Tenure diversity and dependent causation in the effects of regional integration on land use: evaluating the evolutionary theory of land rights in Acre, Brazil

ABSTRACT

In the complex causation behind land change, dependent causation can play a central role. A case in point concerns land tenure diversity, where contrasting use rules for different lands may affect the impacts of other drivers on land use outcomes. We therefore evaluate the evolutionary theory of land rights (ETLR), which assumes homogeneous private property rights, in order to test for dependent causation due to distinct use rules among various types of private lands. In the present analysis, we focus on whether land tenure type modifies the effects of highway infrastructure on key outcomes highlighted in the ETLR framework. We take up the case of rural settlements along the Inter-Oceanic Highway in the eastern part of the Brazilian state of Acre, where there is considerable land tenure diversity. Findings from multivariate models for land titling, the castanha nut harvest, and cattle pasture all indicate that the effects of infrastructure depend on land tenure type. These results confirm the importance of dependent causation behind land use and bear implications for theory on land change, infrastructure impacts, and land system science.

KEYWORDS:

• Land use
• land tenure
• dependent causation
• transport infrastructure
• Brazil
• Amazon

Introduction

Among the various manifestations of institutional diversity (Ostrom, 2005), land tenure constitutes a textbook example. In recent years, various models of land tenure with particular sets of rules have been proposed and implemented in Latin America (e.g. Richards, 1997; Zoomers & Van Der Haar, 2000). Many such models arise from societal demand and governmental recognition that local conditions require modifications in the hope of better fitting local practices for land access and use. The result is an increasingly complex panoply of specific types of parks, reserves, concessions, and communal and private lands, which reflect in varying degrees of local institutional structures, stakeholder groups, and natural resources.

The Amazon basin offers a case in point, where governments pursued frontier expansion and colonization with standardized settlement models (e.g. Moran, 1981; Schmink & Wood, 1984). The result was a history of problems in sustaining colonies due to logistical difficulties and other hindrances to sustainable resource management. As a consequence, proposals from community associations, local government agencies, and social movement leaders proliferated (e.g. Castellanet & Jordan, 2002; Hall, 1997). In many cases, proposals sought to modify tenure rights to recognize local histories and stakeholder priorities. This resulted in complex political negotiations and shifts in relations between local peoples and government, manifest in new tenure categories for lands with distinct rules for occupation and use.

A key issue of tenure diversity is whether distinct rules of access and use actually lead to differences in the behavior of landholders as intended. This becomes especially important when different lands in a given region are affected by the same drivers of change, such as public policies, population growth, or shifts in market prices. In particular, ‘dependent causation’ may occur, as when a given driver affects landholder behaviors differently depending on the tenure rules to which they are subject. That is, under conditions of tenure diversity, differences in land tenure rules may lead to differences in land use outcomes even under conditions of the same shared external driver. Indeed, whether landholder behavior differs in the presence of a given driver becomes a crucial test of the robustness of distinct tenure models with differing use rules.

There has been considerable attention to the various driving forces behind land use, both in terms of multiple causation among different kinds of drivers and in terms of multi-step causation defined by the causal proximity of drivers to land use decisions (e.g. Lambin & Geist, 2006; Wood & Porro, 2002). More recently, we have detailed discussions of the nature of causal mechanisms and causal effects behind land use (Meyfroit, 2016). Such treatments however rarely pay attention to the issue of dependent causation, the circumstance where the effect of one driver is modified by another in terms of land use. Given the importance of land tenure for land use, tenure diversity offers an important case for examination of dependent causation as tenure may modify the effects of other drivers on land use.

This paper therefore focuses on the case of private property, where such properties fall in lands with distinct tenure rules. We draw on the evolutionary theory of land rights (ETLR), which makes specific predictions about landholder behavior under conditions of private property (cf. Feder & Feeny, 1991; Feder, Onchan, Chalamwong, & Hongladarom, 1988). The ETLR assumes that use rules in private property are homogeneous. Tenure diversity thus presents a complication to the ETLR account and raises the possibility of dependent causation as an alternative to the homogeneity assumption. An evaluation of landholder responses to shared drivers but under conditions of differing use rules due to tenure diversity thus offers an evaluation of the ETLR’s homogeneity assumption in the face of possible dependent causation. Observation of distinct landholder behavior on lands with differing tenure rules would indicate the efficacy of contrasting tenure models in the presence of shared drivers and require a theoretical emendation of the ETLR via recognition of dependent causation.

We take up the case of eastern Acre, a state in the western Brazilian Amazon. Acre experienced rural land settlement in the 1970s and 1980s, which led to violent conflicts, social mobilization, and diversification of land tenure (Governo do Acre 2006; Rêgo, 2002; Tocantins, 1979). More recently, Acre incurred large-scale infrastructure investments and urbanization, both of which are hallmarks of regional development (Perz et al., 2012; Schmink & Cordeiro, 2008). Acre thus offers a useful study case to evaluate the impacts of regional development on the behavior of landholders with distinct kinds of lands. Following the ETLR, drivers like highway paving should yield specific responses in landholder decisions; but with tenure diversity, those responses may also reflect dependent causation by varying among lands with different tenure rules.

This paper provides a systematic analysis of several key arguments of the ETLR. We focus on three key questions, each of which centers on a key relationship in the ETLR: (1) the influence of regional integration on land tenure formalization, (2) the effects of integration and formalization on market participation, particularly access to credit, and (3) the importance of integration, formalization, and market participation on land use. In each step, we compare the effects of integration in terms of market access among land tenure types which fall within the general category of private individual tenure presumed by the ETLR, but which have distinct use rules. The findings largely confirm the ETLR’s expected effects of integration and other priors on the outcomes, but many findings also document dependent causation in the form of statistically significant differences in those effects among landholders with different use rules. We conclude with a discussion of the implications of dependent causation for the ETLR and other theories of land change, the manifold impacts of infrastructure, and the dynamics of telecouplings in land system science.

Background

The ETLR

The ETLR emerged as a response to Hardin’s (1968) problem of the ‘tragedy of the commons.’ Conditions of open access to natural resources lead to resource overexploitation and a crash in user well-being. The ‘property rights’ school thus arose to posit that resource scarcity makes it worthwhile to formalize private property rights (e.g. Alchian & Demsetz, 1973). With private rights, landowners have secure control over their holdings, permitting investments to make resources more productive over time, particularly in the agricultural sector. Private property rights were thus asserted to be vital to sustainable development in agriculture.

The property rights argument stimulated theoretical development on land titling, held to be the key form of property formalization and the means of attaining tenure security (Feder et al., 1988). Tenure security is a key precondition for realizing the benefits of private property rights. In particular, in many countries, land titles are a concrete form of substantiating property rights as collateral for obtaining bank credit in order to make investments in resource productivity.

