1

DOES COMMUNITY-BASED SELF-SELECTION OF LAND REFORM BENEFICIARIES WORK? AN ASSESSMENT OF THE BRAZILIAN

CÉDULA DA TERRA PILOT PROGRAM1

Hildo Meirelles de Souza Filho (*) Antônio Márcio Buainain (**)

Carolina Junqueira Homem de Mello (***)

José Maria da Silveira (**) Marcelo M. Magalhães (****)

[Draft Version]

This article addresses a particularly relevant issue for the

sustainability and effectiveness of agrarian reform policies: the selection

of beneficiaries. The study is part of a thorough evaluation of the Cédula

da Terra Community-Based Pilot Program, which presents a radical

shift from the traditional approach of expropriation-distribution in land

reform. The program’s rationale and the expected results of its

governance structure towards targeting are presented here. In addition,

the article provides empirical evidence on how the program is actually

reaching the target population in Brazil.

1. INTRODUCTION Poverty is still a striking problem in Brazil. A recent study by

Maletta, Buainain and Villalobos (1999) based on the 1997 PNAD

(National Households Sample Survey) indicated that about 15 million

households (37% of the households) and 67 million people (44% of

1 This document would have not been possible without the enthusiastic collaboration of Daniela Silva Pires and Marcelo Francisco Melo, undergraduate students at the Institute of Economics/UNICAMP; Celeste María Díaz Cónsul, M.Sc. student of statistics at UNICAMP; Dr. Rinaldo Artes, Assistant Professor at the Mathematics and Statistics Institute of the University of São Paulo. We would also like to thank the anonymous referees for their comments, which have contributed so greatly to improve this paper. The authors are: (*) Lecturer at the Federal University of São Carlos and researcher of NEAD; (**) Lecturer at the State University of Campinas and researcher of NEAD; (***) Undergraduate student of economics at the State University of Campinas; (****) Researcher of NEAD and IICA consultant. Respective e-mail addresses are: hildo@power.ufscar.br; buainain@eco.unicamp.br; carolmel@eco.unicamp.br; jmsilv@eco.unicamp.br; mmagalhaes@nead.gov.br.

2

Brazil’s population) lived below the poverty line. A more comprehensive

view of poverty, using indicators of basic needs, revealed that 48% of

the households and 56% of the population suffer from deprivation.

According to the same study, the figures are higher in rural areas. Low-

income poverty affects 57.5% of the rural households and 66.6% of the

rural population, i.e. about 4.2 million households, comprising 20

million people.2 The percentages rise to nearly 72% and 78%,

respectively, when basic needs are also considered along with income.

In other words, from two-thirds to three-quarters of the rural dwellers

are poor.

Agrarian reform has been selected as a government priority in its

intervention to combat poverty. Since 1994, approximately 300,000

families have been settled under the Federal agrarian reform program.

This figure is greater than the total number reached during the

preceding thirty years (1964-94). The simple expansion of the land

reform program was enough to raise a debate about its efficacy and

sustainability and to generate new policies to restructure land tenure.

Significant efforts have been made to improve government

intervention in this area. INCRA (The National Institute of Colonization

and Agrarian Reform) has been modernized and legislation has also

been updated in some key issues.3 In fact, the legislation recently

adopted has made it possible to speed up expropriation procedures and

to cut down on settlement delays. The cost of expropriated land has also

been brought down, as land prices have steadily decreased during the

decade (Reydon & Plata, 1998).

In spite of considerable improvements over the last years, several

studies have highlighted a number of problems that actually hinder the

2 These figures, based on the PNAD, do not consider rural areas in the Northern Region, where an additional 2 million people (some 600,000 households) are suspected to live below the poverty line. 3 Legislation allowed landowners to delay the process for years through court litigation. They also managed to obtain a compensation payment above market prices for the take-over of the land. Thanks to these legal quirks, the Land Reform process has benefited many landowners. Many traditional absentee landowners ended up richer after their land was taken for Land Reform purposes.

