Paper park performance: Mexico's natural protected areas in the 1990s
Introduction
Tropical deforestation remains a pressing concern. For example, in both Latin America and Africa, deforestation rates averaged one-half of one percent per year in the first decade of the 2000s, five times the global rate (FAO, 2011). This deforestation contributes to a host of global and local environmental problems including biodiversity loss, climate change, soil erosion, and flooding (Harris et al., 2012, Gibson et al., 2011, Chomitz, 2007). Protected areas, or parks, are a cornerstone of the policy response. Although the number of parks in tropical countries has expanded dramatically in the past three decades, in general, insufficient financial and human resources have been devoted to managing them (Balmford et al., 2003, Bruner et al., 2004, Wilkie et al., 2001, WDPA, 2014). The term paper parks has been used to characterize extreme cases of insufficient funding and management capacity (Bonham et al., 2008, Carey et al., 2000).
Given the proliferation of severely under-resourced parks in tropical countries, it is important for policymakers to understand whether and under what circumstances they help stem deforestation. Ex ante, the answer is not obvious. On one hand, paper parks could help stem forest clearing if loggers, ranchers, and other agents of forest-cover change are deterred by the threat of future regulatory enforcement, nonregulatory sanctions, or social mores. But on the other hand, establishing paper parks could contribute to deforestation by creating de facto open-access regimes where extractive activities can be pursued with impunity.
Rigorous evidence on this issue is thin. This is not to say that studies of park performance in developing countries are lacking. A considerable literature uses remotely sensed (satellite and airplane), survey, and other quantitative data to gauge the effect of a park or set of parks on forest-cover change (Geldmann et al., 2013, Nagendra, 2008, Naughton-Treves et al., 2005). However, many of these studies have methodological limitations that tend to bias their estimates of effectiveness upward (Blackman, 2013, Joppa and Pfaff, 2010). Many measure effectiveness by simply comparing rates of deforestation inside and outside parks, without controlling for the fact that parks are typically sited on land with preexisting characteristics—such as inaccessibility and rough terrain—that inhibit land-cover change (Joppa and Pfaff, 2009). As a result, these evaluations give parks credit for stemming deforestation that is actually due to the characteristics of the land on which they are sited. In addition, many studies do not control for leakage, the tendency of some parks to shift forest-cover change to nearby areas.
A newly emerging literature relies on matching and other statistical (“program evaluation”) techniques to correct for these problems (Blackman, 2013, Joppa and Pfaff, 2010). Because this approach is relatively new, however, questions remain. To our knowledge, no such study has focused explicitly on the issue of paper parks. In addition, rigorous analysis of Mexican parks is scarce.
To help fill those gaps, we evaluate the effectiveness of all 56 protected areas established in Mexico prior to 1993 in stemming deforestation between 1993 and 2000. As discussed below, financial and human resources allocated to parks during this period were minimal. We use high-resolution land-cover data derived from satellite images along with statistical techniques (covariate and propensity score matching) that control for nonrandom siting. We address three questions about the 56 parks: What effect did they have on deforestation inside their borders? What effect did they have on deforestation outside their borders? Were parks with certain characteristics particularly effective in stemming deforestation?
Mexico's forests, more than half of which are primary, comprise 65 million hectares, one-third of the national territory (FAO, 2011). The majority are governed by more than 2000 communal forest management units called ejidos and communidades, an artifact of the land reform that accompanied the Mexican revolution (FAO, 2011, Madrid et al., 2010, Bray et al., 2006). Historically, deforestation has been a severe problem in Mexico. Between 1990 and 2000, clearing of all types of forests averaged just over half of 1 percent per year, generating the seventh-highest net annual forest loss of any country in the world, and the clearing of primary forests averaged more than 1 percent per year (FAO, 2011).
Although Mexico established more than 30 NPAs between 1917 and 1979, most were quite small. The total land area in NPAs did not grow substantially until the 1980s (Fig. 1). By 1993, 56 NPAs comprising more than 6.5 million hectares had been created.
