Introduction

The population in the eastern area of the Valley of Mexico has increased considerably. According to data from the 2020 Population and Housing Census (Instituto Nacional de Estadística y Geografía [INEGI], 2020), the region has 6 959 035 inhabitants, accounting for nearly 40 % of the total population of Estado de México. In this context, it is unsurprising that water resources have been diminished. According to the Comisión Nacional del Agua (CONAGUA, 2023), eight of the nine aquifers in Estado de México are overexploited; that is, their average annual availability of groundwater shows a deficit. This is the case with the Texcoco aquifer (1507), which supplies the municipality of the same name and its surrounding areas. The aquifer has a deficit of 149.805 hm³∙yr^-1, representing 100 % extraction relative to its recharge capacity (SEMARNAT, 2023). Considering this scenario, the San Diego Tlailotlacan Drinking Water System, located in the community of the same name in the municipality of Texcoco, faces a challenging situation.

The reduction in quantity, frequency, and quality of water from an aquifer, resulting from overexploitation, directly impacts the quality of life of households. In the study area, this has fostered a culture of nonpayment, which generates deficits in effective governance of the water system (Silva Rodríguez, 2015). The problem of non-payment is so severe that, according to the report from the San Diego Tlailotlacan Drinking Water Committee, A.C. (2018), half of the 2 706 users have outstanding debts or payment delays of up to 15 years. Consequently, the lack of revenue prevents adequate maintenance of the network and wells, leading to deterioration, leaks, and water wastage. Moreover, uncontrolled water distribution facilitates illegal connections, which exacerbate economic losses and system inefficiency (Comité de Agua Potable San Diego Tlailotlacan, A. C., 2018).

Currently, in the community of San Diego, the monthly service fee is 70.00 MXN, totaling 840.00 MXN annually. Users who pay for the entire year in advance receive a one-month service waiver, reducing the fee to 770.00 MXN. Additionally, there is a subsidy covering half of the service cost (420.00 MXN annually) for vulnerable groups such as senior citizens, single mothers, and persons with disabilities (Comité de Agua Potable San Diego Tlailotlacan, A. C., 2018).

To improve the system’s sustainability, measures have been proposed to optimize water usage. These include the installation of flow-regulating dosing valves and user-specific valves to restrict supply to users with outstanding debts, as an incentive for timely payment. Rehabilitation of the well and repair of leaks in the hydraulic network are also planned, both of which are key actions to reduce waste and improve efficiency in water extraction and distribution. These improvements aim to optimize use through a more equitable distribution, thereby alleviating pressure on the aquifer and promoting more sustainable management.

Efficient water management could contribute to the financial sustainability of the water system; for example, in the study by Huaraca et al. (2021), users indicated willingness to pay as long as they receive permanent service. In this context, improved management would enable financing for operations, maintenance, and infrastructure improvement. Commonly reported measures to improve water services include pipeline maintenance, leak reduction, installation of water meters (Sandoval et al., 2016; Valdivia et al., 2022), wastewater treatment, and river rehabilitation (Avilés et al., 2010; Valdivia et al., 2011).

To determine appropriate increases in drinking water service fees, economic valuation methods can be employed. Among these, the Contingent Valuation Method (CVM) is particularly useful for estimating the economic value of environmental resources that lack a real market-such as water. CVM involves presenting respondents with the current state of a non-market good, outlining a potential improvement, and then asking whether they would be willing to pay for that improvement (Perez-Verdin et al., 2016).

In Latin America, and specifically in Peru, Tudela (2017) applied CVM to assess a wastewater treatment project in Puno and found an average WTP of 1.46 USD per person per month. Similarly, Cahui et al. (2019) studied key factors influencing WTP for water and sanitation projects in Peru, highlighting the importance of socioeconomic variables such as income, age, and education level.

Most studies related to water use CVM to estimate economic benefits (Pérez-Verdín et al., 2016). In general, economic valuation models for water resources show that factors such as price, age, education level, household size, and gender, as well as service frequency and proximity to the resource, significantly influence WTP (Huaraca et al., 2021; Parillo, 2022; Quispe et al., 2021). Despite the widespread application and advantages of CVM, the method also faces several limitations, including strategic bias, design bias, payment vehicle, information, hypothetical bias, starting point bias, and operational bias (Avilés et al., 2010). Additionally, since WTP only reflects an intention to pay rather than an actual financial commitment (Ramírez et al., 2022), it is often overestimated. To minimize these biases, it is essential to ensure the careful design and implementation of the questionnaire instrument.