3

current agrarian reform program and the full sustainable development

of its beneficiaries (see Buainain et al., 1999; Leite & Palmeira, 1997;

and Guanzirolli, 1998). The key points are: (a) the cost of agrarian

reform; (b) the natural resources endowments and infrastructure; and

(c) the selection of land reform beneficiaries.

In the past, most beneficiaries of agrarian reform projects were

selected directly by the government. This rather bureaucratic approach

was subject to a series of distortions: a sluggish pace, the political

appointment of beneficiaries, high costs, and others. In recent years,

the government's role in the selection of beneficiaries has been

strikingly redefined. In most cases, it truly legitimates land takeover by

landless rural workers, or by candidates that have previously registered

for a plot of land. In spite of the substantial and desirable participation

of potential and actual beneficiaries in the selection process, several

controversial issues have come up regarding the selection of agrarian

reform beneficiaries.

Agrarian reform opponents argue ―without producing

evidence― that the doors are open to opportunists who are neither in

need nor have anything to do with agrarian reform. Additionally, as

beneficiaries have nothing to lose in the case of failure, this process of

selection may attract people that are not familiar with agricultural work.

Unprepared and inexperienced beneficiaries have certainly undermined

the chances of success of many settlements.

While a series of measures have been taken to improve the

traditional intervention in agrarian reform, new instruments and

approaches have recently been introduced to both improve and expand

poor rural families' access to land. One of such programs is the Cédula

da Terra, a pilot project funded by the World Bank, which uses real-

estate market mechanisms rather than expropriation as its instrument

of land redistribution.

The introduction of the Cédula da Terra program has raised a

series of criticisms on the viability and sustainability of market-oriented

4

social interventions. In the particular case of the Cédula da Terra,

questions regarding the poor communities' weak bargaining power and

lack of negotiation skills have frequently been raised to invalidate

decentralized land purchase as a valid instrument of land tenure

intervention. In addition, considering the rural poor’s secular

subordination to local oligarchies, critics cast doubts as to whether the

decentralized community-based structure of governance may provide a

sound framework for poverty combat interventions. They argue that it is

more likely that eventual benefits will accrue to members of local elite

groups than to the rural poor.

Buainain, Silveira and Teófilo (2000) sustain that the design of

the Cédula da Terra Program defines a governance structure ex-ante,

from which it is possible to analyze the program’s dynamics, delineate

expectations regarding its performance and evaluate its results and

socioeconomic impacts. This structure of governance includes several

control and incentive mechanisms that affect the behavior of the agents

involved in all different phases of the program, from the selection of

beneficiaries, choice and acquisition of properties to the community

investments, and the decisions about production as well (how and what

to produce).4

In other papers Buainain et al. (1999a; 1999b) have revised the

relevant issues at stake and provided a preliminary assessment of the

functioning of the community-based Cédula da Terra Program. They

have focused on land selection (Buainain et alii, 1999b), land

negotiation (Silveira, Buainain and Magalhães, 2000) and investment

patterns (Silveira, Buainain and Magalhães, 1999) as established under

the Cédula da Terra Program. In all cases, their conclusions do not back

up the common criticism that the program is not working and that poor

communities are not prepared to take relevant economic decisions on

their own, and thus require government tutelage. 4 Also, efficiency of the process, from a broader criterion of poverty alleviation, may be compared to the other forms analyzed by the literature (see Bardhan & Udry, 1999, for a synthesis) or as part of social programs.

5

The selection process has not been revised so far.5 In this paper

we will deal with the targeting and selection mechanism of the Cédula

da Terra Program, and provide a preliminary assessment of the

socioeconomic profile of the beneficiaries of the Cédula da Terra

program.

