We assessed the intended behavior of snorkelers and divers using a Discrete Choice Experiment (DCE), which is a multivariate method that aims to identify the drivers of an individual's choice behavior - in this case, their choice of hypothetical coral reefs to visit. Random utility theory (RUT) is the basis of the DCE approach. It postulates that the total “utility” (i.e., sense of satisfaction) of a given alternative is a function of its deterministic and random components (McFadden, 1974; Ben-Akiva and Lerman, 1985). Based on RUT, the utility of a good chosen by individual n can be described by the function U_in = V_in + ε_in, where U_in is the overall utility of a good i, which is composed of V_in, a deterministic vector of attributes, and ε_in, the random component of an individual's choice. An alternative i is chosen over alternative j if U_in > U_jn for all j ≠ i. In other words, RUT assumes that individuals always act to maximize their utility or sense of satisfaction. We first defined the snorkeling or diving experience in Cozumel as the overall good. We then selected an array of environmental, economic and management attributes, with their respective levels, to create different reef profiles. Different combinations of scenarios with varying attribute levels constituted the choice sets presented to respondents.

We created coral reef profiles and choice sets by using a 4²⁰ orthogonal fractional factorial design, which allowed the systematic variation of all coral reef attribute levels in the choice sets (see Raktoe et al., 1981). The DCE consisted of 120 scenarios in 60 choice sets. Each scenario contained eight attributes presented in text or photographs, and each attribute had one of either two or four levels. The attributes described ecologically realistic aspects of the reef environment based on empirical data from the invaded range. Specifically, the attributes presented in digitally calibrated photographs included: lionfish density, native grouper density, native prey fish densities, reef relief, percentage coral cover, and the presence or absence of a Caribbean reef shark. Two attributes were presented in the text: a hypothetical lionfish control fee (in USD, to assess tourist willingness to pay the cost of removing lionfish), and percentage change in a hypothetical excursion price (Table 1). We digitally calibrated photographs of coral reef attributes in Adobe Photoshop CS5 (Table 1, Figure S1). Digitally calibrated images are powerful tools to represent complex systems (Orland et al., 2001). They have been used repeatedly in choice experiments (e.g., Arnberger and Haider, 2007; Landauer et al., 2012; Ryffel et al., 2014), and they present choice attribute levels in a format that is easy for respondents to evaluate (Bateman et al., 2009). During DCE construction we excluded scenarios that contained ecologically unrealistic combinations of attributes such as reefs with very high densities of lionfish and native predators, but very few prey fish, low coral cover and low vertical relief. We also always presented choice sets where the alternative with the most expensive control fee included at least one high-value of an ecologically desirable attribute (e.g., high native predator density) to ensure that the scenarios were realistic (Reed Johnson et al., 2013). Our final experimental design had a D-efficiency score of 87.95.

Our survey consisted of four main sections: (1) questions to assess satisfaction with the snorkeling/diving excursion; (2) questions to assess snorkeling/diving experience as well as environmental and lionfish invasion awareness; (3) questions to establish the socio-demographic characteristics of respondents; and (4) the DCE to determine tourist reef preferences under different lionfish invasion scenarios. Prior to conducting our study, we tested (and subsequently modified) our survey questions and DCE in focus groups comprising coral reef ecologists, recreational dive shop owners, and tourist divers. We administered the DCE to respondents on electronic tablets using a web-based survey (fluidsurveys.com). We selected only adult respondents (i.e., 18 years or older) and administered only one survey per group to minimize pseudoreplication. We approached respondents haphazardly at the Cozumel International Airport in summer, and at dive stores, and dive tour docks in winter, immediately after their return from a diving/snorkeling outing. We asked whether respondents had engaged in snorkeling or diving in Cozumel, and people responding in the affirmative were invited to participate in our questionnaire. Each choice set presented consisted of a pair of hypothetical coral reef locations (Figure S1). Each choice set also included the option “not to dive on either reef if these were the only two reef sites available,” to estimate potential economic losses to the local snorkeling and diving industry. We asked respondents to choose one of the two locations described or neither of them. Each respondent evaluated six choice sets.

