World Development, Vol. 19, No. 9, pp. 1145-1151, 1991. 0305-750X/91 $3.0t) + 0.110 Printed in Great Britain. © 1991 Pergamon Press plc

The Causes of Tropical Deforestation in Ecuador: A Statistical Analysis

DOUGLAS SOUTHGATE, RODRIGO SIERRA, and LAWRENCE BROWN* Ohio State University, Columbus

S u m m a r y . - - This paper reports a statistical analysis of the causes of agricultural colonization and land clearing in eastern Ecuador. The results bolster many of the arguments that have been made about the social forccs driving I~ropical deforestation. In gcneral, the prospect of capturing agricultural rents stimulates settlement of tree-covered hinterlands. Accordingly, the presence of roads and urban centers attracts rural settlement. Colonists also prefer to go where soils are suitable for crop or livestock production, all else remainiug the same. In addition to being a function of population pressure, the extent of land clearing is shown to be a consequence of tenure insccurily. These results suggest that institutional reform is nccdcd to encourage conservation of tropical forests.

1. I N T R O D U C T I O N

Widesp read concern over the impacts of tropi- cal de fores ta t ion has s t imula ted examina t ion of the factors inducing people to sett le and clear t r ee -covered h in te r l ands in the Thi rd World. For the most par t , r esearchers have u n d e r t a k e n case studies, analyzing individual f ront ier c o m m u n - ities ' r esponses to a var iety of gove rnmen ta l policies and marke t forces (Macdona ld , 1981; Moran , 1983: Rudel , 1983; Lisansky, 19891. While this approach yields valuable insights, a series of case studies does not cons t i tu te a solid basis for assessing the relat ive s t rength of the var ious forces responsible for defores ta t ion . Tha t kind of assessment requires an analysis with an e x p a n d e d geographic focus.

The research repor ted in this paper addresses this need . Our s tudy of the causes of t ropical de fores ta t ion was conduc ted in eas te rn Ecuado r ' s A m a z o n i a n lowlands. Work ing with data col lected in a large area has al lowed us to use statistical analysis, r a the r than a case s tudy approach . Tha t analysis is well sui ted to clari- fying re la t ionships a m o n g hind clearing, rural popu la t ion pressure , local d e m a n d for agricul- tural commodi t i e s , inf ras t ruc ture deve l opm en t , and tenure insecurity.

The paper begins with a br ief descr ip t ion of the study area. Basic fea tures of the physical en- v i r o n m e n t are no ted , as are character is t ics of the h u m a n popula t ion . Next , a model explaining t ropical de fores ta t ion is presented . Af te r that ,

da ta used to e l abora te the model are descr ibed and the results of regression analysis are summar - ized. We conclude with a discussion of the needs for addi t ional research as well as the impl icat ions of this study for policy govern ing the develop- men t of the A m a z o n basin.

2. E A S T E R N E C U A D O R ' S A M A Z O N I A N L O W L A N D S

With 26,266,000 ha of ter r i tory on the South A m e r i c a n ma in land and 801,000 ha in the Gala- pagos Islands, Ecuador (Figure 1) is slightly larger than the fo rmer West G e r m a n y or the state of Colorado. Never the less , its geography is as tonishingly wtried. The coastal plain, or li t tor- al, ends abrupt ly 150-200 kill in land from the Pacific Ocean . Nest led in the Andes , which are domina t ed by volcanoes rising as much as 6,310 m above sea level, are t empe ra t e i n t e r m o n t a n e valleys. Beyond the eas te rn range of the Andes , a rapid descent into the A m a z o n i a n rainforest of

*The authors thank the Jesse Smith Noyes Foundation for supporting Mr. Sierra',', graduate studies and the Graduate School of Ohio State University for providing funds to complete the research described ill this paper. Also, Mr. Roberto U[Ioa greatly facilitated data collec- tion in Ecuador and Profs. Paul Colinvaux, Earl Epstein, and John Lyon of ()hit) State University provided many useful insights to the authors, who arc, of course, responsible for all errors and omissions ill this paper.

1145

1146 WORLD DEVELOPMENT

0 0

,¢o~ 50 k.

provincial boundary

duction around Nueva Loja, began during the 1970s. Improved infrastructure, in turn, has stimulated rapid settlement. Between the 1974 and 1982 censuses, 92,700 individuals moved into Ecuador's four eastern provinces. The 1982 census reveals that a relatively small part of the labor force found employment in the oil industry or in the rapidly expanding service and construc- tion sector. Instead, 611% of the region's econo- mically active population worked in agriculture (INEC, 1985).

