Technical Comments

The Effectiveness of Parks

Science  10 Aug 2001:
Vol. 293, Issue 5532, pp. 1007
DOI: 10.1126/science.293.5532.1007a

Bruner et al. (1) examined the effectiveness of parks in the tropics, drawing on survey data to support three conclusions: (i) parks have been effective; (ii) parks need more support; and (iii) parks should remain a central component of conservation strategies. These conclusions remain equivocal.

The Bruner et al. study focused on 520 parks in areas categorized as “partly natural” or “human dominated”; the subject parks had to have been established for at least five years and to exceed 5000 hectares in area. “Directors of conservation organizations and protected area agencies helped identify a representative” subset of 82 parks, Bruner et al. note. A further 11 supplementary samples—5 of 15 conservation areas in Laos and 6 parks from outside the original sphere of interest—were also included in the study, which brought the total sample size to 93.

A questionnaire was used to gather information on conditions within and surrounding the park. Only one response was obtained for each park. “Park managers or park staff constituted 56% of respondents,” according to Bruner et al. [note 13 of (1)]; morever, nine of the parks in the sample were ones that receive “some level of support from Conservation International,” the organization that conducted the survey. Correlations suggest that the 11 supplementary samples and the 9 Conservation International–supported parks do not differ from the general trend of park effectiveness (r = −0.16,P = 0.06, and r = 0.05, P = 0.3, respectively), but there is no way to test for bias in the selection of a “representative” subset or among respondents completing the survey. Respondents are not identified in the data file, so it is not possible to examine whether park managers differed from researchers in their views on park conditions. More crucial, it is not clear whether responses reflect only the opinion of respondents or whether respondents made a genuine attempt to gauge the incidence of fire, clearing, grazing, hunting, and logging within and surrounding the parks. Because of these deficiencies, the data should be considered anecdotal rather than substantive.

Bruner et al. (1) presented analyses that suggested that the incidence of clearing, grazing, and burning is lower and the abundance of game and commercial tree species is higher within parks than in the adjoining 10-km-wide surroundings. The authors attribute these differences to the effectiveness of parks, but other factors, such as natural features and cultural traditions that discourage disturbance, may be at work as well. A more parsimonious suggestion is that the apparent differences may always have existed or may be area-dependent (a 10-km buffer represents twice the median park area). The data suggest that park creation tends to reduce clearing, foster regrowth, or both, within the protected area (pairedt = 1.8, P = 0.03), but they offer no reliable insights into current clearing patterns because the survey employed different categories for clearing in areas within and surrounding the parks (2).

The index of park effectiveness developed by Bruner et al. involved rescaling the clearing data from six categories onto four. Such a rescaling of categorical data involves subjectivity, results in a loss of information, and is not necessary to examine correlations. I computed an alternative index of effectiveness, E, by averaging the (unadjusted) difference between values of the various anthropogenic threats reported inside and outside the parks,E = [(ci co ) + (bi bo ) + (hi ho ) + (li lo ) + (gi go )]/5, where c, b,h, l, and g represent values for clearing, burning, hunting, logging, and grazing, respectively; the subscript i denotes the value for inside the park boundaries, and the subscript o represents the value for outside the park area. The resulting index offers better insights into relative performance of the parks, but not into their absolute performance (3).

My assertion that responses denoted “not applicable” and “not available” both denote missing values resulted consistently in fewer degrees of freedom than reported by Bruner et al. (1). Rather than use Spearman's rank correlations, I plotted all the possible predictors against park effectiveness and selected appropriate transformations to linearize any trend. Correlations obtained in this way differed substantially from those reported in table 2 of (1); significant correlations and selected departures from Bruner's findings are reported in Table 1. It is noteworthy that a park's budget—whether total, national, or donor, whether recently increased or decreased, and whether absolute or per unit area—had no significant correlation with effectiveness in this study or in the Bruner et al. study.

Table 1

Correlation of predictor variables with index of park effectiveness E developed in this study. Second column shows transformation used to linearize data trend. rs value reported in (1) is shown for comparison. Details on predictor variables are given in text and in (1).