Avidity bias is important to consider when studying visit behavior or recreational demand since probability distributions might differ between more frequent users and the general population (Moeltner and Shonkwiler, 2005; Hynes and Greene, 2013). We did not specifically ask respondents how many times they had visited Cozumel in the past. However, we did ask whether they had been to Cozumel before the arrival of lionfish in 2009. Approximately 39% of respondents had done so. This hints at a high return rate of snorkelers and divers to Cozumel, but is in line with the high level of loyalty (i.e., >40% repeat visitors) reported by the Ministry of Tourism for Cozumel tourists in general (SECTUR - Secretaría de Turismo, 2010).

Based on RUT, it is possible to explain the behavior of respondents by estimating the probability of choosing alternative i over j for utility V:

where C is the set of all possible alternatives. Choice models can be analyzed using a multinomial logit model (MNL; McFadden, 1974; Boxall and Adamowicz, 2002; Train, 2003) to produce regression estimates (i.e., part-worth utilities) for each attribute, which when combined represent respondent choice probability P as a whole:

where (M) indicates all scenarios present, X is the vector of explanatory variables, and β is the parameter vector to be estimated. We followed a mixed logit form extension of the MNL, the latent-class model (LCM), which explains preference heterogeneity in respondent choice. In latent-class models it is possible to estimate separate sets of choice model parameters for each latent class c, which will account for preference heterogeneity within one statistical model:

where β is the class-specific vector of the jth alternative, chosen among J_i alternatives by individual n observed in T_i choice situations, i and j (Equation 1), M and X (Equation 2) are defined above (see Greene and Hensher, 2003; Morey et al., 2006 for a detailed explanation of LCM). Thus, the latent-class model divides the sample into classes characterized by relatively homogeneous within-class preferences, assessing the probability that individuals belong to a certain class as a function of their unobserved social, attitudinal and motivational characteristics (Birol et al., 2006).

We used the software Latent GOLD 4.5 (Vermunt and Magidson, 2005) to analyze the DCE results using a LCM analysis following the three-file system procedure. Given the need to better understand how management strategies affect different user groups (Légaré and Haider, 2008), we analyzed the DCE responses using (1) a MNL on all responses (Table S1), (2) latent-class models LCM, exploring different number of classes (however models >2 classes did not converge so were not explored further; Table S2) and (3) a modified latent-class analysis (Table S3) that included two known classes defined a priori (i.e., snorkelers and divers) because of expected differences in motivation and behavior (Vermunt, 2003, 2008). We will refer to the latter model as the segmentation model. During model exploration, we used effects coding to examine the levels for all attributes, and we also examined the linear effects of continuous attributes (i.e., grouper and prey densities, reef relief, and excursion price). The latter was done because levels of these continuous variables were not equidistant. To linearize attribute levels, we centered each level by subtracting from it the overall mean level value for that attribute, and then divided the centered value by the average interval between each pair of successive levels (Table 1). Lionfish density and control fee were not linearized because we were specifically interested in the effect of each level of these two attributes, rather than in the average effect across attribute levels. Given that control fee values could be correlated with lionfish numbers or with the reef attributes likely to be affected by lionfish management, we explored solutions with interacting attributes; however, these resulted in non-converging, uninformative models, which were not explored further. We used the BIC (Bayesian Information Criterion), and AIC (Akaike Information Criterion) as information criteria indicators to select the most parsimonious model, which is reflected by the lowest values of all of these indicators (Akaike, 1974; Vermunt and Magidson, 2005, p. 88). The segmentation model with the lowest information criteria indicators included linearized attribute levels.

