illicit trade in many listed species has called into question its effectiveness (Challender & MacMillan 2014). A better understanding of conservation rule-breaking and illegal activity is an important step toward encouraging compliance (Solomon et al. 2015), and improving knowledge on noncompliance has been noted as a priority to strengthen CITES (Phelps et al. 2010). However, social desirability and threat of prosecution can make people unwilling to discuss their involvement in noncompliance.
To identify potential drivers of noncompliance and provide recommendations for policy implementation, we present the first study of CITES noncompliance among traders and buyers in a specific international This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. wildlife trading community. We choose orchids as a case study, a group that comprises approximately 73% of CITES-listed species and 9.8% of Appendix I (CITES 2013), meaning any transnational trade of orchid species must comply with the Convention's rules (CITES 2004). Orchids are also protected from wild-collection and trade under national policy in many countries (e.g., Commonwealth of Australia 2015). Between 1997 and 2014, reported seizures of illegally traded orchids resulted in fines of up to $257,000 and prison sentences of up to 6 years, although punishments were reported in few cases (TRAFFIC 2014). Although orchid hybrids are top-sellers in mass-market trade (USDA 2015), there exist specialist collectors who demand rare species (Hinsley et al. 2015), which can lead to overcollection and trade that can threaten species (Thomas 2006). Highly demanded groups such as tropical Asian slipper orchids Paphiopedilum spp. are particularly at risk, with many species threatened with extinction due to collection or trade (IUCN 2015). These factors make the orchid trade analogous in many ways to other wildlife trades, such as those with parallel legal and illegal markets (e.g., fur: Wyatt 2009; reptiles: Auliya et al. 2016) and those with coexisting mainstream and specialist consumers (e.g., songbirds: Jepson & Ladle 2009).
We use an online questionnaire to gather data on knowledge and opinions of CITES, and prevalence of CITES noncompliance among orchid growers. As noncompliance is illegal, we use a specialized questioning method, the unmatched count technique (UCT) to encourage truthful reporting (Nuno & St John 2015). In UCT, respondents are presented with a list of statements and asked to report the total numbers that apply to them. This is either a control list of innocuous "non-sensitive" statements or a treatment list with the control items plus an additional "sensitive" statement ( Figure 1). In conservation, UCT has been used to study behaviors such as unauthorized hunting (Nuno et al. 2013) and forest resource use (Harrison et al. 2015). However, in order to assess suitability of different techniques, a better understanding of the limitations of these methods in different conservation contexts is essential but underresearched (Nuno & St John 2015). We employed both direct questions (DQs) and UCT to explore potential trade-offs between techniques, with a focus on statistical efficiency and comparison of study findings.

Methods
We designed pilot and final questionnaires using www.SurveyGizmo.com, which were translated into French, German, Indonesian, Japanese, Malaysian, and Spanish and checked by native speakers with orchid expertise. Final survey links were emailed to all hobbyist societies listed in the American Orchid Society and British Orchid Grower's Association 2014 Directories, and to national or regional hobbyist organizations in Australia, Canada, Europe, Japan, Latin America, South Africa, and Southeast Asia. The pilot survey was sent to a subset of societies.

Pilot study
UCT lists should be carefully constructed, as poorly designed control lists can cause bias (Glynn 2013). A mixture of high and low prevalence statements reduces the likelihood of respondents agreeing with all or zero items, an outcome that removes the protection that UCT provides and decreases the likelihood of truthful answers by rule-breakers (Droitcour et al. 1991). Further, negatively associated statements reduce the variability of answers, increasing statistical efficiency (Glynn 2013). We piloted 32 control statements to assess their prevalence (see Supporting Information); all were related to orchid growing to ensure that the sensitive statement did not stand out (Glynn 2013). We asked respondents to select all true statements and provide feedback. Statement order was randomized for each respondent to avoid presentation order bias. We calculated the prevalence of each statement (% of respondents selecting it) and association between all pairs of statements, using chi-squared and odds ratio tests.

Main survey questions
Based on observations of the orchid community and consultation with the UK CITES Scientific Authority, we constructed statements to investigate four sensitive behaviors: r Smuggling ("I have personally sent or carried an orchid across an international border without obtaining the required CITES paperwork"); r Laundering ("I have personally sent or carried an orchid across an international border using the wrong CITES paperwork for that plant [e.g. paperwork for a different species]"); r Buying online ("I have bought an orchid online that was sent to me without the correct or required CITES paperwork"); r Wild plants ("I have an orchid in my collection that I know or strongly suspect was wild-collected").
These range in sensitivity from active CITES rulebreaking ("smuggling" and "laundering") to passive Please read the following statements and tell us how many are true for you. You do not need to tell us which statements are true for you, just the total number. ("buying online"), and finally socially undesirable but not necessarily illegal ("wild plants"). We included these statements in four UCT treatment lists ( Figure 1). Respondents answered all four questions but were randomly assigned by SurveyGizmo to either the control or treatment list for each. All respondents then answered each sensitive statement in a DQ.
Respondents were asked to rate their knowledge of CITES orchid rules on a five-point Likert scale, and presented with an open-text question about their opinions on the efficacy of the CITES for orchid conservation (see Supporting Information). All UCT questions required an answer before respondents could continue the survey; other questions could be skipped but were not marked as optional.