Research on tenure security in turn led to work on the ETLR in the context of economic development (Deininger & Feder, 2001; Feder & Feeny, 1991). The ETLR thus became a framework with a series of relationships running from regional development to sustainable land use via several causal steps. First, exogenous drivers of development, such as regional integration via new infrastructure and population growth, stimulate rising land values. Where property rights are not clearly stipulated, rising land values prompt formalization of property rights for tenure security, as via titling of demarcated land parcels. Second, formalization prompts the emergence of land markets, since titled lands become fungible and land sales and rentals and other such transactions become more common. Meanwhile, formal sources of credit emerge, and formalized land claims become important for access to credit. At the same time, the State finds it increasingly worthwhile to invest in land titling and enforcement of property rights to reduce agrarian conflicts and thereby ensure tenure security. Third, tenure formalization, emerging land markets, and formal credit motivate landholders to make greater investments in long-term resource productivity. Such investments not only contribute to increased agricultural land use but also to sustainable regional development. A longer discussion of these causal linkages with figures is available in Perz et al. (2016).

In sum, regional integration prompts tenure formalization; integration and formalization catalyze the emergence of land markets and credit and the reduction of land conflicts; and finally, these factors combine to motivate investments in expanded resource use that is sustainable. The property rights school, ETLR, and formalization arguments have supportive empirical research for agriculture in various countries (e.g. Feder et al., 1988; Otsuka, Suyanto, Sonobe, & Tomich, 2001) including in the Brazilian Amazon (Alston, Libecap, & Mueller, 1999; Sills & Caviglia-Harris, 2008).

While the ETLR has driven empirical research and frequently found support, it has also led to various criticisms (e.g. Platteau, 1996; Sjaastad & Cousins, 2008). There has been dispute about several aspects of the ETLR: (1) whether integration leads to tenure formalization as via titling, (2) whether titling begets tenure security, (3) whether credit is necessarily available, (4) whether credit begets investments in improved resource productivity, (5) whether formalization fosters land marketability, and (6) whether resource use among titled properties is ecologically sustainable. A summary of these criticisms can be found in Perz, Barnes, Shenkin, Rojas, and Vaca (2014).

In the present paper, we raise the additional theoretical issue that the ETLR presumes that use rules are homogeneous among all private lands. Where use rules differ, it is likely that landholders may vary in their decisions, including with regard to land use. Under such conditions, the effects of regional integration on a given area may differ on adjacent land parcels with differing use rules. This amounts to a case of dependent causation, where the effect of regional integration on various outcomes depends on land tenure type. As a result, landholders may not respond the same way in terms of seeking formalization, procuring credit, engaging in market transactions, and land use.

Study case: Acre, Brazil

To evaluate the importance of tenure diversity for the relationships posited by the ETLR from regional integration to land use, we take up the case of the Brazilian state of Acre. Acre was a key locus of forest extractivism beginning in the mid-19th century during the rubber boom (Santos, 1980; Tocantins, 1979; Weinstein, 1983). Rubber barons attracted impoverished people from the Brazilian northeast with promises of prosperity by rubber tapping in the western Amazon. Upon arrival, families found themselves indebted to the barons for the costs of transport and thus bound to a form of indentured servitude tapping rubber in order to work off the debts. This system was highly profitable for rubber barons until the rubber boom went bust in the early 20th century.

Beginning in the 1970s, the Government of Brazil opened new highways connecting consolidated portions of the country with new frontiers in the Amazon (Moran, 1981; Schmink & Wood, 1984). This effort at regional integration was successful in Acre insofar as colonists migrated to the state. At the same time, the Government of Acre sought to attract investment to the state (Bakx, 1988; Silva, 1990). Investors bought up rubber estates for cattle ranches and sent in work teams to clear forest for pasture. However, work teams encountered rubber tapper communities still living in the forest. This prompted conflicts over land claims and land use in the 1980s (Paula & Silva, 2006; Sobrinho, 1992). Such conflicts turned violent and culminated in the murder of rubber tapper leader Chico Mendes in 1988, a year of record deforestation in the Amazon.

The association of deforestation and human rights violations gave rubber tappers a political opening to demand official recognition of their land claims and forest-based resource management. This spurred development of the ‘extractive reserve’ tenure model (Allegretti, 1990; Schwartzman, 1992), which stood in contrast to the agricultural colonization model by highlighting forest extractivism. Extractive reserves were managed under the aegis of the federal environmental agency IBAMA (now the Chico Mendes Institute). In eastern Acre, the Chico Mendes Extractive Reserve was officially recognized in 1990.

In addition, private properties were established along the roadsides. Beyond large private ranches established by investors, agricultural communities arose. In the 1990s, colonists promulgated various proposals for clarifications and modifications in their land tenure rules (Governo do Estado do Acre, 2006). As a result, the federal land agency, INCRA, began to delineate various types of lands with distinct tenure rules. Aside from the agricultural projects (Projetos de Assentamento, PA), INCRA designated agro-extractive projects (Projetos de Assentamento Agro-extractivista, PAE) in part as a response to IBAMA’s extractive reserve model. INCRA also designated sustainable development projects (Projetos de Desenvolvimento Sustentavel), agroforestry poles (Polos Agroforestais, PAFs), and other lands. Each INCRA model involved private individual tenure and featured agricultural land use, but each instituted variations in use rules. Whereas PAs generally allotted each owner 50–100 ha and permitted 20% of the land to be deforested, PAEs provided each family with 100–500 ha and only permitted 10% deforestation. This was by design: each tenure type had distinct purposes with regard to the intended resource use. Whereas PAs were designed for small-scale agriculture, PAEs were intended for forest extractivism with some agriculture (Governo do Estado do Acre, 2006).

In 1998, the regional elections brought leaders of social movements including rubber tappers into power (Kainer, Schmink, Leite, & Fadell, 2003). This consecrated the ‘Forest Government’ of Acre, which promulgated an ambitious array of new policies to support sustainable resource use. A key priority of the Forest Government was an economic-ecological zoning plan (Governo do Estado do Acre, 2006). The zoning plan drew on maps of soils, climate, vegetation, settlement, and other factors. The final product, a zoning map of land tenure types, was nothing less than a rainbow mosaic due to the numerous different types of lands designated in the state (Governo do Estado do Acre, 2006).

Table 1 provides a summary of some of the lands recognized in the zoning plan. We differentiate among broad categories defined by the key social actors who manage them and make them more refined distinctions within those categories based on differences in use rules. We note the various types of protected areas (national parks, ecological stations, etc.), sustainable use areas (state forests, extractive reserves, etc.), and private lands (agricultural projects, etc.). The lands included in Table 1 do not include private lands outside of settlement projects, such as cattle ranches, or undemarcated lands. Nonetheless, tenure diversity is abundantly evident.

Table 1. Land tenure diversity in Acre, Brazil, 2006.

Land tenure category^a	Number of areas	Land area (Ha)
Protected areas (state management)
National parks	1	784,079
Ecological stations	1	84,387
State parks	1	695,303
Total	3	1,563,769
Sustainable use areas (concession areas)
Environmental protection areas	3	36,137
Areas of ecological interest	1	2576
Extractive reserves	5	2,673,692
National forests	7	831,662
Total	16	3,544,067
Indigenous Territories (Tribal Communal Areas)	30	2,390,112
Settlement Projects (Private Individual Lands)^b
Agricultural (colonization) projects	72	1,430,557
Agro-extractive projects	11	286,395
Sustainable development projects	6	148,670
Agroforestry projects	3	86,183
Agroforestry poles	12	3412
Peri-urban projects (Projetos Casulos)	3	651
Total	107	1,955,870
Grand total^c	156	9,453,818

^aSource: Governo do Acre (2006: 101–119).

