How to Prioritize the Attributes of Water Ecosystem Service for Water Security Management: Choice Experiments versus Analytic Hierarchy Process

Abstract

The various ecosystem services related to water form a complex structure that impacts on human well-being so it is necessary to know the relationships between their attributes to support decision making for water security. Our work investigates individual preferences for water ecosystem service attributes in the Tilacancha River Microbasin, in northern Peru, using two methods of a different nature. In that context, prioritization results using Choice Experiments and Analytic Hierarchy Process are compared for their abilities to represent purchase preferences and theoretical preferences, respectively. Both methods reveal that in a context of abundant water resources, the public has a higher preference for the attributes Quality Maintenance and Water Regulation, over Sediment Control and Water Yield, which were less valued. The differences allowed us to identify possible applications of the results useful for water security management. Additionally, we conclude that it is possible to combine the results of both methods to support decision making, and we highlight the specific cases in which it is appropriate to use the methods individually.

1. Introduction

Water is part of a complex structure within the environment and constitutes a complex service related to different types of ecosystems [1], which is crucial for the survival and development of living organisms [2]. According to the definition of the UN Water mechanism related to human development, water security should minimally comprise the quantity of water available and its acceptable quality, although water-producing ecosystems are not the only ones that produce water [3]. However, water-producing ecosystems and water in general provide a wide variety of ecosystem services that impact human well-being, such as water for human consumption, water for irrigation, flood protection, biodiversity protection, and others [4]. In this framework, the water ecosystem service (WES) represents all water-related benefits, e.g., water yield, regulation of the hydrological cycle, maintenance of water quality, aquifer recharge, scenic beauty, and others [5,6]. Given this diversity, we refer to the various WES provided by water-producing ecosystems as “attributes”.

Due to anthropogenic activities, water-producing ecosystems require conservation and restoration to ensure water security [7]. In this regard, there is extensive literature on the attributes linked to water, but they show general trends only; therefore, it is necessary to deepen the relationships between the attributes linked to WES [2]. In addition, WES attributes need to be weighted to support decision making, e.g., in the face of water scarcity situations [8]. In situations of scarcity, it is possible to identify optimized solutions for water supply and irrigation based on multi-criteria models [9].

Many of the services provided by ecosystems do not have a market value so it is difficult to establish relationships or weightings. To solve this problem, there are various methods that facilitate the partial or integral quantification of the economic value of ecosystem goods or services [10]. Two methods stand out for their ability to disaggregate the economic value among its attributes, Choice Experiments (CE) and Analytic Hierarchy Process (AHP). CE is recognized for its ability to estimate the individual preferences of the users of a service [11]. It allows estimation of the value of attributes or characteristics of a service [12]. The method is performed by means of surveys carefully designed to be clear and understandable [13]. Individual user preferences are collected when users choose a scenario from a list of scenarios proposed to the respondents [14]. These scenarios present different levels of the attributes [15], and the respondent chooses the alternative that is most useful to him or her [16]. This method has been applied under different approaches in different areas such as health economics, migration, energy economics, forestry economics, or environmental economics, where it is used to estimate the utility of the alternative [17] or the economic value of ecosystem services [15]. On the other hand, AHP is used in prioritization processes [18]. It is a very reliable method [19] used in the organization and analysis of decisions based on multiple criteria [20]. It systematizes and structures the decision-making process based on numerical scales [21]. AHP collects information through paired comparison surveys using standardized tables or judgmental decision matrices to identify priorities [22,23] and obtain relative weights indicating which attribute is more important than another [21]. The method must be applied to a panel of experts on the asset attributes to be weighted [24].

Both methods rely on very different approaches [25]. They yield partially equivalent results, but there are also specific differences [11]. Based on their specific objectives, researchers rely on the advantages of one method for their studies [11]. However, the combination of both methods is rare and there are insufficient case studies to estimate the potential benefits [26]. And to our knowledge, both methods have not been applied jointly in the prioritization of environmental assets.

Jointly, AHP and CE have been applied, compared and contrasted in studies of individual preferences to complex goods (

Table 1. Main studies comparing CE versus AHP.

References	Target	Location	Main Results
[25]	Comparing preference-based weights for age-related macular degeneration treatment attributes.	Germany	20% of coincidence between estimated weights
[11]	Comparing individual preferences of attributes and levels of an agri-food product: rabbit meat.	Spain	Coincidence of 55.6% in their ranking between methods. The variation in utility between levels in both approaches follows a similar shape for two of the three attributes analyzed.
[29]	To assess patient preferences regarding different antiplatelet medication options after an acute coronary syndrome	Germany	50% match.
[28]	Exploring whether expert judgment can substitute for citizens’ preferences in determining public goods investment strategies	England	60% coincidence between estimated weights

), e.g., in analyzing policy priorities in agriculture [27], in the identification of investment strategies for public goods [28], in public health when investigating patients’ preferences for medication options after acute coronary syndrome [29], and in determining treatment levels for macular degeneration [25]. They have also been used to assess preferences for assets that are highly visible to the public in markets, such as agri-food products [11]. These studies have in common that they show partially equivalent results with 20–60% similarity in the prioritization of attributes. Finally, in non-market assets, AHP and CE have been applied to value coastal and marine habitats, combining the techniques in a single analysis [26]. According to Table 1, it is evident that there are partial similarities, but in no case total similarity in the prioritization results. Therefore, it may be necessary to apply both techniques to weight attributes of non-market goods such as water-related attributes, and from this, explore similarities, differences, and possible applications.