Studies on the valuation of water resources, both in terms of quality and supply for urban use, have shown variations in WTP depending on the context and the socioeconomic characteristics of the population. In some cases, WTP is high, indicating strong support for proposed measures and their financial feasibility (Sandoval et al., 2016; Valdivia et al., 2022). However, other studies have reported that fewer than half of users are willing to pay a significant additional amount (Briseño & Macedo, 2021), or that WTP values are too low to cover the costs of project implementation (Suárez et al., 2023). In such cases, researchers have highlighted the need to combine the resources of utility operators with funding from various levels of government to make valuation projects viable (Aguilar & De la Rosa, 2018).

In this context, the main objective of the study was to estimate the WTP of potable water service users in San Diego, Texcoco, Estado de México, for the rehabilitation of the water distribution network to improve service quality and supply. The specific objectives were: (a) to develop an econometric model for WTP, and (b) to compare the results obtained through parametric and non-parametric estimation methods.

Materials and Methods

Study area

San Diego is a community located in the municipality of Texcoco, Estado de México, situated at approximately 19.5° latitude and -98.8° longitude (Figure 1). The community has an estimated population of 7 266 residents and a total of 2 083 households (INEGI, 2020).

The community is served by a local water utility organization known as Sistema de Agua Potable San Diego Tlailotlacan, A.C., which consists of an administrative committee and technical staff. The utility's registry includes 2706 users with a registered water connection.

Sample design and questionnaire

Based on the population of 2 706 registered users reported by the Comité de Agua potable (water committee) de San Diego, the sample size (n) for a finite population was determined using the formula proposed by Anderson et al. (2008):

where,

N = population

Z = statistical parameter depending on the confidence level 1.96 (α = 0.05)

e = margin of error of 0.05

p = probability of occurrence of the studied event (success) 0.5

q = probability that the event does not occur (failure).

Based on the data presented, the sample size was 337. For reasons of practicality and given the payment amounts explained in later sections, 340 questionnaires were applied.

A pilot survey was conducted to adjust some confusing or redundant questions and especially to define the WTP amounts (Ccasani et al., 2023). In the pilot test, 20 questionnaires were applied. The WTP question was in open-ended format “What is the maximum amount you would be willing to pay for the improvement of the water network?”. With this, the five referendum amounts were obtained: 10.00, 20.00, 30.00, 40.00 and 50.00 MXN.

Regarding the final questionnaire, this was divided into three sections:

a) General Information Section. Users were asked about various aspects of the water service. The characteristics of frequency, pressure, quality, and odor of water were evaluated using a five-point Likert scale: very poor, poor, fair, good, and very good, according to each user’s perception.

b) Willingness to Pay (WTP) Section. This section included a description of the problems in the potable water network, as well as a proposed rehabilitation project as a solution. The purpose was to ensure that users were fully informed when responding to the question: “Would you be willing to pay __ pesos per month (in addition to your current water bill) to contribute in the rehabilitation of the water network?” Users were also asked to provide the reason for their response.

c) Finally, the demographic data section included the socioeconomic characteristics of the users.

The WTP question for the final version of the instrument was formulated in a dichotomous referendum format, where all possible proposals of amounts from the surveyor are randomly distributed among the respondents (Tudela, 2008).

Willingness to pay (WTP) estimation using the parametric method

The parametric analysis involves two key steps: selecting the functional form of the utility function and specifying the distribution of the error term. Based on the literature review, the most commonly used distribution is the logistic distribution (Cahui et al., 2019; Parillo, 2022):

Assuming a linear functional form in the income (Cahui et al., 2019; Ccasani et al., 2023), the mean willingness to pay can be estimated using the formulation provided by Haab and McConnel (2002): P * = W T P M E A N = - α β ; where α is the sum of the coefficients of the independent variables multiplied by their respective means, including the intercept, and β is the coefficient of the WTP variable.

Econometric model of willingness to pay (WTP)

The socioeconomic variables considered in the valuation model included: income (INC), gender, age, education (EDU), and number of economic dependents (ND). These variables are commonly found to be significant in studies related to water services (Cahui et al., 2019; Perez-Verdin et al., 2016).

Fee discount was an additional variable incorporated into the model. Fee discount consists of a subsidy covering half the cost of the service for senior citizens, single mothers, and individuals with disabilities. This variable was included because 30 % of the registered users benefit from this subsidy (Comité de Agua Potable San Diego Tlailotlacan, A. C., 2018).

The function assumed to explain WTP of the residents of San Diego for the improvement of the potable water distribution network can be represented as follows:

Where, Pr(IF) is a dichotomous dependent variable representing the probability of responding affirmatively YES = 1 or negatively NO = 0, while the remaining variables are independent (Table 1).

Table 1. Variables included in the parametric model to estimate the willingness to pay of potable water service users in San Diego, Texcoco, Estado de México, for the improvement of the water distribution network.