^bAll values refer to official values. Some have incurred adjustments.

^cExcludes other private lands such as cattle ranches and lands still being demarcated or designated in a tenure category.

While the Forest Government was promulgating sustainable development policies and consolidating the ecological-economic zoning plan, other changes were afoot. Arguably the most notable was the conclusion of paving the BR-317 highway through eastern Acre. This effort advanced under a national integration strategy featuring large-scale infrastructure projects. The final sections of the BR-317 were paved in Acre to the Peruvian border by the end of 2002.

BR-317 in turn became linked to an even larger program, the Initiative for Integration of Regional Infrastructure in South America (IIRSA) (CEPEI 2002; IIRSA 2008). IIRSA was constituted by a meeting of presidents from 12 South American countries in 2000. IIRSA was conceptualized as an explicitly spatial development strategy organized around ‘axes of integration’ targeted for infrastructure investments. Such axes constitute strategic growth corridors for international commerce.

Under IIRSA, BR-317 was considered part of the Inter-Oceanic Highway (IOH). Under the first phase of IIRSA, paving of the IOH occurred across the border in the Peruvian region of Madre de Dios from 2006 to 2010. Hence, the BR-317/IOH in Acre was in turn linked to other segments of the same corridor which stretched all the way to Atlantic ports in southern Brazil and Pacific ports in Peru. The road connections provided the physical basis for telecouplings of distant causal drivers to local land use in Acre (cf. Boillat et al., 2017).

Eastern Acre thus became a theater for the proliferation of land tenure models in the 1990s and a key segment in a road corridor of strategic importance for international economic integration in the 2000s. The paving of the BR-317/IOH in a region of tenure diversity thus raises questions about how highway paving impacted landholders with different types of lands. We can thus examine whether the impacts of highway paving as a key form of regional integration are dependent on tenure model.

Methods, data, and analysis

To test whether the causation from regional integration to formalization and its consequences is dependent on land tenure, we drew on information from surveys of land tenure units and households conducted in 2008 and 2009 along the IOH in eastern Acre. Faculty and students at the University of Florida (UF) coordinated the data collection effort with counterparts at the Federal University of Acre (UFAC). Fieldwork proceeded in two phases. The first phase involved a survey of 93 leaders of 25 distinct land tenure units in Acre in 2008 (Perz et al., 2012). These lands included PAs, PAEs, and other tenure categories. We selected land tenure units geographically, sampling at more or less regular distance intervals along the entire length of the IOH in Acre. Consequently, our sample includes land tenure units with a broad range of characteristics not only in terms of tenure rules but also distances to key markets.

For the second phase of the data collection, we revisited a representative subset of land tenure units in order to conduct more detailed household interviews. In 2009, personnel from UF and UFAC visited several distinct land tenure units along the IOH in Acre and conducted interviews with 266 households. We sampled communities for revisits geographically, based on their location and their land tenure type. For the revisits, we included three PAs, and two PAEs, and two PAFs. In each category, we revisited cases at different distances from the Acre state capital city of Rio Branco, in order to ensure variation in market proximity within each tenure type. Figure 1 shows the locations of the land tenure units and households we revisited for household interviews in Acre. Within each land tenure unit, we pursued random sampling of households. We worked from government population estimates of land tenure units (Governo do Estado do Acre, 2006) to establish the universe for sampling. We used larger sampling ratios in smaller land tenure units to obtain useful sample sizes. Within each tenure unit, we then used cadastral maps of roads and property boundaries to systematically sample households.

Figure 1. Map of households and communities visited in lands with distinct tenure rules along the Inter-Oceanic Highway in Eastern Acre, Brazil, 2008–2009.

For the interviews, we developed a questionnaire based in part on other instruments applied previously in the study region. We ensured that the questionnaire was culturally appropriate in local argot and institutionally relevant for the land tenure regime in Acre. The questionnaire was translated and reviewed by Brazilian collaborators in Acre prior to application. The questionnaire included sections with questions on concepts related to regional integration, land tenure, market participation, and resource management. Sections included items on household location (including GPS coordinates and information on distance and travel times along roads to reach markets), migration history, family labor, household assets, social capital (via absent kin working in nearby towns and organizational memberships), capital inputs (including formal sources of credit), resource tenure (including tenure type, titling status, land sales and rentals, and access to other lands via third parties), health and well-being, and future plans. The section on location provided information to calculate travel times to nearest markets and the Acre state capital city of Rio Branco, both as measures of regional integration. The sections on migration, labor, and various capitals permit measurement of household productive capabilities. The questions on land tenure allow evaluation of tenure type, titling status, and related aspects of formalization, including the incidence of land conflicts. In addition, the questionnaire featured items on agricultural activities (annual and perennial crops, livestock and related products such as milk) and forest extractive activities, which permit analysis of natural resource management, as well as off-farm income sources (such as wage jobs, retirement pensions, and others).

Methods of analysis

We drew on these data to pursue multivariate statistical models of households in order to evaluate specific tenets of the ETLR. We begin by focusing on determinants of formalization. Analyses of formalization in the context of ETLR focus on land titling and presume private individual rights. For the case of settlement projects in Acre, this is appropriate since all such projects (PAs and PAEs) involve private individual tenure. However, many households had not yet received titles for various reasons. We therefore differentiate among degrees of tenure formalization by distinguishing households by titling status, among landowners with a definitive title, those with a preliminary title or title application, and those without any tenure document.

The first part of the multivariate analysis therefore evaluates the importance of various determinants of tenure formality. We focus on factors highlighted by the ETLR as affecting formalization, notably regional integration and population growth. We highlight highway paving as a catalyst of regional integration. Key to infrastructure in Acre has been the paving of the IOH, which reduces travel times to markets. Eastern Acre has several towns which serve as market nodes in distribution networks; so, we consider travel time to the nearest market as well as travel time to Rio Branco. To calculate travel times, we used GPS points for household and market locations along with GIS coverages of road maps, combined with estimates of travel speeds on paved highways (90 km/h), unpaved highways (60 km/h), and unpaved secondary roads (30 km/h). The GPS points and road coverages permitted measurement of distances along roads from houses to markets; the estimated travel speeds by types of road permitted calculation of travel times along routes from landholdings to towns. Travel times thus account for both market distance as well as paving status. For population growth, we observe land tenure unit net migration in the 5 years prior to the land tenure unit survey. If the ETLR applies to Acre, we anticipate that households with shorter travel times and in land tenure units with faster population growth will exhibit more formal tenure. However, we also consider whether the effect of regional integration is itself dependent on land tenure type. We therefore include interactions between travel time with land tenure type to see if shorter travel times spur formalization more in one tenure type than another.

While we feature the integration and population variables, we also control for the effects of other factors. We draw on Bebbington’s (1999) capitals and capabilities framework, which highlights the importance of land, labor and various capitals (cultural, human, social) as assets which influence rural livelihoods. For land, we consider hectares claimed; for labor, we account for household composition (labor availability via the number of adults, and dependency via the number of children and elderly); for cultural capital, we include region of birth (Brazil’s ‘north’ region, which covers the traditional Amazon, as differentiated from the rest of the country, to reflect distinct regional identities) and time on the property (in years); for human capital, we account for place of birth (rural or urban) and respondent education (years of school completed); and for social capital, we consider absent family members (who reside in towns) and the number of types of organizational memberships. More details on the construction of these control variables are available elsewhere (Perz et al., 2014, 2016). Per the ETLR, households with greater assets will invest more in formalization in order to secure their assets and thereby make them more productive.