Based on the above, our main objectives were (a) to determine the relative weight of water ecosystem service attributes with two methods, CE and AHP, for estimating individual preferences of complex goods, (b) to the describe similarities and differences between these methods, and (c) to identify possible particular features, which would recommend their application in specific cases.

2. Materials and Methods

2.1. Context

Chachapoyas is a small city in northern Peru, South America. The city is supplied with water from the Tilacancha River Microbasin, located in the territories of the rural communities of the districts of Levanto and San Isidro del Maino, adjacent to Chachapoyas (Figure 1). The Tilacancha River Microbasin is made up of two watersheds, Osmal-Tilcancha and Cruzhuayco. However, the micro-watershed is named after the Tilacancha River, the natural boundary between the two districts.

Part of the Tilacancha River Microbasin is within the Tilacancha Private Conservation Area (Figure 1), created in 2010 with the objective of conserving the grasslands in the upper parts of the watersheds, the montane forests and the biological diversity of the area to contribute to the proper functioning of the hydrographic system and guarantee the provision of ecosystem services [30]. The ecosystems present in the micro-basin are major producers of water resources. The main water users are the inhabitants of Chachapoyas. The company that administers the drinking water service distributes water to 8747 families that have a home connection, benefiting approximately 33,854 inhabitants living in the city [31].

As mentioned above, we refer to the various ecosystem services provided by water-producing ecosystems as “Attributes”. A rapid hydrological diagnosis in the Tilacancha River Microbasin identified four priority attributes of WES [5]: (a) Water regulation, (b) Sediment control, (c) Water yield and (d) Quality maintenance.

Table 2. Attributes and levels of water service used in CE.

Attributes of the Water Ecosystem Service		Brief Conceptual Description [5]
Symbol	Name
A1	Water regulation	Capacity of the ecosystems in the micro-watershed to store water during the rainy season and to supply it to the lower parts even during the dry season. Factors on which it depends: precipitation intensity, vegetation cover, and depth of surface soil.
A2	Sediment control	Capacity of ecosystems to prevent soil erosion by buffering rainfall shocks. Factors on which it depends: intensity of precipitation and vegetation cover.
A3	Water yield	Capacity of ecosystems to produce water. Factors on which it depends: precipitation intensity, evapotranspiration, and the ways in which water leaves the basin.
A4	Quality maintenance	Capacity of ecosystems to purify water. Factors on which it depends: filtration and absorption of soil particles, and of living organisms present in the water and soil.

shows the definition of the attributes, and to facilitate their operationalization, they are represented by the symbols A₁, A₂, A₃ and A₄, respectively.

2.2. Choice Experiments

CE is a method that helps to disaggregate a non-market good or service into its different characteristics so that people can express the value they place on each of its attributes. In this way, the respondents’ choice indicates the influence of the attributes on their willingness to pay (WTP) for changes in each of them [32]. This method consists of presenting users of a good or service with a list of “Choice Sets” that illustrate different scenarios for the good of interest. Respondents are asked to choose the alternative they consider the best. Traditionally, choice sets are composed of three alternatives, where one of them is the current scenario, or status quo scenario. It is possible to review the methodological aspects of this method in [33]. In addition, the methodological framework developed here is schematized in Figure 2.

Table 3. Description of a choice set.

Attributes	Alternatives
	Status Quo Scenario	Improvement Scenario A	Improvement Scenario B
Attribute 1	Combination of levels for attributes in the current scenario	Combination of levels for attributes in scenario A	Combination of levels for attributes in scenario B
Attribute 2
Attribute 3
...
Attribute n
Monetary attribute
What would you buy?

illustrates the general design of the choice sets. The first column lists the attributes of the environmental asset. The remaining columns list the alternative scenarios for the attributes. These scenarios are designed based on levels for each attribute. In addition, a monetary attribute is incorporated to estimate the economic value in monetary units of the attributes and the entire choice set.

CE is theoretically based on the Theory of Value [34], which proposes that the utilities of goods can be separated into utilities of their attributes, and econometrically, in the Theory of Random Utility [35], which proposes that a perfectly rational individual always chooses the alternative that can produce the highest expected utility. In this context, for the econometric analysis, the data collected are analyzed using a random utility model. The respondent’s choice for one of the scenarios is represented through the discrete choice of a set of alternatives.

