Natural Resource Management in the Hillsides of Honduras Bioeconomic Modeling at the Microwatershed Level Bruno Barbier Gilles Bergeron RESEARCH REPORT 123 INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE WASHINGTON, D.C. Copyright © 2001 International Food Policy Research Institute All rights reserved. Sections of this report may be reproduced without the express permission of but with acknowledgment to the International Food Policy Research Institute. Library of Congress Cataloging-in-Publication data Barbier, Bruno, 1962– Natural resource management in the hillsides of Honduras : bioeconomic modeling at the microwatershed level / Bruno Barbier, Gilles Bergeron. p. cm. — (Research report ; 123) Includes bibliographical references. ISBN 0-89629-125-1 1. Natural resources—Management—Economic aspects—Honduras—Mathematical models. 2. Honduras—Economic conditions—1980—Environmental aspects—Mathematical models. 3. Honduras—Environmental conditions—Economic aspects—Mathematical models. I. Bergeron, Gilles. II. Title. III. Research report (International Food Policy Research Institute) ; 123. HC145.Z65 B37 2001 333.7′097283—dc21 2001055538 Contents Tables v Figures vii Foreword ix Acknowledgments xi Summary xiii 1. Conceptual Framework 1 2. Agriculture in Honduras 4 3. The Microwatershed of La Lima 5 4. Modeling Method 10 5. Model Simulation Results 18 6. General Conclusions 44 Appendix: Mathematical Statement of the Microwatershed-Level Linear Programming Model 46 References 58 Tables 4.1 Simulated effects of soil erosion and fertilizer use on maize yields on three different slopes 16 4.2 Simulated effects of soil erosion and fertilizer use on potato yields on three different slopes 17 Figures 1.1 Conceptual framework 2 4.1 Labor market for a submodel 12 4.2 Labor market for the community 13 5.1 Deflated prices of the main products 19 5.2 Simulated land use 19 5.3 Historical land use 19 5.4 Simulated land use by slope 20 5.5 Simulated land use by zone 20 5.6 Simulated land use by type of farm 21 5.7 Simulated income per person 21 5.8 Simulated income per person by farmer group and zone 22 5.9 Simulated sources of income in the whole microwatershed 22 5.10 Simulated yields 23 5.11 Simulated fertilizer use 23 5.12 Simulated inorganic fertilizer use per hectare 23 5.13 Simulated use of crop production 23 5.14 Simulated livestock herd size 25 5.15 Simulated transportation by mule 25 5.16 Simulated erosion aggregated by slope 25 5.17 Simulated soil depth in three land units 25 5.18 Simulated soil depth and soil conservation structure in one land unit 26 5.19 Simulated water volume in the outflow of the streams by zone 27 5.20 Shadow prices of land and labor 27 5.21 Simulated shadow prices of labor by season 27 5.22 Shadow price of the marketing constraint 29 5.23 Shadow price of water by zone 29 5.24 Simulated per capita income: Alternative scenarios for population growth 29 5.25 Simulated erosion: Alternative scenarios for population growth 30 5.26 Simulated per capita income: Alternative scenarios for access to markets 31 5.27 Simulated erosion: Alternative scenarios for access to markets 31 5.28 Simulated per capita income: Alternative scenarios for technologies 32 5.29 Simulated erosion: Alternative scenarios for technologies 33 5.30 Simulated per capita income: Alternative scenarios for natural resource constraints 34 5.31 Simulated erosion: Alternative scenarios for natural resource constraints 35 5.32 Simulated per capita income: Scenario without market liberalization 36 5.33 Simulated erosion: Scenario without market liberalization 36 5.34 Simulated per capita income: Scenario without inorganic fertilizer 37 5.35 Simulated erosion: Scenario without inorganic fertilizer 37 5.36 Simulated per capita income: Alternative scenarios for land reform 38 5.37 Simulated erosion: Alternative scenarios for land reform 38 5.38 Simulated per capita income: Scenario with dairy production 38 5.39 Simulated erosion: Scenarios with new markets 39 5.40 Simulated land use: Long-term baseline scenario 40 5.41 Simulated erosion by slope: Long-term baseline scenario 41 5.42 Simulated conservation structures: Long-term baseline scenario 41 5.43 Simulated per capita income: Long-term baseline scenario 42 5.44 Simulated per capita income: Alternative scenarios for prices 43 5.45 Simulated erosion: Alternative scenarios for prices 43 viii FIGURES Foreword Almost 2 billion people worldwide live on less favored or marginal lands, where they face increasing poverty, food insecurity, and environmental degradation. According to the pathways of development concept employed in this study, improving the stan- dard of living in such areas is a function of a complex set of conditioning factors, the most fundamental of which is agroecology. In Honduras, one of the poorest countries in the Amer- icas, most of the impoverished rural population lives on marginal hillsides. Previous IFPRI research in the central hillside region of Honduras identified five potential development path- ways, including market-induced intensification of vegetable production on lands otherwise considered marginal and of low economic potential. Barbier and Bergeron explore several hypotheses about the dynamics of natural resource management in the hillsides of La Lima and further explore the causes and consequences of the transition to vegetable production. To fully integrate agroecological factors, such as for- est, water resources, and topography, the authors use a bioeconomic model that links farm- ers’ resource management decisions to biophysical models. This captures production processes as well as the condition of natural resources. The model was used to run different scenarios over the period 1975 to 1995 and then to project into the future. The authors conclude that agroecological conditions are the most important factors de- termining incomes for villages with comparable agroecological conditions. The simulations indicate that recent policy interventions, such as market liberalization, road construction, and crop improvements, have all helped to increase incomes. Roads, for example, provide better access to regional markets, allowing farmers to benefit from the advantageous prices offered for nontraditional crops. The study finds that incomes would have been much lower, and degradation much higher, if vegetable production had not been possible for ecological rea- sons, since farmers would then have been reduced to adopting less sustainable strategies. Yet one strategy does not fit all. Keeping all else constant, it appears that alternative poli- cies that one might assume to be beneficial, such as a land reform or developing dairy pro- duction, would probably not have been successful in La Lima. The authors thus demonstrate that the complexity of marginal environments must be fully considered before implementing policies because the same policy may have dissimilar impacts in different areas. The path- ways of development approach can also help predict future outcomes of policy interventions and guide decisionmakers in targeting resources appropriately. Per Pinstrup-Andersen Director General Acknowledgments This research benefited from the financial support of the French Foreign Ministry and the Inter-American Development Bank. The authors would especially like to thank Peter Hazell for his time and advice in the preparation of this report. The authors are also grateful to John Pender, Sara Scherr, Chantal Line Carpentier, and Natasha Mukherjee in Washington, and to Juan Manuel Medina, Javier Tamashiro, Polly Eriksen, Guadalupe Duron, Fernando Mendoza, and Roduel Rodriguez, and the Panamerican Agricultural School of Zamorano, in Honduras, for their comments and field support. Any remaining errors are ours. Summary The objective of this study is to simulate the effect of population pressure, market in- tegration, technological improvement, and policy decisions on natural resource man- agement in the hillsides of Honduras. To do so, a bioeconomic model was developed and applied to a typical microwatershed. The bioeconomic model is dynamic linear pro- gramming fed with data from a biophysical model. Over recent years, farmers from the se- lected microwatershed have followed a “vegetable intensification” pathway of development. Different scenarios were run with historical data over the period 1975 to 1995 and then pro- jected 25 years into the future from 1995 to 2020. The results of the bioeconomic model presented in this report aid the discussion of a num- ber of induced innovation hypotheses. Many of our model’s results conform with the hy- potheses, but some of the results challenge conventional wisdom. The simulation results suggest that technology improvements such as irrigation and new varieties can help over- come diminishing returns to labor due to population pressure. Population increases in La Lima had only a small effect on the condition of natural resources because the cropped area increased only slowly thanks to the intensification of production. The model shows that the relationship between population growth and natural resource condition can have a U-shaped structure. In the long term, population pressure is likely to lead to continuing improvement in the condition of natural resources. The results also show that improvements in access to markets increase per capita incomes but do not necessarily promote land conservation be- cause land values do not automatically increase. The hypothesis that agroecological condi- tions are the most important factors determining incomes and natural resource condition is illustrated by the results. Past policy interventions such as market liberalization, road construction, construction of the potable water distribution system, crop variety improvement, and extension services have all helped to increase incomes. However, the simulations suggest that replacing inorganic fer- tilizer with organic fertilizers would not maintain incomes at the same level. Dairy produc- tion is a viable option. A land reform would have had a low but positive impact on incomes. The forward-looking baseline scenario suggests that erosion will continue to increase if prices remain as they were in 1995. If commodity prices decline, however, erosion will lessen because farmers will reduce their production of vegetables during the rainy season. Con- versely, an increase in inorganic fertilizer prices will lead to more erosion because farmers will use less fertilizer, obtain lower yields, and increase their cropped area. C H A P T E R 1 Conceptual Framework Population growth, market integration, and new technologies affect rural communities in many different ways, but a finite number of development pathways may be iden- tified within homogeneous agroecological regions. A study by the International Food Policy Research Institute (IFPRI) of the Central Hillside region of Honduras empirically identified five different pathways (Pender, Scherr, and Durón 2001), one of which corre- sponds to market-induced transitions to intensive vegetable cropping. This pathway was con- sidered particularly interesting because it illustrates the potential for intensive commercial agriculture in lands normally viewed as marginal. The microwatershed of La Lima, located on the hillsides close to the capital, Tegucigalpa, was selected as a case study to facilitate understanding of the causes and consequences of this type of transition and to generate hypotheses about possible policy actions in similar con- texts (Bergeron et al. 1997). The study covered a period of 20 years (1975–95) during which time the subsistence-oriented farming strategies that prevailed at the outset in that village evolved toward a semi-commercialized strategy including the production of vegetables using high-input practices. This report shows how a bioeconomic model linking linear programming and a biophys- ical model is developed and used to examine the outcomes of this type of development over the past 20 years and the likely outcomes over the coming 25 years. Multiple scenarios are introduced that indicate the likely effects of specific measures on production, incomes, and environmental conditions. The conceptual framework underlying this work is a stylized explanation of the mecha- nisms occurring in rural areas, and draws principally on the theory of induced innovation in agriculture (Boserup 1965; Ruthenberg 1980; Ruttan and Hayami 1990; Binswanger and McIntire 1997; Pingali, Bigot, and Binswanger 1987; Lele and Stone 1989; and others). Simply put, this theory argues that people endogenously adapt to changes in the conditions they confront, and that these adaptive responses are the main source of technical and institu- tional change in agriculture. Boserup used this idea to argue that population growth is the dominant cause of agricultural development in underdeveloped countries. The logic behind this proposition is that population growth increases the scarcity of land relative to labor, thus increasing the net return to intensifying labor use on a given piece of land by reducing fal- low periods and investing labor effort in increasing land productivity. Boserup also argued that technological innovations (for exam- ple, adoption of organic fertilization) and institutional innovations (for example, de- velopment of private property rights), which are implicitly endogenous, are caused by population growth and resulting changes in land use. Other authors have expanded on Bose- rup’s model by incorporating other exoge- nous factors that also stimulate endogenous agricultural change. Binswanger and Mc- Intire (1997) and Pingali, Bigot, and Bin- swanger (1987) argue that increased access to markets, as may result from development of roads or other infrastructure, also causes agricultural intensification. Lele and Stone (1989) argue that government policies play an important role in shaping the nature and impacts of agricultural change, particularly the impacts on natural resources and the environment. Smith et al. (1994) argue that, because of diminishing returns to labor, ex- ogenous technological advancement is nec- essary to avoid declining output per capita in the process of intensification, an argument reminiscent of neoclassical growth theory as pioneered by Solow (1974). Our conceptual framework incorporates these exogenous (at the community level) factors of agricultural change under the term “pressure/shift variables” (Figure 1.1). Ex- ogenous causal factors include the natural rate of population growth, changes in market access and development of markets for in- puts and outputs, exogenous technological change (such as development of new crop varieties), and changes in government poli- cies affecting property rights, land tenure, access to resources, prices, and other factors of agricultural production. These factors are the primary external drivers of change in a given community. The impacts of these factors in a particu- lar setting will be affected by local “condi- tioning factors” as shown in Figure 1.1. These include the community’s resource en- dowments (land quantity and quality, forests and other vegetation, water resources, cli- mate, topography, and other biophysical characteristics of the environment) and its “social capital” endowments (local institu- tions and organizations). These conditioning variables can be thought of as determining the constraints on decisions at the commu- nity and household levels. The modeling described in this report follows this conceptual framework to simu- late the aggregate behavior of farmers in the La Lima microwatershed. The model high- lights population, markets, and technology 2 CHAPTER 1 Figure 1.1 Conceptual framework as the main forces driving change in agri- culture. Accordingly, four hypotheses are postulated about the dynamics in natural re- source management in the hillsides: • Population pressure leads to greater total income but lower per capita incomes be- cause of decreasing returns to labor. • Population pressure has negative effects on natural resources until the productivity of the resource base declines to critical levels. Only then will farmers improve natural resource management.1 • Market access increases per capita in- comes. Market access also promotes land conservation because land-value increases make investments in land improvements more profitable. • Technological improvements have a sig- nificant positive effect on per capita in- comes but an ambiguous effect on natural resources. A fifth hypothesis verifies our assump- tion about the importance of agroecology in “filtering” outcomes at the local level: • Agroecological conditions are important factors in determining incomes and re- source conditions. The model will aid the discussion of each hypothesis. CONCEPTUAL FRAMEWORK 3 1 This process can be represented by a U-shaped function where the natural resource stock decreases to the point where farmers start to invest in its enhancement (Scherr and Hazell 1994). C H A P T E R 2 Agriculture in Honduras Honduras is one of the poorest countries in the Americas, with a gross domestic prod- uct (GDP) of US$640 per capita (World Bank 1997). The economy relies heavily on agriculture, which generates 70 percent of total export earnings and employs 58 per- cent of the labor force (World Bank 1997). Slopes greater than 12 percent account for 85 percent of the landmass, and the spatial distribution of land is as follows: the fertile lowlands and valleys are owned by large farmers, ranchers, or international fruit companies, while the majority of the rural population lives on hillsides. Population growth is rapid, and some view this as the main cause of deforestation and erosion (Silvagro 1996). It is estimated that soils have experienced moderate to strong degradation on 10–25 percent of the hillsides of the region (Olderman, Hakkeling, and Sombroek 1990), and that erosion ranges from 9 to 50 metric tons per hectare per year on slopes of 15–40 percent (Wouters 1980). Various sus- tainable cultivation techniques have been developed for such environments, but their adop- tion is low in Honduras (Bunch and López 1995; Pender and Durón 1996; Valdés 1994; Léonard 1989). Studies of recent agricultural policy in Honduras usually distinguish between three dis- tinct periods (Durón and Bergeron 1995). Until 1980, industrialization, driven by import sub- stitution, favored the urban zone to the detriment of agriculture (Williams 1986). Beginning in the early 1980s, the government abandoned this strategy and focused instead on increas- ing agricultural exports, mainly of high-value crops such as fruit, cattle feed, and shrimp. Those policies were expanded in 1989 with the adoption of a structural adjustment program, which removed most tariffs and barriers to international trade. This favored local producers, as a higher exchange rate promoted exports and reduced imports. These demographic, economic, and policy changes have had diverse effects on rural areas. In a study of the Central Hillside region of Honduras, five distinct pathways of change were identified: (1) the extension of maize production; (2) the intensification of vegetable production; (3) an increase in coffee production; (4) the intensification of livestock produc- tion; and (5) the intensification of forest product extraction (Pender and Durón 1996). This study was undertaken in La Lima, a microwatershed located close to the capital city and viewed as representative of the “vegetable intensification” pathway. C H A P T E R 3 The Microwatershed of La Lima Data were collected in the microwatershed of La Lima from January 1995 to October 1997, as follows:2 1. A household census was conducted detailing land, labor, capital, livestock, trees, access to water, and on-farm and off-farm activities (Bergeron et al. 1997). 2. Twenty households were interviewed in-depth to determine incomes, expenditures, con- sumption, and investment, as well as crop and livestock budgets. 3. Aerial photographs of the microwatershed from 1955, 1975, and 1995 were analyzed to de- termine current and past land use patterns. The aerial photographs were digitized and im- ported into geographical information system (GIS) software to analyze spatial patterns of land use, soils, and topography. 4. A history of plot and farm changes over the 20 years from 1975 to 1995 was undertaken based on recall interviews with farmers (Bergeron and Pender 1996).3 5. Price time-series data were collected and information on the evolution of the market struc- ture was elicited from key informants (Mendoza 1996). 6. Detailed soil analysis was performed (Eriksen 1996); erosion was monitored on one repre- sentative cultivated plot; and water flows and sedimentation were measured during one en- tire rainy season (Tamashiro and Barbier 1996; Flores Lopez 1996). 2 Data collection was a collaborative effort between the Escuela Agrícola Panamericana of Zamorano, Inter-American Institute for Cooperation on Agriculture (IICA), and IFPRI. 3 The plot history method consisted of randomly selecting 100 points within the microwatershed. For each point, the owner of the plot was asked to recall the land use over the past 20 years. The quality of the recall declines with time, but the results are similar to aerial photography from 1975. General Characteristics The microwatershed of La Lima covers ap- proximately 10 square kilometers. Only 18 percent of the microwatershed has land with slopes under 15 percent; 52 percent has slopes of over 30 percent. The microwater- shed is typical of the hillsides of Central America with respect to its soils, altitude (1,000–1,800 meters), and climate (bimodal rainy season with mean yearly rainfall of 1,200 millimeters). Production systems are also typical of Central America, with mixed farming of vegetables, maize, coffee, and livestock. Most of the land is under extensive use: pine forests dominate (25 percent of the area), followed by pastures (21 percent), and mixed pastures and trees (20 percent). Fal- low land occupies 17 percent and the per- manently cultivated area covers 14 percent. This land use did not change much over the 20-year period. Pastures have remained almost constant because major owners pre- fer large ranching over cultivation. Forests, mainly located on steep slopes, also re- mained largely unaltered. The main change was the development of intensive irrigated cultivation of vegetables but the area af- fected has remained relatively small. In 1995, the microwatershed supported 507 inhabitants living on 80 farms. Between 1975 and 1995 this population increased at a mean rate of 2.5 percent per year. Although the current population density is low in ab- solute terms, at 56 inhabitants per square kilometer, it is high compared with the Hon- duran average of 26 rural inhabitants per square kilometer. The average income in 1995 among a sample of 20 farmers was US$800 per worker4 and the average daily income was higher than the official daily minimum wage of US$1.20.5 However, income variations were large: only 30 percent of the workers earned more than the community’s average; 10 percent had an agricultural net income per worker three times higher than the average. Detailed expenditure data show that only 40 percent of the interviewed farmers made any agricultural investments in 1995. Property acquisition accounted for 70 percent of in- vestment and oxen acquisition for 19 per- cent. There were no reports of investment in cattle, calves, or conservation practices, and lower-income households did not purchase chemical inputs. Labor is a limiting factor in La Lima. The average family farm has 1.7 workers but, since women work little in agriculture, the average consumer/producer ratio is only 0.3. Extended families with several married adults within the household are rare. How- ever, collaboration between family mem- bers remains high, especially through share-cropping practices: 50 percent of the vegetables and a sizable proportion of the maize grown locally are produced under share-cropping arrangements. These help farmers pool labor, land, and capital, and also provide an effective strategy for mini- mizing risk. The labor market is active in La Lima: 60 percent of all farms employ paid labor dur- ing peak periods at an average of 40 days per farm per year. Wage labor is the primary source of income for 25 percent of the work- ers, and is a secondary activity for 8 percent of the workers. Temporary out-migration is rare, probably because demand for labor is high in the microwatershed. Returns to labor in farming in La Lima compare favorably with typical unskilled wage levels in the nearby capital city. This helps explain the low rates of out-migration in recent years. Formal land titling is still rare in the com- munity, and land is usually held in usufruct. The market for usufruct rights is active, par- ticularly within kin groups. Of all the plots currently used by farmers, 66 percent had been purchased, usually from parents, whereas only 33 percent were inherited. 