In the second part of the analysis, we focus on the consequences of formalization. Following the ETLR, formalization begets tenure security; while that is difficult to observe directly, the ETLR asserts that titling should beget tenure security and result in more market engagement as via access to bank credit, as well as via land transactions and fewer land conflicts. We evaluate formal credit because it is central to the ETLR account as a means to increase the productivity of land use. We also evaluated land market transactions including land rentals and plans for land sales. While renting land from others suggests the expansion of productive activity, renting to others and plans to sell instead imply transitions from on-farm production for income, whether to off-farm activities or retirement. Finally, we analyzed land conflicts as tenure security should result in fewer conflicts according to the ETLR. We feature the effects of integration via infrastructure, population growth and the formalization variables, net of the effects of the control variables. In particular, we highlight interactions between the integration variables and land tenure type, to see if the effects of travel time on market interactions differ among households with different land tenure rules.

In the third and final step of the analysis, we evaluated the ramifications of integration, formalization, and market-based transactions for land use. We consider three indicators of land use with regard to sustainability in the Amazonian context: castanha nut harvesting, cattle pasture, and tree crops. Each requires distinct investments and has very different environmental ramifications. Castanha (‘Brazil nut’) is an important non-timber forest product (NTFP) in the southwestern Amazon (e.g. Cronkleton & Pacheco, 2010; Wadt, Kainer, Staudhammer, & Serrano, 2008). Castanha harvesting requires fewer capital investments but permits retention of standing forest and thus affords sustainability in terms of conserving forest cover. By contrast, cattle pasture requires investments for clearing extensive tracts of forest and pasture maintenance against weeds and soil degradation. Ranching has frequently been vilified for causing forest destruction and land degradation in the Amazon, though practices are changing (e.g. Walker et al., 2009). Tree crops constitute long-term investments and provide vegetation cover that secures soil against erosion. Given these contrasts, one would expect that integration, formalization, and its consequences would better account for pasture and tree crops than castanha harvesting. That said, the effects of integration on resource management are also likely to vary among tenure types since tenure rules are tied to land use. Hence, as in previous steps in the analysis, we include interaction terms to evaluate the dependency of effects of integration on resource management on particular tenure rules.

We developed the models in Stata version 12.1 (StataCorp, 2011). Model development involved incrementally adding groups of independent variables, beginning with indicators of integration, then adding the controls. We then added the interaction terms to evaluate their effects and thus the importance of dependent causation of integration with land tenure type. In later steps of the analysis, we incorporated dependent variables from earlier models as additional explanatory factors (cf. Perz et al., 2014, 2016). Hence, we model credit and later use credit to model resource management. We weighted cases based on the population of each land tenure unit and our sampling ratios, to offset larger sampling ratios employed in smaller tenure units. Hence, tenure units with larger populations but smaller sampling ratios were restored to their correct proportional sizes in terms of the populations of the tenure units from which we sampled. This was important to ensure accurate inferential tests based on how representative each household was of the overall population from which we sampled. We used robust standard errors that adjust for heteroscedasticity in the data. We clustered our standard errors by land tenure unit. We checked for multicollinearity among explanatory variables using variance inflation factors (VIFs) and excluded variables with high VIFs (over 10.0 for individual variables; StataCorp, 2011).

Findings

Descriptive statistics

Table 2 provides descriptive statistics for indicators of tenure formalization and its consequences per the ETLR, for the two land tenure types included in the household survey: PAs and PAEs. Preliminary analysis indicated that when weighted to correctly represent the relative predominance of households among tenure types, there were relatively few households in the PAFs (n = 11). This is because PAs and PAEs tend to cover large areas and PAs harbor numerous families. We therefore focus our analysis on the PAs and PAEs.

Table 2. Descriptive statistics for tenure formalization and its consequences, in agricultural projects and agro-extractive projects, Acre, Brazil, 2009.

Indicator	Agricultural projects	Agro-extractive projects	Total	F-test
Weighted n	208	46	254	N/A
Formalization
Preliminary title (0 = No, 1 = Yes)	0.37	0.67	0.43	22.57**
Definitive title (0 = No, 1 = Yes)	0.54	0.20	0.48	14.43**
Market transactions
Formal bank credit (0 = No, 1 = Yes)	0.27	0.22	0.26	0.56
Land rentals to others (0 = No, 1 = Yes)	0.13	0.05	0.11	2.36
Land rentals from others (0 = No, 1 = Yes)	0.07	0.11	0.08	0.41
Plans to sell (0 = No, 1 = Yes)	0.15	0.04	0.13	1.11
Conflict
Conflicts, last 12 months (0 = No, 1 = Yes)	0.03	0.08	0.04	2.24
Natural resource management
Kg Castanha nut harvest, 2008	113.74	600.69	201.12	15.27**
Ln kg Castanha nut harvest, 2008	1.05	3.24	1.45	27.99**
Ha under cattle pasture	21.91	22.89	22.09	0.44
Ln ha under cattle pasture	2.38	2.61	2.42	1.09**
Ha under tree crops	0.58	0.52	0.57	0.01
Ln ha under tree crops	0.17	0.28	0.20	2.10

*p <0.15; **p <0.05; ***p <0.01.

Table 2 offers a look at the ETLR indicators, as well as whether there are differences therein among tenure models (indicated by F-tests). Formalization definitely differs. Whereas overall 43% of the sampled households had preliminary documents and another 48% had definitive titles, proportionally more households in PAs had definitive titles while many more in PAEs only had preliminary titles.

In contrast, none of the four indicators of market transactions show significant differences by land tenure type. Formal bank credit, land rentals to, rentals from, and plans to sell were all similar. This is because all forms of market transactions were relatively rare, with 26% of households reporting access to credit and only roughly 10% reporting rentals or plans to sell. The frequency of land conflicts was also similar among the tenure categories, because conflicts were rare.

Differences however appeared in terms of land use. The castanha harvest (measured in kg) and hectares of cattle pasture differed significantly. Because distributions of both measures were skewed, we transformed raw values to natural logs to reduce the effects of outliers in tenure group comparisons, and in both cases, differences became more significant. Significant differences are not surprising given that PAEs are intended for NTFP harvesting, and mean castanha harvests are indeed larger there than elsewhere. PAEs also showed slightly more pasture than PAs, but that is due to the much larger areas per family in PAEs; in proportional terms, families in PAs had more pasture. Hectares planted with tree crops did not differ significantly among the tenure categories. As with the market integration measures, small differences reflected low values overall.

Table 3 presents descriptive statistics for determinants of these key outcomes, highlighting variables important in the ETLR account, beginning with indicators of regional integration. As in Table 2, Table 3 presents group means among tenure categories to permit evaluation of differences. Among the ETLR drivers, mean travel times to nearest markets did not vary among tenure categories, while travel times to the state capital did. The latter finding likely stemmed from our inclusion of a third PA which was closer to Rio Branco than other tenure units visited. Similarly, population growth in terms of net migration did not vary among tenure category. These findings are important as they indicate that differences in formalization and its consequences cannot merely be explained by differences in the location of types of tenure units. This permits a more robust evaluation of whether the effect of integration differs among households in differing tenure types.