Then, each scenario is represented by a utility function, which consists of a deterministic element ($V_{ij}$) and a stochastic element ($e_{ij}$), which represents the unobservable influence on the individual choice of scenario j. Therefore, the utility of respondent i who chose scenario j is represented in Equation (1).

$$U_{ij}=V_{ij}\left(X_{ij}\right)+e_{ij}$$

(1)

The deterministic element is specified as a function of the WES attributes ($X_{ij}$). With the assumption that the random error term follows an independent and identical distribution, the probability that household head i chooses scenario j is represented in Equation (2).

$$P_{ij}=exp\left(\beta X_{ij}\right)/{\sum }_{j=J}exp\left(\beta X_{ij}\right)$$

(2)

In Equation (2), J represents all possible scenarios and β represents a set of estimated parameters. The probability shown in Equation (2) is estimated using the conditional logit model [36], most frequently used because of its adaptability to various econometric situations [37]. In addition, it is the most widely accepted model for dealing with CE results [38] and provides a convenient closed form for the underlying choice probabilities without any requirement for multivariate integration [37]. To validate the assumptions of independence of irrelevant alternatives (IIA), which guarantees the consistency of the model, the test of [37] is performed. The utility function of the model, with the exception of the error term, is expressed as a linear function of the vector of WES attributes (A₁, A₂, A₃ and A₄) and a fifth monetary attribute (A₅), which is “Price”. Additionally, two alternative-specific constants (ASC) are added, which symbolize the dummy responses of the choice of scenarios A and B in the choice sets. According to [39], the ASCs capture the utilities of the scenarios that the attributes failed to capture. This means that they capture the effect of the status quo bias, in addition to improving the model fit [40] Therefore, the utility function is defined in Equation (3).

$$V_{ij}=AS{C}_{ij}+{\beta }_1{A}_{1,ij}+{\beta }_2{A}_{2,ij}+{\beta }_3{A}_{3,ij}+{\beta }_4{A}_{4,ij}+{\beta }_5{A}_{5,ij}$$

(3)

The β terms are the estimated parameters for each attribute that influences the respondents’ utility. The estimation of the marginal willingness to pay (MWTP) for each WES attribute is performed assuming that all other variables remain constant from the results of Equation (3) with the individual Equations (4)–(7):

$$MWT{P}_{A_1}=-\frac{\partial V/\partial A_1}{\partial V/\partial A_5}=-\frac{{\beta }_1}{{\beta }_5}$$

(4)

$$MWT{P}_{A_2}=-\frac{\partial V/\partial A_2}{\partial V/\partial A_5}=-\frac{{\beta }_2}{{\beta }_5}$$

(5)

$$MWT{P}_{A_3}=-\frac{\partial V/\partial A_3}{\partial V/\partial A_5}=-\frac{{\beta }_3}{{\beta }_5}$$

(6)

$$MWT{P}_{A_4}=-\frac{\partial V/\partial A_4}{\partial V/\partial A_5}=-\frac{{\beta }_4}{{\beta }_5}$$

(7)

The MWTP of each WES attribute constituted the marginal rate of substitution between the price to access the scenario and the referred attribute.

2.3. Analytic Hierarchy Process

AHP is a multi-criteria method useful in decision making on complex problems with multiple conflicting and subjective criteria [41]. It has been used in almost all applications related to decision making [42]. It is even used to value environmental assets, specifically, in the prioritization process [18]. In this method, a panel of experts is presented with a “Paired Comparison Survey” on the environmental asset, which requires weighting the relative importance of the attributes of the environmental asset.

In

Table 4. Paired Comparison Survey.

Attribute	Extreme Importance	Very strong Importance	Strong Importance	Moderate Importance	Equal Importance	Moderate Importance	Strong Importance	Very Strong Importance	Extreme Importance	Attribute
	9	7	5	3	1	3	5	7	9
Attribute 1										Attribute 2
Attribute 1										Attribute 3
…										…
Attribute n − 1										Attribute n

, the general design of the paired comparison surveys is illustrated. The first and last columns list the attributes of the environmental asset, so that they are compared one by one. The columns numbered with odd numbers correspond to Saaty’s Fundamental Scale of Comparisons, whose explanation and intensity of importance is shown in

Table 5. The Fundamental scale of paired comparisons.

Intensity	Definition
1	Both elements are equally important
3	Moderate importance of one over the other
5	Essential or strong importance
7	Very strong importance
9	Extreme importance
2, 4, 6 and 8	Intermediate values used to shade the response between two adjacent values

Source: Adapted from The fundamental scale [23].

. It is possible to review the methodological aspects of this method, as reported by [18,21,23,43]. In addition, the methodological framework developed here is schematized in Figure 2.

The first step is to provide the expert with concise conceptual information on the WES attributes, in order to motivate the recognition of the importance of each one of them. Subsequently, the expert is presented with the paired comparisons survey (Table 4) and is asked to compare the two attributes in each row of the survey, for which he/she is provided with the fundamental scale (Table 5). With these resources, the expert decides which of the two attributes is more important, the one on the left or the one on the right. The expert marks in a box, on the left or right, according to his or her determination. One box is checked for each row. The box chosen depends on the intensity of superiority of one attribute over the other, according to the fundamental scale. If the expert considers that both attributes are of equal importance, the expert should check the middle column.

The fundamental scale was designed by Saaty with odd numbers from one to nine to reduce the difficulty for humans to make complex decisions. However, the even numbers from two to eight can be used to indicate that the superiority rating is between the lower and upper odd number, in which case the expert has to mark on the line separating the two boxes.