The analysis of the beneficiary selection process has a very broad

scope. One can think of the opportunity of favoring a determined target

group (as a first step, according to Ravallion, 1993) and compare it with

other alternatives, or even compare its impact on the government

budget or on any other relevant variable. Other ways of going through

this process can also be taken up and discussed: whether it is

convenient to choose regions, communities or persons; to consider the

cost/benefit relation that could define a trade-off between monitoring

costs and gains in terms of poverty alleviation. Finally, the analysis and

the results from the selection process are not independent from the

parameters used to measure poverty, and thus trigger the need to

evaluate, case by case, which measurements are more adequate. It can

also be questioned whether or not there are any mechanisms that lead

to efficient self-selection processes that reduce monitoring costs as well

as other costs associated with the existence of opportunistic behavior,

asymmetric information and even the seizure of part of the benefits by

interest groups.

The present work does not aim to address this whole set of issues.

Its goal is much more modest: it aims to characterize the beneficiary of

the program and, with this information, to verify if the proposed

governance structure is adequate from the point of view of a target

group selection. Its main goal is to verify the hypothesis that the

program reaches a certain type of target group which has apparent

conditions to face the challenges created by this type of land tenure

intervention, particularly those of implementing and exploiting

economically sustainable businesses. This work does not, by any

5 Deininger (2000) analyzes a similar program in Colombia and South Africa.

6

means, exhaust the complex set of questions raised by the analysis of

the governance mechanisms related to the Cédula da Terra Program.

However, it does make room for further investigations into its efficiency

in the context of social and redistribution programs that seek to

alleviate poverty in Brazil.

2. THE STRUCTURE OF GOVERNANCE, EFFICIENCY AND TARGETING OF PUBLIC INTERVENTION IN THE COMBAT AGAINST POVERTY

2.1 TARGETING, EFFICIENCY AND GOVERNANCE STRUCTURE

Brazilian policy designers have largely neglected the issues of

targeting and the selection of beneficiaries in the poverty combat

programs. A target population has been traditionally defined by a set of

desirable characteristics, such as low income, rural poor, landless rural

workers etc. Actual selection and benefits coverage have been subject to

bureaucratic procedures and controls. Underlying issues that have been

raised by the specialized literature (see Kanbur (1987) and Ravallion

(1993) for instance), such as cost implications, program performance

and sustainability, equity and efficiency have seldom been assessed. Yet

there is evidence that carefully designed targeting is a powerful tool for

enhancing the overall performance of public intervention, particularly

with regards to health and nutrition.

As a first approximation, targeting may be understood as a mere

selection instrument or process which allows policy designers to select,

focus and target on specific groups, regions, problems and so on. In the

context of interventions to combat poverty, the general focus is likely to

be directed to those regions and population groups that have either

lagged behind or have been excluded from the benefits of economic

growth. The idea is to channel public resources towards the selected

"targets” (Homem de Mello, 2000).

It is necessary to draw attention to the confusion commonly

created between a "targeting process” (or simply “targeting”) and

"targeting techniques". A targeting process defines the rules and

7

conditions governing selection and access to any public program

whereas targeting techniques are used to implement a specific selection

process and lay down a series of pragmatic steps which should be

followed or accomplished by a beneficiary in order to be selected or join

the program.

From the point of view of statistics, targeting (or "statistical

targeting", according to Besley, 1993) would correspond to truncating

the population distribution according to desirable variables. Based on

relevant information regarding the population, the targeting process,

that is, the discrimination of subgroups in the population, according to

specific objectives, would be accomplished, technically speaking, by the

use of appropriate available targeting techniques. Although this may

seem to be a very straightforward and simple task, appropriate targeting

is far from trivial, as we are to see.

One alternative is to go into ex-ante very expensive research, data

collection and information analysis before the launch of the

intervention.6 This information would be used to map out the potential

beneficiaries of public programs. Due to a series of reasons, this is

seldom a recommended procedure. Resources are not always available;

the time required for data collection is neither compatible with the

actual and underlying justification for the intervention nor with its

political timing; to encounter institutions that are qualified to carry out

the preparatory data collection and analysis may also be a problem;

and, last but not least, expensive and time consuming targeting

processes are usually associated with considerable diversion of the

scarce financial and human resources from the intervention’s ultimate

goal: that of benefiting the targeted vulnerable population/region.

