Survey’s design

The survey was designed in four sections: 1) design of introductory questions regarding the perception of the main problem of the south of the province of Córdoba, and about knowledge of the calden forest; 2) presentation of an explanatory brochure on the proposed change situation: a conservation program that, over the course of 10 years, will increase the area of ​​the calden forest by 50 %, based on the control of deforestation, recovery and reforestation of the calden forest, regarding the ES provided (Figure 2); 3) consultation of the willingness to pay; and 4) sociodemographic characteristics of the household. In section 3, the WTP was consulted through the payment card format with random order. The amounts of WTP consulted, in Argentinian pesos, were: $ 5, $ 35, $ ​​60, $ 80, $ 100, $ 125 and $ 155. The response options for each amount were: a) Definitely Yes, b) Probably Yes, c) Not sure, d) Probably Not and e) Definitely not. Each of the amounts were presented on a card that contained the previously indicated response options.

The process begins by submitting the seven cards with their face down and asking the respondent to choose one of them and then turning it back. According to the response given, the interviewer proceeded differentially. If the interviewee stated a complete certainty answer - definitely yes or definitely not - they proceeded to remove those cards with respectively lower or higher amounts. Otherwise, respondents were asked to remove another card, repeating the process until no more cards remained.

Figure 2 Descriptive brochure of the current situation and the conservation proposal of the caldén forest in the south of the province of Córdoba. Source: Own elaboration. 

Later, in order to identify possible biases, we analyzed the respondents' reasons for manifesting an affirmative or negative WTP, through the following questions:

a) Which of the following statements describe more appropriately the reasons why you are willing to pay, to contribute to the financing of a land-use planning program in the south of Córdoba? Please, choose only one option.

I think it is important to control the clearing of the calden forest in the south of Córdoba and, therefore, we must take measures to order the territory.

I feel satisfied for contributing to any good cause.

Actually, I believe we will not have to make any payment.

Some other reasons

b) Which of the following statements describe more appropriately the reasons why you are NOT willing to pay to contribute to the financing of a land-use planning program in the south of Córdoba? Please, choose only one option.

I am satisfied with the current conditions of the rural environment.

I agree with a land-use planning program, but I am not in a position to make a monthly payment.

I agree with a land-use planning program, but I refuse to make a payment of any amount.

Some other reason.

The respondents who said they were willing to pay and selected options 2 and 3 were considered purchase of moral satisfaction bias (^{Kahneman & Knetsch, 1992}) and hypothetical bias respectively (^{Azqueta, 1994}). In the case of respondents who are not willing to pay and selected options 1 or 2, those were considered valid negative responses; while options 3 and 4 were considered protest responses.

Derivation of the empirical model

The data to estimate the WTP were obtained with payment cards chosen by the interviewee in a random way (^{Carthy et al., 1998}) with a polychotomous response format (^{Welsh & Poe, 1998}). For the treatment of the data, this is transformed into a double dichotomous format (^{Hanemann & Kanninen, 1998}), which considers three values: Aic is the first value consulted to the respondent; Aiin it is a lower amount in case the answer is negative; Ai,su it is a higher amount if the answer is positive. In this sense, if the first and second answers are negative, we conclude that 0<C<Atin ; if the second answer is positive then Aiin<C<Aic . When the answer is positive it is consulted by the Aisu, if the answer is negative it is concluded that Aic<C<Aisu and if the answer is positive Aisu<C<∞ . In the case of the polychotomous format, Aic , Aiin and Aisu are translated in intervals defined by the maximum amount that the respondent is willing to pay ( Aisimax ) and the minimum amount that he does not accept to pay Ainomin .

If yi1 and yi2 are defined as dichotomous variables that capture the first and second answers, then it is assumed the linear function DAP _i (z _i, µ _i) = z _i ß _i + µi, where z _i is a vector of explanatory variables, β _i is a vector of parameters and µ _i an error term that is distributed N(0, σ ²) (^{Hanemann & Kanninen, 1998}). In the double bounded dichotomous format, the answer is given by a probability with the following expressions, where ϕ is the cumulative distribution function of the WTP:

Piyesyes⁡=Pr (yi1 i=1;yi2 i=1 | zi)= ϕ(z'iβσ-Aisuσ)

Pinono⁡= Pr (yi1 i=0;yi2 i=0 | zi) =1-ϕ(z'iβσ-Aiinσ)

Piyesno⁡=  Pr (yi1 i=1;yi2 i=0 | zi)=ϕ z'iβσ-Aicσ-ϕz'iβσ-Aisuσ

Pinoyes=Pr (yi1 i=0;yi2 i=1 | zi)=ϕz'iβσ-Aiinσ-ϕz'iβσ-Aicσ

These estimates are obtained by the maximum likelihood method, assuming a joint density function given by:

