2. Wastewater management in San Andres

We conducted our study in SAI, the largest and most populated island of the Archipelago of San Andrés, Providencia and Santa Catalina (Colombia), with an area of 27 square km and a permanent population of about 62,000 inhabitants according to the latest census in 2018 (DANE, 2020). SAI hosts about one million tourists annually, most of whom are Colombians (James et al., Reference James, Saleme, Romero, Forbes, Blandon, Stephens, Bravo, Espitia, Torres, Castillo and Cantillo2022). Tourism recovered quickly after the pandemic, reaching the highest number of visitors in 2022, almost 1.34 million (see online appendix figure A1). In the months when this study was conducted, August–September 2022, 220,788 people visited the island, 83.1 per cent of whom were Colombians (Secretary of Tourism, personal communication, 2024).

SAI's urban center is located in the north end of the island and contains the tourism infrastructure that provides the main source of employment and economic activity on the island. Southern areas of the island are less developed (Gavio et al., Reference Gavio, Palmer-Cantillo and Mancera2010) and mostly populated by native communities known as Raizals (Londoño and González, Reference Londoño and González2017). The urban center has separate systems to collect domestic wastewater and rainwater; those systems are operated by the private company Veolia Aguas del Archipiélago S.A.S E.S.P and the local government, respectively. However, according to the last census conducted in 2018, only 17.3 per cent of households in SAI and 16.6 per cent in the archipelago are connected to a sewer system (DANE-ENHAB, 2020). In SAI, wastewater is collected by three stations and then disposed into the sea for its dilution through a submerged outfall (CDM Smith and INGESAM/FINDETER, 2017). Wastewater treatment is limited to separation and retention of floating large solids (e.g., plastic bags, textiles, trunks, among others) by self-cleaning grids located in one of the three sewerage stations. Wastewater is not chemically or biologically treated before being discharged into the ocean.

In rural areas where public sewage infrastructure is not available, hotels and other businesses use their own private wastewater treatment plants, including activated sludge technology that allows for the use of treated wastewater for toilets and irrigation of green areas. Solid waste isolated from wastewater is then disposed of in drying beds for subsequent usage as organic fertilizer (CDM Smith and INGESAM/FINDETER, 2017). Local communities do not have access to this kind of technology, and instead approximately 80 per cent of households in the island use septic tanks (DANE-ENHAB, 2020). However, these tanks tend to be built with low technical specifications and are poorly operated and maintained, failing to prevent soil and groundwater contamination (CORALINA, 2012). The rest of the households dispose of their wastewater directly into the ground or ocean if they are located in littoral zones.

As part of the Works and Investments Plan, included in the Directive Plan of Water Resources, CDM Smith and INGESAM/FINDETER (2017) proposed a series of projects aimed at collecting and treating the wastewater produced in SAI. First, in the urban center, the Plan proposed the renewal of 46 per cent of wastewater pipes (about 5 km) due to leakages and a network expansion of about 438 m with an estimated cost of approximately 18.4 billion (2017) Colombian pesos (2017 US$6.13 million). Second, in rural areas, the Plan proposed a wastewater pipe system of 35 km with an estimated cost of 24 billion (2017) Colombian pesos (2017 US$8 million). Additionally, five pumping stations would be needed to transport wastewater to a new mechanical pretreatment plant at a total cost of 8.6 billion (2017) Colombian pesos (2017 US$2.87 million). Finally, the Plan included a new marine outfall that would cost approximately 2.8 billion (2017) Colombian pesos (2017 US$0.93 million). The Plan also included proposals for improved drinking water and rainwater drainage systems, and identified public funding to cover up to 77.1 per cent of all those investments (CDM Smith and INGESAM/FINDETER, 2017). Unfortunately, to date, the proposed projects are yet to materialize.