Analysis
We analyzed data using R version 3.2.1 (R Core Team, 2015) and, unless specified, the list package version 8.0 (Blair & Imai 2010) designed specifically to analyze UCT.
We used Blair & Imai's (2010) test for design effects to investigate the assumptions of the UCT that the presence of the sensitive item does not influence answers to the control items, and that the treatment group is randomly assigned.
For each behavior, we calculated a prevalence estimate for DQ (proportion of people admitting to behavior) and UCT (difference in mean between treatment and control groups). Associations between answers to each pairwise Table 1 Pilot and final results for the control list of each of the four UCTs, showing list mean and proportion of respondents giving each answer (e.g., "4 of the statements apply to me") List mean % "0" % "1" % "2" % "3" % "4" DQ combination were calculated using chi-squared and odds ratio tests.
To estimate the prevalence of sensitive behavior as a function of respondent characteristics, we fitted logistic regression models to the DQ response and ordinary linear models to the UCT score. Demographic variables and self-assessed knowledge scores were included as potential covariates, while interactions of the group variable (treatment or control) with each potential covariate were also included for UCT models. We selected, ranked, and averaged the most parsimonious models (with corrected Akaike's information criterion: AICc) using the MuMin package v.1.13.4 (Barton 2015), considering only models with interactions for the UCT. Models with AICc <4 were used for final model averaging (Burnham & Anderson 2002). The frequency of each self-knowledge level was calculated and opinion statements were manually categorized into "positive," "negative," and "neutral."

Pilot study
We received 409 completed pilot surveys, mainly from hobbyists (86.6%), with a small number of casual (14.9%) and professional growers (5.6%) in the UK (31.1%), United States (22.1%), and Japan (18.6%). Statement prevalence ranged from 88.0% ("I use fertiliser on my orchids") to 1.2% ("I have been growing orchids for less than one year"), and 102 statement pairs showed significant negative association (see Supporting Information). Four UCT lists were constructed with means between 1.6 and 1.9 for use in the final experiment ( Table 1). The control statement list for the "laundering" UCT demonstrates the combination of low and high prevalence, and negatively and positively associated statements (Table 2).

Main study
In total, 1,354 people started the survey, with most survey abandons occurring at the first UCT (n = 144) or the DQ page (n = 129). We used data from respondents who completed all UCT, DQ, and main demographic questions (n = 814) for the multivariate analyses, of which 56.4% (n = 459) were males and the mean age was 60.2 (SD 14.1) (see Supporting Information).
No UCT showed design effects and three had random assignment; the "buying online" UCT had significantly more males in the treatment group (P < 0.01) (see Supporting Information). Strong positive associations between answers to all DQ pairs (all P < 0.01) suggested the same respondents broke several rules, rather than rulebreaking being widespread throughout the sample. The DQ found that 9.9% of respondents had smuggled, 4.8% had laundered, and 10.8% had received orchids from an online purchase without required paperwork. UCT estimates did not differ significantly (Figure 2). In the multivariate analysis with the exception of "Latin America," "Paphiopedilum," and "Phragmipedium," all covariates were significant predictors of answers to at least one DQ (Table 3). The UCT produced fewer significant covariates for all behaviors, even at 90% confidence intervals (Table 4). Australian respondents were more likely to admit to smuggling via UCT than DQ.
Of 893 respondents who self-rated their CITES knowledge, 38.7% (n = 346) rated it as "good" or "complete," and 40.8% (n = 264) as "no" or "some" (see Supporting Information). Of respondents who gave opinions regarding CITES (n = 649), 61.2% (n = 397) reported negative and 21.0% (n = 136) reported positive views. Positive opinions were generally short with little detail (e.g.,    Shading denotes significance, all at P 0.05 except those marked * , which are significant at 0.05 ࣘ P ࣘ 0.1NA = not included in top models. Reference levels: U.S.; < median knowledge; no experience; male; < median genera; no Phragmipedium; nonprofessional; no Paphiopedilum; nonhobbyist; < median age (see Supporting Information for relative variable importance and confidence intervals).
"Good" or "Adequate"), and 14.0% (n = 19) expressed support of CITES aims but negative views of its current application. Negative statements were generally detailed, allowing them to be split into themes, the most frequent being "CITES hampers orchid conservation" (26.7% of negative statements), e.g., "Orchids could be rescued when their habitat is cut down but CITES won't allow this." Others included "Too strict" (22.2%), "Too complicated" (21.4%), and "Not enforced uniformly" (18.9%). In addition, 9.3% (n = 37) of negative statements alluded to the ease of noncompliance, e.g., "Stupid, counter-productive and generally ignored or got around."

Discussion
This study represents the first in-depth investigation into CITES noncompliance among a specific international wildlife trading community, using a case-study of 606 horticultural plants, illegal trade of which is widespread but largely ignored (Phelps & Webb 2015). Understanding noncompliance with conservation rules is a priority (Solomon et al. 2015), providing information that can inform policy and assist in encouraging compliance (St John et al. 2011).