The paired comparison survey is completed by the selected experts individually for an appropriate period of time. The data are then transported to spreadsheets and arranged in paired comparison matrices of order n × n, by number of attributes (

Table 6. Matrix of Paired Comparisons.

Attributes	Attribute 1	Attribute 2	Attribute 3	…	Attribute n
Attribute 1	1	$u_1$	$u_2$	…	$u_{n-1}$
Attribute 2	${\left(u_1\right)}^{-1}$	1	$u_n$	…	$u_{2n-3}$
Attribute 3	${\left(u_2\right)}^{-1}$	${\left(u_n\right)}^{-1}$	1	…	…
…	…	…	…	…	$u_{\frac{n\left(n-1\right)}{2}}$
Attribute n	${\left(u_{n-1}\right)}^{-1}$	${\left(u_{2n-3}\right)}^{-1}$	...	${(u_{\frac{n(n-1)}{2}})}^{-1}$	1

).

In Table 6, $u_i$ denotes the value assigned by the expert according to the fundamental scale of comparisons (Table 5). The boxes on the diagonal have the value of unity because they represent the comparison of the same attribute. The data with negative power are the inverse of the attribute comparisons, hence their mathematical representation $\frac{1}{u_i}$.

Subsequently, the consistency of each matrix is calculated with the procedure of [43]. This procedure consists of normalizing by the sum of the matrix of paired comparisons, replacing each element by the quotient of that element and the sum of all the values of the column where it is located. Expressed using notation, $u_{ij}$ is replaced by $\frac{u_{ij}}{{\sum }_{k=1}^n{u}_{ij}}$. With the normalized matrix, the rows are added and the average of each sum is obtained, which represents the vector of global priorities. Then, the product of the original matrix and the vector of global priorities results in the total row vector. This vector is divided by the vector of global priorities and a quotient called column matrix is obtained. Next, the average of all the elements of the column matrix is calculated, and the resulting value is denoted as λmax.

The Consistency Index (CI), proposed by Saaty, was evaluated with Equation (8):

$$CI=\frac{{\lambda }_{max}-N}{N-1}$$

(8)

where N is the number of attributes. Finally, the Consistency Ratio (CR), also proposed by Saaty, is evaluated, as in Equation (9).

$$CR=\frac{CI}{Random consistency}$$

(9)

The random consistency value is chosen from

Table 7. Random consistency values.

Matrix Size (n)	1	2	3	4	5	6	7	8	9	10
Random consistency	0.00	0.00	0.52	0.89	1.11	1.25	1.35	1.40	1.45	1.49

Source: Adapted from [43].

according to the size of the matrix (n). The results of Equation (9) are compared with the values in

Table 8. Maximum percentages for consistency ratio.

Matrix Size (n)	Consistency Ratio
3	5%
4	9%
5 or more	10%

Source: [43].

. A matrix is consistent when its CR is less than the values in Table 8.

The CR results define which matrices are consistent and which are not. With the consistent matrices, the eigenvector of each of them is estimated. The eigenvectors show the weights of the attributes. To estimate the eigenvector, the original matrix of each respondent is multiplied by itself and a second matrix is obtained. With this matrix, a third matrix called “column vector” is constructed, which is normalized as follows: the elements of this matrix result from dividing the sum of elements of each row of the second matrix by the sum of all its elements. The process is repeated as many times as necessary, usually four or five times, until the elements of the last two column vectors are the same. The column vector that satisfies this condition is called the eigenvector [43].

The process ends by generating a matrix representing all the consistent matrices. This final eigenvector is generated by calculating the geometric mean, considering that it is the most appropriate method to aggregate individual weights for collective decision making [44]. The geometric mean of the elements of all the eigenvectors is then estimated, and the values are normalized. The final eigenvector represents the weights for the attributes assigned by the expert panel.

Finally, the economic valuation of the WES attributes (A₁, A₂, A₃ and A₄) is calculated based on the fifth monetary attribute (A₅); which is the “Cost of service” (

Table 9. Monetary attribute (A₅) for valuation methods.

Description of the Monetary Attribute	Valuation Method
	CE	AHP
Name	“Price”	“Cost of service”
Conceptual definition	Amount offered in the scenarios (USD/month per household). It allows us to simulate the purchase preference of individuals and also estimate the economic value in monetary units of the WES attributes and of the entire choice set.	Cost incurred by the company that manages the service to produce 1 m3 of water, the word “produce” being understood as the treatment and management of the resource. It represents the pivot value, responsible for monetizing the WES attributes. To estimate the monthly service cost for a household, the unit value would have to be multiplied by the monthly household water consumption.
Values	0.13, 0.38, 0.64 and 0.90 USD/month per household	0.63 USD/m3 of water

). In this process, the monetary attribute plays the role of pivot value, that is, the role of assigning economic value to the other attributes that do not have economic value. The estimation is based on the importance weights of the WES attributes and the weight of the monetary attribute. Thus, by means of the proportionality rule, the economic value of the monetary attribute assigns economic value in proportion to the importance weight of the WES attributes [43].

2.4. Empirical Application

The CE and AHP methods were applied to estimate the importance weights and the economic values of the WES attributes of the Tilacancha River Microbasin. The results were then compared to evaluate similarities and differences and to establish conclusions according to the methodological scheme (Figure 2). Table 9 describes the characteristics of the monetary attribute for each method.