Given the population density, mass tourism, and lack of systematic environmental monitoring, current sewage infrastructure, or the lack thereof, may have substantial effects on population health as well as coastal and marine ecosystems (e.g., coral reefs, seagrasses, mangroves and beaches). From 2000 to 2005, Gavio et al. (Reference Gavio, Palmer-Cantillo and Mancera2010) found that levels of ammonia, nitrites, nitrates, phosphates, and fecal coliforms were higher than recommended for coral reefs and seagrass ecosystems. Similarly, in the last 10 years, several studies have shown that total and fecal coliforms constantly exceed the permissible limit for primary contact waters for recreational use (swimming and diving; 200 NMP/100 ml) as stipulated in Colombian legislation (Obando-Madera and Martínez-Campo, Reference Obando-Madera and Martínez-Campo2019; REDCAM, 2020; Fussalba-Carreño and Aguas-Meza, Reference Fussalba-Carreño and Aguas-Meza2021). Hence, extension and improvement of sewerage are crucial to prevent soil, groundwater, seabed, and seawater contamination and reduce health risks for locals and visitors.

3. Elicitation method

There are multiple approaches to elicit preferences and estimate willingness-to-pay measures for proposed service improvements that have not materialized yet. Those approaches are usually classified in two categories: (1) revealed preference methods and (2) stated preference methods. Revealed preference methods rely on observational data to infer use values for goods and services. In contrast, stated preference methods, such as discrete choice experiments (DCEs) and the contingent valuation (CV) method, rely on hypothetical scenarios to estimate both use and non-use values, including altruistic and option values.Footnote ² In the context of this study, it is crucial to elicit non-use values given that tourists are asked whether they would be willing to donate to improve local sanitation services that they might not use in the future. They may be willing to donate so others can have access to improved sanitation (i.e., altruistic value), and/or because they want to protect coastal resources they might use if they return to the island (i.e., option value).

In DCEs, respondents state their preferences by choosing among two or more hypothetical alternatives with different levels of prescribed attributes. The inclusion of a payment allows researchers to estimate marginal values for multiple attributes of the good, service, or policy under analysis. Depending on the number of attributes in each alternative, alternatives in each choice task, and choice tasks performed by respondents, DCEs may require a substantial amount of time to respond to and may impose a cognitive burden on respondents (Hensher, Reference Hensher2006; Hoyos, Reference Hoyos2010). This may be an issue when eliciting preferences from tourists who may want to spend as little time as possible on a survey.

Relative to DCEs, the single-bound CV method requires less time and cognitive skills from respondents given that they are only asked to make a decision between paying a proposed amount for a hypothetical good or not. Moreover, the CV method may provide more conservative estimates than DCEs (Ryan and Watson, Reference Ryan and Watson2009; Danyliv et al., Reference Danyliv, Pavlova, Gryga and Groot2012), including estimates for specific attributes when combined with split-sample treatments (Haab et al., Reference Haab, Lewis and Whitehead2020). Given these advantages of the CV method over DCEs, we designed a CV survey to elicit tourists’ willingness to donate (WTD) to improve wastewater management in SAI. Vásquez (Reference Vásquez, Thompson, V and Teran2022) and Vásquez et al. (Reference Vásquez, Mateus, Loyola-Plúa, Torres-Suárez and Ochoa-Herrera2024) are recent examples of CV studies eliciting tourists’ WTD for improved public services.

4. Analytical framework and empirical modeling

Andreoni (Reference Andreoni1989, Reference Andreoni1990) provides a utility-theoretic framework to analyze individual donations to a public good. In this model, an individual is assumed to derive utility from income ( $Y$) and the donation to a public good ( ${g_i}$). Given that an individual is assumed to donate due to altruistic and egoistic motivations, the donation ${g_i}$ enters the utility function $V$ in three forms: (1) by reducing income available for other goods and services, (2) through the public good to which the person altruistically contributed ( $G\, = \,{\sum _{j \ne i}}\,{g_j} + {g_i}$), and (3) directly into the utility function due to “warm glow” reasons, i.e., ${V_1}(Y\!-\!{g_i},{\text{ }}G,\,\,{g_i})$. Consequently, an individual will donate as long as the utility lost due to income reduction is compensated by utility gains from contributing to the public good (i.e., altruism) and from egoistic motives, i.e., ${V_0}\left( {Y,{\text{ }}{G_{ - i}},{\text{ }}0} \right) = {V_1}\!\left( {Y\!\!-\!g_i^*,{\text{ }}G,{\text{ }}g_i^*} \right)$, where ${G_{ - i}} = j{\sum _{j \ne i}}\,{g_j}$ and $g_i^*$ is the optimal amount to be donated.

