To analyze individuals' decisions, a survey was used employing questions which reflected the choices to be made in a PGG. This game is usually played in a laboratory by n individuals. This game has been simulated in the context of structured surveys with actual CPR owners. In a lab setting, all individuals i receive an endowment e and simultaneously and independently decide whether to keep this endowment for themselves or invest a given amount gi ε [0, e ] in the CPR in question. The total amount contributed by all n participants together equals g = Σ gi , where 0≤ gi ≤ e is multiplied by a number m (marginal per capita return), with 0 < m < 1 < nm . Therefore, when the PGG is played once, and considering that players are rational and selfish in maximizing utility, contributing 0 is a dominant strategy. However, the joint group outcome is maximized when everybody contributes with the full endowment. The payoff function of users can be expressed as:

A survey was carried out in North-west Spain, in the region of Galicia, from winter 2013 until summer 2014, interviewing a sample of 75 forest owners. In order to facilitate the first contact with forest owners, information from the websites of communal forests was obtained. After an initial phone call or contact e-mail, we visited the place where they live (a total of 29 different places) conducting face to face surveys at their homes; or after attending their annual associa­tion meetings. Some participants preferred to answer via on­line rather than face to face, petition that was accommodated. Information from 29 communal forests belonging to 31 different councils was collected. The four provinces of Galicia are represented in this small sample; with 55.71% of the observations from the province of A Coruña, 22.86% from Lugo, 17.14% from Pontevedra and 4.29% from Ourense. The survey requested information about the characteristics of communal forests, from which they are members, with the objective to reflect members' knowledge. In addition, questions related to their opinion about public authorities were also included, and the way in which decisions related to forest are made. Furthermore, the survey also included questions to test whether PCA were functioning. PCA related questions can be seen in Table 1 and a full description is added in the Table S1 [suppl.]. The PGG was presented in another section of the questionnaire.

Table 2 shows an example of the game presented to participants. Basically, forest owners had to decide about the sharing of endowment that the community receives as a consequence of the actions carried out in order to keep forests in good conditions. As it can be seen, each common owner has to decide the sharing of three different endowments. In this way, we can test whether the size of the endowment is important as has been found in other previous literature. In addition, the survey also included some questions to understand the degree of effort made by members to maintain a forest in good condition, and therefore assesses their opinion of other forest owners' efforts (these questions were asked before participating in the PGG). It is important to highlight that the PGG was conducted individually. The same strategy was employed by Braaten (2014) with the goal of promoting comprehension amongst participants. Finally, socio-demographic characteristics were elicited in the last part of the survey.

Individuals' allocations made in the stated preference exercise were modelled as a function of the size of the endowment (reward), the forest characteristics of individuals and the most relevant social and socio-economic characteristics, among other variables. Furthermore, the PCAs were also incorporated into the regression model in order to test the effect of other perceptions towards the collective management.

A first Ordinary Least Squares (OLS) regre­s­­sion ­was estimated. Using the Breusch-Pagan (1979) test, it was assessed the presence of heteroskedasticity, finding that the test carries a value of 24.93 for a Chi-squared with 1 degree of freedom and an associated p -value of 0.00. Therefore, we reject homoscedasticity. To deal with this problem, an OLS model clustered by individual was estimated ⁴ . In addition, and in order to take into account the panel nature of our data, based on the fact that three different responses from each individual were collected, the between and within variability was also analyzed using a Random Effects (RE) model ⁵ . The dependent variable in this regression equation is the vector of the total distributions (the amount of the endowment shared with regards to the total amount of endowment available), labeled as (Y). This vector represents the percent of endowment that individuals decide to send to the public fund. The vector of explanatory variables ( Xi ) is grouped into four categories that include: the size of the endowment to be shared Xs , the forest characteristics Xf , the socio-economic characteristics Xse , while Xpca represents the PCA.

The estimated model corresponds to the estimation of the following equation:

One of the main aims of this paper is to analyze the impact of PCA in sharing decisions. It is important to note that with the aim of including the effect of PCA several questions were included in the survey (see Table 1); and in order to measure the impact of PCA, we constructed indicators to measure the existence of these principles. Table 1 provides information about the type of questions used to estimate each individual PCA. Questions were answered with a "yes" or "no", and recoded as 1 and 0, respectively. Thus, in order to create the indicators, the affirmative responses were added up.

It is expected that PCA may have a significant impact on explaining the sharing of the endowment. The aim was to investigate whether individuals feel that they are involved in the decision-making process; whether they are aware that their rights are respected; and whether there are controls to assure them, among others factors. Thus, the hypothesis proposed is to test whether the fulfillment of PCA enumerated by Ostrom (1990) influence sharing decisions in the community.

The second question of analysis is whether the size of the endowment matters in sharing decisions. To study this hypothesis, a variable which represents the size of each individual endowment has been included. In this sense, Anderson et al. (1998) found that contributions increase with the endowment; while Andersen et al. (2011) concluded that stakes are important but in the case of ultimatum games. Therefore, this is a novelty aspect in the literature of PGG, due to the fact that no previous studies were found that analyzed the effect of the size of endowment in this type public good game. Specifically, the hypothesis to be tested is:

In terms of the dataset composition, 64% of the respondents were men, with an average age of 55 years and the average number of household members was around three persons. With respect to the level of studies, 52% of the respondents have a basic level of studies or no formal studies ( basic studies ), and with regards to the income-related questions, about 24% of participants earn less than €1,000/monthly ( lower income ).