With the agricultural population expanding rapidly, there has been substantial deforestation in the Oriente. For example, although forests continue to cover well over half of northeastern Ecuador, 36% of the region has been claimed by agricultural colonists, all of whom arc clearing land (MAG, 1987: Brown et al., 19921.

Figure 1. Map of Ecuador and study area. 3. A MODEL OF T i l E CAUSES OF TROPICAL DEFORESTATION

the Oriente begins. An Amazonian frontier town, Nueva Loja (formerly called Lago Agrio) is only 400 km from the Pacific and 300 m higher than the surface of the Atlantic Ocean, which is 3,0011 km due east.

In Ecuador, dramatic topography is matched by climatic and biological diversity. Twenty-five of the 30 life zones identified by Holdridge (1967) are found in the country (Cabarle e ta l . , 1989, p. 5). In addition, the number of plant and animal species living in or endemic to Ecuador ap- proaches similar counts for larger countries in the Western hemisphere (Cabarle et al., 1989). Nowhere are natural resources more varied than in the Amazonian lowlands of the country's four eastern provinces.

Regardless of former President Plaza Lasso's warning that "the Oriente is a myth," many Ecuadorians and their political representatives continue to think that the eastern part of their country has great agricultural potential. Cer- tainly, climatic conditions are promising. Aver- age temperatures vary from 20°C to over 24°C and precipitation is plentiful throughout the year. Ongoing surveys of the Oriente, however, indi- cate that limited soil fertility, and poor drainage render sustained crop production uneconomical in most of the region. In 1987, the Ministry of Agriculture and Livestock concluded that no more than 16'7o of northeastern Ecuador should be in cropland or pasture (MAG, 1987).

Agricultural colonization of the Oriente was confined to a few lower Andean valleys until the late 1960s. Exensive road construction in the region, undertaken primarily to service oil pro-

Although population growth and deforestation are closely related, land clearing is not an exclusively demographic phenomenon. As Bromley and Cernea (19891 point out, tangible environmental problems in the developing world, including tropical deforestation, are often a manifestation of underlying institutional crises.

A crisis is obvious when public sector claims on forests far outstrip the government's capacity to manage "its" properties. In Ecuador, for exam- ple, practically all tree-covered land is designated as "forest patrimony" or national parks. The government's effective control of that land is negligible, however. No rangers work in the 211(111,000 ha of forest patrimony delimited in the northwest and northeast (MAG, 1987) and, as of 1987, a mere two administrators, 25 technicians, and 119 permanent and seasonal rangers had been assigned to the 2,111(1,000 ha of parks in continental Ecuador (DINAF, 19881.

A discrepancy between public sector claims on forests and limited governmental capacity to control access would normally bring about what Hardin (1968) calls a "tragedy of the commons." That is, individuals encroaching on the forest patrimony or national parks (e.g., to collect fuelwood) would neglect the costs society as a whole associates with forest degradation. Tacitly recognizing this, Ecuador's government allows individuals and firms to make claims on tree- covered land. Private parties interested in timber management, however, cannot acquire forests: logging concessions were banned in 1982. In- stead, formal property law, as executed by the Ecuadorian Institute for Agrarian Reform and

CAUSES OF TROPICAL DEFORESTATION 1147

Colonization ( IERAC), has required that those settling in tree-covered hinterlands, clear their respective parcels in order to acquire formal property rights.

This tenurial regime induces a cycle of exces- sive land clearing and inadequate soil conserva- tion in the vicinity of agricultural frontiers (Southgate, 1990). In addition, bureaucratically induced tenure insecurity contributes to defore- station. Like other land reform agencies estab- lished during the era of the US-sponsored Alliance for Progress, IERAC requires a long time, often years, to adjudicate claims for formal property rights. Delays are explained in part by administrative constraints; IERAC's record- keeping system is cumbersome and the agency did not acquire its first computer until the late 1980s. Furthermore, the complexity of formal property law lengthens the adjudication process; IERAC is obliged to execute 10 separate proce- dures during the course of settling a tenure claim (Seligson, 1984).