For empirical modeling, we assume that the maximum amount that individual $i$is willing to donate ( $g_i^*$) follows a log-linear specification,Footnote ⁷

(1)\begin{equation}\ln\!\left( {g_i^*} \right) = {X_i}\beta + {e_i}\end{equation}

where $X$is the vector of covariates, $\beta $ represents the corresponding vector of coefficients to be estimated, and $e$ is the idiosyncratic error term. While $g_i^*$cannot be observed in the dichotomous format of our CV question, it can be traced given that the probability of a positive response is equivalent to the probability that the natural logarithm of $g_i^*$ is greater than or equal to the natural logarithm of the proposed donation, i.e., P(Yes) = P(ln( $g_i^*$) ≥ LNBID) = P(Xβ + e ≥ LNBID). Consequently, rather than estimating equation (1), we model the probability that an individual will donate ( $P$) using a logit specification:

(2)\begin{equation}LN\left( {\frac{P}{{1 - P}}} \right) = \alpha {\text{ }}LNBID + X{\text{ }}\delta + u\end{equation}

where $\alpha $ and $\delta $ are coefficients to be determined using a maximum likelihood estimation approach, under the assumption that the stochastic error term $u$ follows a logistic distribution. Subsequently, estimated coefficients (i.e., $\hat \delta $ and $\hat \alpha $) are used to compute the amount that the median tourist would donate (i.e., the WTD) to improve wastewater management as follows:

(3)\begin{equation}WTD = {\text{ }}{e^{ - \left( {\bar X\hat \delta /\hat \alpha } \right)}}\end{equation}

where $\bar X$ is a vector of the average of covariates.

Table 1 shows the variables included in vector $X$. The variable LNBID is included to depict the effect of the donation amount presented in the contingent scenario on the likelihood of giving to improve wastewater management. This effect is expected to be negative because the marginal utility of giving diminishes as the donation amount increases. Moreover, the utility of giving also decreases with each additional dollar that will not be available for the individual’s own needs. The binary indicators COVERAGE-75 and COVERAGE-100 are included to conduct a scope test as they depict whether respondents were told that the proposed project would treat 75 and 100 per cent of the wastewater generated in the island, respectively. Given that the base of comparison is 50 per cent of the wastewater, we expect those indicators to show a positive effect on the likelihood of respondents donating. Our experimental design also allowed us to investigate time preferences. Binary indicators YEAR-2030 and YEAR-2035 depict time preferences relative to the year 2025, which was presented as the most immediate time to complete the proposed project. Additionally, three binary indicators (INFO-ENV, INFO-HUMAN and INFO-BOTH) are included to assess whether information regarding the effect of inadequate sanitation on the environment and population health influences tourists’ WTD. Those information effects remain to be empirically estimated. In addition to the experimental indicators mentioned above, we control for the survey mode given that some respondents needed assistance to complete the survey, mainly due to lacking access to Internet or an electronic device when they were invited to participate in our study.

Table 1.

Variables definition and descriptive statistics

Note: Observations = 318.

^a These indicators are excluded from logit models to be used as base of comparison.

We also control for altruistic motives, environmental preferences, personal experiences in the destination, and sociodemographic characteristics of respondents. Standardized indices ALTRUISM and WARMGLOW depict tourists’ altruistic and egoistic motivations to donate, respectively. Therefore, both are expected to have positive coefficients. Similarly, ENVPREF, a standardized index on environmental preferences, is expected to have a positive effect on the WTD given that poor sanitation may adversely affect coastal and marine ecosystems that provide tourist and recreational services to visitors. We also included a binary indicator, LOCALCOMM, depicting whether the respondent visited local communities. It is expected that those who observed the conditions in which local communities live are more likely to donate as a demonstration of empathy. It can also be assumed that tourists who plan to come back to SAI are more likely to donate to improve wastewater management, as they could benefit from those services in the future. Finally, we include variables depicting respondents' citizenship, sex, age, education and incomeFootnote ⁸ to further control for heterogeneity across respondents. A priori, we do not have expectations of the effects of those characteristics on tourists’ WTD.