CE was applied to water resource users in the city of Chachapoyas. Based on the levels identified in

Table 10. Attributes and levels of water ecosystem services for the design of improvement scenarios in the choice sets.

Symbol	Attribute	Level Description	Level
A1	Water regulation	Effect of programs and projects that guarantee effective protection and assurance of the provision of the attribute Water regulation compared with a scenario without intervention (unit: %)	Level 1: 0% * Level 2: 50% Level 3: 100%
A2	Sediment control	Effect of programs and projects that guarantee effective protection and assurance of the provision of the attribute Sediment control compared with a scenario without intervention (unit: %)	Level 1: 0% * Level 2: 50% Level 3: 100%
A3	Water yield	Effect of programs and projects that guarantee effective protection and assurance of the provision of the attribute Water yield compared with a scenario with no intervention (unit: %)	Level 1: 0% * Level 2: 50% Level 3: 100%
A4	Quality maintenance	Effect of programs and projects ensuring effective protection and assurance of the provision of the attribute Quality Maintenance compared with a scenario without intervention (unit: %)	Level 1: 0% * Level 2: 50% Level 3: 100%
A5	Price	Monthly amount derived from the implementation of programs and projects for the conservation of the attributes of the water ecosystem service (unit: USD/month per household)	Level 1: 0.13 Level 2: 0.38 Level 3: 0.64 Level 4: 0.90

* Refers to the base level in the status quo.

, the choice sets were designed. The levels of improvement were defined with focus groups of professional experts from the Tilacancha water ecosystem service. As can be seen, three levels were defined for the four WES attributes (0%, 50% and 100%). For the fifth monetary attribute, four levels were defined in the Peruvian currency (0.5, 1.5, 1.5, 2.5 and 3.5 PEN/month per household), equivalent to (0.13, 0.38, 0.64 and 0.90 USD/month per household). At the time of the study, USD 1.0 was equivalent to approximately PEN 3.91.

Choice sets are made up of alternatives, or scenarios. These are designed by combining the defined attributes and levels [45]. Traditionally, CE researchers design choice sets with three scenarios [40,46,47]. Accordingly, choice sets are composed of three alternatives (see example in Figure 3):

Status quo scenario: Describes the current state of the WES attributes. In this scenario no improvements are formulated, i.e., the level for all attributes is “unchanged” and the price is zero. For these reasons, it is an unchanged scenario.

Scenario A: This is an improvement scenario designed based on the combination of attributes and levels in Table 10. For this reason, the price for accessing the proposed attribute improvements is greater than zero, and also because it will support, in a hypothetical situation, the cost of the attribute improvements.

Scenario B: A scenario designed in the same way as Scenario A with a combination of attributes and levels.

When a choice set (Figure 3) is presented to the water user, the respondent is faced with choosing the alternative that best represents his or her interests as if he or she had to buy it. It is possible to design a large number of choice sets, with different scenarios A and B, with the status quo scenario considered invariant. To design an improvement scenario, we multiply the number of levels of the five attributes (Table 10). This is 3 × 3 × 3 × 3 × 4 $=324$ combinations of attributes and levels. If scenarios A and B are combined, the number of choice sets is high, and executing them would be impractical. To solve this problem, a fractional factorial design is chosen which ensures the characteristics of the full factorial design by means of an orthogonal arrangement [39,48]. The chosen factorial design consists of 54 scenarios. Then, to form a choice set, two of the 54 scenarios are chosen at random including one scenario A and one scenario B (see example, Figure 3). Therefore, the two scenarios are chosen with the restriction that they must be different. With this procedure, 37 choice sets were designed.

The choice sets were provided to water users located in the city of Chachapoyas, who provided the monthly MWTP of their household’s improvements in WES attributes.

The choice sets were provided as a survey to randomly selected heads of household in the city of Chachapoyas. In order to proportionally distribute the 37 types of choice sets designed, the sample was increased to 370 heads of household, meaning that each choice set was provided to 10 different randomly selected respondents. To enhance the understanding of the attribute levels, the respondents were presented with photographic diagrams to aid in the understanding of the scenarios. The respondents are then asked to choose the scenario that best represents their preferences. The survey was carried out by previously trained interviewers. The households were randomly selected to maximize the scope of the questions.

At the same time, AHP was applied with the same attributes as used for CE, as described in Table 2. The difference lay in the fifth attribute, the monetary attribute. For the AHP technique, the “Cost of service” described in Table 9 was chosen. According to the methodological scheme (Figure 2), the paired comparison survey was designed defining only one evaluation criterion “ecological importance of the attribute to provide indispensable services for water users”. This criterion was evaluated by the expert from a theoretical point of view, as opposed to the preferences of water users in CE, who base their choice on the competitive prices they face as if they had to buy it, and therefore, report their stated preference or purchase preference.

The survey was constructed following a one-to-one comparison format [18], as shown in

Table 11. Paired comparisons survey.