Using Latent Gold, we then added to the most parsimonious model 15 covariates that could explain class membership. We obtained these covariates from survey Sections 1–3. Sociodemographic covariates included general demographic information such as gender, country of residence, age groups (in three categories: 19–35, 35–50, and >50 years), yearly income (<$20, $20–39, $40–59, $60–79, $80–99, $100–149, $150–250, and >$250, figures in thousands USD) and education (high school or less, trade technical or college education, university degree and postgraduate degree). We also explored covariates reflecting motivation and engagement in snorkeling or diving (i.e., number of snorkeling/diving trips per year, certification level, total number of logged dives, number of tropical regions of the world visited to snorkel/dive, ownership of snorkeling or diving equipment), and commitment to snorkeling/diving (assessed through a question asking the annual frequency and engagement level with snorkeling/diving). We explored perception covariates such as owning field identification guides, knowledge and involvement in environmental causes, awareness of invasive species and impacts, and awareness of the lionfish invasion and related impacts. Lastly, we assessed tourist satisfaction with their trip to Cozumel using a five-point Likert scale. To minimize the number of dimensions, related covariates (e.g., motivation and engagement in snorkeling/diving) were incorporated as principal components derived from a Principal Component Analysis (PCA). We kept in our final model only significant explanatory covariates that helped us to describe class membership (Vermunt and Magidson, 2005). Once we obtained our final model, we identified significant differences in part-worth utilities between attribute levels by interpreting z-values, and between classes by using Wald II statistics (Vermunt and Magidson, 2005, p. 88).

Differences between the socio-demographic characteristics of respondents surveyed in the summer and winter months were assessed using unpaired t-tests. We evaluated the significance of these differences against a Bonferroni-adjusted critical alpha level of 0.014 (unadjusted alpha of 0.1/7 tests performed) to account for multiple testing (Armstrong, 2014) in R version 3.3.2 (R Core Team, 2016).

We used the results of the DCE and segmentation model analysis to develop a Decision Support Tool (DST), in which the demand for different scenarios could be estimated for each tourist group based on segmentation model parameters. Such a tool can be used to estimate tourist support for hypothetical management scenarios—in this case, we modeled tourist support for various options to control lionfish density. To do this, we used the part-worth utility values from each attribute for each class in Equation (2) and then estimated the choice probability for a given class under a given scenario. The resulting values are probabilities interpreted as percentage of support by a class for a scenario (i.e., a given set of attribute levels), which we refer to as market shares (Hensher et al., 2005; Vermunt and Magidson, 2005). We structured our DST in the same way as the choice sets in the survey, thus estimating the probability of choice between two hypothetical alternatives (e.g., status quo and management actions) and neither of the two (i.e., dive elsewhere). By changing the levels for relevant attributes in each of the scenarios, one can calculate the likelihood of choice for many possible situations that could arise as lionfish continue to invade the region and management actions are implemented or not. Changes in probabilities based on the different part-worth utilities of two management scenarios result in changes in class market shares, which are then interpreted as changes in demand by a class for a given scenario.

For our DST analysis, we assessed how the different classes of tourists reacted to four hypothetical lionfish invasion scenarios (Table 2). Scenario 1 was a status quo scenario, with coral reef and lionfish status as seen in Cozumel in 2014; Scenario 2 (short-term, no management) depicted coral reef and lionfish status expected after 2 years with no lionfish management; Scenario 3 (long-term, no management) considered coral reef and lionfish status expected after 5 years with no lionfish management; and Scenario 4 showed coral reef and lionfish status if management actions were in place to control lionfish.

We estimated the short and long-term no-management scenarios based on reported impacts of lionfish in The Bahamas by Green et al. (2014) and Smith et al. (in press). We constructed the management scenario (Scenario 4) based on tourist preferences toward the smallest fee tested ($ 5 USD; see Results section), and on the assumption that this scenario would produce a substantial reduction in numbers of lionfish (i.e., we assumed lowest lionfish density tested on DCE) and positive effects on native prey and predator fishes (i.e., the intermediate prey and grouper densities tested on DCE) (Table 2). These effects, at least in terms of lionfish and native prey numbers, are consistent with empirical results of experimental lionfish removals by Green et al. (2014). Analysis of the four scenarios allowed us to compare tourist reaction to lionfish control fees being implemented or in consideration elsewhere in the Wider Caribbean region. We maintained reef relief value (1.8), percentage of coral cover (5–15%), reef shark absent, and no change in excursion price (i.e., snorkeling or diving trip) constant across the four scenarios. We performed a sensitivity analyses to assess how variation in these four attributes affected class-specific market shares when other attributes of interest were kept constant (Figure S2). We acknowledge that other management scenarios (e.g., the establishment of no-fishing areas, mooring buoys, coral aquaculture, etc.) could benefit coral reef attributes; however, given our lionfish management focus we did not explore these possibilities. All DST estimates were calculated in Excel.

