Maximum Yield?:
Sustainable Agriculture as a Tool for Conservation

Richard Margoluis
Vance Russell
Mauricia González
Oscar Rojas
Jaime Magdaleno
Gustavo Madrid
David Kaimowitz

CONTENTS

Introduction

Doing Conservation Better: BSP's Analysis and Adaptive Management Program

One of the major direct threats to biodiversity in terrestrial ecosystems is deforestation caused by the conversion of forests to pasture for livestock and the expansion of agricultural lands. To address this threat, conservation organizations have been struggling for decades to find the magic formula that balances the socioeconomic development needs of local populations with the urgency of conserving natural resources.

One of the major obstacles to determining which conservation strategies are most suitable to which sites, is that there exists relatively little guidance written for conservation practitioners to enable them to make wise choices about best practices. Many times, valuable scientific research results are communicated in a way that is incomprehensible and, therefore, of little value to project managers.

In the 1980s, sustainable agriculture gained popularity among international conservation organizations as a tool for project managers to combat deforestation, but there was little concrete evidence of its conservation utility. As conservation organizations gained experience in implementing sustainable agriculture projects, they began to ask: Under what conditions does sustainable agriculture work to help reach conservation goals? To what extent does sustainable agriculture decrease rates of deforestation?

Two of the organizations asking these questions — Defensores de la Naturaleza in Guatemala and Línea Biósfera in Mexico — approached the Biodiversity Support Program (BSP) for assistance in analyzing the benefits of their sustainable agriculture programs. BSP, Defensores de la Naturaleza, and Línea Biósfera worked together to design and implement a learning process that would not only gauge the utility of sustainable agriculture as a conservation tool but also build the capacity of the two participating nongovernmental organizations (NGOs) to do this type of research in the future.

Our results shed light on the conditions under which sustainable agriculture may be used as an effective conservation strategy and on an approach to practitioner-designed and implemented research that produces concrete, practical results. In this publication, we present guiding principles for the use of sustainable agriculture in conservation and process lessons on how to best undertake this type of action-inquiry. We hope you find this information useful in your work.

Maximum Yield?: Sustainable Agriculture as a Tool for Conservation

Deforestation is one of the primary threats to biodiversity in tropical forests around the world. Deforestation has many direct causes, including conversion of forests to pasture for livestock, expansion of agricultural lands, commercial logging, and urbanization. Indirectly, deforestation is influenced by a host of other factors, including road construction, technological change, agricultural prices, household incomes, and land tenure and security.

In recent decades, the destruction of tropical forests has been a primary concern of conservation organizations. These organizations have tried many different approaches to reduce deforestation, including direct protection, restoration, education, policy changes, and the use of various incentives. However, relatively little practical guidance exists for conservation project managers in the field to compare different conservation tools to determine which one has the highest probability of success at their site. What is missing are clear, useful, and practical principles for designing, managing, and monitoring conservation strategies to reduce threats to biodiversity.

To make wise choices about best practices — what works, what doesn't, and why — we must learn about the conditions under which specific strategies are most effective. This is no easy task. To gauge the appropriate use of a given conservation tool, learning must be systematically and routinely incorporated into project implementation, and it must be done across a suite of projects to determine the conditions under which the tool works.

In recent years, sustainable agriculture has been promoted as an effective tool to reduce deforestation in tropical areas. This analysis explores the conditions under which sustainable agriculture works to achieve conservation in tropical forest settings.

Why Study the Link Between Sustainable Agriculture and Conservation?

In many tropical areas of the world, farmers practice swidden agriculture, which is sometimes referred to as "slash-and-burn." In this traditional approach to agriculture, farmers typically cut down a forested area, let it dry, and then burn it. Ash from the fire increases soil fertility, and fields normally maintain crop yields for about two to three years. After a few years, however, weed infestation becomes problematic and soil fertility declines to the point that farmers are forced to start the cycle of cutting forest, burning, and planting again. In areas with vast amounts of available land and low population densities, swidden agriculture may not pose a major threat to biodiversity. But such places are becoming increasing difficult to find.

In the 1980s, sustainable agriculture projects gained popularity among international conservation organizations that were attempting to control deforestation. Sustainable agriculture has since been promoted as a conservation strategy in much of the tropical world, including Latin America, the Caribbean, Asia, the Pacific, and Africa. For many decades before the advent of sustainable agriculture, development organizations had promoted household-level agricultural intensification as a strategy to increase family farm yields while decreasing required labor inputs.

The term sustainable agriculture is used by many people to mean many different things. In the context of some conservation projects, and for the purposes of this study, sustainable agriculture programs are designed to promote farmer-based technologies that intensify production and that, according to implementing conservation organizations, will reduce deforestation. These programs typically incorporate a number of techniques such as those listed in the following box.

The type of sustainable agriculture that is the focus of this research has been promoted by both conservation and development organizations to increase the production of subsistence crops, such as maize and beans, that farmers grow primarily for household consumption. Many of these techniques, however, can also be applied to cash crops such as coffee and cardamom. One major assumption of the conservation community has been that expansion of subsistence crops — not cash crops — is the major cause of deforestation in fragile tropical areas.

Sustainable agriculture, as we define it in this publication, is meant to decrease the need to cut and burn new lands every few years. According to implementing conservation organizations, the major underlying assumption is that, by increasing investments in land and increasing yields, farmers will be less likely to move as frequently and will, ultimately, need less land to produce the amount they require to feed their families.

As conservation organizations have gained experience in implementing sustainable agriculture projects, they have discovered some of the challenges in making it work as a conservation strategy. Given their various experiences, conservation project managers want to know: Under what conditions do sustainable agriculture projects work to help reach conservation goals? To what extent does sustainable agriculture decrease rates of deforestation? How do sustainable agriculture projects affect recovery of damaged and fragmented forestlands? To what extent do sustainable agriculture projects enable conservation organizations to win the confidence and trust of community members so they will be more open to conservation messages and programs in the future? What specific tools and techniques are most useful in promoting sustainable agriculture as a conservation strategy? Are there other conservation benefits to sustainable agriculture projects that have not been previously contemplated? What do we know at this point about sustainable agriculture that will allow us to enhance its effectiveness for future conservation efforts? These questions, which must be answered to gauge the utility of sustainable agriculture as a conservation tool to address biodiversity loss around the world, drove our research.