To test the null hypothesis of a relationship among the variables, the Chi-square test was applied. Furthermore, the marginal effects of the independent variables on the dependent variable were analyzed to assess how each factor influences the probability of agreeing to pay. Thus, the marginal rate of change in the probability of the event occurring with respect to changes in the explanatory variables was calculated as: dP _i /dX _i = ß _i P _i (1-P _i ).

Estimation of willingness to pay (WTP) using the non-parametric method

The non-parametric model analyzes the proportion of negative responses (k _i ) to the proposed payment amounts (A _i ) in the survey, distributed in subgroups to represent various WTP levels, where A ₁ < A ₂ <, …, <A _k-1 and k ≥ 2. Responses are obtained from m ₁ , m ₂ , ..., m _k-1 individuals. For each subgroup i(i = 1, 2, …, k-1) the percentage hᵢ of individuals accepting the proposed amount Aᵢ is recorded, generating a series of decreasing proportions (pᵢ), which ideally follows a downward-sloping demand curve (López et al., 2017): p ^ i = h i m i .

If the data do not follow this trend, they are adjusted using the Pooled Adjacent Violator Algorithm (PAVA), which creates an ordered sequence of responses. This sequence becomes a maximum likelihood estimator for the probability of accepting the price, in accordance with Ayer’s theorem (Flores, 2017).

The procedure yields a number of points based on the unknown WTP in the sample; that is, the demand curve, p ^ ( A ) = D ^ ( A ) , which is related to the empirical cumulative distribution function (CDF), F ^ ( A ) , such that p ^ ( A ) = 1 - F ^ ( A ) . The function p ^ ( A ) is generally referred to as a survival function. To estimate the mean WTP, it is necessary to have additional information about the behavior of p ^ ( A ) between the observed points and in the tails of the distribution. For this purpose, Boman et al. (1999) apply linear interpolation. In this way, the expected value of WTP, µ, can be estimated by (Boman et al.,1999; López et al., 2017):

; where,

The above relationship suggests an estimator of the mean WTP:

Which is equivalent to: μ ^ = ∫ 0 A k p ^ ( A ) d A - ∫ A 0 0 1 - p ^ ( A ) d A

To ensure consistent values, the lower limit assumes p(0) = 1 (a positive willingness to pay), and the upper limit is adjusted until affirmative responses approach zero at high levels of A _i . According to Boman et al. (1999), the mean and variance can be calculated using the adjusted proportions shown in Table 2.

Table 2. Mean and variance formulas from Boman et al. (1999) used in non-parametric estimations.

Table 3. Haab and McConnell (2002) equations for non-parametric means and variances.

Comparison of parametric and non-parametric estimates

To compare welfare measures, confidence intervals were employed, following the methodology of López et al. (2017) and Vilca and Coila (2018) for both parametric and non-parametric procedures. The confidence intervals for the non-parametric method are presented in Table 2. For the parametric method, the Krinsky and Robb simulation procedure was used (López et al., 2017). The WTPCIKR command in the STATA software was employed, with 5 000 iterations and a 95 % confidence level (Jeanty, 2008).

In the analysis conducted for the present study, the following software tools were used: Statistical Analysis System (SAS) version 9.1, NLOGIT version 4.0.1 (Econometric Software Inc., 2007), and Stata version 14.0 (StataCorp LLC, 2015) for the parametric procedure, and Microsoft Office Excel 2016 for the non-parametric procedure.

Results and Discussion

Willingness to pay (WTP) by parametric estimation

Of the total sample, 56 % confirmed WTP for improvements to the water supply network, while the remainder responded negatively. From the pilot survey, it was observed that more than half of the respondents expressed no willingness to pay for the proposed improvements, even though the amounts were no more than 50.00 MXN.

In the global test of the model, the parameters were statistically significant at the 99 % confidence level. The likelihood ratio test, with a Chi-square value of 76.127 and 7 degrees of freedom, was also significant, allowing for the rejection of the null hypothesis (Table 4). The model demonstrated an overall accuracy of 70.29 %, with a sensitivity of 74.73 % and a specificity of 64.66 %. It was able to predict willingness to pay with 72.82 % accuracy and unwillingness to pay with 66.89 % accuracy.

Table 4. Statistical fit criteria of the logistic model for willingness to pay (WTP).

Of the seven variables used in the model, five were significant (P < 0.05): WTP, education level and economic dependents are significant at 99 % confidence and the variables fee discount and age at 95 %. The income variable was significant at a level of 10 %. Gender was not significant, so it was discarded from the final model. All variables presented the expected signs (Table 5).

Table 5. Econometric model estimates of willingness to pay based on individual tests.

Regarding the value of McFadden’s pseudo-R², the model yielded a low value (0.1631) for goodness of fit. Quispe et al. (2021) state that a value between 0.2 and 0.4 represents an optimal fit for discrete choice models. Nevertheless, values above 0.1 are considered to indicate a good fit in this type of model (Bateman et al., 2002).