We collected a total of 312 surveys, with a response rate of 74 and 72% in summer and winter, respectively. Of these, 302 respondents fully completed the DCE section. Snorkelers and divers surveyed in the summer and winter months did not differ in age, income, education level, motivational attitudes, commitment to snorkeling/diving, trip satisfaction, environmental awareness or awareness of the lionfish invasion (Bonferroni-adjusted alpha level: 0.014; Table S4). Respondents were 42 years old, on average (range: 19 to 70 year), and were mainly male (70%). The majority (84%) of respondents resided in the USA, with 7% hailing from Canada, 4% from Mexico, 3% from European countries, and the remaining 2% from other countries. The income of most respondents (40%) ranged from $60-100,000, with 24% having incomes above $100,000, and 17% below $60,000; the remaining 19% did not know or chose not to disclose their income. The majority of respondents (77%) held at least one university degree. The demographic profile of our respondents appears to be representative of Mexican Caribbean tourists (Anaya-Ortíz and Palafox-Muñoz, 2010; Güemes-Ricalde and Correa-Ruíz, 2010; SECTUR - Secretaría de Turismo, 2010; Table S5).

We obtained a three-class model for marine tourists visiting Cozumel (Table 3). It included an a priori identification of a snorkeling class, and divided divers in two latent classes: casual and committed divers (Table 4). Members of these three classes differed in age, outdoor activity commitment, and awareness of the lionfish invasion and its related impacts (Table 5). The other eight covariates examined did not differ systematically among classes. Part-worth utilities indicated that all but two of the reef attributes had an effect on excursion utility for all three classes. The exceptions are native prey density, which was important for committed divers only, and changes in excursion price, which had an effect on snorkelers only. Wald II statistics indicated that lionfish density, lionfish control fee, and percentage coral cover preferences were different among the three classes (Table 4). Casual divers differed significantly from committed divers in terms of their attitude toward lionfish presence on the reefs and willingness to pay control fees (see below), while snorkelers and casual divers behaved similarly to each other (Table 4).

Exploring the segmentation model results through a decision-support tool (DST) helped us to identify tourist class preferences for different lionfish management scenarios (Table 2). Market shares are based on class size (i.e., 33, 44, and 23% for snorkelers, casual and committed divers, respectively). The status quo scenario always represents the conditions that visitors could experience on Cozumel reefs in 2014.

In the absence of management action taken in the short term (i.e., two years), we would expect a large increase in lionfish density and a moderate decrease in native prey fishes (Table 2). These changes result in decreases in market shares for all three classes of marine tourists, compared to the status quo scenario (Figure 2A). The market share of snorkelers would decrease by 7%—the largest decrease in market share—while those of casual and committed divers would decrease by 5 and 4%, respectively (Figure 2A). Committed divers would experience a market share loss of 1% (i.e., 1% of divers would prefer not to dive in Cozumel at all; Figure 2A).

If the lionfish invasion were allowed to progress unchecked for a longer period (i.e. ~5 years.), a further moderate increase in lionfish abundance and now severe declines in native prey and grouper densities would be expected (Table 2). Market shares for snorkelers and committed divers would decrease by 7 and 19%, respectively, compared to the status quo scenario, with a market share loss of ~2% for committed divers (Figure 2B). On the other hand, casual divers show a small increase in market share (~2%; Figure 2B).

If a small control fee ($5 USD) were implemented to aid management actions, which would prevent increases in lionfish densities (compared to the status quo year) and minimize impacts on native prey and predator densities (Table 2), market shares for all three classes would increase when compared to the status quo scenario. Snorkelers and casual divers market shares would increase by 2 and 6%, respectively, while an increase of nearly 7% would be observed for the committed divers class (Figure 2C). However, while implementing a higher control fee would benefit the reef environment even more and increase the market shares of committed divers, it would decrease the market shares of snorkelers and casual divers, given their unwillingness to pay >$5 USD per visit to control lionfish numbers (Figure S2).