What We Wanted to Know

Numerous studies have assessed the socioeconomic benefits of sustainable agriculture projects. These studies have looked primarily at variables such as changes in household agricultural productivity and yield, returns to labor, and income (in particular, see Buckles et al. 1998; Faris 1999; Lutz et al. 1994). Very few studies, however, have addressed the conservation benefits of sustainable agriculture projects per se. Even fewer studies have attempted to quantitatively measure the effects of sustainable agriculture on conservation goals. We also found relatively little practical guidance for conservation project managers on how to implement successful (in terms of conservation outcomes) sustainable agriculture projects. Although the insights and conclusions of many of the studies we reviewed are useful to practitioners, they are for the most part communicated in a way that most practitioners find difficult to interpret and use.

Much of the Biodiversity Support Program's (BSP) programmatic work focuses on areas of high biodiversity that are, to some extent, formally protected. Many of the local or national partner nongovernmental organizations (NGOs) with which we have worked use sustainable agriculture as a conservation tool around protected areas. In recent years, some BSP partners have expressed skepticism concerning the efficacy of sustainable agriculture as a conservation strategy. In 1996, BSP and two of its local NGO partners in Latin America — Defensores de la Naturaleza in Guatemala and Línea Biósfera in Mexico — decided to collaborate to learn about the conditions under which sustainable agriculture is effective in achieving conservation success.

Based on the scarcity of practical guidance for using sustainable agriculture as a conservation tool, we developed two main goals for this study.

In addition to these two collective goals, BSP had a third goal that focused on the process of doing partner-based, applied research in the context of adaptive management. BSP wanted to learn about some of the requirements of designing and implementing an effective learning portfolio. This approach is designed to bring together multiple project partners to learn about the conditions under which a specific conservation tool or strategy (in this case, sustainable agriculture) works and does not work.

To this end, we included the following additional goal:

Results in Brief

The main purpose of this study was to determine the conditions under which sustainable agriculture functions as an effective conservation tool. We used a list of conventional wisdom, distilled from the literature and based on the perceptions of project managers, to guide our inquiry into sustainable agriculture. Following is a brief summary of what we found, organized around the main themes of the conventional wisdom.

Area planted to subsistence crops and its relationship to deforestation

Sustainable agriculture, as defined in this study, does not necessarily contribute to decreases in area planted to subsistence crops. In Guatemala, farmers who used sustainable agriculture techniques planted more area to maize than farmers who did not use sustainable agriculture. In Mexico, farmers who used sustainable agriculture techniques planted less area to maize than their non-user counterparts. This leads us to conclude that sustainable agriculture, as defined in this study, does not always lead to decreased pressure on new lands for subsistence agriculture.

Sustainable agriculture, as defined and used in this study, was associated with increased investments in labor per hectare in Mexico and decreased investments in labor per hectare in Guatemala. Farmers who used sustainable agriculture in Guatemala put their saved labor to use in ways that worked against conservation goals by increasing the amount of area planted to maize or establishing cash crops in forest areas. Involvement in sustainable agriculture programs, therefore, does not necessarily lead farmers to labor input savings or motivate them to act in ways that are supportive of conservation.

Access to land is an important determinant of area planted and, thus, deforestation. In Guatemala, where land is relatively available, farmers lacked appropriate incentives to be efficient in their use of land and increased their maize production by increasing area planted. In Mexico, where land access is restricted, farmers were much more efficient in their use of land and increased maize production by increasing yield. We conclude that sustainable agriculture programs that promote the same techniques farmers used in our study sites are unlikely to contribute to decreased rates of deforestation where access to land is not restricted.

Based on the results of this study, reduction in the use of fire was perhaps the greatest conservation benefit of the sustainable agriculture techniques farmers used at our two sites. In traditional agriculture, fire is used to prepare agricultural plots and control weeds and pests. Sustainable agriculture discourages the use of fire, which is one of the primary threats to habitat in the Sierra de las Minas and El Ocote Biosphere Reserves.

Fallow and its relationship to forest recovery

In our study, sustainable agriculture did not contribute to fallow amount or duration and, therefore, had no effect on forest recovery.

Use of chemical inputs and contamination of the environment

The farmers included in our study use very little, if any, chemical fertilizers or pesticides. Because of these small numbers, there was no evidence that sustainable agriculture contributed to decreased contamination from chemical inputs. However, as sustainable agriculture use contributed to decreases in the use of fire to prepare fields, we can conclude that it reduces pollution from smoke.

Attitudes about the environment

In general, farmers who used sustainable agriculture techniques were more aware of the importance of biological resources and their relationship to agricultural practices. In addition, sustainable agriculture programs proved to be crucial to building trust and confidence in the communities in which Defensores de la Naturaleza and Línea Biósfera work.

Community organization as a mechanism to contribute to conservation

Community organization played different roles in the sustainable agriculture projects at our study sites. In Guatemala, Defensores de la Naturaleza's sustainable agriculture program served as a mechanism to encourage farmers to participate in subsequent conservation activities. In Mexico, the highly organized nature of communities provided the foundation for the adoption and diffusion of sustainable agriculture throughout the project area.

The Conventional Wisdom on Sustainable Agriculture and Conservation

BSP, Línea Biósfera, and Defensores de la Naturaleza wished to test some of the major underlying assumptions related to the use of sustainable agriculture as a tool for achieving conservation. The assumptions that we include here come from our review of the available literature and discussions with researchers, project managers, and other professionals in the fields of conservation and development. Based on our review, we generated a list of key hypotheses that we have summarized into our list of conventional wisdom or presently held beliefs and assumptions related to the linkages between sustainable agriculture interventions and biodiversity conservation.

We divide the conventional wisdom into two main sections: conventional wisdom related to the direct impacts of sustainable agriculture on biodiversity conservation and conventional wisdom related to its indirect impacts. We first define the main variable we wish to investigate and the conventional wisdom most associated with this variable.

Direct Impacts of Sustainable Agriculture on Biodiversity

The literature and conservation professionals generally assume that sustainable agriculture will have a direct impact on conservation by decreasing rates of deforestation through reduction of demand for new agricultural lands by farmers. This linkage seems simple enough: increase crop production per unit of land around areas of high biodiversity, and rural poor farmers will not need to deforest more land for agriculture to meet household demands. The belief is, in essence, that if you can attract farmers living around areas of high biodiversity to a development intervention that has such high economic returns to labor, then the conservation benefits will flow naturally. As a result of these sustainable agriculture interventions, it follows that project managers could expect to see a deceleration of the advance of the agricultural frontier into areas of high biodiversity and, therefore, decreased rates of deforestation. In addition to affecting rates of deforestation, the conventional wisdom holds that sustainable agriculture programs have a direct impact on forest regeneration and contamination of the environment.