The estimated probability model [P(SI)] was:

The average WTP for the improvement of the potable water network, based on the parametric estimation, was:

Therefore, if the fee increase were applied to the 2 706 registered users, it would result in a total annual amount of 1 210 880.00 MXN available for investment in potable water service infrastructure.

Each parameter of the model was evaluated by considering its marginal effects (Table 6). The most important variable in economic valuation studies is WTP for the improvement or hypothetical improvement scenario. In this study, the WTP coefficient indicates that an increase in price reduces the probability of a positive response-by 1 % for every additional 10.00 MXN. This finding is consistent with previous studies (Avilés et al., 2010; Cahui et al., 2019; Sandoval et al., 2016; Trujillo & Perales, 2020). Education level was also statistically significant; the probability of a positive WTP response increases by 10 % for each additional level of education, similar to the findings of Valdivia et al. (2022) and Quispe et al. (2021).

Table 6. Marginal effects of the independent variables in the econometric model of willingness to pay (WTP).

The number of economic dependents showed a negative effect. The probability of a positive WTP response decreases by 9 % for each additional dependent, which contrasts with some studies where household size has a positive effect (González et al., 2016; Valdivia et al., 2022). Age has a positive influence on WTP, with the probability of a positive response increasing by 0.5 % for each additional year of age. This agrees with the findings of Sandoval et al. (2016) and Godoy et al. (2019), who report that older users tend to place greater value on the stability of basic services.

Similarly, household income increased the probability of a positive response by 4 %, consistent with studies by Zavaleta et al. (2020), Parillo (2022) and Ramírez et al. (2022).

Finally, fee discount (discounts for water service payments) had a negative effect of 21 % on WTP. This may be attributed to the lower or unstable incomes of vulnerable groups who receive such discount. In the study area, this discount is targeted at vulnerable populations such as the elderly, individuals with disabilities, and single mothers. Therefore, it can be inferred that the income levels of these groups tend to be lower or more variable, which may explain the negative effect of this variable.

Willingness to pay (WTP) by non-parametric estimation and comparison with the parametric method

The WTP values obtained through the non-parametric method are summarized in Table 7. The estimated function was monotonically decreasing; therefore, the PAVA algorithm was not applied (Vilca & Coila, 2018). Likewise, the non-parametric variance values are also presented.

Table 7. Comparison of parametric and non-parametric estimates of the mean and variance of willingness to pay (WTP) for the improvement of the potable water system.

Estimates obtained through the non-parametric method yield lower means compared to the parametric method. Although the confidence intervals overlap, the non-parametric means do not fall within the parametric interval. This is consistent with the findings of López et al. (2017) and Vilca and Coila (2018), where non-parametric estimates were more conservative, tending to underestimate WTP; moreover, lower bounds were observed in the non-parametric method, even though both approaches produced overlapping confidence intervals. Nevertheless, since the means are considered statistically equivalent, either the parametric or non-parametric WTP can be used for economic valuation estimates. Future research is encouraged to compare results using semiparametric methods and to analyze the impact of subsidies on WTP for improvements in water or public services more broadly.

Regarding the financial sustainability of the system, annual expenditures amount to 1 931 033.44 MXN, while revenues generated from water services total 1 500 561.00 MXN, excluding municipal subsidies. This indicates an operational deficit that compromises long-term viability (Comité de Agua Potable San Diego Tlailotlacan, A. C., 2018). Although municipal funding has allowed the system to function, relying on these subsidies is not a sustainable long-term solution.

The economic valuation of the project estimated that, with an average WTP of 37.30 MXN per month, over 1 200 000 MXN in additional revenue could be generated. These additional funds could potentially finance the first phase of rehabilitation, which includes repairs to a well and a section of the water distribution network, estimated at 1 100 000 MXN. However, due to the high rate of payment delinquency, actual revenue collection would be lower than estimated, threatening the project's financial viability.

Conclusions

Just over half of the residents in San Diego, Texcoco, Estado de México, expressed willingness to financially contribute toward infrastructure investments aimed at improving the quality and availability of drinking water. Incorporating this additional willingness to pay (WTP) into the monthly water bill could generate over 1 000 000.00 MXN in additional revenue for the local water committee. Nevertheless, 44 % of users indicated they are unwilling to contribute, which poses a significant challenge to the implementation of proposed system improvements, as the actual amount collected would be lower than estimated. To address this issue, a user awareness campaign is recommended to foster a sense of shared responsibility in management and sustainable use of water resources. Regarding the estimation method, either the parametric or non-parametric approach may be used without affecting the outcome; however, the non-parametric method may serve as a useful complement, offering a more conservative estimate of WTP.