6 CHAPTER 3 4 The fraction of production directly consumed by the family is not included. 5 The four largest ranchers of La Lima were not included in the sample of 20 farmers from whom detailed consumption and expenditure data were collected. The capital market is deficient in La Lima. There is no formal credit and farmers appear reluctant to use the informal credit market. In 1995, the informal real interest rate in a neighboring community was be- tween 0.97 and 2.47 percent per month, giv- ing an annual rate of between 11 and 30 percent. Saving rates being low and inflation high, farmers prefer to reinvest their surplus in land. Maize Production Maize is the most important crop in terms of area in La Lima. The average maize yield is 2.1 metric tons per hectare, which is higher than the national average. This above- average performance is apparently due to the adoption of new varieties and the increasing use of inorganic fertilizers. Maize produc- tion does not require pesticides because pest damage is low. Farmers generally plow their maize fields using hand tools but, if the to- pography allows, they may also use a tradi- tional plow pulled by two oxen. Small farmers produce maize mainly for consump- tion purposes, whereas ranchers have larger fields of maize for sale that are share- cropped with small farmers. Vegetable Production Farmers from La Lima were already pro- ducing vegetables in the 1960s (particularly potatoes), but the production of vegetables intensified and diversified rapidly after the opening of an all-weather road in 1985. Until then, farmers had to carry their produce by mule to the nearest road, located 6 kilo- meters from La Lima. In 1993, the La Lima road was improved and intermediaries now come to buy vegetables directly from the community. Vegetables are produced by 85 percent of the farmers, and represent 75 per- cent of farmers’ cash income. Potatoes and onions are the main commercial crops. Dur- ing the dry season, irrigation is required but installation costs are low, especially since the government installed a potable water system that some farmers tap for their pro- duction needs. Pesticide and inorganic fertil- izer use has increased rapidly with vegetable cropping, but analysis by the Escuela Agrí- cola Panamericana (Panamerican Agricul- tural School) of Zamorano found no trace of contamination in the water or soils of La Lima. Livestock Production Cattle ranching is relatively important in La Lima. In 1995, there were 486 head of cattle, or one per person. This is similar to the Hon- duran average (FAO 1997). Cattle density is low, at 0.6 animals per hectare for the whole microwatershed, but there is considerable variability in this figure because of the het- erogeneity of pastures. Pastures are found both on sloped areas and in some water- logged flat areas. Some areas were over- grazed, and large grazing areas were found to be invaded by shrubs. Forests are also fre- quently used for grazing animals. In addi- tion, farmers let livestock graze on maize residue during the dry season, partly because of the shortage of forage during the dry sea- son but also to increase soil fertility through manure deposits. Local ranchers usually keep their female calves and sell the males to ranchers in the valley, or locally as oxen. A small local mar- ket for milk and cheese has developed but production is low: the local breed produces less than 700 liters of milk per cow per year. Natural Resource Management in La Lima Water Management Water in La Lima is relatively abundant and in 1993 the government installed a potable water distribution system in the lower part of the microwatershed. As part of this study, water volumes were measured using a flow meter at different locations within the micro- watershed to assess existing and future water shortages (Tamashiro and Barbier 1996). In April 1996, at the end of the dry season, the springs of the microwatershed produced 869 liters per minute. Of this, 16 percent was captured by the potable water distribution THE MICROWATERSHED OF LA LIMA 7 system, 56 percent was used directly by in- dividuals, and 26 percent was left for down- stream users. The remaining 2 percent was lost in evaporation. Access to the main streams is unequal: 10 percent of the farms in the microwater- shed own 51 percent of the total irrigated area and 56 percent of the total water is cap- tured by those few individuals whose fields have access to the stream. These individuals hire farmers who do not have access to the stream as wage laborers (260 days of wage labor per farm per year, which is six times more than the village average). Additional water could be used by the community to in- crease the irrigated area by 20 or 30 percent, but this would increase water scarcity for the downstream users. Vegetable producers who live downstream of the microwatershed have already had to reduce their production dras- tically owing to the increasing water use in La Lima (Mendoza 1996). Forest Management The forest consists mainly of pine trees and some patches of broad leaf trees. Aerial photo comparisons show that the forested area was reduced between 1955 and 1975, but stayed almost constant between 1975 and 1995. Three hypotheses may be sug- gested to explain the maintenance of tree cover in that period. First, the Forestry Law of 1973 gave control over national forests to Corporación Hondureña de Desa- rollo Forestal (Honduran Corporation of Forestry Development), a government agency, which took away cutting permits from local farmers. Second, most of the for- est remaining after 1975 was located in areas with slopes greater than 30 percent and was therefore too difficult to cut. Third, many of the forests in the microwatershed are used for grazing. Large ranchers have taken advantage of the fact that forest own- ership is poorly defined and acquired de facto control of large areas of forested land that they do not officially own. These areas are still considered to be forest, even though their tree density is low and continues to de- crease because of cattle grazing. Soil Nutrient and Organic Matter Management Subsoils in La Lima are basaltic and rhy- olitic and are part of the Ojojona soil series that is common in Honduras (Canales 1994). Of 243 soil samples selected randomly, 49 percent were classified as loamy clay, which is soil with more than 28 percent clay con- tent (Eriksen 1996). The clay content and the kaolinite nature of this soil makes for poor drainage. Because these soils become water- logged when it rains, and compact after a few days without rain, they are difficult to plow. Farmers in La Lima classify their soils ac- cording to texture and fertility (Ardón 1996), but they do not consider soil characteristics to be important factors in crop choice, be- cause they see them as relatively homoge- neous throughout the microwatershed. The soils are acid, with a pH of about 5 (Eriksen 1996). The soil analysis, com- pared with the norms of the soil laboratory of the Panamerican Agricultural School at Zamorano, suggests good soil nutrient management in the cultivated plots, with no major phosphorus, potassium, or nitrogen deficiencies. Soil fertility management has changed over time in La Lima. The traditional tech- nique was based on fallow cycles, alternat- ing livestock rearing with crop production, and using slash and burn techniques to per- form periodic clearing. Now, 90 percent of farmers use chemical fertilizers, and slash and burn has been abandoned. On 30 percent of the cultivated plots, maize residues are grazed and incorporated into the soil. No manure is applied other than that left by roaming animals. Erosion Management Erosion was measured on one representative maize plot (160 meters long) using the “nails and washer” technique (Burpee 1997) dur- 8 CHAPTER 3 ing the rainy season of 1996. The average slope on the plot was 25 percent, and erosion measured 4 millimeters. With a soil density of 1.4, this translates into an erosion rate of approximately 56 tons per hectare. Such a rate, although not statistically representative and probably not sustainable in the long term, is what one might expect given the characteristics of this plot, and so we use it as indicative of soil loss rates in La Lima. Measuring sedimentation at the margins of the microwatershed, a rough extrapolation suggests an erosion level of 6 tons per hectare for 1995 for the entire microwater- shed (crop, pasture, and forest).6 This level is low for hillside conditions, but possible given the relatively good soil cover within the microwatershed. Accordingly, farmers in La Lima do not perceive soil erosion as a threat. They are more concerned about the lack of rain and about water-logging of the soil than about erosion. There is evidence that farmers even prefer to plant maize on slopes: 29 percent of the total maize area is planted on slopes greater than 30 percent, and 34 percent is planted on slopes of 15–30 percent; only 36 percent is planted on areas with a slope of less than 15 percent. This situation does not seem to be due to limited access to flat areas by small farmers because even large ranch- ers cultivate maize on slopes and keep flat areas for productive pastures. Few techniques have been adopted to control erosion in cultivated areas despite persistent efforts by extension services.7 The main methods are stone walls, found on 15 percent of plots, and contour plowing, used on 14 percent of plots. Live barriers, ter- races, tree planting, and drainage ditches are rare. Soil erosion is reduced by three other means: 1. Weeds in the maize field reduce runoff. 2. Almost all plots are fenced with stone walls, hedges, or barbed wire and the eroded soil accumulates behind these fences or at the lower end of the fields. In the long term this creates a relatively ter- raced landscape, which characterizes the microwatershed today. 3. Plots are small and water flow does not have time to accelerate within the field. These traditional techniques, common in Honduras, are not explicitly used to reduce erosion but are effective and should be considered as part of microwatershed management. THE MICROWATERSHED OF LA LIMA 9 6 These numbers should be treated with caution because soil erosion is an irregular process that occurs during a few rain events in the course of a year. 7 For a review of soil conservation experiences in Central America, see Lutz, Pagiola, and Reiche (1994) and Sims and Ellis-Jones (1994). For a review of agroforestry in Central America, see Current, Lutz, and Scherr (1996). C H A P T E R 4 Modeling Method Linear programming has been widely used to predict the supply response and farmers’ incomes under different agricultural policy scenarios (Hazell and Norton 1986). More recently, a new type of model, called a bioeconomic model, has been developed. A bioeconomic model links mathematical programming formulations of farmers’ resource management decisions to biophysical models that describe production processes as well as the condition of natural resources. The objective is to address both agricultural production and environmental concerns. In developed countries, these models focus on environmental pollution, whereas in developing countries the focus is on land degradation. The development of bioeconomic models in developing countries has been slow be- cause the situation is more complex than that in developed countries. First, farming sys- tems in developing countries are less specialized and tend to combine a larger range of interlinked activities such as crops, livestock, forestry, and off-farm activities (Ruthenberg 1980; Beets 1990). Second, most farming systems in developing countries include livestock and tree management. Modeling these activities necessitates the use of a dynamic frame- work and requires information about the length of the planning horizon, the discount rates, the returns to investment, the depreciation of capital, and loan repayments. Third, the farm- ing systems of developing countries rely more directly on the condition of local natural re- sources than