Rates of Deforestation

Definition: Deforestation is the loss of primary or mature secondary forest through cutting or burning. In the case of much subsistence agriculture — the target of the sustainable agriculture programs that are the focus of this study — farmers cut down and burn forests for agriculture. Once land is depleted of its nutrients and weed infestation becomes difficult to control after only a few years, farmers slash-and-burn more to open new agricultural fields. As population pressures intensify, or technologies, market forces, and policies change, this process of cyclical forest destruction leads to increased rates of deforestation.

Forest Regeneration

Definition: Forest regeneration refers to the extent of forest regrowth after land is no longer used for agriculture.

Contamination of the Environment

Definition: Contamination of the environment from agricultural practices is evident in many forms, including pollution from chemical fertilizers and pesticides, erosion of agricultural lands that causes sedimentation of rivers and streams, and production of smoke from burning during field preparation.

Indirect Impacts of Sustainable Agriculture on Biodiversity

Many project managers believe that sustainable agriculture's greatest value to biodiversity conservation is the indirect benefit it provides by functioning as a way for conservation organizations to win the trust and confidence of community members. Project managers believe that by focusing on issues that are most important to rural poor populations (such as agriculture), community members are more inclined to work with conservation organizations on future projects more directly linked to conservation (such as strict protection or environmental education). This approach has been recently classified as the bridge strategy for integrating conservation goals and development activities.

The conventional wisdom holds that community members' attitudes will become more supportive of conservation activities and messages once a conservation organization wins the trust of the community. In addition, organization of the community around an issue unrelated to conservation will increase a community's capacity to organize itself for other conservation-related activities in the future.

Attitudes Concerning Conservation

Definition: Attitudes concerning conservation include perceptions by farmers related to the relationship between biodiversity and the quality of their agriculture, their family's health, and the environment in which they live, including water and air.

Participation in Community Organizations

Definition: In many rural communities, local organizations are an important part of the social structure and management of community affairs. Community members may participate on development, education, or infrastructure committees. Organizations such as cooperatives and religious groups also play a vital role in community life. In some countries, communities participate in broader regional or national organizations as well. Many conservation organizations view community organization as a mechanism to work efficiently with dispersed communities.

This conventional wisdom serves as our basic framework to better understand the conditions under which sustainable agriculture is an effective strategy to reach conservation goals.

What Did We Do?

This study was designed primarily to determine the conditions under which sustainable agriculture is effective as a conservation tool and to determine principles for its implementation. But the study also was designed to determine the best way to promote learning within implementing organizations and how to effectively share lessons learned with the broader conservation community. Our approach to this project can, therefore, be divided into two main sections: (1) determining conditions and principles and (2) helping project partners to answer their own questions.

Determining Conditions and Principles

Determining the conservation effects of sustainable agriculture projects was a much more difficult task than we originally expected. Because proponents of sustainable agriculture argue that adoption of sustainable agriculture techniques slows the advance of the agricultural frontier and, therefore, slows rates of deforestation, our initial response was to examine this relationship at a broad geographical or regional level.

At first, it seemed to be pretty easy. All we had to do was determine where farmers were using sustainable agriculture techniques, measure the changes in the movement of the agricultural frontier, and presto! we would be able to measure the effects of sustainable agriculture on deforestation! Unfortunately, the puzzle of determining causality is much more complex.

We had to look for a different approach to making the link between sustainable agriculture and deforestation. Trying to do it on a broad geographical scale was impossible for a variety of reasons, including the following:

• Precise rates of deforestation are difficult to calculate and the information did not exist for our sample. In the two sites that were a part of this study, no reliable data existed to calculate rates of deforestation either before or after the sustainable agriculture projects began. Therefore, there was no way we could measure how rates changed over time. Although some aerial photography existed for one of the sites, it was incomplete for the years included in the project. Finally, on a regional scale, the mechanism through which the expansion of agriculture leads to deforestation is a relatively slow process. Although the immediate effects of clearing of forest for agriculture are readily seen on a scale of a couple of hectares, rates of deforestation are much more difficult to calculate and measure over short periods of time (for example, two or three years) when examining large geographic areas.

• Determining control or comparison groups is difficult. Some adoption of sustainable agriculture techniques by farmers occurred in many of the communities in both study sites. Finding comparison communities, or areas in which no sustainable agriculture was adopted but that were otherwise similar to sites included in our study, proved nearly impossible. In other words, we could not control for differences between communities, something we would have had to do if our unit of analysis was the community.

• The proportion of farmers who adopted sustainable agriculture varied over space and time. In some communities in our two study sites, only a handful of families used sustainable agriculture techniques. In other communities, adoption was close to 100%. Likewise, in some communities, adoption rates increased over time while in other communities, adoption rates decreased over time. This variability made it challenging, if not impossible, to link adoption of sustainable agriculture at the community level to changes in deforestation rates.

• Many other variables affect deforestation. Although expansion of subsistence agriculture is a direct cause of deforestation, it is not the only cause. Deforestation may also occur because of clearing of land for pasture, construction of homes and new communities, and commercial logging. In addition, many underlying or indirect factors influence the relationship between agriculture and deforestation, including market, sociodemographic, political, and cultural factors (Kaimowitz and Angelsen 1998). We had no way to control for these factors across communities and regions in order to look at the sustainable agriculture-deforestation link on a broad geographic scale. This particular issue proved to be the biggest hurdle to overcome — one that in the end was insurmountable.

So, if we could not look at how sustainable agriculture influenced rates of deforestation on a regional level, how then could we measure this relationship? How could we precisely and specifically measure causality between sustainable agriculture and conservation if we could not do it by looking across a large region where sustainable agriculture is practiced? The answer required a fundamental shift in the way we had conceived the study. We decided that, if we could not measure the effects of sustainable agriculture on deforestation at a regional scale, then perhaps we could measure this relationship at a different scale.

The conventional wisdom we outlined in the previous section is clear about the mechanism through which sustainable agriculture influences conservation. Although the expected impact is regional in nature, it starts with individual farmers and their families making decisions about land-use management — where and how they carry out agricultural activities. In its most basic form, therefore, the effects of sustainable agriculture on conservation should be detectable in individual household-level plots of agricultural lands.

Understanding deforestation attributable to subsistence agricultural expansion at a regional scale can be simplified by understanding deforestation from agricultural expansion at a family farm scale. In essence, regional deforestation attributable to subsistence farming is the sum of all deforestation that occurs at the household level for agricultural purposes, assuming no changes in other variables that affect the number of farmers or their behavior. Deforestation at the household level is a reflection of the amount of land farmers need to clear to plant crops to provide for their families. It is relatively easy to measure changes over time in area planted and, thus, the demand for new land, at the household level.