Throughout this period of rapid expansion, the creation and administration of NPAs were not coordinated at the national level. Mexico's 1988 comprehensive environmental law provided the legal underpinnings for the National System of Natural Protected Areas (Sistema Nacional de Áreas Protegidas, SINAP), administered by the Ministry of the Environment. The system was formally inaugurated four years later, in 1992. The same year, the Global Environmental Facility (GEF) provided US$25 million to support an elite group of 10 NPAs within SINAP (World Bank, 2002, Pérez Gil Salcido, 1995).
Despite those positive developments, during the 1990s—the period during which we measure NPAs’ effect on deforestation—virtually all Mexican NPAs were paper parks. SINAP was a system in name only: central management and coordination were minimal (Pérez Gil Salcido, 1995). Moreover, funding, staffing, planning, and enforcement for all but the elite group of 10 NPAs receiving GEF funds were grossly insufficient. In fact, the vast majority of parks lacked any financial support, personnel, or management planning (Rivera and Muñoz, 2006, Cervigni and Brizzi, 2001, Pérez Gil Salcido, 1995). From 1990 to 1994, Mexican funding for all nonelite NPAs was carved out of budgets for other programs and averaged just $60,000 per year, roughly US$0.01 per hectare per year (Rivera and Muñoz, 2006) (Fig. 3). In 1995, for the first time, the national budget included a specific allocation for SINAP. Over the next four years, that allocation increased somewhat but still averaged less than US$0.85 per hectare per year, far less than what was needed for the most basic management (Rivera and Muñoz, 2006, Cervigni and Brizzi, 2001). By comparison, the United States, Canada, and European countries currently spend an average of US$28.00 per hectare per year on protected areas, and Mesoamerican countries spend an average of US$4.59 (Flores, 2010, Bovarnick et al., 2010).
It was not until the 2000s that SINAP funding and administration improved significantly. In 2000, administration of SINAP was transferred to a new, semi-autonomous institution called the National Commission for NPAs (Comisión Nacional de Áreas Naturales Protegidas, CONANP), consisting of a headquarters in Mexico City and nine regional offices (Fig. 2). SINAP's budget and activities have increased markedly since that time (Rivera and Muñoz, 2006).
Finally, it is important to note that in addition to chronic shortages of resources, Mexican NPAs face a second significant challenge: less than one tenth of NPA land is state owned. The majority is owned by ejidos, communidades, and other communal tenure institutions and at least 10 percent is owned by private concerns (World Bank, 2002). Therefore, NPA effectiveness in stemming deforestation depends critically on creating incentives for communal forest management units to adopt sustainable practices (Cervigni and Brizzi, 2001, Pérez Gil Salcido, 1995). However, most of these units, particularly the smaller ones, lack the capacity to do that (Anta Fonseca, 2006). Indeed, previous research indicates that some communal tenure was associated with higher rates of deforestation during the 1990s (Bonilla-Moheno et al., 2013). Given the role of land tenure in NPA management, it will be important to control for this characteristic in our statistical analysis.
What evidence do we have on the effectiveness of Mexican NPAs in stemming deforestation during the 1990s? As noted above, evaluations of protected areas using matching and other program evaluation techniques to control for both nonrandom siting and leakage have only recently begun to appear. Studies using other methods have reached varying conclusions. For example, Mas (2005) evaluates the effect on deforestation of a single large NPA in southeastern Mexico (the Calakmul Biosphere Reserve) by comparing 1993–2000 rates of forest clearing inside the NPA and rates in similar adjacent areas. He concludes that NPAs halved the annual deforestation rate. Using similar methods, Foo and Sánchez-Cordero (2008) examine the 69 NPAs established prior to 1997 and find that the majority were “effective” in stemming deforestation from 1993–2002. By contrast, Durán-Medina et al. (2005) find that the average 1980–2000 rate of deforestation in a national sample of more than 50 NPAs was significantly higher than that for selected community forestry enterprises in two states in southern Mexico (Guerrero and Qintana Roo).