Tourist groups visiting Cozumel reefs differed greatly in their reaction to lionfish. Our study shows that seeing at least one lionfish adds to the experience of snorkelers and casual divers (Table 4, Figure 1). This reaction is expected from tourists seeking to enjoy natural attractions when initially faced with an arguably beautiful and exotic fish (Moore, 2012; Hoag, 2014). In contrast, committed divers showed great aversion to lionfish, even in low numbers, and preferred reefs with no lionfish.

In spite of a key divergence in preference in relation to lionfish, snorkelers and divers shared an overall preference for reefs in good condition. This result is consistent with previous studies showing that various attributes correlated with reef state are important for snorkelers and divers (Rudd and Tupper, 2002; Uyarra et al., 2005; Dinsdale and Fenton, 2006; Shideler and Pierce, 2016). Indeed, snorkelers and casual divers value reef attributes such as high coral cover, high reef relief and high grouper abundance as much as, or more than, they value lionfish presence (Table 4). Perhaps not surprisingly, the preference for higher levels of all natural reef attributes was stronger for committed divers than for the other two classes. The non-significant attraction or dislike of some of the tourist classes to attributes such as changes in native prey density, reef shark presence and excursion price change might arise because of opposing attitudes within a class. For example, a subgroup of snorkelers might like to see sharks during their visit while the rest of their class might fear such encounters (Dobson, 2007). It is possible that our sample size prevented us from identifying subgroups of snorkelers, which comprised a smaller class (33% of respondents) than divers.

The divergent preferences of different classes of marine tourists covaried with demographic and motivation characteristics. Environmental attitudes have been found to vary with age (Bremner and Park, 2007; Lee, 2011; Sharp et al., 2011), involvement in outdoor activities (Luo and Deng, 2007; Nisbet et al., 2009), and environmental awareness (Luo and Deng, 2007; Peters and Hawkins, 2009). Similar factors were useful to define our model and explain class divisions in our study. We expected to observe latent classes mainly as a result of diver specialization, as reflected by experience level and commitment to the diving activity (Dearden et al., 2007; Anderson and Loomis, 2011). Not surprisingly, commitment to snorkeling or diving did explain class membership, but age and knowledge of the lionfish invasion also contributed. These three covariates were useful to distinguish between snorkelers and committed divers, but were not useful to define casual divers. It is possible that the casual diver class was not well described because respondents within this class included people at the extremes of the covariates tested, but the low sample size prevented us from identifying further subgroups.

Tourists engaging in environmental activities are often sensitive to excursion prices (Dellaert and Lindberg, 2003; Saayman and Saayman, 2014). We found that this was not the case for divers. Only snorkelers reacted to changes in excursion price, stating a dislike for increases. It is possible that the sensitivity of snorkelers to higher excursion prices is a result of their average trip price, which is lower (~$50 USD) than that of divers (~$100 USD) (see below). While this finding is not directly relevant to the issue of lionfish control, it can have implications for the willingness of snorkelers to contribute to management actions that require financial contributions from users.

Individuals and groups who are actively involved in outdoor activities often display strong support for conservation initiatives. People who spend more time doing outdoor activities, such as visiting nature reserves or interacting with wildlife, generally show more environmental concern and endorse pro-environmental attitudes (Luo and Deng, 2007; Sorice et al., 2007; Nisbet et al., 2009; Semeniuk et al., 2009; Lee, 2011). This stance extends to the control of terrestrial invasive species. In the USA (Sharp et al., 2011), Scotland (Bremner and Park, 2007), and Spain (García-Llorente et al., 2011), individuals who are environmentally engaged, aware of the impact of invasive species and/or familiar users of terrestrial parks strongly support management actions against invasive species. In general, WTP and DCE studies focused on marine tourism indicate divers and snorkelers value healthy ecosystem attributes highly (Rudd and Tupper, 2002; Uyarra et al., 2005; Dinsdale and Fenton, 2006; Shideler and Pierce, 2016), and readily support the implementation of conservation management initiatives and management fees (Depondt and Green, 2006; Casey et al., 2010; Emang et al., 2016). Our results show that this attitude also prevails in relation to marine invasive species, at least among some marine tourists.