Similarly, it is difficult to accurately measure regional rates of forest regeneration. We can, however, measure the extent to which individual farmers allow forest recovery to take place. We can measure this by looking at the amount of land farmers have in fallow and the length of time they leave the land in fallow. The reasoning was similar to that for the relationship between rates of deforestation and household area planted: the greater the amount and duration of land left in fallow by a farmer, the greater the contribution to forest regeneration.

Decisions regarding land use in rural subsistence societies occur principally at the household level. It is also at this level where the myriad factors that affect land-use patterns have their greatest impact. For these and the above reasons, our best option for measuring the association between sustainable agriculture and conservation outcome proved to be at the household level. At this level, we could compare the conservation outcome of those farmers who used sustainable agriculture techniques with those of farmers who did not use such techniques. For some of our analysis, it was necessary to disaggregate the household-level data even further into agricultural plot units, because some farmers had more than one plot of land and their agricultural practices sometimes varied between plots. This approach allowed us to compare plots in which farmers used sustainable agriculture techniques with plots in which farmers did not use sustainable agriculture techniques. These scales — household and plot — enabled us to deduce the impacts of sustainable agriculture on conservation at a regional scale.

Our Sample

The sample for this study was determined, in large part, by the organizations that were involved with BSP in the initial conceptualization of this research project. Línea Biósfera and Defensores de la Naturaleza have historically worked in two protected areas that, in many respects, are very similar. In addition, since about 1991, they have been involved in promoting sustainable agriculture as a conservation tool in and around protected areas. The two organizations were, in fact, two of the original NGOs in Central America and Mexico that were involved in a World Wildlife Fund (WWF-US)-supported project designed to promote sustainable agriculture as a conservation tool. Additionally, organizations from Brazil, Peru, and Honduras were originally involved in this WWF-US project that lasted until about 1998. Serving as the primary trainer and facilitator for the project was a Honduran NGO, COSECHA, founded to promote sustainable agriculture in Latin America and around the world.

To address the first two goals of this study, we had to carefully select the sites and farmers to include in our sample. If we selected wildly different sites, with completely different environmental, social, and cultural factors influencing sustainable agriculture adoption and conservation, then it would have been virtually impossible for us to determine useful guiding principles for project managers working under similar conditions. If we selected sites that were very similar to each other in many respects, then we would run the risk of producing principles that were applicable only to those sites and not generalizable to other sites. The challenge was to come up with a sample of sites and households that were similar enough to control for some of the major confounding factors that could influence the sustainable agriculture-conservation relationship, but different enough that we could compare the influence of specific factors and conditions between sites.

The trade-off was clear. We could either include a wide range of projects under widely varying conditions and generate very general guiding principles, or work with a small, focused subset of projects to establish precise and specific principles that could also be applied to other projects under similar conditions. We decided to pursue the latter because there is little concrete guidance that project managers can use to select and implement various interventions under different conditions. In addition, we thought it prudent to test out our assumptions and methods on a smaller sample, and then possibly include other sites in a subsequent study. Finally, for reasons related to available budget and staff time, working with a limited number of sites that were close to each other was the best option.

In order to strike this balance, we developed the following list of criteria that we used in selecting sites for the study:

Environment and Geography Factors

• Project is located in Mesoamerica. Primarily for logistical reasons, we needed to find projects that were relatively close to each other. By selecting only projects in Mesoamerica, we also controlled for a variety of social and cultural factors.

• Project takes place in a mountainous area. Different sustainable agriculture techniques have different uses depending on the environmental conditions in which they are applied. Many techniques are used solely to combat some of the challenges to farming, such as erosion, in mountainous areas. Selection of techniques and their utility is thus often dependent on slope.

• Project is located in tropical moist forest. Selection of sustainable agriculture techniques is also dependent on rainfall and other climatic conditions. The techniques most used in arid conditions, for example, are often different from those selected for areas that receive much rain.

Social, Cultural, and Economic Factors

• Farmers live in communities that are rural and agrarian, situated next to or in a protected area. Land-use patterns often are determined by the socioeconomic situation of the people who live in a given region. People living in urban areas will use land differently from people living in rural areas. As sustainable agriculture use is related to agricultural practices in general, it is important to select farmers who are similar in this respect. Location near a protected area is important because, for our purposes, sustainable agriculture must be implemented as a tool to achieve biodiversity conservation goals.

Subsistence farmers in our study include those farmers who plant maize and beans primarily to feed their families. These farmers may, however, sell some of their harvests to earn cash income to buy household items and services needed by family members. In addition, these farmers may plant cash crops for additional income.




• Farmers own small family farms. Farmers who plant crops almost exclusively to feed their families are different from farmers who plant crops for primarily commercial reasons. In addition to employing different agricultural practices, amount and types of inputs are usually different between these two types of farmers, as is the area of land that they cultivate. Only subsistence farmers whose primary crops were maize and beans were included in the study because the planting of these crops has been the primary focus of sustainable agriculture projects in the past.

• Farmers are relatively poor and have access to limited resources. Socioeconomic status has some bearing on how farmers work their fields and on their willingness to try the sustainable agriculture techniques that are promoted in these types of projects. Likewise, access to resources will have some bearing on adoption rates and conservation outcomes.

Management Factors

• Sustainable agriculture is used as a biodiversity conservation tool in and around a protected area. For our study, the goal of the sustainable agriculture intervention must be conservation. Because sustainable agriculture is believed to have both socioeconomic and environmental impacts, the outcome of these two factors would probably be different depending on the primary goal of the implementing organization.

• Project is managed by an NGO and is implemented in multiple communities. Implementation of a sustainable agriculture project by different types of institutions will influence outcome as well. For example, the conservation impact of implementation by a national agriculture agency focused on family production will probably be different from that of a local NGO focused on biodiversity conservation.

• Implementing NGO has worked in the relevant sustainable agriculture extension program for five years. The effects of sustainable agriculture projects do not happen overnight. Time is needed to determine how adoption of sustainable agricultural techniques influences factors in both the socioeconomic and conservation realms.

After an extensive search, we found three sites that fit these criteria. In the end, we included only the El Ocote Biosphere Reserve and the northern side of the Sierra de las Minas Biosphere Reserve described above.

Because we wanted to measure the conservation impacts of sustainable agriculture and we had decided to use households and agricultural plots as our units of analysis, we needed to compare farmers who used sustainable agriculture techniques with farmers who did not use these techniques. We were particularly careful to define precisely what it meant to be included in the study as a farmer who uses sustainable agriculture (referred to here as "SA User") and what it meant to be a farmer who did not use sustainable agriculture ("SA Non-User"). If farmers used any of the sustainable agriculture techniques that had been promoted by the participating NGOs, then they were classified as SA Users.