5. Survey and estimation results

The descriptive statistics shown in table 1 provide a profile of the average respondent. About 82 per cent of our respondents were Colombians and the rest were foreigners. This rate is statistically similar to the official report that 83.1 per cent of the individuals who visited SAI at the time of this study (August–September 2022) were Colombians (t = 0.476; p-value = 0.635). Almost 62 per cent were females, 52.8 per cent had an undergraduate degree, and 26.4 per cent had a graduate degree. The average respondent was 33 years old and had a monthly income of 3.96 million Colombian pesos (US$879).Footnote ⁹ The average profile of respondents is statistically similar across subsamples obtained in each of our exogenous treatments, i.e., percentage of wastewater to be treated, year of project completion, and information on effects of poor wastewater management (see tables A3, A4 and A5 in the online appendix).

Our survey results indicate that 44 per cent of respondents were willing to donate the amount proposed in the CV scenario (see table 1). The WTD to improve wastewater management decreases with the donation amount (see online appendix figure A2), yielding support to the construct validity of our CV scenario. While almost 85 per cent of the respondents who were presented with a donation amount of 10,000 Colombian pesos (US$2.22) responded positively to the CV question, only about 22 per cent did so when proposed a donation amount of 250,000 Colombian pesos (US$55.55). The most popular reason for a negative response to the CV scenario was that the proposed donation amount was too high (34.3 per cent of respondents who would not donate), followed by the belief that donations would not be used for improving wastewater management (30.9 per cent), and earning too low an income to donate (14.6 per cent). Less popular reasons include making the government responsible for wastewater management, having paid an entrance fee, and perceiving corruption (see online appendix figure A3).Footnote ¹⁰

We also estimate logit models to identify factors underlying the decision to donate or not. However, before presenting those results, we continue to show results from factor analyses of altruistic motives and environmental preferences conducted to compute indices that are included in logit models.

5.2. Logit models of willingness to donate

Table 2 shows two logit models of tourists’ WTD to improve wastewater management. Marginal effects are reported to facilitate the interpretation of the influence of corresponding covariates on the WTD. Model 1 includes experimental variables only. Model 2 further controls for heterogeneity across respondents by including variables depicting altruistic motives, environmental preferences, interactions with local communities, and personal characteristics of respondents. Pseudo R² coefficients and Akaike Information Criterion (AIC) suggest that the extended model (Model 2) fits the data at hand better than the restricted model (Model 1). However, the Bayesian Information Criterion (BIC) favors the parsimonious specification of Model 1 over Model 2. Results on experimental variables are robust across model specifications.

Table 2.

Logit models of willingness to donate for sanitation services (marginal effects)

Notes: Observations = 318. Robust standard errors are reported in parentheses.

Estimated coefficients on LNBID are negative and statistically significant, suggesting that the likelihood of giving decreases (increases) as the donation amount requested increases (decreases). This result supports the construct validity of our contingent scenario. Other experimental indicators are statistically insignificant, with the exception of INFO-HUMAN. Based on Model 2, estimated marginal effects of INFO-HUMAN suggest that the likelihood of giving would increase by approximately 10 percentage points when tourists are informed that poor wastewater management may adversely impact the population health.

Results also indicate that respondents who were assisted to take our survey are 16 percentage points less likely to state that they are willing to donate than those who took the survey online on their own. Additionally, among the attitudinal variables, only WARMGLOW was statistically significant (at a 5 per cent level). This suggests that respondents who experience emotional rewards from giving are more likely to donate in order to improve local wastewater management than those respondents with less intense warm-glow motives. In contrast, purely altruistic motives and environmental preferences do not seem to influence the decision to donate. Consistent with egoistic motives, respondents who plan to go back to SAI are more likely to donate than those who do not intend to do so. The differential in the likelihood of giving between those groups is 19 percentage points. According to the estimated marginal effect of the indicator COLOMBIAN, the likelihood of giving is lower among domestic tourists relative to foreigners by more than 14 percentage points. Finally, we found that tourists with an undergraduate degree would be less likely to donate than respondents with a lower level of education. Other personal characteristics are statistically insignificant.