on external inputs. Natural resource conditions result from complex biophysical processes that are difficult to quantify. Fourth, it is more difficult to validate accurate bioeconomic models in developing countries because land degradation has a stronger impact on yields. Because this impact was not well known, productivity modeling was not accurate. Several linear programming models for simulating land degradation have recently been developed, including variables such as soil erosion (E. Barbier 1988) and soil nutrient or organic matter depletion (Parikh 1991; Kruseman et al. 1995; Barbier 1998). Fifth, natural resource management in developing countries usually includes problems that go beyond farm boundaries. This feature has been included in recent community-level models (Kebe et al. 1994; Taylor and Adelman 1997; Barbier and Benoît-Cattin 1997). Fi- nally, rapid changes in population and markets in developing countries limit the value of static analysis. The Linear Programming Model In response to the above challenges, a bio- economic model was developed for the La Lima community that is both dynamic (with a five-year planning horizon) and recursive (over the 45-year period 1975–2020). The model, designed at the microwatershed level,8 includes two social groups (ranchers and small farmers), who are spatially disag- gregated by nine different segments in the landscape (defined by topography and soil type). This gives 18 farm submodels within the community model. Farm submodels in- teract through seasonal labor markets. The model includes soil erosion and dynamic in- teractions with livestock, crops, and forest. Yields and erosion parameters are given by the biophysical Erosion Productivity Impact Calculator (EPIC) model developed by Wil- liams, Jones, and Dyke (1987). The model maximizes the aggregate utility of the whole microwatershed over a five-year planning horizon.9 Utility is de- fined as the discounted value of future net monetary incomes plus the closing value of livestock and trees, plus the value of leisure taken. Leisure is valued using a fixed reser- vation wage, and a 15 percent discount rate is assumed in the baseline scenario. A sensi- tivity analysis showed that changes in the discount rate had only modest effects on land use. The model maximizes aggregate utility for the entire community. This approach gives the same solution as if optimizing each of the individual farm submodels separately, subject to the common labor market con- straints, providing there are no externality problems cutting across farm subgroups. If such externalities exist, this would lead the aggregate model to achieve a higher value of utility than is possible when the submodels are solved separately. We shall return to this issue after describing the structure of the model in more detail, but our general con- clusion is that serious externality problems do not arise in the model given the way we have formulated it. Although the model is dynamic and op- timizes over a five-year planning horizon, it is also solved recursively each year to gen- erate a series of annually updated plans. This is done 20 times for the period 1975 to 1995 and 80 times for the period 1995 to 2075. In this framework, the optimal solution for the first year of the planning horizon becomes the initial resource constraint of a new model that is solved for the following five-year period, and the process is repeated each year. The resources carried over in this manner are population, livestock, tree volume of differ- ent aged trees, soil depth, soil conservation structures, and plows. The recursive method allows us to track much longer periods than the five-year planning horizon, and to shock the model each year for exogenous changes in prices. Note that technological parameters and prices are the actual exogenous parame- ters prevailing just before the planning period, and these are maintained over the planning horizon. The key assumptions of the modeling process are described in the following section (a full description is in the Appendix). Population and Labor The available farm family labor in the watershed is constrained by the active popu- lation residing there each year. The number of residents is given exogenously and fol- lows from projections of population growth and permanent net migration out of the vil- lage. Farmers can hire or sell seasonal labor both within and outside the watershed. These transactions are defined in day units, and encompass labor transactions between the different subgroups within the watershed, the hiring in of day labor from outside the watershed, and the selling of community MODELING METHOD 11 8 For microwatershed-level analysis, see Thurow and Juo (1995), as well as Inter-American Development Bank (1995). 9 The program is written in General Algebraic Modeling System (GAMS) (Brooke, Kendrick, and Meeraux 1988). labor outside the watershed. Transactions outside of the watershed occur at exogen- ously given wage rates, whereas those within the watershed are driven by the shadow price of labor in each season, which is endogenous to the model. The supply of labor each season within a submodel comprises a stepped approxi- mation of a conventional upward-sloping function (Figure 4.1). The cheapest source of labor is own-family labor at a wage equal to the reservation price of leisure (wr). This price is fixed at 80 percent of the local wage rate. Once all the own-family labor has been employed (L0), then the next least-cost source of labor is to hire workers from other farms (submodels) in the community. The price for this labor is determined endog- enously in the community labor market. However, because workers can be brought in from outside the community at a fixed wage (wh), the endogenous wage for family labor transactions within the community cannot exceed that wage, nor can it fall below the reservation value of leisure. Once all the available community workers are employed, at L1 (where L1 is total family labor in the community), then the final source of labor is to bring in workers from outside at an exogenously given wage (wh). These three sources of labor define the three segments in the labor supply function in Figure 4.1. Note that the middle segment is not a single step but comprises lots of little steps that depend on increases in the opportunity cost of labor in other commu- nity submodels. The number and length of these steps depend on the complexity of the model and the number of basis changes that occur in the solution as the wage rate increases (see Hazell and Norton 1986, Chapter 7). The aggregate labor supply function for the community is simpler, comprising just two steps (Figure 4.2). The first step is the aggregate supply of family labor (L1) at the reservation wage (wr) or above. The second step is the supply of outside labor at wage wh. The demand functions for labor are also shown in Figures 4.1 and 4.2. In each case it is a step function approximation and, as in any linear programming model, comprises a series of steps, the number and length of which are determined by basis changes in the model solution as the wage is increased. Both functions become perfectly elastic at wage ws, the wage at which community workers can sell their own labor outside the community. The labor market equilibrium for the community always leads to a wage that falls within the range w0 and wh. If the labor de- mand intersects the supply function at either w0 or wh, then the wage rate will in effect be exogenous to the model. But if the intersec- tion occurs between these two values, then the wage is determined endogenously and is equal to the shadow price of labor. There are four seasonal labor markets in the model, each of which has the structure just described. The Division of the Microwatershed The simulated microwatershed is delin- eated into three zones based on altitude. The objective of the disaggregation is to as- sess the effect of water use by one zone on the zone located below. Springs are used for human, livestock, and crop consumption. 12 CHAPTER 4 Figure 4.1 Labor market for a submodel The unused water runs out of the micro- watershed. The first zone, which is located above 1,500 meters, benefits from a limited amount of water during the dry season and is controlled by a few ranchers. The zone located between 1,350 and 1,500 meters has an irregular topography, but benefits from several abundant springs during the dry season. The lowest zone, between 1,100 and 1,350 meters, is flatter and benefits from good access to the main stream of the microwatershed. Each zone has a different initial endow- ment of three types of soil with varying slopes, soil depths, and productivity. Each zone is managed by two groups of farmers— the small farmers and the ranchers—with different initial population densities. Three zones, three types of soil, and two groups of farmers lead to 18 land units. The optimal choice of land use for each of the four sea- sons is a combination of forest, pasture, crops, and land conservation structures. Each zone is located at a different distance from the main road and from the other zones, giving rise to varying transportation costs in terms of time from zone to zone. This spatial definition and disaggrega- tion of the watershed also avoids any major problems with externalities that could cut across the 18 submodels and invalidate use of a single maximand for the entire com- munity. Soil erosion is tracked for each sub- section of the watershed, and the model does allow for erosion to affect yields within the subsection in which it occurs. Movement of soil and associated plant nu- trients largely stays within the subsection in which it occurs (because of contouring, fencing, and so on) and moves from the cropped to the non-cropped area. The soil and nutrient runoff that does escape exits via streams that carry the soil out of the wa- tershed entirely. Since the runoff is not de- posited in neighboring sections of the watershed, this externality does not affect the model solution. Water flows within the watershed are more complex. Because there are springs in the upper segments of the watershed that feed the main stream that passes through the community, water use by farms in the upper watershed during the dry season has an im- pact on the amount of water available for ir- rigation by other farms in the community. This is potentially a problem for the model because maximization of aggregate utility for the community will lead to rational allo- cation of this water between different sub- models, even though this would not occur if the utility of each submodel were maximized separately. Fortunately, the problem may not be very serious. In the first place, the springs in question supply only a small part of the total water passing through the main stream. Second, not all the water in the stream is used by the community anyway; as reported in Chapter 3, about a quarter of the water is al- lowed to pass downstream for use by other communities, even in the dry season. Third, there is evidence that the community has taken some steps to solve this problem in a rational way. There seems to be an informal rule prohibiting any one farmer from taking excessive amounts of water for irrigation, and farmers downstream have successfully lobbied in the past to stop other farmers from overusing the springs in the upper water- shed. MODELING METHOD 13 Figure 4.2 Labor market for the community Another potential externality problem is pesticide contamination of water supplies. However, a study by the Zamorano Pan- american Agricultural School found no trace of agricultural pollutants within the water- shed. The real problem is contamination of the water that flows out of the La Lima community to other communities down- stream. Although this is a serious externality problem, the damage is caused outside the area encompassed by the model and hence does not affect the validity of the model’s objective function for simulating farmers’ behavior. Crop Production The model offers a selection from four crops: maize, potatoes, onions, or tomatoes. Onions are produced during the dry season, maize and tomatoes are produced during the first rainy season, and potatoes are produced during the second rainy season. The model also distinguishes four labor periods: the dry season, the first half of the first rainy