Table 3. Descriptive statistics for determinants of tenure formalization and its consequences, in agricultural projects and agro-extractive projects, Acre, Brazil, 2009.

Indicator	Agricultural projects	Agro-extractive projects	Total	F-test
ETLR drivers
Travel time to nearest market (min)	87.35	102.72	90.30	1.72
Travel time to state capital (min)	168.34	234.34	181.03	21.25**
Community net migration (In − Out)/Families × 100)	−0.62	0.42	−0.43	0.33
Household capitals and capabilities
Land
Ha private land area	65.74	180.38	86.96	32.86**
Ln ha private land area	3.51	4.94	3.77	53.37**
Labor/Dependency
Number of adults	2.71	2.72	2.72	0.00
Number of children	1.12	1.37	1.17	1.45
Number of elderly	0.24	0.22	0.23	0.28
Days of hired labor paid	25.72	21.01	24.87	0.31
Ln days of hired labor paid	1.84	1.97	1.86	0.20
Cultural capital
Region of birth (0 = Other, 1 = Amazon)	0.73	0.82	0.75	1.72
Years of residence	15.82	16.39	15.92	0.80
Human capital
Place of birth (0 = Rural, 1 = Urban)	0.22	0.10	0.20	3.27***
Years of education completed	5.01	4.02	4.83	2.32***
Social capital
Number of absent family members	1.66	1.24	1.58	1.32
Types of organizational memberships	1.17	1.36	1.21	1.71

*p <0.15; **p <0.05; ***p <0.01.

The rest of Table 3 provides statistics for the household capabilities variables. Most do not show significant differences across tenure types. The main exception concerns land areas, which of course depend on tenure type. Households in PAs averaged roughly 66 ha each, whereas those in PAEs had 180 ha. These values correspond to the design intent of these tenure models in Acre. Beyond land, households averaged nearly three adults and one child and frequently hired labor. Most household heads originated in Brazil’s north (Amazon) region and had lived on their land for roughly 15 years. In terms of human capital, most respondents reported a rural background and nearly 5 years of education completed, neither of which varied strongly among tenure types. We used two social capital indicators, and they showed no significant variation by land tenure. The number of absent family members was roughly 1.5, and organizational memberships were somewhat greater than 1.2. Hence, overall, households varied in terms of land assets as a reflection of differing tenure models, but they did not vary systematically across tenure types in terms of most other indicators.

Multivariate models

In light of the small n for PAFs, we focus our multivariate analysis on the PAs and PAEs. This is still a useful comparison in light of the ETLR. Whereas PAs permit a higher proportion of land to be deforested, the PAEs reflect a commitment to NTFP harvesting and forest conservation. Consequently, whereas the PAs represent an agricultural land settlement model that corresponds to the ETLR’s focus on agricultural land use, the PAEs constitute an alternative that features forest-based extractivism.

As noted above, Step 1 in the analysis involves multivariate modeling of formalization in terms of land titles. Since there are three categories which differ in degree of formalization (0 = no title, 1 = preliminary document, and 2 = definitive title), we employ an ordered logit model. Table 4 presents the results, in a base model and an interactions model. The base model includes the ETLR and household capability determinants, and the interactions model adds interaction terms to evaluate whether tenure type modifies the effect of travel times on tenure formality. Since the title variable has ordered categories, and since definitive titles are the most formal category, the coefficients indicate whether an increase in the value of a determinant raises the probability of a household having a definitive title.

Table 4. Ordered logit models of formalization (Land Titles), Households in agricultural projects and agro-extractive projects, Acre, Brazil, 2009.

	Base Model	Interactions Model
Pseudo-R^2	0.38	0.40
n	196	196
Model df	16	18
ETLR drivers
Travel time to nearest market (min)	−0.031 (0.012)**	−0.041 (0.012)**
Travel time to state capital (min)	0.029 (0.012)*	0.041 (0.013)**
Community net migration
(In − Out)/Families × 100)	0.148 (0.073)*	0.196 (0.079)*
Household capitals and capabilities
Land
Tenure type (0 = AP, 1 = PAE)	−4.587 (1.558)**	−1.223 (1.954)
Ln ha private land area	0.633 (0.365)***	0.512 (0.368)
Labor/Dependency
Number of adults	2.290 (0.830)**	2.158 (0.817)**
Adults squared	−0.342 (0.127)**	−0.321 (0.128)*
Number of children	0.034 (0.114)	0.044 (0.115)
Number of elderly	1.117 (0.401)**	1.163 (0.361)**
Ln days of hired labor paid	0.238 (0.022)**	0.223 (0.034)**
Cultural capital
Region of birth (0 = Other, 1 = Amazon)	0.420 (0.403)	0.413 (0.303)
Years of residence	0.019 (0.011)***	0.018 (0.014)
Human capital
Place of birth (0 = Rural, 1 = Urban)	0.733 (0.348)*	0.911 (0.329)**
Years of education completed	0.026 (0.066)	0.017 (0.067)
Social capital
Number of absent family members	0.349 (0.102)**	0.384 (0.095)**
Types of organizational memberships	−0.230 (0.324)	−0.212 (0.310)
Interactions
Travel time to nearest market × PAE		0.016 (0.004)**
Travel time to state capital × PAE		−0.023 (0.008)**
Cut points
1	3.347 (1.296)	3.830 (1.302)
2	9.017 (2.102	9.590 (2.000)

*p <0.15; **p <0.05; ***p <0.01.

Standard errors are clustered by community and appear in parentheses after coefficients.

The base model is highly significant due to several determinants. All of the ETLR drivers are significant: households were more likely to have definitive titles if they had shorter travel times to the nearest market, longer travel times to Rio Branco, and faster population growth. Whereas the first and third findings confirm expectations of the ETLR, the second begs for an explanation. A key secondary regional market in eastern Acre is Brasiléia–Epitaciolândia–Cobija, three sister cities that sit next to each other across local rivers. Located roughly 230 km from Rio Branco, this is an important market center located near some of the tenure areas we visited. Hence, the positive coefficient for distance to Rio Branco likely reflects proximity to the tri-city area and thus indicates integration of rural households with local markets. Space limitations prevent a full discussion of the effects of the capabilities indicators, but we note that all significant effects run in the directions anticipated.

The interactions model considers our key question of whether tenure type modifies the effect of regional integration on the formalization of tenure. We therefore incorporated interactions for each travel time measure with PAE tenure status, using the PA category as the reference. The findings show that the negative effect of longer travel times to markets is slightly weaker for households in PAEs than PAs. Conversely, the positive effect of distance to Rio Branco is significantly lower for households in PAEs than PAs. Thus in both instances, the effect of regional integration in terms of travel time is greater in PAs than PAEs. Put another way, the importance of regional integration is greater in tenure units more focused on agriculture than tenure units that place greater emphasis on forest extractivism. These findings confirm that the ETLR account better fits agricultural lands than those intended for extractivism. They also confirm that dependent causation is important to fully capture the effects of integration and land tenure on formalization.