In the face of tenure insecurity, which is especially severe for those who lack the funds or influence needed to "move the system along," agricultural colonists in the Amazon basin tend to safeguard land claims by continuously using their parcels for crop or livestock production (Rudel, 1983). Indigenous groups respond identi- cally to the same institutional incentives. Mac- donald (1981) reports, for example, that the periodic fallowing scheme long practiced by the Amerindian community of Pasu Urcu, in eastern Ecuador, was abandoned during the 1970s after IERAC agents informed the community that fallow lands could be claimed by agricultural colonists, who were 50 km away at the time. For colonists as well as indigenous inhabitants, then, continuous agricultural production is a way to assert and to protect property rights.

To describe the demographic and institutional forces responsible for deforestation, our model treats settlement in tree-covered hinterlands and land clearing as two separate activities.

(a) Agricultural colonization

Because frontier tenure regimes impair the ability of individuals and firms to collect forestry and other nonagricultural income, settlement in tree-covered hinterlands is stimulated by the prospect of capturing agricultural rents. That is, there is a direct and positive relationship between demographic pressure at any particular site and agricultural (as opposed to nonagricultural) rents at the same site.

Agricultural rents and, hence, settlement are

an increasing function of three factors: the scale of nearby urban development, soil productivity, and access to outside markets. The presence of cities and towns enhances agricultural income by strengthening local demand for crops and lives- tock. Soils underlying tropical forests are usually infertile and often erodible (Nicholaides et al., 1983), however, which tends to impede agricul- tural colonization. Finally, improved access is critical for agricultural development in develop- ing country hinterlands. For example, Bromley (1981) points out that primary forests covered most of northwestern and northeastern Ecuador before all-weather highways were constructed in those two regions.

These influences on agricultural rents are expressed by the three right-hand side terms of the first regression in our model, which explains the extent of agricultural colonization in the 20 cantons of the four provinces in eastern Ecuador:

A G P O P = bl0 + bit URBPOP + b12 SOILS + bj3 ROADS, (I)

where: A G P O P = the canton's agricultural labor

force, URBPOP = the canton's urban population, SOILS = hectares in the canton covered

with soils that do not have major limitations for crop production, and

ROADS = kilometers of all-weather roads in the canton.

A G P O P is an indicator of rural population pressure. As indicated above, the size of the canton's urban population, URBPOP, is a satis- factory proxy for local demand for agricultural products. Discussed in the next section are the criteria used to measure the variable SOILS. The extent of all-weather roads in a canton, ROADS, indicates the canton's accessibility. Regression coefficients bll through bL~ are" all expected to be positive.

Some explanation of the exogeneity of the three factors influencing agricultural rents is in order. Urban expansion in the Amazon basin tends to be weakly related to agricultural produc- tion in the surrounding countryside. The growth of northeastern Ecuador's towns, for example, has been closely linked to petroleum industry development. At an extreme, the large size of Manaus, Brazil is explained by its status as a duty-free zone, which has greatly stimulated industrial development. Given this characteristic of urban development in the region, it is approp- riate to treat URBPOP as exogenous in equation (1).

1148 WORLI) DEVEL()PMENI'

The second right-hand side variable in equa- tion (1) is independent because the natural capacity of land to support crop or livestock product ion is beyond human control.

The history of the A m a z o n basin generally and eastern Ecuador specifically demonst ra tes that the penet ra t ion of highways into forested hinter- lands usually leads to agricultural colonizat ion, not vice versa. The governments of Brazil and o ther countries have constructed roads more to establish "human frontiers, '" which safeguard terri torial claims, than because they are con- vinced of the Amazon basin's great agricultural potential (Landau, 1980). Set t lement of north- eastern Ecuador began in earnest only after infrastructure had improved, following the dis- covery of pe t ro leum in the region (Bromlcy, 1981). Brazilian and Ecuador ian exper ience con- firms that the wlriable R O A D S can be t reated as exogenous in this analysis.