season, the second half of the first rainy season, and the second rainy season. The crop production function in the lin- ear programming model represents the aver- age expected response to different factors of production. The production functions are linear-segmented approximations of non- linear functions.10 The production functions are specified for each type of crop, each zone, each type of soil, each type of farm, and each year of the planning horizon. The total production of each crop is an average yield multiplied by the cropped area, with the effects of the amount of organic and in- organic fertilizers used and of plowing instead of hoeing added, and the effects of inadequate irrigation during the dry season, soil erosion, and insufficient soil depth sub- tracted. The effects are non-linear and are approximated by linear segments. An exo- genous parameter that varies the response to organic and inorganic fertilization enables the model to simulate the effects of crop variety improvements over time. Product Allocation The marketing of vegetables is constrained in the model. In the 1970s, most farmers from La Lima could not produce vegetables because they did not have marketing out- lets. Only a few farmers who had special ties with the few traders who came to the closest town were able to sell. This is re- flected in the model by constraining the sales of products during the first years of the simulation. This constraint is progres- sively relaxed over time to reflect an in- creasing demand for vegetables, and after 1985, when the road was built, the con- straint is removed altogether. In the model, maize may be stored, con- sumed by the population and livestock, or sold during each season of the year, with different activities programmed for each season. The population consumes a fixed amount of grain during each period. Grains may be produced by the household or bought. The model seeks the best moment to sell, buy, and store grain depending on sea- sonal prices and family grain needs. Livestock Production The model simulates the size and manage- ment of herds of cattle, oxen, and mules that are owned by the small farmers and ranchers. Herd growth is determined by weight gain and by birth and mortality rates. If it is economically attractive, cattle can be bought or sold. Each livestock unit requires labor time, veterinary expenses, and forage throughout the year. Oxen are used for land preparation, mules for trans- portation, and cattle for producing milk, which is sold in some scenarios or is con- sumed on the farm. 14 CHAPTER 4 10 A complete development of the equations is available in the Appendix. The quantity of forage produced by pas- tures differs by season, by type of soil, and by altitude. A fraction of the unused forage is carried over from one season to the next. Livestock can also be fed with crop residues and with purchased feed. Cattle access to pastures in the microwatershed is controlled by market transactions in the form of graz- ing fees. Soil Erosion Soil erosion per hectare is modeled as a function of the area of each crop and the presence or absence of conservation struc- tures. Erosion can affect yields in two ways. First, runoff affects yields by reducing the amount of nutrients and water available to the plant (this loss is referred to as the “nu- trient effect”). Second, erosion cuts yields by diminishing soil depth, which reduces root growth once a minimum soil depth is reached (the “soil depth effect”). The linear programming model, based on the data generated by EPIC (Tables 4.1 and 4.2), shows how the nutrient effect is captured simply by specifying that yields decrease as a function of the quantity of soil eroded. Modeling the soil depth effect is more complex. In each of the 18 land unit areas, there are two initial volumes of topsoil, one planted with crops and one under forest, grass, and soil conservation. There is much less erosion under forest, pastures, and soil conservation structures than under crops. Over time, the soil volume under crops decreases whereas the soil volume under pastures, forest, and soil conservation struc- tures remains constant. However, when the model expands the cropped area at the ex- pense of the noncropped area, soil volume is transferred from the noncropped area to the cropped area. Conversely, when the cropped area is abandoned, a transfer of soil volume occurs from the noncropped area to the cropped area. This transfer provides for the possibility of abandoning cultivation on eroded plots and reclaiming pastures and forests. In each land unit, the topsoil volume has to be greater than a minimum volume per hectare of crop. If the soil volume falls below the minimum level, a variable representing insufficient soil depth takes on a positive value. This variable has an effect on yields in the production function. An equation is then added limiting each ton of soil deficit to the cropped area. This equation allocates the soil deficit to each crop within the land unit. The model can adopt soil conservation tech- niques such as terraces, live barriers, grass strips, or fertilization to reduce erosion, but only if these techniques are profitable. Forest and Perennials There are three types of trees in the model: pines, coffee in traditional plantations, and coffee in intensive plantations. Each land unit has different initial areas and volumes of pine groves by age group (1–4 years and older) and different wood productivity lev- els. If a cropland or pasture is abandoned, it returns to forest. Dead wood is collected for domestic consumption. When a plot is cleared to become a field or a pasture, dead wood can be used as fuelwood. Coffee trees are assumed to start producing beans three years after planting. The model can plant two varieties of coffee, a traditional variety and a more productive variety. The Biophysical Model EPIC Characteristics of the Model The biophysical model EPIC is used to de- scribe how land use practices affect yields and soil quality and how land quality in turn affects future crop yields. EPIC simulates hydrology, erosion, sedimentation, phos- phorus and nitrogen cycles, plant growth, and soil temperature. The interactions of these simulations are calculated on a daily basis, with the weather for each day gener- ated by a random weather generator. In EPIC, yields are expressed as a fraction of biomass, which in turn is a function of solar MODELING METHOD 15 active radiation and leaf area. Leaf area is simulated as a function of heat unit accumu- lation, crop development stage, and crop stress. Stress factors that reduce biomass growth are lack of nitrogen, phosphorus, and water, as well as inadequate temperature, soil compaction, excessive soil acidity, and aluminum toxicity. Soil erosion decreases biomass growth by leaching nutrients and by reducing root growth when roots reach more compact soil layers. Erosion levels were estimated using the Modified Univer- sal Soil Loss Equation (MUSLE) adapted for small microwatersheds (Williams, Jones, and Dyke 1987). Scientists have applied EPIC to many tropical conditions (Abruna, Rodriquez, and Silva 1982; Pavan, Bingham, and Pratt 1982; Williams, Jones, and Dyke 1984). The argu- ment that EPIC has been developed in the United States and thus is not adapted to trop- ical conditions is not exact. Many of the components of EPIC have been calibrated with data collected under tropical conditions in the United States and elsewhere. The main criticism that can be made of EPIC is that it has been developed for agriculture based on fertilizers and improved germplasm. The model is less reliable in extensive systems where crop performances depend upon the natural fertility of the soil. In La Lima, the agriculture is intensive enough to be mod- eled with EPIC. Results EPIC was parameterized to the soil condi- tions, climate, and cropping pattern found in La Lima and the model enacted yields of maize, onion, potatoes, and tomatoes grown in different rotations. Each scenario in- volved keeping the same climatic sequence, in order to compare different scenarios. Yields used in the linear programming model are the average of the yields obtained from 12 years’simulations with EPIC. Agronomic characteristics of the maize included in EPIC were adjusted to obtain yields similar to local yields. EPIC does not take into ac- count the competition between crops and weeds typical of farming systems where herbicides are not used. To evaluate the simulated effects of soil erosion on crop yields, erosion and non- erosion scenarios for three different types of slopes and fertilizer practices were com- pared. The results for maize and potatoes planted in a deep soil (30 centimeters) are re- ported in Tables 4.1 and 4.2. These simula- 16 CHAPTER 4 Table 4.1 Simulated effects of soil erosion and fertilizer use on maize yields on three different slopes (tons/ha) Slope Scenario 10% 22% 35% Yields with no erosion and no NPK 1.585 1.588 1.601 Yields with erosion and no NPK 1.573 1.536 1.471 Erosion 12.980 59.480 171.000 Yields with no erosion and NPK = 100 kg 2.135 2.134 2.130 Yields with erosion and NPK = 100 kg 2.127 2.103 2.043 Erosion 7.700 34.320 93.590 Yields with no erosion and NPK = 300 kg 2.887 2.874 2.848 Yields with erosion and NPK = 300 kg 2.880 2.864 2.811 Erosion 4.650 14.320 44.900 Note: NPK indicates inorganic fertilizers. tions illustrate that the use of fertilizers in- creases soil cover, which reduces erosion. The model also simulated scenarios with various soil depths to obtain the long-term effects of soil erosion instead of the short- term effects reported in Tables 4.1 and 4.2. When soil depth becomes insufficient, the yield decline is significant. The results from the EPIC simulations are incorporated into the economic model. MODELING METHOD 17 Table 4.2 Simulated effects of soil erosion and fertilizer use on potato yields on three different slopes (tons/ha) Slope Scenario 10% 22% 35% Yields with no erosion, NPK = 500 kg 14.364 14.320 14.215 Yields with erosion and NPK = 500 kg 14.060 13.440 13.130 Erosion 39.000 163.000 612.000 Yields with no erosion and NPK = 700 kg 16.446 16.390 16.272 Yields with erosion and NPK = 700 kg 16.215 15.720 15.156 Erosion 28.000 123.000 451.000 Note: NPK indicates inorganic fertilizers. C H A P T E R 5 Model Simulation Results The primary purposes of constructing the model were to discuss induced innovation hy- potheses and to explore the consequences of alternative policy scenarios for the La Lima microwatershed. However, before presenting the relevant simulations, we first articulate the baseline results for 1975–95 and how they are validated against the actual his- tory of the La Lima microwatershed over this period. Baseline Scenario The baseline scenario compared land use generated by the model with the historical infor- mation obtained from farmer interviews in La Lima. The result of this comparison establishes the validity of the model and its ability to replicate correctly the decisions taken by farmers in the community, with particular focus on the evolution of incomes, crop yields, commer- cialization, land management, erosion trends, water management, and shadow prices. The historical events known to have had an impact in La lima were introduced progres- sively into the simulation. These included the diffusion of sprinkler irrigation in 1979; the construction of an all-weather road in 1985; its improvement in 1993; and the construction of a water distribution system in 1992. Changes in historical prices were also fed into the model, because they were determined exogenously (Figure 5.1). Prices had been under strict government control until 1989, but this changed dramatically after the structural adjustment program of 1990, which comprised changing the system of export and import taxes and de- valuing the local currency. Land Use The simulated land use shows an evolution similar to that recalled by farmers in the plot his- tory survey (Figures 5.2 and 5.3). In both cases, land use changed only slightly despite pop- ulation growth, which increased steadily from 37 inhabitants per square kilometer in 1975 to 56 inhabitants in 1995. The main change in land use is the progressive development of vege- table production, induced by exogenous events such as the introduction of irrigation sprinklers, the road, market liberalization, and the potable water