Attribute	Extreme Importance	Very Strong Importance	Strong Importance	Moderate Importance	Equal Importance	Moderate Importance	Strong Importance	Very Strong Importance	Extreme Importance	Attribute
	9	7	5	3	1	3	5	7	9
Water regulation										Sediment control
Water regulation										Water yield
Water regulation										Quality maintenance
Water regulation										Cost of service
Sediment control										Water yield
Sediment control										Quality maintenance
Sediment control										Cost of service
Water yield										Quality maintenance
Water yield										Cost of service
Quality maintenance										Cost of service

. The survey presents the 10 possible combinations among the five attributes. The attributes are located in the first and last column of the survey. The top of the survey presents Saaty’s fundamental scale of comparisons, the middle column of which is the value of unity or equal importance. This range increases symmetrically to the left and right in odd numbers up to the number nine.

AHP was provided to a panel of experts composed of 16 professionals. They were chosen for their extensive knowledge of Tilacancha’s WEE, considering that AHP requires the selection of a panel of experts who have detailed knowledge about the environmental asset under study [43]. In the context described above, the experts were presented with the pairwise comparisons survey (Table 11) and weighted the attributes based on the evaluation criteria.

The survey was conducted in person and individually with each expert. After an induction process on the methodology of filling out the survey and relevant information about the study, the experts were allowed time to consider their weightings. The data were then entered into spreadsheets and arranged in 5 × 5 paired comparison matrices by the number of attributes according to the scheme in Table 6. The surveys with consistent weights were aggregated by geometric mean to obtain final weights for the attributes [44]. Finally, using the proportionality rule, the economic value of the monetary attribute was assigned in proportion to the importance weight of the WES attributes [43] of the WES attributes.

3. Results and Discussion

3.1. CE Results

Table 12. Results of the conditional logit model.

Variable	Estimated Coefficients	Error Standard	p-Value
ASCA	0.059 (0.19)	0.3077	0.8488
ASCB	0.089 (0.29)	0.3050	0.7711
Water regulation	0.360 * (3.36)	0.1073	0.0008
Sediment control	0.358 * (3.33)	0.1075	0.0009
Water yield	0.032 (0.28)	0.1172	0.7818
Quality maintenance	0.376 * (3.61)	0.1041	0.0003
Price	−0.498 * (−5.41)	0.0921	0.0000
Number of observations	370
Log simulated likelihood	−371.4077

* Indicates statistical significance at the 1% level.

presents the results of the conditional logit model without interactions. With a confidence level of 95%, the null hypothesis that all coefficients jointly are equal to zero is rejected. Therefore, the model is accepted. The test Prob [chi squared > value] = 0.0000 evidences the overall consistency of the model, and the test of [37] evaluated the IIA, showing that the alternatives considered are relevant or consistent.

The analysis shows that “Water regulation”, “Sediment control”, “Quality maintenance” and “Price” are statistically significant. The signs of all the coefficients of the attributes, with the exception of “Price”, are positive, which means that the higher the level of the attributes, the higher the utility of the respondents, and vice versa. The negative sign of “Price” is expected, meaning that the higher the cost of accessing the improvements in the rest of the attributes, the less likely people are to opt for the improvements. This can be interpreted as the higher the tariff, the lower the respondents’ utility services.

The marginal value of the water resource components was calculated using Equations (4)–(7). The results are shown in

Table 13. Economic value of the attributes estimated with CE.

Attribute	MWTP (USD/Month per Household)
A1: Water regulation	0.185
A2: Sediment control	0.184
A3: Water yield	0.017
A4: Quality maintenance	0.193
Aggregate WTP	0.579

.

The results indicate that the monthly WTP of households is, on average, USD 0.185 for the conservation attribute “Water regulation”. For “Sediment control” it is, on average, 0.184 USD. In addition, households are willing to pay USD 0.017 monthly for “Water yield”, and USD 0.193 for “Quality maintenance”. Overall, households’ monthly WTP is, on average, USD 0.579 for the Tilacancha WES attributes.

3.2. AHP Results

The AHP technique allowed us to obtain weights, determined by the panel of experts, for each of the WES attributes. Of the 16 experts, 8 had paired comparisons that passed the consistency test, that is, having a consistency ratio (CR) < 10%, for 5 × 5 order matrices. The results of the aggregation of weights for the five attributes are shown in

Table 14. Aggregate weights and economic valuation of attributes with AHP.

Attribute	Aggregate Weight *	Pivot Value (USD/m3 Water)	Economic Value ** (USD/m3 Water)
A1: Water regulation	0.3411	-	2.83
A2: Sediment control	0.0952	-	0.79
A3: Water yield	0.2189	-	1.81
A4: Quality maintenance	0.2683	-	2.22
A5: Cost of service	0.0764	0.632	0.63
Economic value added			8.29

* Calculated using the geometric mean from the individual ratings. ** Estimated using the direct proportionality rule. Note: To obtain monthly values per household, the estimated values in the third column should be multiplied by the monthly water consumption of the households.

.

Regarding the weights of importance, the results suggest that the attribute “Water regulation” is the most important with an aggregate weight of 34.1%. The attribute “Quality maintenance” is in second place with an aggregate weight of 26.8%, in third position is “Water yield” with an aggregate weight of 21.9% and, in last place is the attribute “Sediment control” with an aggregate weight of 9.5%. The fifth attribute “Cost of service”, as explained, is not a WES attribute, but it was included in the model because it fulfills the function of pivot value, as it is an attribute with value in the market. It has an aggregate weight of 7.6%.