(b) L a n d c learing

The second equat ion of tile regression model reflects the institutional environnlent facing set- tiers in the A m a z o n basin. That is, it relates tile extent of land clearing in a canton, D E F O R , to demographic pressure (specifically, the first equat ion ' s dependen t wlriable) as well as to a non-negat ive index of relative tenure security, T E N S E C :

D E F O R = b > + bel A G P O P + h~.2 T E N S E ( ' . (2)

A G P O P ' s coefficient, bel, is expected to be positive. T E N S E C is constructed so as to indicatc differences in formal tenure security among cantons:

T E N S E C - adjudicated % of a canton 's agricultural land

adjudicated % of agricultural hind in study area (3)

If the wtlue of T E N S E C for any particular canton were greater than 1.00, then that canton 's far- mers would be more likely than others in eastern Ecuador to have lk)rmal tenure. Conversely, a wdue less than 1.00 indicates that a canton 's farmers are less likely to have formal tenure. T E N S E C ' s coefficient bee, is expected to bc negative because enhancing formal tenure secur- ity reduces the need for continuous exercise of informal agricultural use rights. Because the pace of land adjudicat ion in different parts of thc country is the prerogat ive of the national govern- merit, T E N S E C is an exogenous variable.

Arguably, tenure security could influence far-

mers" Iocational decisions, t towever , the correla- tion between A G P O P and T E N S E C is only 0. 133. Accordingly, it is reasonable to conclude that formal tenure security influences colonists ' resource deve lopment decisions (i.e., choices about hind clearing) once they have settled m a particular area. Cohmizat ion itself is driven by forces expressed in equat ion (1).

4. D A T A

The first equat ion 's dependent wmable , A G P O P , and one of its independent variables, U R B P O P , were obtained from a 100% sample of individual returns collected in eastern Ecuador during the country 's most recent census, con- ducted by the National Institute for Statistics and the Census ( INEC) in 1982. The agricultural labor force was found to be highly correlated with other measures of set t lement in the country- side (e.g. rural populat ion or rural employment) , Similarly, urban populat ion is highly collinear with other proxies for local demand for agricul- tural conlmodit ies (e.g., nonagricultural work force).

Thc variable SOILS, which equals thc area in each canton with "good" agricultural potential , was assessed through interpretat ion of soil maps prepared by the National Program for Agrarian Regionalizat ion ( P R O N A R E G ) . Recognizing the principal envi ronmenta l constraints on crop and livestock production in eastern Ecuador , wc classified the region's soils according to drainage and fertility. Our procedurc yielded results very similar to those obtained by the country 's Natio- nal Developnlcnt ( 'ouncil (CONADI: . ) and the Unitcd Nations Envi ronmenta l Program ( t IN- EP), which also relied on P R O N A R E G maps. Whereas C O N A D E - U N E P (1986) found that 5.69% of eastern Ecuador is suitable for agricul- ture, we de te rmmed that 5.04% of the same region was free of serious drainage or fertility problems.

Road maps prepared by INE(" as well as an inventory conlpleted by thc Ministry of Public Words (MOP, 1987) were used to detcrnl ine how nlany kihmletcrs of all-weather roads existed in each canton in 1982.

Tile National Institute for tile Colonizathm ol tile Ecuadorian Anlazon ( I N C R A E ) and the Center for Integrated hwentory of Natural Re- sources ( ( ' I+IRSEN) have used aerial photo- graphs and satellite inlages to assess land use change in different parts of eastern Ecuador between 1977 and 1985 ( I N C R A E - ( ' I A R S E N , 1987). The results of that study allowed us to calcuhttc the extoll! of tropical deforestat ion in

CAUSES OF TROPICAL DEFORESTATION 1149

the early 1980s (i.e., the second equation's dependent variable) in 11 of eastern Ecuador's 20 cantons. No such estimates are available for the other nine cantons.

Finally, records showing cumulative land ad- judication in eastern Ecuador were collected from IERAC. Those records, along with IN- CRAE land use data, were used to determine the index, TENSEC, for each canton.

Correlations among each regression equation's right-hand variables are, in general, quite low. The major exception is the 0.636 coefficient for URBPOP and ROADS, which is understandable since urban settlement is rare where all-weather roads have not yet penetrated. Collinearity between the two variables impedes reliable estimation of their respective standard errors. As indicated above, the two right-hand variables in equation (2), which explains land clearing, are not collinear.

5. REGRESSION RESULTS

Equations (1) and (2) comprise a recursive system of regressions. Koutsoyiannis (1977, pp. 340-342) points out that ordinary least squares (OLS) estimates of such a system's coefficients are efficient.

OLS estimation of equation (1), which ex- plains agricultural settlement, yields the follow- ing results (standard deviations are shown in parentheses):

AGPOP = 469.144 + 0.263 URBPOP (251.334) (0.061)

+ 0.008 SOILS + 3.350 ROADS (0.002) (2.527)

F = 26.856 adj R 2 = 0.803 n = 20.