5.3. Estimates of willingness to donate

Using the estimates from Model 2 in equation (3), we estimate how much money the median tourist would donate to improve local wastewater management. Table 3 shows those WTD estimates for every type of information regarding effects of poor wastewater management in the contingent scenario. The highest WTD estimate is observed when tourists are informed about the effects of poor wastewater management on the population’s health. The median tourist who received that information is willing to donate 40,468 Colombian pesos (about US$8.99). On the other hand, the lowest WTD estimate of 24,468 Colombian pesos (about US$5.44) is observed in the scenario with no information. Other communications (i.e., information on environmental effects and information on both health and environmental consequences) yield higher WTD estimates than the no information treatment; however, those WTD differentials are statistically insignificant.

Table 3.

Median willingness to donate and 95% confidence intervals (thousands of Colombian pesos)

We also provide more conservative estimates by correcting for respondents’ uncertainty regarding their positive response to the CV question (Blumenschein et al., Reference Blumenschein, Blomquist, Johannesson, Horn and Freeman2008; Champ et al., Reference Champ, Moore and Bishop2009; Ryan and Watson, Reference Ryan and Watson2009). Specifically, positive responses were recoded as negative ones if the certainty reported was below seven in a 0–10 scale. Then, we used recoded responses to estimate equations (2) and (3). Vásquez and Trudeau (Reference Vásquez and Trudeau2022, Reference Vásquez and Trudeau2024) followed the same approach in their studies on WTD for COVID vaccines. Table 4 shows the certainty-corrected estimates by information treatment. Again, the highest WTD estimate is observed when tourists are informed about (human) health effects of poor wastewater management, with a median WTD of 8,116 Colombian pesos (about US$1.80). The lowest WTD estimate is observed among tourists who did not receive information about effects of poor wastewater management, a one-time donation of 2.885 Colombian pesos (US$0.64).

Table 4.

Median willingness to donate and 95% confidence intervals after correcting for uncertainty (thousands of Colombian pesos)

6. Discussion and conclusions

In this paper, we used the CV method, along with split-sample treatments and psychometric scales, to estimate tourists’ WTD for local wastewater treatment facilities and identify factors underlying stated preferences to donate. Split-sample treatments allowed us to conduct a scope test, investigate time preferences, and observe responses to information regarding effects of poor wastewater management on the environment and population health. Psychometric scales were implemented to distinguish between pure altruism and impure (warm-glow) motivations to donate, as well as environmental preferences. While recent studies have estimated tourists’ WTD for managing protected areas, improving local water services, and transitioning to renewable energy (e.g., Alpízar et al., Reference Alpízar, Martinsson and Nordén2015; Vásquez, Reference Vásquez, Thompson, V and Teran2022; Vásquez et al., Reference Vásquez, Mateus, Loyola-Plúa, Torres-Suárez and Ochoa-Herrera2024), this is the first study on tourists’ preferences for improving wastewater management in their destination, and the first one to investigate how tourists’ WTD responds to year-of-completion and information treatments.

Our results suggest that tourists would voluntarily donate to improve wastewater management in their destination. WTD estimates vary between US$5.44 and US$8.99, depending on the information provided to respondents. These estimates are below 1 per cent of the average monthly income reported by respondents, and less than one-third of the 2022 entrance fee (US$27.55). Compared to related studies conducted in the Galápagos Islands, our estimates are considerably lower than the WTD of US$45 for improving drinking water (Vásquez, Reference Vásquez, Thompson, V and Teran2022), and US$85 in support of the transition to renewable energy (Vásquez et al., Reference Vásquez, Mateus, Loyola-Plúa, Torres-Suárez and Ochoa-Herrera2024). Conservative WTD estimates, obtained by correcting for respondents’ uncertainty regarding positive responses to the CV question, vary between US$0.64 and US$1.80 depending on the information provided to respondents. These estimates are 0.07–0.21 per cent of the average monthly income, and 6.5 per cent of the 2002 entrance fee.