In addition to classifying farmers, we also classified individual plots because some farmers had more than one plot (although most had only one plot) and SA Users did not necessarily use sustainable agriculture in all of their plots. If any sustainable agriculture technique was used in a plot, we classified it as an "SA Plot." If no sustainable agriculture techniques were used in the plot, we classified it as a "Non-SA Plot." During preliminary interviews with candidate farmers, we determined user status in order to immediately classify each household. We classified plot status later during farmer interviews.

It turned out that not of all the techniques initially promoted by the implementing NGOs were adopted by participating farmers. Of the 10 techniques originally promoted by Defensores de la Naturaleza in the Sierra de las Minas, 3 were used by farmers: planting a cover crop known as velvetbean (Mucuna pruriens), minimum tillage, and live barriers. Línea Biósfera originally promoted more than 15 techniques and then focused its efforts on 6. In the end, farmers in El Ocote adopted primarily three of these, including planting velvetbean, minimum tillage, and integrated pest management. In both sites, the technique used most frequently by farmers was planting velvetbean.

In each site, we selected communities in which the implementing NGO had promoted sustainable agriculture for at least five years. From each of these communities, we selected our sample of SA Users and SA Non-Users. We used a sampling technique called quota sampling, which required the selection of a predetermined number of individual cases (in this case, SA Users) and an equal number of comparison individuals (in this case, SA Non-Users) to provide sufficient statistical power to discern a difference, if in fact one exists, between the two groups. The accepted practice is to collect at least 100 representatives in each group, giving a sample of at least 200 at each site. In fact, both Línea Biósfera and Defensores de la Naturaleza exceeded this minimum, with Línea Biósfera sampling 300 and Defensores de la Naturaleza sampling 308. The even split between SA Users and SA Non-Users can be seen in the following table. With these samples, we were able to analyze the data for each site separately, and then combine the samples to do analysis across our two sites.

Number of SA Users and SA Non-Users Included in the Study ­ Guatemala and Mexico

We selected the two groups for our quota sampling using a technique called frequency matching. This step in the data-collection phase was extremely critical because it provided us with a sample that made it possible for us to isolate the effects of sustainable agriculture use. Using a sheet that profiled a typical household found in the study site — including household, demographic, and socioeconomic factors — we matched SA Users to Non-Users to ensure that the two groups were as similar as possible, except for their user status. We controlled for the following potentially confounding variables: gender of primary farmer in the family (all primary farmers in the study communities were men), family size, access to goods and services, and family wealth. If an equal number of families could not be selected from the same community, SA Non-User families were selected from another community that was most similar to the SA User community with respect to environment, infrastructure, socioeconomic status, and access to goods and services.

Results of Matching SA Users and SA Non-Users ­ Guatemala

Data Collection

Field teams that spoke the local languages (Q'eqchí in Guatemala and Tzotzil in Mexico) were recruited and organized by the two implementing NGOs and trained by BSP. All data-collection instruments were developed and field-tested jointly by the three participating organizations. In this way, we were able to standardize the instruments so that both quantitative and qualitative data were collected using the same questionnaire or topic guide at each site. All field data collection took place during the fall of 1998.

We developed the following four instruments to collect quantitative data:

Direct Observation Checklist. This checklist allowed interviewers to quickly assess the socioeconomic status of the interviewee and ensure that the family fell within established general selection criteria.

Family Matching Sheet. This instrument allowed the interviewers to appropriately match SA Users and SA Non-Users. On each form, a profile of the SA User was filled out and an SA Non-User was then sought that matched this profile, except for user status. We matched (and therefore controlled for) the following variables: primary occupation of father, observed socioeconomic status, family size, and access to electricity and a potable water system.

Household Questionnaire. Interviewers asked each farmer a series of questions from this form to determine his (all interviewees were men) knowledge, attitudes, and practices related to agriculture and conservation. In addition, interviewers recorded household characteristics, including socioeconomic status, level of education, age structure of the family, and sources of income.

Plot Survey. Some farmers had more than one plot of land. For each plot, the interviewer recorded the size and age of the plot, what crops were being planted; techniques used, including sustainable agriculture; inputs; problems with agricultural pests; and yields. This instrument also was used to collect historical data on each plot. In addition to answering questions about the year in which the survey was conducted (1998), farmers were asked about area planted, production, and inputs for the three previous years (1995-1997).

Qualitative data were collected using two types of instruments: focus group topic guides and key informant interviews. The results of these sessions were used primarily to complement the quantitative results.

Focus Group Topic Guides. These topic guides were developed primarily to explore the knowledge, attitudes, and practices of farmers in the study sites. The guides covered general agricultural practices, use of sustainable agriculture techniques, and perceptions of the relationship between agriculture and the environment. Focus groups were conducted only with male farmers who were actively engaged in subsistence agriculture. At each of the two sites, focus group interviews were conducted with both SA User and SA Non-User groups.

Key Informant Interviews. Informal interviews were conducted with key informants in each of the two sites. These interviews were used primarily at the beginning of data collection in the communities to help orient the interviewers and to serve as an "ice breaker" with community leaders. The questions asked included many of the same topics covered in the focus group topic guide.

Helping Project Partners to Answer Their Own Questions

To address the third goal of this project, BSP worked with Defensores de la Naturaleza and Línea Biósfera to design and implement this research project and analyze and communicate the results. One of the coauthors of this publication, a representative of the Center for International Forestry and Research (CIFOR), provided additional assistance in the conceptualization and design phases of the project. In October 1997, BSP facilitated a meeting of experienced researchers and practitioners to discuss the concept of investigating the conditions under which sustainable agriculture works as an effective conservation tool. The project was formally launched with a design workshop in June 1998 that included members of BSP, Defensores de la Naturaleza, Línea Biósfera, CIFOR, and WWF-US. This meeting provided us the opportunity to develop a learning framework that included the specific operational questions we wished to address and the process we would use to answer those questions.

In August 1998, BSP facilitated a training workshop during which the data-collection instruments were finalized and field-tested. At the same time, BSP trained project staff in data-collection techniques and interviewing. Fieldwork continued through the fall of 1998 at each of the two sites.

In early 1999, BSP hired a statistician to assist in analyzing the data. Once data were collected, BSP worked with Defensores de la Naturaleza, Línea Biósfera, and the statistician to clean the data and input them into a database. In August 1999, we conducted the first in a series of analysis workshops to develop the findings from each site and to begin cross-site comparisons. BSP staff worked with both organizations as they interpreted and began to write up their results.

In August 2000, we had a final meeting to discuss findings. The purpose of this meeting was to look across both sites to determine the conditions under which sustainable agriculture works, develop guiding principles for practitioners around the world, and document our analysis of the learning process.