In spite of limited degrees of freedom, applying the F-test with a 99% confidence interval leads to rejection of the hypothesis that all regression coefficients equal zero. This test result is consis- tent with the hypothesis underpinning this equa- tion, that the prospect of capturing agricultural rents stimulates colonization of forested hinter- lands. At the same time, application of a t-test with a 99% confidence interval leads to rejection of the null hypothesis for the coefficients of the first two right-hand variables. The somewhat lower t-statistic for the coefficient of ROADS is possibly explained by multicollinearity between that variable and URBPOP. The signs of the right-hand coefficients are all positive, as ex- pected.

OLS regression of the second equation, which

explains land clearing, yields the following re- suits:

DEFOR = 16,111.728 + 51.073 AGPOP (23,977.621) (6.241)

- 26,989.394 TENSEC (12,321.915)

F = 36.097 adj R z = 0.875 n = 11.

Applying the F-test with a 99% confidence interval leads to rejection of the hypothesis that the right-hand coefficients of equation (2) all equal zero. AGPOP's and TENSEC's coeffi- cients are positive and negative, respectively, as expected. AGPOP's coefficient is significantly different from zero when a 99% confidence interval is used. If a confidence interval of 94% or less is used, then TENSEC's coefficient is also significantly different from zero.

The preceding estimates of regression coeffi- cients are efficient. Had a larger sample been available, then confidence intervals around true coefficient values would have been smaller. In particular, more conclusive testing of the significance of TENSEC's coefficient would have been possible. Also, the problem of multicolli- nearity is frequently addressed by using a larger sample. Consequently, high covariance between URBPOP and ROADS, which might explain acceptance of the null hypothesis for the latter's coefficient, might be symptomatic of the small sample used in this research.

In spite of the small data set available to us, the preceding results of regression analysis provide conclusive evidence about the causes of tropical deforestation.

6. DISCUSSION

To date, case studies have constituted the principal point of empirical reference for discus- sions of the causes of tropical deforestation. That literature provides useful insights into the market forces and public policies that induce the clearing of tree-covered land near the equator. However, a series of case studies alone does not constitute a satisfactory basis for general statements about linkages among land clearing, tenure security, rural population pressure, and the factors in- fluencing agricultural rents.

The research reported in this paper constitutes a methodological departure from previous analy- ses of the causes of tropical deforestation. We have shown that statistical analysis can indeed clarify the human context of land clearing and resource management across broad stretches of an agricultural frontier. Our results bolster the

1 1511 WORLD DEVELOPMENT

argument that se t t lement of forested hinterlands is driven by the prospect of collecting agricultural rents. In addition, it suggests that deforestat ion is a consequence of demographic pressure. Defor- estation also occurs as colonists a t tempt to safe- guard their tenuous legal hold on frontier lands.

These results, obtained with a small sample, should encourage parallel research with the larger data sets that can be assembled in o ther countries. Also, as remote sensing technology advances, it will be possible to under take this type of analysis with samples comprising smaller geographic units.

In and of itself, however , our analysis yields insights for policy making. Developing country governments can do two things to resolve the institutional crises contributing to tropical de- forestation. First, they can streamline tenure adjudication procedures. If this is done, people living in t ree-covered hinterlands will no longer be obliged to deplete resources just to safeguard weak proper ty rights. Second, land clearing should no longer be a prerequisi te for formal tenure. Where land clearing is a prerequisi te , no one has a clear right to the income associated

with forest conservation. Consequent ly , every- one is obliged to respond immediately to any opportuni ty to capture the rents generated by land use conversion (Southgate, 199(/).

Location theory informs us that where there are nonat tenuated property rights agriculture's extensive margin is flmnd where per-hectare agricultural rents equal per-hectare rents associ- ated with forestry or other alternative land uses (Southgate, 1990). Unfor tunate ly , property rights at tenuation is the norm in Latin America. Tenure insecurity is a problem in all countries with extensive tropical forests. In these coun- tries, property rights are a reward for land clearing, as Mahar (1989) documents for the case of Brazil. Under these circumstances, the rela- tionship among agriculture's extensive margin, property rights, and rents associated with differ- ent land uses cannot be determined. While this paper 's analysis furnishes no specific insights into the likely impacts of institutional reform in Ecuador or any other country, it nevertheless bolsters the argument that institutional crisis contributes significantly to deforestat ion in tropi- cal America .