Utility of the UCT
The UCT has found increasing conservation application (e.g., Nuno et al. 2013), out-performing DQs and specialized methods (e.g., randomized response technique) when used face-to-face (Glynn 2013). Our UCT estimates were not significantly higher than DQ, even at broadened confidence levels, possibly because self-complete online surveys offer greater anonymity than face-to-face administration (Holbrook & Krosnick 2010). In addition, widespread negative opinions of CITES in our sample may have reduced social desirability pressures to conceal rule-breaking. This highlights the need to better understand and consider the limitations of specialized questioning techniques, and recognize the utility of direct questioning when designing surveys. We call for further research into the limitations of these techniques, how their use affects findings, and contexts in which they are most appropriate.

Enforcement priorities
Strengthening CITES enforcement is a priority (Phelps et al. 2009), supported by little concern about the possibility of punishment for CITES noncompliance among our respondents. Rule-breaking in our sample was not widespread but focused on subgroups breaking several rules, supporting the recognized need for a targeted approach to enforcement (Res. Conf. 11.3 [Rev.CoP16]). We suggest that one priority for enforcement improvements should be online wildlife trade, which is growing and hard to control (Lavorgna 2014). The prevalence of noncompliant online purchases by professional growers in our sample is of particular concern as this group is likely to buy in higher volumes. As online orchid buyers have a preference for rarity (Hinsley et al. 2015) and threatened wild plants are sold widely online (Shirey et al. 2013), our results demonstrate that the internet may be allowing traders to bypass CITES rules for rare species.
In particular, enforcement efforts should focus on improving the detection of protected species at the point of sale on various platforms (Sajeva et al. 2013), and on intercepting orchids transported via the postal system, only one seizure of which has been reported to date (TRAFFIC 2014).
The role of consumer countries in preventing illegal trade has been recognized by CITES (Res. Conf. 11.3 [Rev. CoP16]), and our findings suggest that a second priority should be to identify enforcement gaps at the point of import. In particular, we found that noncompliance among EU respondents (excluding the UK) was significantly more likely, supporting findings that the EU is a destination for smuggled wildlife (Auliya et al. 2016), potentially due to variable enforcement between Members States (Reeve 2006). Specific weak points of entry to the EU for orchids and for other CITES taxa should be identified, and the best ways to address these enforcement gaps assessed.

Engaging traders and end-consumers in policy decisions
While increased enforcement should be a priority, many countries do not have the capacity to adequately enforce wildlife trade legislation (Phelps et al. 2009) and we note that enforcement alone may not automatically reduce illegal trade (Challender & MacMillan 2014). Reluctance of Australasian respondents in our study to admit to smuggling via DQ could be linked to the region's strict enforcement of wildlife trade rules (e.g., Commonwealth of Australia 2015). Although potentially a chance result, it may suggest that stricter enforcement increases secrecy about noncompliance, rather than preventing it from occurring completely. We therefore recommend that enforcement improvements are complemented with efforts to address the widespread disengagement and distrust of CITES among orchid growers. While greater knowledge of rules has been linked with higher compliance at a community level (Nkonya et al. 2008), we found that better knowledge of CITES rules was linked to more active forms of noncompliance, suggesting that simply raising awareness of the rules is unlikely to be effective. We suggest that noncompliance may be linked to widespread negative opinions of CITES. For example, those stating that CITES is bad for conservation may be using "neutralization," a process by which rule-breaking can be justified as being for the greater good (Sykes & Matza 1957). Similarly, small-scale rule-breaking can be used as a form of protest by those who feel powerless to change laws they disagree with (Scott 2008), a practice demonstrated in the case of wildlife poaching (Bell et al. 2007). If these processes are playing a role in rule-breaking, one potential solution may be for those most affected by CITES to be empowered by being more involved in decisions relating to its rules. This could be achieved by better engagement with horticultural networks and traders' associations.
While negative opinions may make them initially unwilling, encouraging the involvement of key stakeholders in these discussions may increase compliance and trust, strengthen legal businesses, and allow the reasons for noncompliance to be better understood. For example, we found that professionals were more likely than nonprofessionals to launder, possibly to bypass high charges such as those of £74 per plant genus for UK imports (UK Government 2013), even for orchids that have over 800 genera. As professionals were also more likely to own wild plants, this laundering could represent illegal movement of wild orchids as artificially propagated plants but is also likely to include plants of all sources transported under the wrong names. This may be undermining the important role that CITES plays in monitoring species in trade, and efforts should be made to understand how these charges may impact small businesses, and how this may affect traders' compliance behavior.
As it was self-selecting and administered online, our sample may have omitted countries without wellestablished orchid networks and internet access, suggesting that further studies of consumer behavior and compliance would be of benefit. Our recommendations for improving compliance are likely to apply to the trade in other wildlife products, particularly those with both legal and illegal trades, and those with subgroups of specialist consumers. In addition, our findings highlight that knowledge of the behavior and motivations of consumers and traders in any wildlife market can provide information about the effectiveness of current approaches, and inform recommendations to strengthen compliance.