L=∏i=1nPiyesyesdiyesyesPinonodinonoPiyesnodiyesnoPinoyesdinoyes

where d ^yesyes , d ^yesno , d ^noyes and d ^nono are indicator variables that have the value 1 or 0, depending on the relevant case for each individual; that is, a given individual contributes to the logarithm of the probability function in only one of its four options (^{Hanemann & Kanninen, 1998}). The logarithm of the likelihood function is given by:

ln⁡L=∑i=1ndiyesyesln⁡Piyesyes+dinonoln⁡Pinono+diyesnoln⁡Pisino+diyessiln⁡Pinoyes

The expected value for the WTP is given by:

E(DAPiziβ)=z´β

Specification of the econometric model

The econometric model was specified with seven explanatory variables that affect the change in utility. In order to design public policies, political decision makers need to know not only the existence of the WTP, but also the differences in this valuation between groups. In this sense, ^{Carson and Hanemann (2005}) recommend analyzing if there are differences between income groups, residents of different geographical areas, as well as between attitudes, knowledge and uses of the valued good or service. In accordance with this recommendation, the explanatory variables shown in Table 1 were selected for the development of the following function:

Pyesyes=β0+β1MPP+β2rio iv +β3edad +β4ing2+β5ing3+β6ing4+β7ing5+β8vinagr2+β9vinamb2+η

where P ^yesyes is the dependent variable that represents the probability of obtaining a positive answer for the WTP. The answer options were: a) Definitely Yes (DY), b) Probably Yes (PY), c) Not sure (NS), d) Probably not (PN) and e) Definitely not (DN). The answers DY and PY were considered positive answers (equal to 1) and the rest were considered negative (equal to 0).

WTP’s parameters, mean and median were estimated applying maximum likelihood method considering significance tests (Z test) and goodness of fit with the Wald tests and likelihood ratio. The data was processed with Stata 12 software, using the command doubleb (^{Lopez-Feldman, 2012}). Stata 12 directly calculates βi and σ (standard deviation) and in this way it is possible to interpret the effects of the explanatory variables.

The amounts obtained are expressed in Argentinian pesos (ARS) at 2012current prices and US dollars (USD, converted with a currency exchange rate 1 USD = 4.54 ARS). In the WTPAdded=N.WTP^i , N is the number of households and WTP^i is the household’s willingness to pay. To annualize the monthly amounts of WTP a social discount rate of 2.4 % per year was considered.

Identification of biases and answer inconsistencies

Table 2 shows the type of answers obtained in order to identify the valid answers for the econometric analysis. In Río Cuarto and Villa Huidobro, 64 and 62 % of WTP positive responses were obtained, of which 13 and 12 % respectively resulted with moral satisfaction purchase biases, and 3 and 11 % with hypothetical bias. Regarding all the positive and negative answers without biases, the valid answers were 79 and 77 % in Río Cuarto and Villa Huidobro, respectively. In the case of inconsistent responses, 11 % of answers were identified in Río Cuarto and 8 % in Villa Huidobro e. g. whose answer was Probably Yes to 35 ARS (7.70 USD) and Probably NOT to 5 ARS (1.10 USD).

Regarding the protest responses, 10 % were recorded in Río Cuarto and 15 % in Villa Huidobro. In those cases, the respondent argued that the government is the one in charge of the land-use planning program, for the conservation of the calden forest. The cases of WTP with inconsistencies and protest responses were removed from the analysis (^{Bateman et al., 2002}; ^{Broberg & Brännlund, 2008}), while the affirmative responses with biases were controlled in the estimation, although they were statistically not significant.

Sociodemographic and attitudinal characteristics of the sample

Table 3 shows the descriptive statistics of the sample obtained, regarding the valid answers. In the case of the continuous variables, the average number of household members was three, coinciding with the population average of both locations (^{Dirección General de Estadística y Censos [DGEC], 2008}); the average age of the respondents was 47 years in Río Cuarto and 53 years in Villa Huidobro. In the case of the categorical variables, a similar proportion was observed in both locations: 1) the predominant monthly household income is the category 1 500-4 000 ARS (330-881 USD); 2) the majority of the heads of family surveyed were women, whose most frequent level of education is primary school. In contrast, there are statistically significant differences (P < 0.1) on calden forest knowledge; in Villa Huidobro, 77 % of respondents have knowledge of this, while in Río Cuarto only 21% expressed they knew it. On the other hand, less marked differences were detected in the use of forest and in the relationship of the respondents with the agricultural and environmental sector. The use of the calden forest is moderate; 18 % of the respondents in Río Cuarto showed some type of use and 9 % in Villa Huidobro. In both locations, the level of connection with the agricultural sector was also moderate: 18 % in Río Cuarto and 35 % in Villa Huidobro; however, the relationship with environmental organizations is even lower: 7 in Río Cuarto and 3 % in Villa Huidobro.