The economic values of the attributes were estimated based on the pivot value “Cost of service” expressed in USD/m³ of water. Therefore, the values for the attributes are also estimates based on 1 m³ of water. The results indicate that “Water regulation” has an economic value of USD 2.83. “Sediment control” has an economic value of USD 0.79, “Water yield” a value of USD 1.81 and “Quality maintenance” an estimated value of USD 2.22 per household. In aggregate, the economic value of all attributes is USD 8.29 per household. This amount represents the economic value of WES per household per 1 m³ of water. To estimate monthly values, the estimated values would have to be multiplied by the monthly water consumption of the households.

3.3. CE versus AHP

In the application of CE and AHP, some technical and methodological differences have been recognized (

Table 15. Technical differences between CE and AHP.

Indicator	CE	AHP
Estimated variables	Coefficients	Weightings (weights)
Group interviewed	Water resource users	Panel of experts
Sample size	Relatively large	Reduced
Type of method	Stated preferences	Multi-criteria analysis
Cognitive load	Lower	Higher
Experimental design	Complex and restrictive	Simple
Interaction between attributes	Allows	Does not allow
Application costs	Expensive to apply (requires field work)	Cheap (does not requires field work)

). AHP is not a traditional technique for economic valuation of environmental assets; however, its ability to determine relative weights and its simplicity has led to its recent application in the last decade to the prioritization of attributes that lack a market such as ecosystem services, or valuation of ecosystem services [26]. Because of these qualities, this study compared the results of AHP with a traditional technique in economic valuation, CE, in which there is extensive applied work [49]. Previously, the comparison of both techniques has been carried out on goods with a market [11,25,28,29].

As can be observed in

Table 16. Normalized weights of attributes with CE and AHP.

CE Results (Declared Payment Preference)		Relative Importance	AHP Results (Theoretical Stated Preference)
Attribute	Weight		Attribute	Weight
Quality maintenance	0.334	1	Water regulation	0.369
Water regulation	0.320	2	Quality maintenance	0.290
Sediment control	0.318	3	Water yield	0.237
Water yield	0.029	4	Sediment control	0.103

, there is no similarity in the prioritization of environmental attributes with both techniques, this is explained by several reasons. When different monetary attributes are used for CE and AHP, the comparison is qualitative and does not focus on the intensity [11]. Possibly, it is also due to the fact that AHP incorporates a consistency check, whereas CE does not [28]. In addition, during the AHP weighting, AHP does not take into account the ranges of the levels, whereas in the estimation of importance weights in CE, they always depend on the level [25].

Because the fifth attribute, which serves the function of monetizing the four WES attributes, is different for both techniques, the values found with both techniques have different meanings. In CE, the fifth attribute “Price” is a vector of payments (0.13, 0.38, 0.64 and 0.90 USD/month per household) and symbolizes the hypothetical payment that households would make monthly for accessing one of the improvement scenarios. Therefore, the values found with CE represent the MWTP for each WES attribute, i.e., the additional values that households are willing to pay per month to retain the WES attributes.

On the other hand, the AHP technique uses “Service Cost” as the fifth attribute. This value is the cost incurred by the company that manages the WES to produce 1 m³ of water per month, the word “produce” being understood as the actions of treatment and management of the resource. Using the proportionality rule, “Cost of Service” is assigned an economic value based on the aggregate weights of each WES attribute. In this case, unlike in the CE technique, it does not symbolize willingness to pay, but represents the economic value that the attribute has for the expert panel. Therefore, it is expected that the values obtained through AHP (Table 14) are higher than those obtained through CE (Table 13).

Theoretically, the MWTP represents the relative weight of the attributes in CE. Therefore, the results of the CE and AHP techniques were compared using the MWTP values and normalized weights. The normalization was necessary for the comparison because in both methods, a fifth attribute was added that is not part of the WEE; therefore, it is understood that the four attributes should add up to 100% of the weights (Table 16). As can be seen, CE presents the weights assigned by the payment preference of the randomly selected water users, and AHP presents the theoretical preference assigned by the panel of experts.

The results in Table 16 should be understood as an indicator of the relative importance for the two groups of individuals, the water users and the panel of experts, expressed as a ranking of the attributes with the highest preference. As can be seen, comparing one to one, there is no consistency in the ranking between the techniques. However, it can be seen that the attributes ranked 1st and 2nd in CE are ranked 2nd and 1st in AHP, respectively. Similarly, the attributes in the 3rd and 4th positions in CE are in the 4th and 3rd positions in AHP. This indicates that there is overlap in the prioritization of pairs of attributes. This means that for both CE and AHP, “Quality maintenance” and “Water regulation” are of first importance and “Sediment control” and “Water yield” are second rank attributes.