Step 2 of the analysis moves on to a key consequence of tenure formalization, namely access to formal bank credit. Since bank credit is a binary categorical variable, we estimated coefficients using a logistic regression model. We focus on bank credit from among several indicators of market integration and avoidance of conflicts. This is because there were relatively rare instances of market integration and conflict, which led to estimation problems using maximum likelihood routines in logistic regressions.

Table 5 reports results from several specifications of a credit model. We begin with a base model, and then a formalization model that accounts for additional effects of land titling. Then, we offer two interaction models that build on the formalization model. The first formalization interactions model incorporates the effects of travel times and tenure type; the second includes interactions of travel times and titling status. These two specifications permit consideration of whether the effects of regional integration depend on two different aspects of land tenure, namely tenure type and titling status. Given that we present several models, we report a subset of the findings and only focus on the ETLR drivers, tenure type, tenure formality, and their interactions. Lists of the coefficients for other variables are available upon request.

Table 5. Binary logit models of formal bank credit, households in agricultural projects and agro-extractive projects, Acre, Brazil, 2009.

	Base model	Formalization model	Formalization + interactions model 1	Formalization + interactions model 2
Pseudo-R^2	0.18	0.18	0.18	0.21
n	206	196	196	196
Model df	16	18	20	22
ETLR drivers
Travel time to nearest market (min)	0.999	0.999	1.003	1.012
	(0.003)	(0.004)	(0.071)	(0.010)
Travel time to state capital (min)	1.001	1.001	0.997	0.988
	(0.004)	(0.006)	(0.010)	(0.013)
Community net migration
(In − Out)/Families × 100)	0.969*	0.951*	0.941***	0.995
	(0.012)	(0.021)	(0.036)	(0.014)
Tenure type (0 = PA, 1 = PAE)	0.766	1.096	0.916	0.686***
	(0.264)	(0.587)	(0.778)	(0.176)
Formalization
Preliminary title (0 = No, 1 = Yes)		1.918	1.880	1.471
		(2.713)	(2.688)	(2.841)
Definitive title (0 = No, 1 = Yes)		2.518	2.534	0.252
		(3.900)	(3.903)	(0.322)
Interactions 1
Travel time to nearest market × PAE			0.992
			(0.007)
Travel time to state capital × PAE			1.004
			(0.007)
Interactions 2
Travel time to nearest market × provisional title				0.990
				(0.011)
Travel time to nearest market × definitive title				0.971**
				(0.010)
Travel time to state capital × provisional title				1.012
				(0.013)
Travel time to state capital × definitive title				1.030**
				(0.011)

*p <0.15; **p <0.05; ***p <0.01.

Standard errors are clustered by community and appear in parentheses after coefficients.

Coefficients for control variables for household capitals and capabilities are not shown but are available upon request.

The base model for credit had statistically significant variables, but most ETLR drivers and tenure type were insignificant. Net migration was significant but exerted a negative effect on access to credit. The formalization model indicated that titling status was also insignificant for bank credit. Adding travel time interactions with tenure type also failed to increase model performance. The base and formalization models run contrary to the expectations of the ETLR, though the insignificance of the interaction models thus far does not. The fourth and final model in Table 5 includes the interactions for titling status and travel times. Both interactions for definitive titles are significant. Households with shorter travel times to the nearest market and a definitive title were more likely to have credit. This squares with the ETLR, as more integrated households with formal titles were more likely to have credit. Interestingly, households with shorter travel times to the tri-city area (and thus longer travel times to Rio Branco) and a definitive title were also more likely to have credit. While this runs contrary to the ETLR insofar as households farther from the primary metropole were more likely to have credit if they had a definitive title, it is not so contrary since those households were actually closer to the secondary metropole. Both interactions show that the causation behind access to credit involves some dependency between different determinants, and not only simple, direct effects.

The remaining models take up Step 3 of the analysis, which considers the importance of regional integration, formalization, and market transactions for natural resource use. Table 6 presents a suite of models for the size of the castanha harvest. Because the distribution for castanha has many zeros, we employed a Tobit estimator to account for the truncated distribution at zero. Because the castanha distribution was non-normal, we logged the raw values.¹ The base model had significant variables, including all of the ETLR drivers as well as tenure type have significant effects. Interestingly, the direction of the effects all run in the opposite direction to those one would expect for agricultural activities. Households with longer travel times to the nearest market, those closer to Rio Branco, those in communities with slower population growth, and those in PAEs all reported larger castanha harvests. The finding for travel time to Rio Branco reflects the spatial distribution of castanha trees, which are abundant close to the capital. The strongest variable (not shown) is for total land area claimed, which was highly significant and positive. At first glance, these findings indicate that ETLR expectations do not hold for NTFP harvesting. The castanha harvest is greater among households farther from local towns in areas with more land per household that were growing slower. That said, the findings imply that agriculture occurs closer to markets and in places with faster population growth, which is consistent with the ETLR. Agriculture and NTFPs thus constitute two sides of the same coin since they occur in different places in the landscape around market centers.

Table 6. Tobit models of the castanha harvest, households in agricultural projects and agro-extractive projects, Acre, Brazil, 2009.

	Base model	Formalization + credit model	Formalization + credit + interactions model 1	Formalization + credit + interactions model 2
Pseudo-R^2	0.17	0.18	0.18	0.19
N	205	191	191	191
Model df	16	21	23	25
ETLR drivers
Travel time to nearest market (min)	0.035**	0.033*	−0.007	0.079
	(0.011)	(0.013)	(0.029)	(0.093)
Travel time to state capital (min)	−0.046**	−0.043**	−0.001	−0.110
	(0.012)	(0.014)	(0.029)	(0.086)
Community net migration
(In − Out)/Families × 100)	−0.344*	−0.260***	−0.085	−0.509*
	(0.137)	(0.136)	(0.164)	(0.226)
Tenure type (0 = PA, 1 = PAE)	6.016**	4.544*	8.966*	8.481**
	(2.042)	(2.189)	(4.038)	(3.179)
Formalization
Preliminary title (0 = No, 1 = Yes)		−3.798***	−3.307	−7.872
		(2.334)	(2.299)	(6.878)
Definitive title (0 = No, 1 = Yes)		−2.944	−3.636	−1.309
		(2.826)	(2.780)	(7.589)
Market engagement
Formal bank credit in 2008 (0 = No, 1 = Yes)		1.099	1.065	1.143
		(1.416)	(1.389)	(1.381)
Land rentals to others (0 = No, 1 = Yes)		−1.634	−1.624	−1.770
		(1.999)	(2.018)	(1.995)
Land rentals from others (0 = No, 1 = Yes)
		−0.976	−1.296	−1.212
		(1.932)	(1.959)	(1.997)
Interactions 1
Travel time to nearest market × PAE			0.043
			(0.030)
Travel time to state capital × PAE			−0.047***
			(0.030)
Interactions 2
Travel time to nearest market × provisional title				−0.053
				(0.092)
Travel time to nearest market × definitive title				0.041
				(0.094)
Travel time to state capital × provisional title				0.064
				(0.084)
Travel time to state capital × definitive title				0.007
				(0.083)

*p <0.15; **p <0.05; ***p <0.01.

Standard errors appear in parentheses below coefficients.

Coefficients for control variables for household capitals and capabilities are not shown but are available upon request.