system. The uptake of irrigation was slower in the plot history data than in the model simulation. This is because the model does not account for the time needed to learn a new technique. The model is also optimistic about the availability of savings among small farmers. MODEL SIMULATION RESULTS 19 Figure 5.1 Deflated prices of the main products Figure 5.2 Simulated land use Figure 5.3 Historical land use Source: Authors’ data. The forest area decreases only slightly over time, both in the model and in reality. This is explained in the model by the con- siderable amount of labor time necessary to clear the forest. This result suggests that the current forest area is stable and that even re- moval of the current prohibition on tree cut- ting might not increase deforestation by farmers. The model does not require trees to be cut for energy needs because dead wood gathered by cleaning the existing forest pro- vides sufficient fuelwood. The coffee plantation area decreases over time in the simulation as it did in real- ity. In the model, this occurs because small farmers cut their coffee plantations in favor of vegetables, which offer better economic returns. The profitability of coffee increases slightly when the planning horizon in the model is extended, but not enough to com- pete with vegetables given the current price conditions in La Lima. Figure 5.4 shows the simulated land use by slope category. Soils with less than 15 per- cent slopes are predominantly in pastures. This is because water-logging on flatter fields results in lower yields than on soils with a steeper slope. The 15–30 percent slope area is covered mainly with crops and pastures, while the forest area decreases over time. The 30 percent sloped land has the more exten- sive forested area and, surprisingly, includes a significant area of cropped land. Figure 5.5 shows the simulated land use for three different zones of the microwater- shed. The upper zone has the largest propor- tion of crops and pastures while the two lower zones still have extensive forests. Figure 5.6 shows the simulated land use for each group of farmers. Small farmers 20 CHAPTER 5 Figure 5.4 Simulated land use by slope (o1 = flat, o2 = medium, o3 = steep) Figure 5.5 Simulated land use by zone (s1 = top, s2 = medium, s3 = lowest) have more than half of their holding under crops with a decreasing area of forest over time, whereas ranchers have mainly pastures and forest. Incomes There are two distinct periods in the evolu- tion of per capita income: first, a period of slow increase before the market liberaliza- tion policies of 1990; and, second, a period of dramatic increase after market liberaliza- tion (Figure 5.7).11 The increase during the first period is due to technological improve- ments, which allow incomes to rise slightly despite worsening prices and continued pop- ulation growth. After 1990, however, simu- lated real income doubles in less than four years because of rapidly increasing vege- table prices. There were sizable differences in the re- sults for different types of farmers (Figure 5.8). Ranchers’ per capita income decreases continuously after 1985 as meat prices de- clined. Small farmers’ income increases slightly in that same period due to exoge- nously introduced varietal improvements. After market liberalization, all incomes in- crease at a similar pace because meat and crop prices increase. Income sources also change over time (Figure 5.9).12 Incomes from livestock de- crease, while onions and potatoes replace MODEL SIMULATION RESULTS 21 Figure 5.6 Simulated land use by type of farm (h1 = ranchers, h2 = small farmers) Figure 5.7 Simulated income per person 11 All costs and incomes are deflated to 1987 prices with the Consumer Price Index (World Bank 1997) and are in lem- piras (in 1987, one lempira was equal to US$0.30). 12 In all graphs that report incomes, the value of leisure is included in income. The value of leisure is small and changes only slightly in the different scenarios. maize as the primary source of income. Cof- fee and off-farm activities remained mar- ginal sources of income. Crop Production The simulated yields of maize, onions, pota- toes, and tomatoes increase steadily through time thanks to technological improvements in the form of improved varieties (Figure 5.10) and increasing application of fertilizers (Figures 5.11 and 5.12).13 The simulated yields for 1995 are close to the actual yields farmers obtained in La Lima. Surprisingly, however, the recent crop price increases do not result in large yield in- creases. This is explained in the model by the limited availability of labor at harvest time.14 The model prefers corralling (where cattle graze maize residues) to compost or manure production because of the latters’ high labor requirements. Commercialization The sale of part of production increases with time, particularly after the adoption of irrigation and fertilization techniques in 1979–80, and after the construction of the road in 1985 and its improvement in 1993 (Figure 5.13). When irrigation is adopted in 1979, the model begins selling less maize 22 CHAPTER 5 Figure 5.8 Simulated income per person by farmer group and zone (s1 = top, s2 = medium, s3 = lowest) Figure 5.9 Simulated sources of income in the whole microwatershed 13 The quantity of inorganic fertilizer per hectare increases in steps because of the linearity of the solver. In reality, fer- tilizer use increases more continuously. 14 Note that the model did not allow for immigration, because it does not occur in reality, as neighboring communities also have a labor shortage. Note also that families provide most of the labor and that wage labor remains relatively marginal. MODEL SIMULATION RESULTS 23 Figure 5.10 Simulated yields Figure 5.11 Simulated fertilizer use Figure 5.12 Simulated inorganic fertilizer use per hectare Figure 5.13 Simulated use of crop production and more onions and potatoes. When the road is constructed in 1985, the model sells even more potatoes, and starts to buy a por- tion of the maize that is consumed locally. When the road is improved in 1993, the model diversifies into fresh vegetables such as tomatoes. Surprisingly, however, maize remains a competitive crop all along. The reasons for this will be explored later. Ranching Improved prices for vegetables and maize create an incentive to convert pasture into cropland. Accordingly, the model slowly de- creases the cattle herd throughout the simu- lation (Figure 5.14). Despite the low return per unit of land, however, ranching remains in the solution because ranching requires limited labor, and fencing can be done dur- ing periods of low activity. The model increases the small farmers’ cattle herd in 1979 because the introduction of irrigation and the increasing intensifica- tion of farming reduced the total need for cropland and freed it up for conversion to pastures. Ranchers do not increase their herd, however, because all their pastures are used and the conversion of existing forest into pastures is labor consuming. Mule numbers decrease over time be- cause the new road makes local transporta- tion less necessary (Figure 5.15). The volume transported by mules decreases twice: first in 1985 when the road is built, and then again in 1993, when the road is im- proved. Oxen numbers increase only in the upper zone of the microwatershed where the cropped area also increases the most. Erosion The average simulated amount of soil ero- sion for the whole microwatershed is close to 6,700 tons per year (or 7 tons per hectare per year) (Figure 5.16). This is almost the same amount as was estimated in 1996 at the out- stream of the microwatershed.15 To obtain this result, however, the nutrient effect of erosion on yields had to be suppressed. If the nutrient runoff effect is maintained, then the simulated erosion becomes less than 3 tons per hectare because the model adopts grass strips on 40 hectares at the beginning of the simulation. This technique was adopted by the model because it reduced erosion while requiring little labor and investment, its only cost being the space occupied in the field. The model compensated for this lost area by expanding the cropland area—the popula- tion density being low enough in La Lima to make such an expansion affordable. In reality, farmers did not adopt grass strips. According to our interviews, farmers were not aware of the effects of erosion on yields. Simulations with EPIC also suggest that the effects of erosion on yields are small where soils are deep enough (less than 3 per- cent yield loss per year on steep slopes). Moreover, incomes increase by only 1.2 percent in the model after removing the nu- trient effect of erosion, again showing a small effect. In other words, the model re- acts to something that farmers do not think is important. Consequently, removing the nutrient effect of erosion in the baseline sce- nario mimicked farmers’ perceptions. How- ever, the results in this report maintain the erosion calculation and the soil depth effect on yields. In the baseline scenario, soil depth di- minishes rapidly on the steeper slopes (Fig- ure 5.17) and quickly reaches the level where roots become affected. The model re- acts by abandoning these plots and reclaim- ing new ones, given this approach is less expensive than the construction of soil con- servation infrastructures. A notable excep- tion to this pattern arises in 1991 in the most populated land unit, which also has a high proportion of steep slopes. In this case, the model invests in the development of terraces on 10 hectares (Figure 5.18). The model does this as the critical soil depth becomes 24 CHAPTER 5 15 The two quantities differ slightly because the total sedimentation at the out-stream is not the sum of erosion at the plot level. A stream can deposit part of its sediment before it reaches the out-stream of the microwatershed; conversely, the stream can carry to the outlet sediments captured in the channel of the stream itself. Figure 5.14 Simulated livestock herd size Figure 5.15 Simulated transportation by mule Figure 5.16 Simulated erosion aggregated by slope Figure 5.17 Simulated soil depth in three land units insufficient, and there are no more pastures or forest lands available for cropping. These results correspond to reality, and today this land unit is the one with the most conserva- tion structures. Water Management The model replicated quite accurately the ac- tual use of water, showing a progressive in- crease from 1979 onward in the use of water for irrigation and human consumption (Fig- ure 5.19). As a result of the introduction of sprinklers into the model in 1979, water out- flows from the microwatershed decreased significantly. Figure 5.19 shows that in the upper microwatershed the small amount of spring water is rapidly used for irrigation. After potable water distribution was intro- duced in 1993, the model used even more water because the distribution system al- lowed the irrigable area to be extended.16 Shadow Prices The shadow price of a factor of production (land, water, labor, or capital) measures the amount by which the utility function would increase if one more unit of this fac- tor became available.17 Induced innovation theory suggests that, if population increases, then the shadow price of labor should de- crease (holding everything else constant) while the shadow price for land should increase, because land becomes scarce rela- tive to labor. However, in our results, the shadow price of land stagnates while the shadow price of labor increases continu- ously (Figure 5.20). This result is due to the increasing profitability of labor-intensive activities such as vegetable production. Con- sequently, farmers have fewer reasons to acquire new land. Land is still abundant in La Lima and extra land would increase current incomes by a small amount. Most small farmers have share-cropping arrange- ments with larger farmers to produce maize and vegetables on larger farmers’ land. These results imply that, for situations simi- lar to those in La Lima, agricultural research and extension should focus on ways to in- crease labor productivity, particularly during peak periods. For example, labor-saving methods for harvesting maize and vege- tables would increase productivity and in- comes. Conversely, techniques that increase yields would have a smaller effect on per capita incomes. The shadow prices of labor vary by pe- riod, and are not uniform through time or 26 CHAPTER 5 Figure 5.18 Simulated soil depth and soil conservation structure in one land unit 16 A lower bound was added in the model for the stream volume, because in reality there is an implicit rule that users cannot completely drain a stream. 