Evaluating the attributes individually, the CE results indicate that for water resource users, “Quality maintenance” is the attribute they value the most and is valued over “Water regulation” and “Water yield”. This means that users value quality over quantity “Water yield” and availability of the service (permanent water supply that is achieved naturally through the attribute “Water regulation”). This result is reasonable considering that in the city of Chachapoyas, there is still no scarcity; therefore, “Water regulation” and “Water yield” are secondary values for the users. It is observed that among the first three attributes, there is little difference between the weights; therefore, the differences in prioritization are minimal. “Water yield” is the least favored attribute with a very low valuation score. This is supported by the fact that the Tilacancha River Microbasin provides large volumes of water, of which only a part is used. Therefore, at present, “Water yield” is not a priority.

The results obtained with AHP show notable differences in the prioritization of attributes by the panel of experts. “Water regulation” is the most valued attribute, which may suggest that the experts recognize its importance in providing permanent water at all times of the year. “Quality maintenance” is the second most important attribute, followed by “Water yield”. These three attributes have similar values. “Sediment control” was assigned by the experts to the fourth position. This attribute is closely related to water quality, and it is notable that it received a very low score.

The CE and AHP techniques represent the relative importance of attributes to individual users and panel experts in two different ways: payment preferences and stated theoretical preferences. There were notable differences in the application of the techniques, with each having distinct advantages and disadvantages.

The CE technique demands greater resources since it is applied to a large sample number to ensure representative results. The comparison task is a complex process, since the interviewee is offered complex products, such as those of real markets, in which people are faced with choosing among many alternatives. From another point of view, the technique has certain advantages, since by providing the respondent with more information in terms of attributes and levels for each scenario, the results are closer to the consumer’s real preferences. The CE technique also allows the integration of socioeconomic variables to explain preferences for the environmental asset.

The AHP technique, on the other hand, requires minimal economic resources to carry out. However, it is difficult but crucially important to identify a group of experts with experience or knowledge of the environmental resource. Another advantage is that AHP allows the representation of individual preferences, an aspect that cannot be achieved with CE, since it shows individual preferences, but in a global way. The task of comparing attributes with AHP, although it follows a simple format, causes more mental fatigue to the interviewee since not as much information is presented to him/her as in CE, making the comparison task more abstract. Another disadvantage of AHP is that it does not allow the representation of utility functions, nor the integration of socioeconomic variables in the analysis. However, both approaches were powerful in determining the relative importance of the WES attributes and establishing their ranking and levels.

4. Conclusions

The research focuses on comparing individual preferences for the four priority attributes of the water ecosystem service of the Tilacancha River Microbasin using two direct valuation techniques: Choice Experiments and Analytic Hierarchy Process. This enables the prioritization of the attributes for management. The differences are based on the conditioning factors in the choice. In CE, the user of the water resource is faced with options with competitive prices, as if he or she had to buy it; therefore, he or she expresses his willingness to pay or his purchase preference. On the other hand, the AHP expert is only required to compare attributes according to a value scale. AHP establishes relative weights based on the opinion of a panel of experts representing society, and CE weights coefficients based on the opinion of users about an ecosystem service.

The preference analysis reveals that, one to one, there are no similarities in the estimated relative weights; however, on a pair-wise basis, with both methods there is a greater preference for the attributes “Quality maintenance” (CE weight: 0.334; AHP weight: 0.290) and “Water regulation” (CE weight: 0.320; AHP weight: 0.369), over “Sediment control” (CE weight: 0.318; AHP weight: 0.103) and “Water yield” (CE weight: 0.029; AHP weight: 0.237). In Chachapoyas, where the abundant water from Tilacancha is consumed, the theoretical preferences determined with AHP and the payment preferences determined with CE indicate that, in a situation of resource availability, the priority is to ensure the ecosystem’s capacity to purify water (“Quality maintenance”) and the capacity to supply it to the lower areas, including during the dry season (“Water regulation”). In this situation, the capacity of the ecosystem to avoid erosion by buffering rainfall (“Sediment control”) and the capacity of the ecosystem to produce water (“Water yield”) are of secondary value.

The relative weights of CE and AHP are monetized with different “monetary attributes”. In AHP, the “Cost of service” represents the expenditure in inputs for chemical treatment to make the resource potable, and in CE, the “Price” for the user represents hypothetical improvements to the attributes of the water resource; therefore, the weights derived from each technique are of different magnitudes. In this sense, it is convenient to use the economic values of CE to make payments for ecosystem services since they reflect changes in the welfare of the users of the ecosystem service through a participatory instrument. Because of their magnitude and nature, the AHP values are appropriate for carrying out cost-benefit analyses, which are required by regulations in many countries to demonstrate the economic viability of establishing conservation areas, and as an input for extrapolating benefits in the evaluation of public investment projects in ecosystem services or natural infrastructure projects. It is evident that the choice of one technique or another is a function of the available budget. AHP is much cheaper because it is applied using a panel of experts with few staff. CE is considerably more expensive due to travel expenses and extensive staff resources required.

Since WES attributes may vary with the inherent characteristics of each ecosystem, we do not generalize how close or transferable the results of the Tilacancha case are to other water producing ecosystems. However, we conclude that it is possible to combine the results of both methods to support decision making, as mentioned in the previous point, and highlight the specific cases where it would be appropriate to use CE or AHP to prioritize WES attributes.