Managing marine invasive species can be expensive (Bax et al., 2003; Williams and Grosholz, 2008). Culling is currently the most common form of lionfish control within the region (Malpica-Cruz et al., 2016), and it can effectively decrease lionfish abundances and limit their ecological impacts at local scales in some situations (e.g., Frazer et al., 2012; Green et al., 2014; but see Dahl et al., 2016). However, these interventions must be maintained over the long term to prevent lionfish populations from rebounding (Arias-Gonzalez et al., 2011; Barbour et al., 2011). Furthermore, culling is likely to be ineffective at large spatial scales and at depths beyond recreational diving limits (Andradi-Brown et al., 2017). Even if lionfish removals were limited to small, spatially discrete areas such as shallow-water coral reef patches, culling is time- and labor-intensive (Usseglio et al., 2017). As such, culling is likely to pose an undue financial burden on marine resource managers, unless sources of sustainable financing are identified.

Our results suggest that some reef tourists would be willing to contribute to lionfish management. Committed divers, in particular, supported high lionfish control fees (US$10–15 per excursion) and disliked options without such fees. This support for control fee and aversion to lionfish presence on reefs by committed divers closely aligns with their keen participation in lionfish derbies—competitive events occurring throughout the Caribbean in which participants can gain monetary or material prizes for capturing lionfish (Malpica-Cruz et al., 2016). The majority of divers participating in these events are aware of the impacts of lionfish and willing to invest time and money in their management (Ali et al., 2013; Hoag, 2014; Trotta, 2014). In contrast, casual divers disliked high lionfish control fees and only supported the smallest fee (i.e., $5 USD). Snorkelers were the least supportive of implementing a control fee, perhaps because such a fee would represent a larger proportion of their total excursion price than it would for divers. It is worth noting that, given the strength of the preferences stated for various reef attributes, the support of snorkelers and casual divers for relatively low control fees appears to be driven more by the beneficial ecological effects of controlling lionfish on reefs, rather than by direct reductions in lionfish numbers. Nevertheless, the positive attitude toward a $5 USD control fee by snorkelers and casual divers, and the keen acceptance of higher fees by committed divers, indicates the potential willingness of many marine users to contribute financially to lionfish control. These findings add to the notion that snorkelers and divers are generally willing to contribute financially to the preservation and conservation of reef environments in marine protected areas (MPAs) (Arin and Kramer, 2002; Green and Donnelly, 2003; Depondt and Green, 2006), and to management actions to restore damaged reef ecosystems (Seenprachawong, 2003; Beharry-Borg and Scarpa, 2010).

To our knowledge, the possibility of charging a lionfish control fee to marine tourists visiting Caribbean marine protected areas (MPAs) has not been explored. Marine reserves and protected areas in the Caribbean are often poorly managed and have limited budgets to regulate recreational snorkeling and diving operations (Bustamante et al., 2014). A lionfish control fee would provide valuable additional financial resources that could be used, for example, to implement lionfish surveys and monitor the state of the invasion, undertake periodic removals in key locations within MPAs (e.g., core locations of high biological diversity) where large-scale derbies are not feasible, and mount awareness campaigns. While all tourists entering Cozumel Reefs National Park already pay a daily visitor fee (~$1.5 USD), these funds are distributed at the federal level among all terrestrial and marine protected areas in Mexico, yielding minimal funding for lionfish-focused interventions (Comisión Nacional de áreas Naturales Protegidas, 2010). Rough estimates indicate that the funds raised through a lionfish control fee in Cozumel could be substantial. In 2014 ~300,000 people paid a visitor fee to enter Cozumel Reefs National Park to take part in aquatic activities (Blanca Quiroga García, pers. comm.). If just 20% of these visitors engaged in snorkeling or diving, an acceptable, modest ($5 USD) lionfish control fee would result in $300,000 USD annually to be used for lionfish management. This estimate is equivalent to 50% of the authorized federal budget in 2014 for biological monitoring programs in all protected areas of Mexico (Comisión Nacional de áreas Naturales Protegidas, 2016).