More recent rigorous evaluations of Mexican NPAs during the first decade of the 2000s—which as noted above, was a period of substantially increased funding and management—suggest that they have been effective in stemming deforestation inside their borders. Pfaff et al. (2014a) find that on average, NPAs avoided about 3 percent deforestation inside their borders between 2000 and 2005 while Sims and Alix-Garcia (2014) find that (a different sample of) NPAs reduced national baseline deforestation by 20 percent between 2000 and 2010.
Section snippets
Matching estimators
As noted above, the main challenge we face in attempting to accurately measure pre-1993 NPAs’ effect on deforestation is that they were not randomly sited. Rather, as we will show, these NPAs were disproportionately located on land with certain preexisting geophysical, climatological, and socioeconomic characteristics that affect deforestation. For example, they tended to be sited relatively far from population centers and at relatively high elevations, both attributes that tend to discourage
Sample
We use a dimensionless plot of land defined by latitude and longitude coordinates as our unit of analysis. Drawing on the spatial data sources listed in Table 1, for each plot, we collected spatially explicit data on outcomes (deforestation), treatment (location in or near an NPA), and control variables—a variety of socioeconomic, geophysical, and climatological land characteristics. In addition, for plots inside NPAs, we collected data on NPA characteristics, which we used to construct
Results
For most of the dozens of combinations of estimators and subsamples discussed below, MSB (our measure of matching quality, presented in square brackets in our tables of results) is below the 3–5 percent threshold for acceptability (Table 2, Table 3). This is particularly true for the covariate matching estimators that we emphasize: in general they do a better job than our propensity score estimators of matching NPA plots to similar non-NPA plots. For the sake of concise exposition, in what
Discussion
In this section, we focus on three aspects of our results: NPAs’ heterogeneous effects on deforestation inside their borders, possible explanations for NPAs’ effects in Region 3, and possible explanations for NPAs’ spillover effects on deforestation outside their borders.
Conclusion
We have used high-resolution land-cover change data derived from satellite images along with matching techniques that control for nonrandom siting and spatial spillovers to measure the effectiveness of Mexico's 56 pre-1993 NPAs in stemming deforestation between 1993 and 2000, a period when these ‘paper parks’ received minimal financial and human resources. Our principal findings are as follows. First, most of our estimates of the magnitude of NPAs’ effects on deforestation inside their borders
Acknowledgments
Funding for this research was provided by the Tinker Foundation, the National Space and Aeronautics Administration through the SERVIR Applied Science Team, the Swedish International Development Corporation Agency, Sida, through the Environment for Development (EfD) Initiative, and the Swedish Research Council, Formas, through the Human Cooperation to Manage Natural Resources (COMMONS) program. We are grateful to Nisha Krishnan and Yatziri Zepeda for expert research assistance; to Zepeda and the
References (70)
Evaluating forest conservation policies in developing countries using remote sensing data: an introduction and practical guide
For. Policy Econ.
(2013)- et al.
The effectiveness of listing under the U.S. Endangered Species Act: an econometric analysis using matching methods
J. Environ. Econ. Manage.
(2007) - et al.
Effectiveness of terrestrial protected areas in reducing habitat loss and population declines
Biol. Conserv.
(2013) - et al.
Governance, location and avoided deforestation from protected areas: greater restrictions can have lower impact, due to differences in location
World Dev.
(2014) Bounding a matching estimator: the case of a Norwegian training program
Oxf. Bull. Econ. Stat.
(2001)- et al.
Large sample properties of matching estimators for average treatment effects
Econometrica
(2006) - et al.
Forest conservation and slippage: evidence from Mexico's National Payments for Ecosystem Services Program
Land Econ.
(2012) - et al.
Measuring the effectiveness of protected-area networks in reducing deforestation
Proc. Natl. Acad. Sci. U. S. A.
(2008) Forest certification in Mexico
- et al.
Global variation in terrestrial conservation costs, conservation benefits, and unmet conservation needs
Proc. Natl. Acad. Sci. U. S. A.
(2003)