The two possible alternatives to consider with regard to the conservation of Pentro horse were the following:
• Alternative A: The respondent’s family, along with all Molisan families, does not make a donation. The ‘‘Pentro horse’’ protection project is not carried out. The ‘‘Pentro horse’’ could become extinct.
• Alternative B: The respondent’s family, together with all Molisan families, makes a donation. The ‘‘Pentro horse’’ protection program is carried out. The Pentro horse is saved from extinction.

No information was provided on the socioeconomic characteristics of the population in the area.

552 completed questionnaires

Sponsorship by means of a single donation

A bio -economic model was used to estimate the costs of conservation. A mixture model was used to estimate the distribution of willingness to pay.

Estimation of costs of the conservation program
The breeding costs revealed were quite low because the horses are raised in a wild state. In fact, the breeding cost for the colts is only 3 €/year, while for the adult horse the cost rises to 82 €/year. If a conservation program is adopted, the breeding costs rise, because intervention is required in order to improve the present breeding techniques. The costs are estimated at about 258 €/horse per year. In particular, the breeders would pay a rent to the municipality to guarantee access to grazing in the Pantano della Zittola (17 €/horse per year) and they would also pay for the prophylaxis (42 €/horse per year). Moreover, in order to avoid that the extra load of livestock impoverishes the Pantano della Zittola resource, a major alimentary supplement will be necessary, which adds an extra 116 €/horse per year. Finally, it is estimated that the manpower requested for this type of breeding is 83 €/horse per year (Table 1).
The estimate of lost revenues is based on the revenues that the breeder would have obtained if the conservation program were not activated. Assuming that the size of horse population remained constant over the years, the breeders would continue to sell 54 horses, 48 colts and six end-of-career horses each year. The age at which the colts are sold is influenced by the climatic conditions and this causes some variation in terms of costs and gains.
In this case, it was hypothesised that 70% of the colts would be sold at 6 months and 30% the following year. As the number of horses increases, the breeders have the possibility of selling the surplus colts (both male and female). The value of these new sales, net of the costs, is subtracted from the lost revenues and from the breeding costs sustained. A break-down of the resulting prediction in cost structure is summarised in Table 2, for a period of 14 years.

Estimation of benefits of the conservation program
The parameter estimates for the WTP distribution are shown in Table 3 (model 1). Table 3 also shows the estimates of mean and median WTP values, with their respective confidence intervals approximated using the Krinsky and Robb (1986) procedure.
The truncated mean value is 33 € while the median value is 19 €. Multiplying the mean and the median by the number of families living in Molise (117 138) an estimate of the aggregate value is derived, which amounts to 3.8 million € for the mean and 2.2 million € for the median. While the first value is to be considered in a Kaldor-/Hicks cost-benefit test, the second value may be important to local politicians, given the interest in median-voter behaviour in a referendum context.
Model 2 in Table 3 differs from model 1 because of the covariates. The number of observations for which the analysis was conducted was reduced to 417 because 17% of the sample (87 interviews) had to be excluded due to incomplete information regarding the income of the respondents. The socio-economic covariates that show significance include the degree of education, family income, knowledge of the Pentro horse prior to the interview (Horse), and the indicator variable for the respondent being a member of environmental associations (Environmentalist). 
Both the mean and median values are higher than those from the model estimated without covariates. The truncated mean is equal to 35 € and the median is equal to 24 €. The estimate for the aggregated mean has a value of 4.1 million € and that for the aggregated median is 2.8 million €.

Results of the cost benefit analysis
The comparison of costs and benefits of the conservation program for the Pentro horse must take into account that the latter are present value estimates, while the estimated costs refer to various moments in time across a 14-year time period.
In the context of CBA for environmental programs, particular attention is given to the discount rate through which the costs and benefits present values are computed. In this specific case study, the rate of choice is the discount rate
suggested by government authorities to evaluate public investments. The present value of costs for the conservation program over the 14 years discounted at both 3.5% (discount rate suggested by the Italian government) and 0% are shown in Table 4, while in Table 5 the data supporting the CBA computed by considering different parameter estimates (mean and median) of the WTP distribution are presented. It can be seen from the NPV that the benefits are always greater than the costs, even when the discount rate is equal to 0.
According to the results even in the worst scenario the benefit/cost ratio equals to 1.67, justifying a conservation policy.