17 The shadow price of a factor is the amount by which global net income will increase if one unit of this factor is added. If the factor is not limiting, the shadow price is equal to zero. MODEL SIMULATION RESULTS 27 Figure 5.19 Simulated water volume in the outflow of the streams by zone Figure 5.20 Shadow prices of land and labor Figure 5.21 Simulated shadow prices of labor by season season (Figure 5.21). From 1975 to 1981, for instance, the shadow price of labor is high- est for the period of land preparation for maize. This means that maize production in that season is most constrained by labor scarcities. Between 1981 and 1989, by con- trast, all four working seasons have the same shadow prices, implying that all peri- ods are equally limiting because the model smoothes labor requirements over time by scheduling some tasks during low activity periods. For example, the transportation of maize or inputs can be postponed or planned in advance. Similarly, the dobla technique18 for maize postpones the harvest until more labor is available. After the market liberal- ization program of 1990, the shadow price of labor by period diverges again (Figure 5.21). The limiting period becomes the middle of the rainy season, when the maize harvest competes with vegetable harvesting and po- tato planting. Methods to reduce labor re- quirements during this period would have a big impact on production. The shadow price of the marketing con- straint on vegetables is positive only at the beginning of the simulation and disappears once the road is built (Figure 5.22). At the beginning of the simulation, the marketing constraint depresses the shadow price of labor by limiting vegetable production. The shadow price of water increases rap- idly in the last few years of the simulation be- cause water becomes a scarce resource when the installation of the water distribution sys- tem allows an increase in irrigation (Figure 5.23). Water then becomes a binding con- straint on expansion of the irrigated area, but the model reacts by producing vegetables during the rainy season. Hypotheses In this section, model simulations over the period 1975–95 aid discussion of the in- duced innovation hypotheses formulated in Chapter 1 about the effects of population pressure, increased market access, and im- proved technologies and market prices, and the impact of agroecological conditions on these relationships. The Effect of Population Pressure The first two hypotheses in Chapter 1 state that increases in population pressure lead to (a) lower per capita incomes and (b) contin- uing degradation, until some critical value of productivity is reached at which point it be- comes profitable to invest in resource im- provement. Two contrasting scenarios help in the discussion of these hypotheses: one with increased population pressure, and one in which no population growth is assumed. To simulate increased population pres- sure, we allowed permanent workers to mi- grate into the village but at an annual rate that cannot exceed the annual population growth of the existing farm population. The model will take in new families if the discounted value of their contributions to the commu- nity’s aggregate utility is at least as great as their costs (additional food requirements, for example). We do not consider the oppor- tunity cost of migrant families outside the community. This is because we are not trying to model migration decisions per se, but only to simulate what will happen in La Lima if the population density were increased. Since labor is initially scarce in the com- munity and additional family workers prove cheaper than hiring in day labor from outside the community, the model allows in all the additional permanent workers that the con- straints permit until a population density of 150 inhabitants per square kilometer is achieved, which is about three times the cur- rent density. Beyond this density, total ag- gregate income starts to decline, and some farmers out-migrate. Per capita income de- clines to 24 percent of its value in the base- line scenario by 1995 (Figure 5.24). All the forest is cut and households have to turn to 28 CHAPTER 5 18 The dobla is a traditional technique whereby the maize stem is bent over right under the cob, which allows the grain to dry in the field and removes the urgency of harvesting. MODEL SIMULATION RESULTS 29 Figure 5.22 Shadow price of the marketing constraint Figure 5.23 Shadow price of water by zone Figure 5.24 Simulated per capita income: Alternative scenarios for population growth alternative energy sources (for example, kerosene for cooking). Most cattle are also sold, and only a few oxen and mules are kept for productive purposes. Almost the entire microwatershed is cultivated with maize, with a limited area being kept under vege- tables. Soil erosion initially rises to more than 25 tons per hectare because steep slopes are cultivated, but then conservation tech- niques are adopted and erosion declines again by 1994 (Figure 5.25). When the population is assumed to re- main constant at its 1975 level, then per capita income is about 10 percent higher than in the baseline scenario (Figure 5.24). Soil erosion is halved (Figure 5.25). The results of the two population simu- lations are similar to our initial hypothesis that, when population density is still rela- tively low, population pressure has negative effects on natural resources. However, when the population reaches a higher density and the productivity of the resource base is threatened, farmers start to improve their natural resource management practices. This process can be represented by a U-shaped function where the productivity of natural resources decreases to a point where farmers start to invest in their enhancement. The results also suggest that, with addi- tional population pressure, farmers are likely to expand their cropland by converting pas- ture rather than forest areas, both because of the low profitability of ranching and because it would be costlier in labor terms to convert forest rather than pasture into cropland. Thus, any future expansion of crops will occur at the expense of pastures, and the for- est area will remain largely intact. The simulations show too that increased population pressure results in lower per capita incomes, unless technological inno- vations or higher prices compensate for the decreasing returns to labor. Population growth leads to lower returns per capita be- cause, although every additional worker in- creases the global income of the community, the income of this marginal worker is lower than the average. This occurs despite the adoption of more labor-intensive activities. Out-migration could theoretically offset these income declines, but outside opportu- nities are not attractive enough to encourage out-migration in the model. The Effect of Access to Markets The hypothesis is that market access in- creases per capita incomes and promotes land conservation because increases in land value make investments in land improve- ments more profitable. To simulate this hy- pothesis, two scenarios simulated the effect of market access, using distance to a road as an indicator of market access. In the first simulation, the equations cap- turing the effect of the construction and sub- sequent improvement of the road in the community are removed. The model re- sponds by transporting products to the next community 6 kilometers away. However, 30 CHAPTER 5 Figure 5.25 Simulated erosion: Alternative scenarios for population growth some of the more perishable crops (such as tomatoes) are no longer produced because they cannot be transported this way. The “re- moval” of the road reduces income by only 11 percent, which is surprising given the im- portance usually attached to market access (Figure 5.26). The model still produces sim- ilar amounts of maize, onions, and potatoes to those in the baseline scenario, and com- pensates for the lack of a good road by de- laying the transport of maize and some vegetables to less busy periods. Figure 5.27 shows that the road con- struction leads to a sharp increase in soil ero- sion; this is because farmers start to produce more potatoes for the market, which are a highly eroding crop. The model simulated different scenarios with respect to the distance of the micro- watershed from the main road—specifically, distances of 20, 30, and 40 kilometers from the village to the main road—plus a scenario in which the distance remains unchanged but the connecting road is removed. All transportation must be made by mule. This simulated situation of remote market access (which is actually quite typical of the Cen- tral Region) has a radical effect on per capita incomes (Figure 5.26). Coffee production increases sharply in the simulation after the coffee price boom of 1979, whereas the maize area declines to the minimum area needed to meet local food consumption. Maize is not intensified and yields remain low because the model does not find suffi- cient labor to transport fertilizers. Surpris- ingly, the model produces potatoes and irrigated onions and transports them to mar- ket by mule. The number of cattle increases to reach 30 percent more units than in the baseline scenario; this is explained by the decrease in maize area. However, cattle MODEL SIMULATION RESULTS 31 Figure 5.26 Simulated per capita income: Alternative scenarios for access to markets Figure 5.27 Simulated erosion: Alternative scenarios for access to markets numbers decrease again after the liberaliza- tion when farmers allocate more labor to vegetables. Soil erosion decreases when the distance to the road increases because the cropped area becomes smaller (Figure 5.27). Land conservation infrastructures are still not adopted, as soil depth never reaches the crit- ical levels where yields are reduced. Erosion increases after the liberalization because farmers plant more rainy season vegetables. These simulations underline the role of roads in determining the development path- way that a community may follow. Coffee is more profitable than maize in remote areas. Onions and potatoes are also prof- itable even if the produce has to be trans- ported by mule to the closest road. The simulations suggest that erosion decreases with the distance to roads because the cropped area decreases as more time is spent in transport. The initial hypothesis held the expectation that better road access would mean more investment in land conservation structures. This does not hap- pen because the model finds it more cost- effective to allocate labor to production than to invest in terraces or live barriers. The Effect of Technological Improvement The fourth hypothesis states that technolog- ical innovation compensates for decreasing returns to labor. To simulate this hypothesis, the model simulated removal of three tech- nologies from La Lima, namely crop variety improvement, sprinkler irrigation, and the potable water distribution system. Crop Variety Improvement. Per capita in- comes fall dramatically after we remove the new crop varieties (this was done by keeping the same crop response to fertilizers as in 1975): in 1995, per capita incomes are 41 percent lower than in the baseline scenario (Figure 5.28). Incomes decrease until 1989, but then begin to increase again after the market liberalization. Despite population pressure and higher commodity prices, yields and the amount of fertilizer used per hectare remain almost the same. The 1995 maize area is 6 percent larger than that in the base scenario because the model uses more extensive production methods with less labor per unit of land. Erosion remains low without technology improvement, because the potato area is much smaller than in the baseline scenario (Figure 5.29). Irrigation in 1979. Sprinklers were introduced by the extension services in 1979. If this technology is eliminated, the model simply stops producing vegetables during the dry season. However, the model compensates for the loss of income by producing more maize and vegetables during the rainy sea- son and transporting maize and some vege- tables to the markets during the then less busy dry season (Figure 5.28). 