View this table:

The data in Table 1 are not orthogonal, and several variables seem to reflect the same trend (for example, the correlation between local educators and guards is 0.58). A stepwise regression suggests that only two variables—local educators and border demarcation—are relevant to effectiveness, under the following relation:Embedded Imagewhere E is the index of park effectiveness,Ledu is the number of local staff (full-time equivalents per 100 km2 of park area) working primarily on educating and building awareness in local communities about the park's goals and mission, and D is the percentage of the park's border that is clearly identifiable, with the data divided into six categories (0 to 10%, 11 to 30%, 31 to 50%, 51 to 70%, 71 to 90%, and 91 to 100%). Standard errors for these parameters are 0.5, 0.06, and 0.25 (P <0.001, P = 0.001,P = 0.3), respectively. This implies that parks are more effective if local people know that the parks are there and know why they are there. It supports the contention that tropical parks are effective and that differences within and surrounding parks may be due to the establishment of a protected area. It is regrettable that the anecdotal nature of the underlying data detracts from this finding.

I conclude that the evidence supporting the first conclusion of Bruner et al., that parks have been effective, is equivocal. Respondents to the survey held that conditions within parks are better than in surrounding areas, but parsimony begs whether these apparent differences have always existed or could have been be a sampling artifact related, for example, to how the parks were selected, how the survey was designed and conducted, and the differing scales of measurement used to compare areas within and outside of the parks in question. There is also no evidence to support the second conclusion of Bruner et al.—that increased support will improve effectiveness of parks. This lack of evidence prevails despite the fact that any personal biases of directors involved in selecting the sample and managers involved in completing the survey would likely have contributed to a more pro-park result than in a more rigorous survey.

The third conclusion of Bruner et al., that parks should remain a central component of conservation strategies, remains untested. Their data and analyses shed no light on the efficacy of protected areas relative to other conservation initiatives. Because both land and financial support for conservation are limited, the issue is not whether parks are good, but whether parks are better than the alternatives to them. The alternative selected by Bruner et al., a 10-km band around a park, is not an informative one. A better test, for considerations of habitat conservation, would be a comparison of habitat quality within protected areas with that in community-managed areas and other multiple-use initiatives. For considerations of species conservation, comparing the success and costs of protected areas against those of ex situ conservation efforts may be more relevant.


Response: Vanclay makes a number of methodological critiques of our study (1) and presents several new analyses of our data. The latter, however, largely support rather than detract from our findings, and Vanclay presents no substantive evidence to back his methodological critiques. We therefore conclude that his claims do not challenge our major findings. We respond to his main points in the following paragraphs.

Selection of parks. Vanclay's first criticism is that our sample may not be representative and that there is no way to test for biased park selection. Instead of random sampling, we asked experts to select a representative sample of parks—both for practical reasons (using existing contacts to obtain information in logistically difficult locations) and to ensure that we captured a wide range of conditions in our small sample from each country. Vanclay's own tests support ours in suggesting that the sample was indeed representative: He showed that the 11 supplementary parks did not differ from the sample with respect to effectiveness. Our original correlation tests, meanwhile, showed that the sample did not differ significantly from the group from which it was drawn with respect to age, geographic distribution, or IUCN (World Conservation Union) category.

Even if the sample were not representative of all tropical parks, the finding that 18 million hectares of protected areas are broadly effective at protecting their resources would remain important. As a point of comparison, another widely promoted conservation option, Forest Stewardship Council (FSC)–certified tropical forestry, currently covers a total area of 2 million hectares of natural forest (2), less than 15% of the area of the parks included in this study.

Bias of respondents. Our study design gave park managers little incentive to overestimate effectiveness. We guaranteed anonymity to respondents and agreed to publish only aggregate findings. Further, recent accounts of park failures in the popular media attest to the willingness of managers to make public information about threats to their protected areas. The fact that there were no significant differences in responses for the six parks in our sample that had multiple respondents (admittedly a small sample) suggests a lack of bias in the responses, as does the fact that many parks in our sample that showed effectiveness against some threats showed a marked lack of effectiveness against others. As to Vanclay's question of whether respondents made a “genuine attempt” to provide accurate responses, we think that respondents, who were universally among the world's experts on the parks for which they provided data, were indeed able to rank the threats and impacts to their parks into broad categories. The numerous missing data points that Vanclay notes attest to the willingness of respondents to leave answers blank where they could not provide accurate information.