Econometric model

The estimated econometric model has good goodness of fit (Table 4). According to Wald's statistical tests and likelihood ratio, the model had good joint significance in both cases (P < 0.01), rejecting the null hypothesis which indicates that the coefficients of the explanatory variables are equal to zero.

Four of the seven parameters estimated were significant and consistent with the expected signs. The estimated coefficient of the variables WTP and age were significant and with a negative sign; that is, the higher the WTP and age, the lower the probability of obtaining a positive answer. The coefficients of the estimators of the variables regarding income (ing _i ) and relationship with the agricultural activity (vinagr ₂) were significant and with a positive sign i. e. the greater the household's ing _i and vinagr ₂, the greater the probability of an affirmative response of WTP. The expected effects of the coefficients of the significant variables were consistent with economic theory. All the coefficients of the variables regarding income had a positive sign and increased to higher income levels.

The signs of the estimated coefficients of forest knowledge (conoc_bo and location (rio iv) were different from those expected and were not statistically significant (P > 0.1). The estimated coefficient of the variable conoc_bo resulted with a negative sign, which would imply that the probability of a positive response of WTP of the respondents with previous knowledge of the existence of the forest is lower than that of the respondents who said they did not know about the forest. Possibly, people with forest knowledge compare it with lands considered more productive commercially as agricultural land and do not take into account all the forest ES. Besides, the question about knowledge was direct and simple and does not allow speculating too much with the result. On the other hand, the estimated coefficient of the variable rio iv was positive; that is, the probability of a positive response of WTP of the respondents from Río Cuarto (most distant city) would be higher than that of the respondents in Villa Huidobro. This coincides with the results found by ^{Mombo et al. (2014}), who estimate a higher WTP in urban households compared to rural households, close to forest areas; however, it differs from the results of ^{Lockwood et al. (1993}).

The estimated coefficients vincamb ₂ and usayafect were also statistically non-significant (P > 0.1). This result is very important, since one of the main critical points of the CVM is the possibility of inducing biases in the WTP assessment responses (^{Bateman et al., 2002}).

The estimated WTP has a relatively high precision regarding the close relationship between the average/median and intervals obtained. The average WTP was 52 ARS (11.45 USD) with a variation of ±8 ARS (±1.76 USD), equivalent to an annual WTP of 615 ARS (135.4 USD) per household with a range between 518 ARS (114 USD) and 714 ARS (157 USD). The estimated WTP is in the range of economic values ​​reported for remnant native forests in private and state establishments. ^{Thompson et al. (2002)} estimated an annual individual WTP of 217 USD (988 ARS) to prevent the conversion of the American oak woodland to agricultural systems in California, United States. ^{Juutinen et al. (2014)} estimated an annual WTP per household between 44 USD (200 ARS) and 71 USD (324 ARS) for maintaining and expanding the forests of the Finnish State for recreational use. ^{Chen and Nakama (2015)} estimated the WTP per household at 4.5 USD (64 ARS) to maintain and expand the forest in three locations of Okinawa, Japan. ^{Lienhoop and Völker (2016)} estimated an annual WTP per household of 9 USD (41 ARS) for an increase in forest cover in the Saxony region, Germany.

Regarding that the south of Córdoba has 115 854 households (DGEC, 2008), the current annual value of the WTP of the urban population of southern Córdoba is 71.3 million pesos (ARS) (15.7 million dollars [USD]), meanwhile, the current value projected at 10 years for the urban population of Río Cuarto is 629.3 million pesos (ARS) (138.6 million dollars [USD]) (Table 5).

The results of the present study, obtained through contingent valuation, differ from the values reported by ^{Carreño and Viglizzo (2007)}. If we consider the average annual aggregate WTP and the area of forest valued (56 472 ha), the value of ES of the calden forest is 1 264 ARS·ha^-1 (278 USD·ha^-1), less than 6 139 ARS·ha^-1 (1 352 USD·ha^-1) estimated by ^{Carreño and Viglizzo (2007)}. This overestimation is expected, because the method used by these authors does not take into account the budgetary constraints of households, but the estimated WTP in this study does consider them.