32 CHAPTER 5 Figure 5.28 Simulated per capita income: Alternative scenarios for technologies Erosion is slightly greater without irriga- tion than with irrigation, leading to soil depth problems and a return to pastures, which in turn results in lesser erosion in 1995 (Figure 5.29). This last result underlines the impor- tance of dynamics in natural resource man- agement. A community may have a current low level of erosion because farmers previ- ously eroded their plots so much that finally they reached a critical soil depth and had to invest in conservation structures. Or a com- munity may currently have more erosion compared with another community because farmers still have deep soils thanks to better soil management. Potable Water Distribution in 1993. Simulat- ing removal of the potable water distribution system in 1993 results in a slightly lower in- come increase of about 2.5 percent (Figure 5.28). The water distribution system has a smaller than expected impact on incomes, because it was not designed for irrigation. In reality, however, the potable water distribu- tion system has a larger impact on equity be- cause it allows almost every farmer to produce at least a few square meters of vege- tables near the house during the dry season, whereas, before, only farmers with plots near the main streams could produce vege- tables. The potable water distribution system has no effect on erosion. Conclusions. Seed improvement had a sig- nificant impact on per capita incomes through its effects on vegetable and maize production. In fact, it more than offset the negative impact of population growth on per capita income, thereby affirming our hy- pothesis. In reality, however, poor farmers in La Lima used inferior varieties of seeds. This suggests that credit and extension programs would likely have an important effect on production and incomes, as well as on the distribution of income.19 The adoption of sprinkler irrigation was also an important source of technological change in La Lima. However, irrigation had a smaller than expected impact on yields because the dry season is short and vege- tables can be produced during the rainy sea- son. The use of a gravity-fed system makes irrigation possible anywhere below water collection points. In practice, however, irri- gation in La Lima is concentrated in areas closest to streams and water points. The MODEL SIMULATION RESULTS 33 Figure 5.29 Simulated erosion: Alternative scenarios for technologies 19 Extension services are often considered to be ineffective. It is true that extension services have limited success when they promote land conservation practices. In La Lima, however, farmers were relatively positive about extentionists’ impact, explaining that extension services brought the new seeds and the sprinklers currently used in the area. In the hillsides of Honduras there is little adoption of new technologies without the help of extension services (Bunch and López 1995). introduction of the potable water distribution system did not markedly increase the pro- duction of vegetables, but it did enable the benefits of irrigation to be spread more equi- tably across farmers. Seed improvement worsened erosion be- cause it increased rainy season potato culti- vation. The adoption of sprinklers first reduced erosion by reducing the area of rainy season crops, but irrigation, by increasing the labor cost during the dry season, makes investment in land conservation less likely. The Effect of Agroecological Conditions The hypothesis is that agroecological condi- tions are the most important factor deter- mining incomes and resource conditions. Three simulations were run to discuss this hypothesis. The first assumes that vegetables can no longer be produced during the rainy season, a situation that is common in many of the lower-altitude areas of Central Hon- duras because of unreliable rains. The sec- ond assumes that vegetable production is not possible at all, again a common feature in many less-favored hillside areas. The third simulation assumes shallower soils. No Rainy-Season Vegetables. Confronted with the absence of a reliable rainy season, the model increases the production of maize and grows a few hectares of irrigated onion during the dry season. Per capita income is 39 percent lower in 1995 than in the baseline scenario (Figure 5.30). The maize area is greater than in the baseline scenario and also has higher yields, because more labor can be devoted to maize production. Despite the greater area of maize, less erosion occurs be- cause rainy season vegetables are the main cause of erosion (Figure 5.31). No Vegetables. If vegetables cannot be pro- duced at all in the microwatershed, income would fall to 30 percent below the baseline in 1995 (Figure 5.30). The model attempts to compensate for income losses by producing more maize. This leads to some decline in soil erosion, particularly after the road is constructed in 1985 (Figure 5.31). Reduced Soil Depth. Given shallower soils, the model has farmers invest earlier in land conservation techniques (terraces, live barri- ers, and grass strips), with the result that soil erosion is rapidly reduced to less than 2 tons per hectare (Figure 5.31). The labor spent in constructing land conservation structures initially reduces per capita incomes, but in- comes return to baseline levels once the con- struction work is completed (Figure 5.30). This important simulation shows that policies to reduce erosion are more likely to succeed in areas where soils are shallow. In regions with deep soils, farmers are likely to be much less responsive to land conservation programs. 34 CHAPTER 5 Figure 5.30 Simulated per capita income: Alternative scenarios for natural resource constraints Conclusion about Agroecological Conditions. The first two simulations show that climate is a major factor in explaining income dif- ferences across communities. If vegetable production is impossible, per capita in- comes are much lower. The simulation for reduced soil depth suggests that soil depth has a small impact on income because land conservation structures are not very costly. However, if these conservation structures are not built, incomes decrease to very low levels. Policy Interventions The bioeconomic model provides a tool for simulating the possible impact of alternative policy interventions on incomes and natural resource conditions. The policy scenarios simulated below were selected because of their relevance to ongoing policy discus- sions in Honduras. Each case simulated what the impact would have been during the pe- riod 1975–95. First, what would have hap- pened if the liberalization of 1990 had not occurred? Second, what would progress have been without the use of inorganic fer- tilizers? Then a land reform simulation forecast how this would have affected the development of the community. The final scenario shows whether dairy markets would have developed in La Lima if there had been access to a processing plant. Market Liberalization This study examines the impact of market liberalization by “canceling” this policy in 1990 and keeping prices at their 1990 level thereafter.20 In the first three years after 1990, the scenario without liberalization produces higher incomes because prices were more favorable to vegetable produc- tion. However, by 1995, incomes without liberalization are 32 percent lower compared with the baseline scenario (Figure 5.32), showing that, in the longer term, liberaliza- tion did increase incomes. Another positive effect of liberalization is that it reduced in- come inequality, as the increased profitabil- ity of vegetables diverted labor from wage work on ranches to vegetable production. The liberalization led to higher fertilizer use, which increased yields, and increased labor demand per hectare. This extra labor re- quirement per unit of land reduced the area planted and hence improved soil erosion slightly (Figure 5.33). MODEL SIMULATION RESULTS 35 Figure 5.31 Simulated erosion: Alternative scenarios for natural resource constraints 20 It is not possible to assert what prices would have been after 1990 without market liberalization, but it appears plausi- ble to assume that the direction of price changes since 1990 is consistent with what one would expect the market lib- eralization and its devaluation policies to have caused. No Inorganic Fertilizer This scenario simulates what would have happened during 1975–95 without the use of inorganic fertilizers. There are regular dis- cussions in Honduras about using inorganic fertilizers because of the environmental con- tamination and economic dependency that imported non-organic inputs create. Accord- ing to the model, a ban on inorganic fertiliz- ers would have reduced net per capita income by 29 percent by 1995, compared with the baseline scenario (Figure 5.34). To compensate for the lost nutrients, the model produces up to 850 tons of compost per year while continuing to corral cattle. However, lower maize yields lead to less crop residue, which in turn reduces livestock feed and livestock manure. The model also brings more land under cultivation to compensate for losses in yields. Furthermore, lower fer- tilization decreases soil cover by crops. These changes lead to a 35 percent increase in soil erosion by 1995 compared with the baseline scenario (Figure 5.35). Land Redistribution In this scenario, the model allows the popu- lation to move freely within the microwater- shed, and spatially relocates farmers so as to maximize total and average community income, as would happen with a well- conceived land reform. This contrasts with a baseline scenario in which farmers are con- strained by the initial land endowments within each submodel, and no land transac- tions (either sale or lease) are allowed be- tween submodels. Under this new scenario, the model suggests moving some of the 36 CHAPTER 5 Figure 5.32 Simulated per capita income: Scenario without market liberalization Figure 5.33 Simulated erosion: Scenario without market liberalization population from the more highly populated areas to the less populated area. The move enables full advantage to be taken of the springs for irrigation during the dry season. The global effect of this measure on total income is small; average per capita income increases by only 4 percent (Figure 5.36). This is because, in the baseline scenario, the development of vegetable production helps small farmers obtain higher incomes, while the inequitable distribution of land is also compensated through the labor market. This simulated land reform leads to slightly more erosion because a larger area is cultivated (Figure 5.37). Dairy Farming The possibility of producing and selling milk requires the organization of a collection sys- tem by a milk processing factory. If these conditions are introduced into the model, specialized dairy farming appears in 1983 when the road is built, to become one of the main production activities. It also signifi- cantly increases per capita income (Figure 5.38). These changes rapidly boost the num- ber of cattle in the microwatershed to 700 units (almost all mules and oxen are re- placed). More than 80 tons of maize and 20 tons of feed concentrate are purchased every year to fulfill local needs. Small farmers’per capita incomes are increased to the same level as ranchers’ incomes, although their different resource endowments foster spe- cialization, with small farmers producing milk and large ranchers producing meat. Much of the present cropland is turned into pasture and the vegetable area is reduced MODEL SIMULATION RESULTS 37 Figure 5.34 Simulated per capita income: Scenario without inorganic fertilizer Figure 5.35 Simulated erosion: Scenario without inorganic fertilizer Note: NPK indicates inorganic fertilizers. Note: NPK indicates inorganic fertilizers. 38 CHAPTER 5 Figure 5.36 Simulated per capita income: Alternative scenarios for land reform Figure 5.37 Simulated erosion: Alternative scenarios for land reform Figure 5.38 Simulated per capita income: Scenario with dairy production considerably, farmers growing only irri- gated onions during the dry season. This new land use produces very low levels of erosion (Figure 5.39). However, in 1992 the