Marine tourism is arguably the most important economic activity in Cozumel. In 2012, 41% of the roughly 4 million tourists visitors to Cozumel indicated an interest in aquatic activities, 75% of which entailed diving or snorkeling (Mota and Frausto, 2014). There is no specific information available on the relative economic contributions of snorkelers and divers in Cozumel. All tourists incur accommodation and meal expenses during their 3-day (average) stay on the island, but the average excursion costs of snorkelers and divers differ ($50 vs. $100 per trip, respectively). The fact that divers predominated (67% of respondents) in our sample, combined with the high diving excursion prices, suggests that factors that discourage visits by divers might have a larger impact on the industry than those that reduce appeal to snorkelers, if our sample is representative of the whole industry.

Our study suggests that not implementing actions to mitigate the lionfish invasion could change the distribution of market shares of divers and snorkelers, with potential negative effects on the economy. In general, marine tourists across all three classes preferred the status quo (2014) reefs than reefs that remained unmanaged in the short and longer term. Lower market shares for unmanaged reefs are partly driven by a lionfish effect: most marine tourists tend to value seeing a few lionfish (e.g., 1 lionfish, in the status quo scenario) more than a higher abundance of lionfish (i.e., 10 lionfish in the 2-year no-management scenario, or 25 lionfish in the 5-year no-management scenario; Figure 1). The single deviation from the overall trend—the higher market share of casual divers for the 5-year no-management scenario—is consistent with the high value they place on abundant lionfish (Figure 1). The market share patterns are also partly attributable to a native fish effect: marine tourists tend to prefer the higher abundances of native groupers (and for committed divers only, native prey fish) in the status quo scenario than the lower abundances offered in the unmanaged scenarios. Unsurprisingly, committed divers showed the largest aversion to reefs that are unmanaged in the long term, with ~20% fewer divers preferring those reefs than the status quo reefs (Figure 2B), reflecting the strength of their preferences for reef attributes that reflect good condition. It was also the only class of users that would choose not to dive in Cozumel when lionfish impacts are large. Such a potential loss could be amplified if dissatisfied experienced divers, who focus on the environmental aspects of their visit, become less likely to recommend Cozumel as a destination to others (Chi, 2010; Morais and Lin, 2010).

Interestingly, more marine tourists of all three classes would prefer to visit managed than status quo reefs. Our DST indicates that the driving force of these higher market shares is the control fee itself. Grouper density did not change between the managed (fee-paying) and the status quo scenarios. Small prey fish density decreased, but snorkelers and casual divers—unlike committed divers—are relatively insensitive to variation in abundance of small native fish. However, across all three classes of marine users, the preference for $5 was higher (or, in the case of committed divers, the dislike was less intense) than the preference for no fee (Table 4). Given that more than two-thirds of respondents preferred lionfish on the reefs, their acceptance of even a small lionfish control fee seems counterintuitive. Nonetheless, this result might reflect the fact that snorkelers and casual divers, like many other tourists regardless of their outdoor involvement, are willing to contribute financially to conservation initiatives (Casey et al., 2010). However, given that snorkelers and casual divers do not tolerate control fees >$5 USD, the implementation of higher fees would be likely to reduce market shares, and perhaps increase market losses, of tourists visiting Cozumel reefs.

It is important to note that the market share changes described here as a result of lionfish management (or lack thereof) cannot be safely extrapolated to predict economic impacts on the Cozumel tourism industry. Our results are based on estimates of marine tourist preferences for various reef features and on potential impacts of the lionfish invasion on these attributes. Tourist preferences could change over time, and the potential impacts of lionfish may not come to pass as expected. We also cannot assess potential avidity bias, and hence the extent to which our pool of respondents represents “average” visitors to Cozumel. Lack of representativeness would weaken any attempt to forecast real economic impacts (Moeltner and Shonkwiler, 2005; Hynes and Greene, 2013). However, the high visitor return rate of our respondents is in line with official tourism statistics (SECTUR - Secretaría de Turismo, 2010), which suggests that our estimated market share changes might be realistic. Finally, local and global factors affecting tourism market growth, tourist return rates, travel security concerns, etc., which are beyond the scope of this study, will also affect market share changes such that the overall tourism industry might grow or shrink independently of coral reef state or management.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.