Some Things to Keep in Mind

As you read through our findings, please keep in mind the following caveats to help you interpret our results as accurately as possible:

• Association is not the same as causality. Our research design is cross-sectional, and our sampling is not random. We can, therefore, say that there is an association between two variables, but not a causal link. If we had wanted to more accurately identify causality, we would have needed to conduct a randomly sampled longitudinal study. Nevertheless, the associations we see in the results provide a fairly compelling description of the possible association between sustainable agriculture and conservation.

• Our research design does not allow us to quantify the regional impacts of sustainable agriculture. For the reasons we mentioned above, our best option to test the relationship between sustainable agriculture and conservation outcome was at the household level. Using the household unit, we were able to study direct effects of sustainable agriculture use on conservation. But we did not attempt to quantify the total impact of sustainable agriculture on a regional scale. To do so, we would have had to spend precious time to determine accurate prevalence rates of adoption, control for myriad additional confounding factors, and ascertain variation in sustainable agriculture use throughout the entire region of each site. We did not have the time or resources to do this. We do, however, discuss some implications of sustainable agriculture for regional deforestation impacts in the Putting the Findings in Perspective section of this publication.

• Our data on crop outcomes and inputs are primarily recall data. Almost all of our data came from farmers' recollection of past agriculture outcomes over a four-year period. These data may, therefore, be influenced by farmers' ability to remember details about past years. We overcame this problem, in part, by using the data for the most recent complete year of harvests for most of our analysis. Some of our results may be biased because farmers who use sustainable agriculture may have been inclined to answer questions in a way they thought would please the interviewers. This potential bias may be particularly noteworthy in our results on the use of fire to prepare agricultural fields. We did, however, attempt to triangulate farmers' responses as much as possible to minimize bias.

• Our results are limited to the characteristics of our samples. By narrowing our sample, we were able to come up with clear and precise findings for the areas included in our study. Our findings are, therefore, particularly useful to other sites with similar environmental, physical, socioeconomic, cultural, and institutional characteristics.

• Our analysis is limited to specific sustainable agriculture techniques. The farmers included in our samples only adopted a limited subset of sustainable agriculture techniques. We can, therefore, say little about the potential conservation impacts of the techniques that were not adopted by farmers. Farmers' willingness to adopt a specific technique, however, can be interpreted as an indicator of the technique's success in terms of its socioeconomic value and, to a lesser extent, its conservation importance. Although farmers chose to adopt only one or two techniques, this does not compromise the representativeness of the study results. In fact, because two independent projects arrived at the same primary technique, there is some evidence that supports the notion that these two sites, and the behavior we observed in farmers related to sustainable agriculture adoption and use, are typical.

• We only included projects carried out by NGOs. Our research is limited to those sustainable agriculture projects that are implemented by local conservation organizations. Results may be different for similar projects implemented by development organizations or government agencies.

• Our sample came from one region of Latin America. The distance between our two studies sites is relatively small and, in many ways, the characteristics of these two sites are very similar. Had we included other sites from around the world, our results might have been different. However, many of the characteristics found in our sample also are found in many other countries, and we believe that our results will be useful to others working under similar conditions.

• We included only subsistence crops in our analysis. We did not look at cash crops because most sustainable agriculture projects with a conservation goal have focused on subsistence farmers in agricultural frontiers. We believe that the results would be different for cash crops, especially those that mimic secondary forests, such as shade-grown coffee and cardamom.

Despite these caveats, the strength of association and consistency in the study results lead us to believe that we arrived at some pretty telling insights. While prudence should be used to interpret and generalize our results — as is the case with all studies of this nature — we believe that the findings can be of great use to conservation project managers around the world who are attempting to implement similar projects.

What Did We Find?

In this section, we present the results of our analysis from our study sites in Guatemala and Mexico. Much of our analysis is associated with agricultural outcomes, and it is a well-known fact that agricultural production and yield often vary from harvest to harvest. At both sites, farmers generally enjoy two harvests annually: the first takes place in April-May, and the second, main, harvest occurs in November-December. In addition to collecting the same data for each of these harvests, we also collected data for four years of harvests from 1995 to 1998. We collected these additional data to control for variation between years. All data were based on farmers' recollections of past outcomes.

In our analysis, we combined area planted, production, and yield data for the two crops for each year to create total annual amounts. For much of our analysis, we wished to link sustainable agriculture use with conservation outcome. Therefore, we wanted to allow as long as possible for project implementation at each site to increase our chances of observing any possible effects. Ideally, we would have used the crop data we collected for 1998. Unfortunately, because of issues related to the timing of the study, we had to complete the data-collection phase just before the second harvest of 1998, so our data are incomplete for that year. Therefore, for those analyses in which we want to observe the maximum effect of sustainable agriculture, we use the latest year for which we have complete crop data — 1997.

We designed the study to examine both subsistence and cash crops. After completing the data-collection phase, however, we concentrated our analysis on maize production because there appeared to be little variation between the SA User and SA Non-User groups with respect to the cultivation of other crops, including beans, pepper, coffee, and cardamom. As we mentioned earlier, most sustainable agriculture interventions, including those that took place at the two sites in our study, focus principally on subsistence crops. At both of our study sites, maize is the primary subsistence crop and is the major target of sustainable agriculture activities. For these reasons, almost all of our analyses of crop characteristics, including area planted, production, yield, and inputs, center on maize.

Although we designed the data-collection instruments to collect information on multiple plots cultivated by each farmer, we discovered during the data-collection phase that most farmers had only one primary plot of land devoted to maize. In our analysis, note that we use either farmers or plots of land as our unit of analysis, depending on the question we are trying to answer.

The data-analysis phase of this study proved to be the most challenging aspect of our work. Given the complexity of trying to isolate the effects of sustainable agriculture projects on conservation outcomes, we needed to use many different types of data analyses and statistical tests. In addition to requiring a sophisticated level of knowledge related to statistics, our analysis also required a high level of proficiency in the use of statistical software. For these reasons, we found it necessary to hire a statistician to assist us with the analysis.

While this specialist was not part of the study team during the conceptualization and design phases of the project, she was integrated into the team soon after data collection began. She worked closely with BSP, Línea Biósfera, and Defensores de la Naturaleza to help analyze the data from each of the sites individually and in combination for our final analysis. During the data-collection phase of the study, each country team met frequently with BSP and our analysis specialist.

In this section, we provide separate analyses from the Guatemala and Mexico sites, and we provide combined analyses from the two sites where the results are insightful. Most of the results we present compare only two factors (bivariate analysis) but, where appropriate, we also present the results of looking across more than two factors (multivariate analysis).