Explanations for park effectiveness.Although Vanclay agrees that our data show that parks are in better condition than their surroundings and that parks have largely resisted deforestation over time, he questions whether management is responsible for these results. He offers no evidence to support his speculations about alternative causes, however, and his own analyses actually contradict his claims. His correlation and regression analyses both found that management activities do indeed correlate significantly with effectiveness. His claim that “area-dependent impacts” could be responsible for the findings in our study (1) is inconsistent with the lack of a negative correlation between park size and effectiveness in both his analyses and ours. Similarly, his suggestion that parks tend to be in better condition than their surroundings from the outset is inconsistent with our finding that 40% of parks increased their natural vegetative cover after establishment. Had these parks been founded in areas that were already protected by isolation or natural features, they would not have demonstrated such an increase. Moreover, even if some parks were originally in better condition than their surroundings, that they have remained so over long periods (a median of 21 years) with major threats at their borders is a strong testament to management effectiveness.

Correlation with effectiveness. Vanclay claims that the results of his correlation analyses differ substantially from ours; actually, the opposite is true. In light of the fact that he used transformations, no rescaling, parametric statistics, and a smaller sample size (see below), the two sets of results are remarkably similar and, indeed, are strongly correlated (r = 0.8,P <0.005; Fig. 1). Vanclay's analyses are therefore consistent with our basic conclusion that—despite great variation in park location, context, geography, and other important characteristics— basic management activities correlate with effectiveness.

Figure 1

Scatterplot of r values reported by Vanclay against those reported by Bruner et al. (1) for factors potentially correlated with park effectiveness (r = 0.8, n = 10, P<0.005). Plot includes nonsignificant r values reported by Vanclay but not presented in (1).

The main distinction between Vanclay's analyses and ours is Vanclay's conclusion that environmental educators are a principal determinant of effectiveness. This and other distinctions may be explained by several differences in methodology. Vanclay did not calculate a measure of effectiveness for individual parks unless data were provided for all five of the impacts examined in our study, and he excluded these parks from his analysis. In contrast, we included parks that had one missing value, in which case the measure of park effectiveness was the average of four impacts, not five. This was the case when respondents could not provide accurate data or when the measure was not applicable to the park in question (e.g., the incidence of logging in a savannah park). We therefore had a much larger sample size for our analyses than did Vanclay. Further, whereas Vanclay used regression, we used nonparametric statistics, which are more conservative and appropriate given the categorical data and the difficulties in achieving linear transformations of some of the variables. These differences in statistics and sample size suggest that our results are more robust. That said, we believe that a range of factors beyond those we found significant may be important to effective management.

The importance of increasing support for parks. Vanclay states that our analyses do not demonstrate that increasing support for parks will increase their effectiveness. This conclusion seems unwarranted, given that his findings mirrored ours in showing that management activities do indeed correlate with effectiveness. That neither analysis found a correlation between absolute funding correlated and effectiveness may be due to the significant variance in local purchasing power of a U.S. dollar across the countries included. A more meaningful comparison is units of personnel or service, which obviously incur a cost. Vanclay's analyses therefore reinforce our conclusion that increased support for management (financial and otherwise) will increase effectiveness.

Vanclay concludes by claiming that our study cannot recommend continued support for parks, because we did not compare them with other conservation strategies. We agree that because funding for tropical conservation is limited, rigorous comparisons of conservation approaches are urgently needed. According to Vanclay's reasoning, however, no conservation strategy should be supported at present, because no comprehensive comparison of approaches exists. We disagree. In the absence of a rigorous comparison, it is sensible and appropriate to choose options that have been shown to be effective. For this reason, our study provides strong support that parks should remain a central component of conservation strategies.


Navigate This Article