REFERENCES

Bromley, D., and M. Cernea, "The management of common property natural resources: Some concep- tual and operational fallacies," Discussion Paper No. 57 (Washington, DC: World Bank, 1989).

Bromley, R., "The colonization of humid tropical areas in Ecuador," Singapore Journal o f Tropical Geo- graphy, Vol. 2, No. 1 (19811, pp. 15-26.

Brown, L. A., R. Sierra, and D. Southgate, "'Com- plementary perspectives as a means of understanding regional change: frontier senlemcnt in the Ecuador Amazon ," Environment and Plannin,~ A , Vul. 24 (1992).

Cabarle, B., M. Crespi, C. Dodson, C. Luzuriaga, D. Rose, and J. Shores, "An assessment of biok)gical diversity and tropical forests for Ecuador," Draft report fl)r USAID/Quito (Washington, DC: World Resources Institute, 1989).

CONADE-UNEP (National Development Council and United Nations Environmental Program), Programa de Manejo Ambiental para el Desarrollo lntegrado de h~ Regi6n Amaz6nica Ecuatoriana: lnforme Final (Quito: CONADE-UNEP, 1986).

DINAF (National Forestry Directorate), Plait de ActiOn Forestal para el Ecuador: Diagn6stico del Sector Forestal, draft (Quito: Ministry of Agriculture and Livestock, 1988).

Hardin, G., "The tragedy of the commons," Science, No. 168 (1968), pp. 1243-1248.

Holdridge, L., LiJO Zone Ecology (San Jose, Costa Rica: Tropical Science Center, 1967).

INCRAE-CLIRSEN (National Institute for the Col- onization of the Ecuadorian Amazon and the Center

lk~r lntcgratcd Inventory of National Resources), Amilisis de los Cambios en el Uso de la Yierra en el Frente de Cohmizaci6n de la RegiOn Amazonica Ecuatoriana (Quito: INCRAE-CLIRSEN, 1987).

INK(? (National Institute for Statistics and the Census), IV Censo National de Poblaci6n (Quito: INEC, 19S5).

Koutsoyiannis, A., Theory o f Econonwtrics, 2nd ed. (London: Macmillan E~tucation Limited, 1977).

Landau, G., "'The treaty for Amazonian cooperation: A bold new instrument for development," Georgia Journal o f International and ( 'omparative Law, Vol. 10, No. 3 (1980), pp. 463-489.

Lisansky, J., Migrants to Amazonia: 3;potttatleott,s Colonization in the Brazilian Frontier (Boulder: Westview Press, 1989).

Macdonald, T., "Indigenous responses tu an expanding frontier: Jungle Quichua economic conversion to cattle ranching," in N. Whitten (Ed.), Cuhural Tran~forntations atrd Ethnici O' in Modern Ecuador (Urbana: University of Illinois Press, 1981).

Mahar, D., "Government policies and delk~restation in Brazil's Amazon region," Mimeo (Washington, DC: World Bank, 1989).

Ministry of Agriculture and Livestock (MAG), lnfi)rnw F)'nal de la Limitaci6n del Patrimonio Forestal del Estado (Quito: Ministry of Agriculture and Lives- tock, 1987).

Ministry of Public Works (MOP), lnventario de la Red Vial Nacional (Quito: Ministry of Public Works, 19871.

Moran, E., "'Government-directed settlement in the

CAUSES OF TROPICAL DEFORESTATION ll51

197(1s: An assessment of Transamazon highway colonization," in E. Moran (Ed.), The Dilemma of Amazonian Development (Boulder: Westview Press, 1983).

Nicholaides, J., P. Sanchez, D. Bandy, J. Villachica, A. Coutu, and C. Valverde, "Crop production systems in thc Amazon basin," in E. Moran (Ed.), The Dilemma of Amazonian Development (Boulder: Westview Press, 1983).

Rudel, T., "'Roads, speculators, and colonization in the Ecuadorian Amazon," Human Ecology, Vol. 11, No. 4 (1983), pp. 385-403.

Seligson, M., "'Land tenure security, minifundization, and agrarian development in Ecuador," Mimeo (Quito: USAID, 1984).

Southgate, D., "'The causes of land degradation along expanding agricultural frontiers in the Third World,'" Land Economics, Vol. 66, No. 1 (1990), pp. 93-101.