For our bivariate analysis, we used two types of statistical tests to see if there was a difference between the two variables we were analyzing. If the data we were analyzing were continuous, we used the t-test of significance. If the data we were analyzing were categorical, we used a x² (chi-square) test of significance. We also include the P value for each of our statistical analyses. And we use the convention of "n" to denote the sample size. In some of the results, you will see "n (%)" in titles or headers, signifying that both numbers and percentages are shown in the corresponding tables. Sometimes the number of farmers or plots in a particular analysis will be lower than the totals we have in the sample. This is most commonly the result of missing data and information.

In addition to statistical significance, we discuss programmatic significance in our analysis. At times, statistical analysis may produce results that, in the real world, have little relevance. In other words, just because a relationship between two variables may be statistically significant, it does not meant that the relationship is noteworthy. Conversely, sometimes an analysis does not turn out to be statistically significant, but the results are extremely important from a practical perspective. We might find, for example, that a certain sustainable agriculture technique consistently saves, on average, 20% of the total amount of labor farmers need to invest in their plots to prepare them for planting. While this relationship may not prove to be statistically significant for a variety of reasons, it is probably extremely important to farmers!

In our bivariate analysis, we sometimes talk about odds ratios. An odds ratio (OR) indicates the increased likelihood one group has over another for a given factor. For example, let us compare yields for farmers who use chemical fertilizers with those of farmers who don't. If we came up with an OR of 2.3 for those who use fertilizer, that means that farmers who use fertilizer are 2.3 times as likely to have a high yield than those who don't.

The purpose of our multivariate analysis was to determine what combination of variables is most predictive of a certain outcome. So, for example, you will see below that the outcome of the amount of area planted to maize by farmers in Mexico is primarily a function of: (1) total amount of labor invested in the plot, (2) number of years a farmer has worked his plot, (3) family size, and (4) user status. The advantage of multivariate analysis over bivariate analysis is that it provides the opportunity to gauge the relative importance of one variable over others.

For our multivariate analysis, we used two types of statistical tests as well. If the variable we were trying to predict (the dependent variable) was continuous, we used linear regression. If the dependent variable was categorical, we used logistic regression. In both of these types of analysis, our goal was to find those variables that best predict the outcome of the dependent variable. Each of these analyses provides information about how changes in multiple variables can be predictive of the dependent variable. The combination of these predictor variables (or independent variables) is often referred to as a "model." The statistic we use that describes the extent to which the model of independent variables accurately describes the dependent variable is called the R².

We evaluate the extent to which our analysis supports each conventional wisdom with the scale and symbols shown below. This design allows you to quickly assess our findings.

Direct Impact of Sustainable Agriculture on Biodiversity

The first section of our framework looks at the direct impacts of sustainable agriculture projects on biodiversity, including amount of area under cultivation, fallow area and duration, and contamination.

Area planted to subsistence crops

According to the conventional wisdom, for sustainable agriculture to affect rates of deforestation, it is necessary for two intermediate outcomes to occur: crop yield (production/unit of area) must improve and this must, in turn, lead to a decrease in the amount of land a farmer needs to plant to feed his family. So, in addition to looking solely at area planted, we need to examine the results of our analysis of farmers' yields at both our study sites. In addition, to understand what influences yield, we need to look at a variety of other factors besides the use of sustainable agriculture. These factors, such as the use of fertilizer and pesticide and the amount of labor the farmer invests in his plots, could disproportionately increase yield between farmers. Other factors, such as pest infestation, could decrease yield. We included these factors and other potentially confounding variables in our data collection and analysis.

Similarly, area planted can be influenced by many different environmental and social factors other than sustainable agriculture. We controlled for many of these variables, including family size, soil quality, rainfall, and slope, in our sampling strategy. We included others, such as the sale of crops, availability of labor, access to credit, and land ownership,in our data collection and analysis.

Bivariate Analysis

If we look at area planted to maize at the farmer level at both sites, we see that Guatemalan farmers who use sustainable agriculture are significantly more likely to plant more area to maize than those who do not use sustainable agriculture — just the opposite of what the conventional wisdom predicted. But Mexican farmers who use sustainable agriculture plant significantly less area than farmers who do not use sustainable agriculture — just what the conventional wisdom predicts. On the surface, these results seem to be contradictory. As you will see later, they are, in fact, completely logical.

Average Area Planted to Maize in Hectares for SA Users and SA Non-Users, for 1997 ­ Guatemala and Mexico

As shown in the following table, the average plot size is significantly different between plots in which sustainable agriculture is used and plots in which it is not used in both Guatemala and Mexico. Note, however, that again the relationship is opposite between the two sites. In Guatemala, SA Plots are significantly larger than Non-SA Plots. In Mexico, SA Plots are significantly smaller than Non-SA Plots. These results are similar to the user-level results because most farmers have only one plot.

Average Area Planted to Maize in Hectares for SA Plots and Non-SA Plots, for 1997 ­ Guatemala and Mexico

When we look at the yield data, we see some even more interesting results. SA Users and SA Non-Users in Guatemala have almost identical yields. But in Mexico, yield is significantly higher for the SA Users than for the SA Non-Users. As far as programmatic significance goes, the difference in Mexico is extraordinary: SA Users yield on average 1.5 times more maize than SA Non-Users.

Average Yield of Maize in Kilograms (kg) for SA Users and SA Non-Users, for 1997 ­ Guatemala and Mexico

The plot-level results confirm these findings. There is really no difference in yield between SA Plots and Non-SA Plots in Guatemala. But in Mexico, the difference is statistically significant on the same order of magnitude as we saw at the user level.

Average Yield of Maize in Kilograms (kg) for SA Plots and Non-SA Plots, for 1997 ­ Guatemala and Mexico

To make sure we were truly looking at the effects of sustainable agriculture use, and not some other factor, we looked at inputs that might affect this outcome. For use of fertilizer and pesticide, and access to credit, there were virtually no differences between SA Users and SA Non-Users and between SA Plots and Non-SA Plots. When we looked at the total amount of labor (family members plus paid labor) invested in maize production, we found no statistical relationship between SA Users and SA Non-Users in either Guatemala or Mexico. But there may be important differences between these two groups and our two sites from a programmatic perspective. In Guatemala, it appears that SA Users use about five days of labor per hectare less than SA Non-Users. In Mexico, it appears that SA Users use about 5.5 days of labor per hectare more than their SA Non-User counterparts.

Average Amount (in Days) of Total Labor Used by SA Users and SA Non-Users, Controlling for Size of Plot (Days/Hectare), for 1997 ­ Guatemala and Mexico

We also looked to see if one type of farmer was more likely to sell surplus maize than the other. Indeed, in both Guatemala and Mexico, SA Users were significantly more likely to sell maize than SA Non-Users. The numbers for Guatemala, however, show only marginal programmatic significance because they are relatively small.