Regarding our split-sample treatments, INFO-HUMAN is the only statistically significant indicator. This suggests that tourists’ WTD increases when learning that population health could be at risk due to poor wastewater management. Previous research indicates that respondents may be more responsive to framing potential environmental problems as public health issues (Maibach et al., Reference Maibach, Roser-Renouf and Leiserowitz2008; Maibach et al., Reference Maibach, Nisbet, Baldwin, Akerlof and Diao2010; Atherton, Reference Atherton2019). In fact, Atherton (Reference Atherton2019) found that respondents with human health concerns had the highest WTP for water remediation in the UK over respondents who had other priorities, including environmental concerns. Our finding that WTD increases with the health information is consistent with their results. This also implies that human health risks are relevant to respondents’ WTD and must be disaggregated from environmental preferences to prevent omitted variable bias in CV scenarios (Guignet and Martinez-Cruz, Reference Guignet and Martinez-Cruz2022).

Similar to results of Vásquez et al. (Reference Vásquez, Mateus, Loyola-Plúa, Torres-Suárez and Ochoa-Herrera2024) from the Galápagos Islands, we found that tourists are indifferent regarding the year of project completion. These time preferences, or lack thereof, may be related to the fact that tourists are unlikely to benefit from improved local wastewater management and, consequently, the year of completion is not relevant to them.

Likewise, indicators depicting our coverage treatment were statistically insignificant, despite using a simple approach to distinguish among project sizes (i.e., percentage of locals covered by the project randomly varied across respondents). This failure to pass a scope test may raise some concerns regarding the construct validity of our CV scenario. However, Nunes and Schokkaert (Reference Nunes and Schokkaert2003) showed that willingness-to-pay estimates may fail a scope test when warm glow values are embedded in those estimates. That is the case for our results as demonstrated by the statistical significance of our WARMGLOW index. Additionally, there is no consensus that failing a scope test challenges the validity of CV scenarios (Heberlein et al., Reference Heberlein, Wilson, Bishop and Schaeffer2005; Johnston et al., Reference Johnston, Boyle, Adamowicz, Bennett, Brouwer, Cameron, Hanemann, Hanley, Ryan, Scarpa, Tourangeau and Vossler2017). In contrast, there is consensus that the statistical significance of the bid presented in the CV scenario supports the construct validity of the contingent scenario. Given the careful design of our survey, we are confident that the CV question used in this study is appropriate to estimate tourists’ WTD to improve local sanitation services.

Another interesting result is that the likelihood to donate is lower for our in-person survey protocol than for the online version. This is unexpected given that prior studies have found that in-person interviews yield greater willingness-to-pay estimates than self-administered surveys due to social desirability bias (e.g., Leggett et al., Reference Leggett, Kleckner, Boyle, Dufield and Mitchell2003). Compared to tourists who responded to our online survey, respondents who participated in person could have expected that the surveyor was going to ask for the donation amount at the end of the interview. This stronger perception of the consequentiality of our study could influence respondents to be more truthful about their WTD. Future studies can investigate survey mode effects on tourists’ WTD, including potential channels such as perceived consequentiality and response certainty. Given that our survey was administered by a single surveyor, we also leave the analysis of the influence of different types of surveyors (e.g., females vs. males) on tourists' expectations that they will actually be asked for a donation for future studies.

Our findings also indicate that tourists would donate primarily motivated by egoistic reasons, such as experiencing emotional rewards from giving (the so-called warm-glow effect), rather than by altruism. They would also donate to have the option of improved service on their next visit to SAI. These results support using charitable campaigns that appeal to egoistic motives and anthropocentric preferences of tourists in order to fund local sanitation projects. Specifically, those campaigns could emphasize emotional rewards from giving and reduced health risks in future visits.

In accord with Vásquez et al. (Reference Vásquez, Mateus, Loyola-Plúa, Torres-Suárez and Ochoa-Herrera2024), we found that environmental preferences and personal characteristics such as sex, age and income are unrelated to tourists' WTD. The statistical insignificance of the environmental preference index (ENVPREF) is consistent with our finding that tourists’ WTD is irresponsive to information on the environmental impacts of inappropriate sanitation. Similarly, tourists’ WTD is income inelastic, which is not surprising given that the intended one-time donation amount is a small share of the average monthly income (less than 1 per cent). Contrary to our expectation, visiting local communities did not seem to make a difference in the tourists’ WTD. This finding may be explained by the fact that current sanitation issues are difficult to notice in a short visit to local communities.