Even when information is available (or made available for the

targeting exercise) and cost considerations are left aside, it is not

possible to neglect the possible negative effects of incomplete

information (which is the rule and not the exception in real life) on the

6 For a similar procedure, see David, et al. (1999).

8

identification and selection of potential beneficiaries. On one hand, the

accuracy of the statistics available in developing countries is usually far

from the desirable standards as far as policy design is concerned. On

the other hand, even ignoring the possible problems in accuracy, which

may be associated with the available statistics, in most cases these

statistics are still incomplete, as they may not portray the actual

information required for carrying out the best selection of beneficiaries.

This may represent a serious problem if the performance of the

intervention rests on unrevealed specific characteristics and attitudes of

the beneficiaries, such as their entrepreneurial skills, compliance with

the rules of the program, their willingness to make efforts, personal or

family history and so on.

In fact, a statistical analysis hardly ever reveals the necessary

information about the effort the candidate is expecting and is willing to

put into the program. Candidates could, for example, simply answer the

question posed with the intention of obtaining some later benefit

without actually taking up a believable commitment to the program’s

goals. How can one ensure, ex post, that a candidate, once selected, will

actually dedicate himself to the effort required for the program to be

successful?

There is no need to go into depth in the conventional economic

theory to raise questions regarding the "distortions" such interventions,

if conducted in the wrong way, would cause on economic efficiency (see

Sadoulet and De Janvry (1995) for a discussion about efficiency and

market intervention). On the other hand, several authors have

convincingly argued that both extreme poverty and socioeconomic

inequality are main sources of inefficiency that hinder the development

and economic growth of poor countries.7 The challenge is to substitute

7 See, for instance, the following papers: Hoff (1999) and Bardhan and Udry (1999), particularly the chapter on Poverty Alleviation. Both works point out the negative effects of high levels of social inequality and poverty on overall economic development. The negative effects drastically reduce the efficiency of the market as well as other mechanisms of governance that may be very efficient under other socioeconomic contexts. Hoff (1999) indicates the importance of considering the presence of market failures when they are relevant and how inadequate the Paretian reference is.

9

the interventions, which focus on the population as an indistinct mass

by programs to combat poverty which have been adequately designed to

match the differentiated profile and potential of poor families.

2.2 FORMS OF TARGETING

As seen, the emphasis given to the connection between targeting

and governance structure does not exempt the analysis of different

forms of targeting in the context of poverty combat interventions. There

are several distinct forms of targeting, and all of them have different

costs and benefits to the poor and to the society in general. The choice

and evaluation of the selection process must necessarily take into

account these differences as well as the implications of the approach

chosen on the political sustainability and the acts of public

interventions (Homem de Mello, 2000).

2.2.1 Broad Targeting , Measures for Redistribution and Welfare8

The first distinction to be made is between the two categories of

targeting: broad targeting and narrow targeting (Van de Walle, 1998).

The broad targeting approach does not introduce socioeconomic

differentiation to identify who may and who may not benefit from a

given public intervention. Public resources are simply allocated to

expenditure items, such as education and health, or to specific regions.

Although vulnerable population and regions are expected to benefit the

most from this intervention, either because they are in more need of the

services provided or due to the high geographical concentration of

vulnerable population, access to the broad target programs is universal.

The whole population is entitled to the programs' benefits, irrespective

of their specific socioeconomic characteristics. Good examples of broad

targeting are expenditures in basic social services such as primary

education, public health, basic sanitation and so on.

Broad target based public programs are usually more popular

than those which "discriminate" in favor (and also, in contrast, against)

8 See Van de Walle (1992; 1994) for a comparative study of targeting processes in education and health.

10

part of the population. Other possible additional advantages of such an

approach are its operational simplicity and low targeting costs. The

universal character of the criterion usually renders it unnecessary to

set up both special selection rules and implementation mechanisms to

supervise the selection of participants and the distribution of benefits.