Number of SA Users and SA Non-Users Who Sold Maize, From 1997 Harvest ­ Guatemala and Mexico

In terms of how much farmers sold, there was virtually no difference between SA Users and SA Non-Users in Guatemala. In Mexico, however, SA Users sold significantly more than SA Non-Users — almost 450 kg more, representing on average an added income of 675 pesos (US$87 at the 1997 exchange rate).

Amount of Maize Sold in Kilograms (kg) for SA Users and SA Non-Users, From 1997 Harvest ­ Guatemala and Mexico

While we were looking for the direct links between sustainable agriculture and deforestation through changes in crop yields and area planted, we came across what is arguably sustainable agriculture's greatest benefit to conservation — fire reduction. This factor has not been addressed widely in previous studies but demonstrates a high degree of association (in the same direction) at both study sites. Fire is one of the major threats to habitat in tropical forests near human settlements. Most often, people set fires that burn large tracts of primary forest. In the traditional preparation of plots for cultivation, farmers burn vegetation before planting to increase soil fertility. Sustainable agriculture discourages burning and instead encourages farmers to turn agricultural waste back into the soil to increase fertility.

In both the Sierra de las Minas in Guatemala and El Ocote in Mexico, fire is one of the biggest threats to the reserves. Often, serious forest fires are started by agricultural fires that burn out of control. We found that, by an overwhelming majority, SA Users in Guatemala and Mexico were less likely to use fire to prepare their plots than SA Non-Users. In fact, in Guatemala SA Non-Users were 7.7 times more likely to use fire than SA Users; in Mexico, SA Non-Users were 16.6 times more likely to use fire!

Number of SA Users and SA Non-Users Who Use Fire To Prepare Agriculture Land, for 1997 ­ Guatemala and Mexico

We were able to cross-check these results because in the first half of our interview with farmers, we asked the simple question: "Do you use fire in the preparation of your agricultural fields?" The results are in the table above. Later on, as we collected data on each of the farmer's plots, we asked how the land was prepared — through use of any combination of the following techniques: simple cutting of vegetation, mixing vegetation into the soil, burning, and use of herbicides. We then compared plots in which fire was used with plots in which fire was not used. At the plot level, SA Plots are 5.4 times less likely to be burned in Guatemala and almost 20 times less likely to be burned in Mexico. The positive effects of sustainable agriculture are clear for this factor.

Number of SA Plots and Non-SA Plots in Which Fire is Used for Preparation, for 1997 ­ Guatemala and Mexico

* This is from a total of n = 110 because there were many missing data for this question.

Multivariate Analysis

In our multivariate analysis, we looked at the combination of factors at each site that were most predictive of four main outcomes: (1) user status (whether a farmer was an SA User), (2) area planted to maize, (3) yield of maize, and (4) whether farmer used fire to prepare his fields. For each factor, variables are listed in order of importance (i.e., the proportion of the outcome variable they describe). From the multivariate analysis, we can also determine the direction (positive or negative) of the relationship. The R² of the model is included as well.

User Status

In Guatemala, the combination of variables that best predicted whether or not a farmer is an SA User included: (1) use of fire, (2) age of the farmer, (3) perception of positive effects of sustainable agriculture, and (4) visits by an extensionist. Our analysis showed that sustainable agriculture users were less likely to burn their plots, older, more likely to perceive benefits of sustainable agriculture, and more likely to receive a visit from an extensionist than non-users. The R² was 0.97.

In Mexico, the variables that best describe user status are: (1) use of fire, (2) age of the farmer, and (3) visits by an extensionist. SA Users were less likely to burn their plots, younger, and more likely to receive a visit from an extensionist than SA Non-Users. The R² was 0.90.

Combining the Guatemala and Mexico data, we found that the variables most predictive of sustainable agriculture use across both sites are: (1) use of fire, (2) visits by an extensionist, and (3) perception of positive effects of sustainable agriculture. SA Users were less likely to use fire, more likely to be visited by an extensionist, and more likely to perceive benefits of sustainable agriculture. Age dropped out of the model because it had the opposite relationship to user status in Guatemala and Mexico. The R² for the combined analysis was 0.88.

Area Planted to Maize

Variables that predict the amount of area planted to maize in Guatemala include: (1) user status, and (2) number of years a farmer has worked his plot. Area planted is greater when the farmer is an SA User and the longer the plot has been cultivated. The R² is a very low 0.052, meaning we could not come up with a model that was very predictive of area planted in Guatemala.

In Mexico, variables in the model for area planted include: (1) total amount of labor invested in the plot (not controlling for size), (2) number of years a farmer has worked his plot, (3) family size, and (4) user status. Area planted is greater with increased investments of labor, the longer the plot has been cultivated, the greater the family size of the farmer, and when the farmer is an SA user. The R² is 0.27.

Combining Guatemala and Mexico, variables that predict the amount of area planted to maize include: (1) total amount of labor invested in the plot, (2) number of years a farmer has worked his plot, (3) age of farmer, (4) visits by an extensionist, and (5) problems with agricultural pests. Area planted is greater with increased investments of labor, the longer the plot has been cultivated, the older the farmer, the greater the likelihood the farmer has been visited by an extensionist, and the more likely the farmer has problems with agricultural pests. The R² is 0.20.

Maize Yield

In Guatemala, maize yield is most predicted by (1) number of years a farmer has worked his plot, and (2) the area of the plot. Yield is higher in older, smaller plots. The R² is a low 0.05.

In Mexico, maize yield is most predicted by (1) the area of the plot, and 2) user status. Yield is higher in smaller plots farmed by an SA User. The R² is 0.13.

Looking at both Guatemala and Mexico combined, yield is described best by (1) the area of the plot, and (2) user status. Yield is higher in smaller plots farmed by an SA User. The R² is a low 0.08. This low R² is to be expected because yield is a function of very different conditions in Guatemala and Mexico, as we will see later.

Use of Fire

In Guatemala, variables that best describe whether a plot is burned are (1) user status, and (2) total amount of labor invested in the plot. Plots that are burned are more likely to be farmed by an SA Non-User with higher total amounts of labor invested. The R² is 0.61.

In Mexico, variables that best describe whether or not a plot is burned are (1) user status, (2) age of the farmer, and (3) use of herbicides. Plots that are burned are more likely to be farmed by an older SA Non-User who uses more herbicides. The R² is 0.90.

Combining Guatemala and Mexico, variables that best describe whether a plot is burned are (1) user status, and (2) age of the farmer. As for Mexico, plots that are burned are more likely to be farmed by older SA Non-Users. The R² is 0.75.

Next File