Our WTD estimates can be used to predict the amount that could be raised annually to improve wastewater management in SAI. While we obtained estimates at the individual level, it is likely that only one individual from the same family or group actually donates. Under this assumption, our WTD estimates can be multiplied by the number of travel groups visiting SAI. We estimate that, in 2023, there were 229,218 groups visiting SAI by dividing the number of visitors by an average group size of four individuals. By aggregating our basic certainty-corrected WTD estimate of 2,885 Colombian pesos (about US$0.64) over 229,218 potential donors, we estimate that approximately 661.3 million Colombian pesos (about US$146,954) could be raised annually to improve wastewater management. This can be increased to 1.86 billion Colombian pesos (almost US$413,407) if charitable campaigns provide information regarding potential impacts of poor wastewater management on population health. These estimates can be compared against the cost of interventions aimed at improving wastewater management in SAI. For instance, installing an anaerobic septic system for a family of 10 members would cost approximately 19 million Colombian pesos (US$4,208) (Local construction company, personal communication, 2024). Hence, the funds raised could be used to install at least 35 septic systems in a year. Our estimates can also be considered in planning tools such as the Works and Investments Plan designed by CDM Smith and INGESAM/FINDETER (2017) as a funding source complementary to official revenues from fees, grants and taxes.

As does any other study, ours has some limitations that future studies can address. First, the indices on altruistic and environmental traits included in our logit models could be related to unobserved factors, and our survey did not gather instruments to test and treat potential endogeneity biases derived from that correlation. Analyzing the potential endogeneity of altruistic and environmental traits would be a logical extension to our analysis of tourists’ WTD. Second, we recommend that future studies gather more information that could have helped us estimate more precisely the funds that could be raised, such as family size. Third, there might be concerns that our sample size could be small to test the effect of our split-sample treatments (318 observations for 3 × 3 × 4 treatments). Our randomization procedure provided equal subsamples across our split-sample treatments, and our finding of a statistically insignificant effect of year-of-completion treatment is consistent with prior findings (Vásquez et al., Reference Vásquez, Mateus, Loyola-Plúa, Torres-Suárez and Ochoa-Herrera2024). Yet, future studies would do well to use a larger sample of tourists to investigate the effects of coverage, time of project completion, and information on their WTD.

Lastly, we would like to address potential concerns about the use of voluntary donations as the payment vehicle in CV scenarios. Potential hypothetical biases could be introduced in our estimates due to the inherent lack of consequentiality of voluntary contributions (Johnston et al., Reference Johnston, Boyle, Adamowicz, Bennett, Brouwer, Cameron, Hanemann, Hanley, Ryan, Scarpa, Tourangeau and Vossler2017). We provided respondents with official information regarding current wastewater management in our study site to enhance the perceived consequentiality of our CV scenario. We also implemented an uncertainty-based correction procedure to provide more conservative and arguably accurate WTD estimates (Blumenschein et al., Reference Blumenschein, Blomquist, Johannesson, Horn and Freeman2008; Champ et al., Reference Champ, Moore and Bishop2009; Ryan and Watson, Reference Ryan and Watson2009). Our estimates are reasonable relative to the average tourists' income and travel expenditures such as the entrance fee each visitor pays when arriving in SAI. Yet, those estimates should be used with some caution when predicting the amount of money that could be raised. Despite those potential shortcomings, our study indicates that crowdfunding campaigns would be an effective mechanism to help improve wastewater management in SAI. Moreover, along with the findings of Vásquez (Reference Vásquez, Thompson, V and Teran2022) and Vásquez et al. (Reference Vásquez, Mateus, Loyola-Plúa, Torres-Suárez and Ochoa-Herrera2024), our results indicate that donations from tourists are a viable funding source to improve public services in tourist destinations. For destinations with low coverage of public services, our study also provides a roadmap to assess how much tourists would donate to improve public services.