With regards to the characteristics of forests, about 44% of them had more than 100 communal members ( over 100 members ). It was also detected that in terms of location, 58% of the communities in this sample were close to the sea ( near to the coast ) (as opposed of being distant rural communities). This geographical indicator is important, since it provides information about the degree of economic dynamism of the area. Previous studies have shown that communities located near the coast are more active and dynamic, mirroring the economic local conditions ( Alló & Loureiro, 2016 ). With the goal to know more about their engagement in forest management, participants were asked about their level of effort compared with the rest of the owners; and 21% stated that their own effort is higher than the rest of the members ( greater effort ).

With the aim of obtaining more information about their social characteristics, additional questions related to whether they were born in locations near to forests were included, with 75% answering in a positive way ( place born ). About 52% responded that their relationships with the rest of owners were very or quite good ( good relationship ).

Finally, it was also taken into account the survey mode: whether the surveys have been completed via online instead of face to face, with 33% preferring the online mode ( online ). Table 3 shows the summary statistics for the variables analyzed.

The distribution of the endowment is presented in Figure 1. It is important to note that each individual had to make three allocation decisions due to the three different sizes of the endowment. The amount of endowment shared can be considered as a proxy of cooperation ( Stoop et al. , 2012 ). Assessing the results, it was observable that when the endowment to share was €10, around 86% decided to allocate the full amount to the common fund, and around 8% decided to invest half of the endowment in the public fund, while the rest (6%) decided to keep the endowment for themselves. Thus, on average the mean contribution was about €9. However, when the endowment was higher, the allocation changed. Specifically, for an endowment of €100, 5% decided not to share this endowment, also another 5% decided to send a positive value below the half to the public fund; while 11% decided to split the endowment at 50%. Moreover, around 3% sent to the public fund more than half, but less than 100% of the endowment, and 76% allocated the total €100 to the fund. On average, the contribution to the fund was about €85. This decision changed slightly when the endowment was even higher. For an endowment of €1000, 3% kept the full amount, around 9% decided to share less than half with the fund, while 28% decided to share half with the public fund. In addition, 7% of the owners share more than half with the public fund, and finally, 53% decided to send the total amount to the fund. On average, the mean contribution to the fund for this endowment was about €759. Therefore, the higher the size of the endowment, the higher the amount of endowment that collective owners keep for themselves. Nevertheless, it is important to highlight that in spite of that, they were quite generous, given that the percentages of endowment kept by themselves was quite small (around €1 on average for an endowment of €10, around €13 for an endowment of €100, and about €240 for an endowment of €1000). In the survey of literature conducted by Chaudhuri (2011) , it was found that average contributions were about 40% to 60% of the optimal level. This is an interesting result in terms of cooperation, particularly in the absence of mechanisms to encourage it (such as penalties, or communication strategies, among others). Stoop et al. (2012) concluded that without any kind of contact among players, cooperation is not guaranteed.

The survey results related to the PCA are presented in Table 1. With regards to the principle of clearly defined boundaries , around 72% of respondents believed that in their communities all members knew each other; 49% thought that residents who were not members of the communal forest knew this fact, while the 69% considered that neighbors respected community members in practice.

When examining congruence between appropriation and provision rules and local conditions , it is important to note that only 53% of commoners thought that the rules that they had to follow were appropriate for resource conservation. With regards to the regulation's effectiveness; only 50% believed that it was effective . Moreover, 59% thought that the existing rules respect local traditions . Analyzing the results for the principle of collective choice arrangements , 85% confirmed that forest decisions were taken collectively with individuals equally represented. In addition, around 50% of participants stated that there were controls to ensure compliance by the forest community, and 44% stated that there were also controls for outside members ( principle of monitoring ). With regards to the principle of graduated sanctions , 51% of commoners affirmed that there were penalties for people who did not comply with the requirements, and 35% stated that these penalties varied depending on the degree of infraction in the case of the endowment. Only 22% expressed that there was a quick mechanism to resolve conflicts and 46% believed that their rights and decisions are respected by non-members and the administration for the endowment treatment.

In order to understand the factors that drive the allocations of endowment to the CPR, we estimated a robust OLS and a RE models. The main results are summarized in Table 4. First, we examined and discussed the estimates with regards to the size of the endowment for both models. We observed that the coefficient obtained for size of endowment was negative and statistically significant. Therefore, when the endowment to be shared was larger, less proportion was shared with the public fund. This means that with lower endowments, individuals are willing to share proportionately a greater amount of endowment with the public fund.

We also considered the effect of different charac­teristics of owners to understand their decisions. Thus, Bechtel & Scheve (2014) concluded that socio-demographic characteristics play an important role in social dilemmas. Specifically, we obtained that the lower income indicator was negative and statistically significant. Moreover, those individuals who thought that they made a greater effort than the rest, were more likely to allocate the endowment to the public fund. These may be more altruistic or more committed commoners.

With regards to the effect of PCA, we observed that the principles related with the congruence between appropriation and provision rules and local conditions and graduated sanctions showed negative and statisti­cally significant coefficients. Therefore, individuals who believed that these principles were being fulfilled are less willing to share the endowment. Furthermore, we also observed that three PCA had a positive impact promoting more cooperative decisions, specifically, the monitoring , the conflict resolution mechanisms and the minimal recognition of rights . These results may be implying that when individuals feel that they are involved in the management process, are more willing to share the endowment towards the CPR. Therefore, results show that PCA fulfilment may encourage more generous decisions, but not always (as in the case of congruence between appropriation and provision rules and local conditions and graduated sanctions ).