The second model adds other ETLR variables. The ‘formalization + credit’ model adds titling status and credit. Again, there are significant effects for travel time, population growth, and tenure type, but titling status has little effect on the castanha harvest. These findings confirm those seen in the base model and indicate that NTFP harvesting is more important in places farther from towns and growing more slowly. In such areas, titling is less important, a finding consistent with the ETLR. The insignificant result for credit suggests that informal credit mechanisms may instead be at work in the castanha harvest.

The final two models add different groups of interactions. The penultimate model considers the travel time interaction with tenure type and finds only weak interaction effects. The final model replaces those interactions with terms crossing travel time with type of title. None of the interactions is significant. These findings suggest that the effects of regional integration in terms of travel times do not in turn depend on tenure type or titling status. Nonetheless, travel time and tenure type have independent direct effects on the castanha harvest.

Table 7 presents a similar suite of models for pasture. Now we expect to see findings consistent with the ETLR, since cattle ranching is a key agricultural activity on deforested lands in the Brazilian Amazon. We employ a WLS estimator since the distribution of pasture values had few zeros. Because pasture exhibited a non-normal distribution, we logged the raw values. The base model is very strong, and the travel time measures and tenure type all have significant coefficients. Households closer to the nearest market but farther from Rio Branco had larger pasture areas. There are two processes at work here: households closer to local towns are older settlements with more land cleared, and those farther from Rio Branco occur near the tri-city area where a strong ranching economy predominates. Both findings correspond to ETLR expectations as they play out in eastern Acre. The base model also shows that pasture areas are significantly smaller in PAEs than PAs. This follows expectations given that PAs focus more on agriculture.

Table 7. Weighted least squares models of cattle pasture, households in agricultural projects and agro-extractive projects, Acre, Brazil, 2009.

	Base model	Formalization + credit model	Formalization + credit + interactions 1 model	Formalization + credit + interactions 2 model
R^2	0.68	0.71	0.73	0.74
N	198	184	184	184
Model df	16	21	23	25
ETLR Drivers
Travel time to nearest market (min)	−0.007*	−0.003***	0.001	−0.010***
	(0.002)	(0.002)	(0.001)	(0.005)
Travel time to state capital (min)	0.005**	0.001	−0.002	0.011*
	(0.001)	(0.002)	(0.002)	(0.004)
Community net migration
(In − Out)/Families × 100)	0.018*	−0.003	−0.015	0.017***
	(0.005)	(0.007)	(0.011)	(0.008)
Tenure type (0 = AP, 1 = PAE)	−1.184**	−0.678***	−1.030***	−0.934**
	(0.169)	(0.330)	(0.550)	(0.126)
Formalization
Preliminary title (0 = No, 1 = Yes)		0.208	0.120	0.804**
		(0.183)	(0.205)	(0.071)
Definitive title (0 = No, 1 = Yes)		0.663*	0.656**	0.893**
		(0.180)	(0.134)	(0.095)
Market engagement
Formal bank credit in 2008 (0 = No, 1 = Yes)		0.257***	0.234***	0.180
		(0.100)	(0.106)	(0.118)
Land rentals to others (0 = No, 1 = Yes)		0.376*	0.323*	0.301***
		(0.103)	(0.104)	(0.136)
Land rentals from others (0 = No, 1 = Yes)		0.251	0.244	0.223
		(0.166)	(0.177)	(0.165)
Interactions 1
Travel time to nearest market × PAE			−0.008**
			(0.001)
Travel time to state capital × PAE			0.005***
			(0.002)
Interactions 2
Travel time to nearest market × provisional title				0.008
				(0.006)
Travel time to nearest market × definitive title				−0.001
				(0.007)
Travel time to state capital × provisional title				−0.010***
				(0.004)
Travel time to state capital × definitive title				−0.004
				(0.005)

*p <0.15; **p <0.05; ***p <0.01.

Standard errors are clustered by community and appear in parentheses after coefficients.

Coefficients for control variables for household capitals and capabilities are not shown but are available upon request.

The second model adds the formalization and credit variables. Definitive title emerges as a key factor for pasture areas: households with definitive titles had more pasture. This is consistent with the ETLR in that tenure formality and thus tenure security is important to impel larger production systems, in this case for cattle ranching. Adding credit and other indicators of market engagement slightly strengthens the model, but the positive effect of credit is weak, though rentals to others coincide with expanded pasture area.

The final pair of pasture models considers the different travel time interactions with tenure. The first of these adds in the travel time interactions with tenure type. The findings are strong for travel time to nearest market and weak for travel time to the capital. Households with longer travel times to nearest markets in PAEs had significantly less pasture than those in PAs. The effects of travel time on pasture thus depend on tenure type. Hence, while most findings for the ETLR on cattle pasture along the IOH in eastern Acre correspond to the expectations of the ETLR, the importance of regional integration for pasture depends on tenure type. The final model considers interactions of travel time and titling status. Here, the results are weak. These results confirm the ETLR.

The final models appear in Table 8, on tree crops. Tree plantations for agricultural production constitute an intensive form of agricultural land use and should thus correspond strongly to the ETLR. We model the natural log of hectares planted under tree crops using a WLS estimator. The base model has significant variables but is not very strong overall. The integration variables overall exhibit weak effects. Interestingly, households with longer travel times to the nearest market reported larger areas planted under tree crops, though households with longer travel times to Rio Branco reported less area under tree crops. The first finding was a surprise that goes begging for an explanation, though the second is consistent with the ETLR.

Table 8. Weighted least squares models of tree crops, households in agricultural projects and agro-extractive projects, Acre, Brazil, 2009.

	Base model	Formalization + credit model	Formalization + credit + interactions 1 model	Formalization + credit + interactions 2 model
R^2	0.16	0.17	0.19	0.18
n	197	184	184	184
Model df	16	21	23	25
ETLR drivers
Travel time to nearest market (min)	0.002***	0.001***	0.002	0.000
	(0.001)	(0.001)	(0.002)	(0.002)
Travel time to state capital (min)	−0.002***	−0.001***	−0.002	−0.001
	(0.001)	(0.001)	(0.002)	(0.002)
Community net migration
(In − Out)/Families × 100)	0.006*	0.008***	0.005	0.007
	(0.002)	(0.003)	(0.004)	(0.004)
Tenure type (0 = AP, 1 = PAE)	0.027	−0.069	−0.096	−0.076
	(0.119)	(0.181)	(0.308)	(0.196)
Formalization
Preliminary title (0 = No, 1 = Yes)		0.168***	0.147	0.111
		(0.090)	(0.119)	(0.160)
Definitive title (0 = No, 1 = Yes)		0.053	0.051	0.052
		(0.105)	(0.099)	(0.110)
Market engagement
Formal bank credit in 2008 (0 = No, 1 = Yes)		−0.026	−0.034	−0.024
		(0.098)	(0.104)	(0.093)
Land rentals to others (0 = No, 1 = Yes)		0.097***	0.078	0.091
		(0.037)	(0.051)	(0.042)***
Land rentals from others (0 = No, 1 = Yes)		0.070	0.064	0.065
		(0.139)	(0.136)	(0.158)
Interactions 1
Travel time to nearest market × PAE			−0.002
			(0.002)
Travel time to state capital × PAE			0.001
			(0.002)
Interactions 2
Travel time to nearest market × provisional title				0.001
				(0.002)
Travel time to nearest market × definitive title				0.001
				(0.002)
Travel time to state capital × provisional title				−0.000
				(0.002)
Travel time to state capital × definitive title				−0.001
				(0.002)

*p <0.15; **p <0.05; ***p <0.01.

Standard errors are clustered by community and appear in parentheses after coefficients.

Coefficients for control variables for household capitals and capabilities are not shown but are available upon request.

The next model adds the formalization and credit variables. Explanatory power does not rise and the additional variables are not strongly significant. These findings were surprisingly weak in light of the strong findings for pasture and the relevance of the ETLR for agriculture. The likely explanation resides in the fact that households along with IOH in eastern Acre did not plant large areas in tree crops; hence, there was limited variation (Table 2). Whereas both castanha and pasture varied considerably overall and among tenure types, tree crops did not. Hence, while one could assert that the case of tree crops in eastern Acre undermines the ETLR, one could also argue that tree crops in this study region are not important enough to adequately evaluate the ETLR.

Discussion

In the context of regional integration and tenure diversity in eastern Acre, the ETLR proved largely applicable. This was due to a combination of strong confirmatory findings as well as results that indirectly confirm by circumscribing the applicability of the ETLR. Households more integrated in terms of shorter travel times to markets were more likely to be titled, have credit, have more pasture, and engage in less forest extractivism. The first three confirm the ETLR; the last confirms that the ETLR applies more to agriculture than forest extractivism. Hence, our findings clarify the scope of applicability of the ETLR.

The other contribution made here is to confirm the importance of dependent causation in the relationships highlighted by the ETLR. While many previous studies have confirmed the importance of land tenure among other factors as having an independent effect on land use, the focus on multiple causation and multi-step causation has obscured the importance of dependent causation. Our findings indicate that tenure type influences the effects of ETLR drivers like regional integration on tenure formalization. Tenure type also mattered for the effects of integration on specific forms of land use, namely cattle ranching and forest extractivism. Hence, tenure diversity modifies the effects of regional integration on formalization and its consequences.

These findings bear implications for the ETLR as a theory of land use, the study of regional integration, and land system science. The main implication for the ETLR is that whereas it has often been criticized for its lack of applicability in various respects, there remains a need to better delineate its scope. Rather than engage in slash-and-burn criticism of the all-or-nothing sort, we find it more useful to attend to theoretical scope and refine the breadth of applicability. The ETLR may well not fit various situations where, for examples, titling is politically rather than economically motivated, titling does not confer tenure security, or informal credit is widely available (cf. Platteau, 1996; Sjaastad & Cousins, 2008). But that is not to say that the ETLR therefore applies never and nowhere. It is more useful to carefully delineate the respects in which it applies – and does not apply – in various contexts. The case of Acre is useful in that regard, as we confirm that the ETLR applies better to agriculture than forest extractivism, and its propositions are subject to emendations in the presence of tenure diversity due to the resulting dependent causation. It would therefore seem useful to revisit previous critical work of the ETLR to more systematically sort out the respects in which it applies and in which it does not under various conditions.

In contrast to the many critiques of the ETLR, there has been little question that regional integration, especially via infrastructure, has significant effects on land tenure and land use. The study of regional integration in terms of infrastructure impacts is however a fragmented one, for the impacts of infrastructure are manifold as well as mixed (Perz et al., 2012). Economic appraisals tend to be positive, in terms of economic growth and poverty reduction (e.g. Bourguignon & Pleskovic, 2008; Straub, 2008). By contrast, ecological research generally yields negative findings, whether they involve habitat loss, watershed degradation, or shifts in species assemblages (e.g. Forman et al., 2003; Trombulak and Frissell 2000). Social research has also tended to be negative, highlighting limited participation in planning (e.g. Robinson 2001), unequal benefits among social groups (e.g. Shriar, 2009), and conflicts over land tenure (e.g. Schmink & Wood, 1992). The findings from the present study underscore the need for analyses to carefully select factors that differentiate the effects of integration as well as the outcome variables. In the present case, we emphasized the importance of tenure type in order to evaluate the dependency of the effect of regional integration. Further, we considered a suite of outcome variables, informed by the ETLR but going beyond it with regard to land use to consider agriculture and forest extractivism. Such a differentiated approach to evaluating infrastructure impacts is badly needed in light of the contrasting findings in different literatures on infrastructure.

The ETLR is an eminently dynamic theory, and system dynamics are similarly underscored in land system science (cf. Boillat et al., 2017). The study of land change has featured frameworks which highlight the importance of causal proximity, from distant to intermediate to proximate determinants of land change outcomes (e.g. Lambin & Geist, 2006; Wood & Porro, 2002). Such frameworks increasingly recognize the feedbacks from land change to the causal drivers and other mechanisms which in effect set causal hierarchies in motion as land cover and its drivers change over time (Rounsevell et al., 2012). Prominent among such drivers are infrastructure and tenure, both featured in the ETLR. In particular, there is growing attention to ‘telecouplings,’ notably from global processes such as commodities markets to local processes such as changes in land use (Boillat et al., 2017; Friis et al., 2016; Seto & Reenberg, 2014). The ETLR is germane to such frameworks in that it too makes arguments about synergies among processes that affect land outcomes, such as infrastructure projects (e.g. Perz et al., 2012, 2016). This is especially the case to the extent that a key multi-scale process featured in land system telecouplings concerns regional integration initiatives, like infrastructure, that facilitate connections with commodity chains and global markets. Whereas land system frameworks highlight global–local telecouplings in dynamic land systems, the ETLR focuses on local dynamics in causation in the synergies among development processes tied to tenure change as a determinant of land change. The two can thus be usefully wedded, though there remains a need for theoretical development about telecouplings in the local–global feedbacks from land change outcomes as a result of tenure evolution.

A key issue that then arises concerns the need to refine the particular mechanics of the processes involved from drivers like economic integration through the key processes underscored by the ETLR, including formalization and market engagement, as they play out in influencing land change in locally specific ways. Insofar as specific local circumstances depart from the assumptions of the ETLR, whether due to tenure diversity or the balance of extractivism and agriculture, the feedbacks from local land use to global markets will vary. Delineation of this variation permits alternative predictions, which provides a framework for identifying ‘dependent telecouplings’ and ‘dependent dynamics’ in land system science. Hence, dynamic theories like the ETLR are thus useful for land system science, and incorporation of dependent causation permits testing for variation in the effects of drivers of and feedbacks from land use.

Acknowledgments

Financial support for this research came from the National Science Foundation, Human and Social Dynamics Program: [Grant Number 0527511], the Coupled Natural and Human Systems Program: [Grant Number 1114924]. The coauthors thank the students and other collaborators who contributed to the community survey fieldwork and data entry in Acre, Brazil (Cleilton Amaral, Arthur Lima Carvalho, João Paulo Ferraz, Raionery Gonçalves, Genildo Macedo, and Robson Belo Nogueira). Errors are the responsibility of the authors.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

Financial support for this research came from the National Science Foundation, Human and Social Dynamics Program: [Grant Number 0527511], the Coupled Natural and Human Systems Program: [Grant Number 1114924].

Notes

1. We added one to the castanha values before the log transformations in order to avoid losing observations with zeros.