However, though the approach provides for operation savings,

broad target programs tend to be somewhat bigger than narrow target

interventions. Overall program costs tend to go up as benefits are

extended to the entire population, irrespective of the participants' actual

need. If available resources are not enough to fund program operations,

a situation which commonly occurs in developing countries,

governments have to decide whether to scale down a program’s scope

in a planned and organized way or to allow the quality of services

provided by the programs to deteriorate. For political reasons, it is

easier to just let a program fall apart than to introduce the unpopular

reforms required in order to close the gap between the availability of

resources and the scope and costs of broad target based programs. The

consequences of insufficient resources to run the programs as they

were conceived are poor quality public services and deteriorating

agencies.9

The validity of the broad target approach has also been

questioned on the grounds of equity. Even though the intervention may

reach all regions and the entire population, it is likely that wealthy

people and affluent regions, which already have access to such services

and assets, will benefit the most from broad targeting interventions. Is it

a sound policy to allocate rather scarce resources to provide subsidized

public services to people who can pay for them, and in many cases

would do so without complaining? Furthermore, is it a sound policy,

when there is evidence that in many cases the participation of people

who are better off is secured through the exclusion of those who are

9 Buainain (1999) sustains that since the 1980s many of Brazil’s public interventions have followed the trend of slow deterioration and loss of efficiency. For the evaluation of broad targeting programs, see Subbarao (1996) and Ravallion (1993).

11

most in need?

Finally, though broad targeting may be recommended in some

policy interventions, such as primary, basic education and health, it

may not be an appropriate instrument to reach groups with specific

needs nor to achieve well-contoured particular objectives. Criticisms of

broad target based public interventions have grown pari pasu together

with the fiscal restrictions most developing countries have been facing

lately.10

When trying to apply broad targeting programs, one can perceive

a clear trade-off between scope and simplicity, on one hand, and better

accuracy and attention to the multiple requirements of reality on the

other hand.

The development of techniques that direct less ambitious

intervention policies may be associated to a better use of the

beneficiaries’ unique characteristics and to the perception of the most

severe restrictions to which they are held, that is, the market failures to

which certain target groups are submitted.

2.2.2 Narrow targeting: exploiting market failures to benefit beneficiaries.

Narrow targeting is defined as a deliberate attempt to concentrate

benefits for the poor (or poor regions), regardless of the category of

spending. This approach rests on two principles or strategies (or a

combination of them): the first is the targeting indicator and the second

is self-targeting.

The first strategy identifies a set of characteristics of poor people

(preferably one characteristic) that is highly correlated with low income

(or any other relevant characteristic of the ideal target population, such

as malnutrition), but one, which is easier to identify and observe than

income itself or the population’s nutritional status. This set of variables 10 Kanbur (1987) uses the Foster, Greer and Throbocke index to assess the results of broad target validity of broad targeting used in regional transfer policies and Ravallion (1993) compares differential of gain in poverty alleviation of several forms of using the making the transfers. Ravallion also analyzes the political convenience of certain forms of redistribution, which though effective on poverty alleviation,

12

(or variable) takes the role of target indicator and is used as a proxy for

income or nutritional status. For example, from the previous available

information (though incomplete, as it will surely be) policy formulators

have found that poor (or undernourished) families live in neighborhoods

located on the outskirts of big cities and away from the roads of main

access. Instead of undertaking costly data collection research and

setting up selection and supervision procedures, it might be simpler to

extend the benefit (subsidized milk or bread, for instance) to all

inhabitants of these neighborhoods. The service or product would be

provided only in this specific area. Outsiders who would like to get the

subsidized product or service (though theoretically not entitled to the

benefit) would have to be willing to travel to the area. Though this would

certainly be considered a leakage of the program’s benefits, from a

technical point of view, the outcome would probably not be undesirable.

On the contrary, if the program were properly designed and subsidies

set at a level that allowed access without creating obvious business

opportunities for the neighborhood inhabitants, the "leakage" would

probably work as a target correction rather than as a distortion. It can

be expected that only people in real need would go to the trouble and

bear the cost of an urban displacement just to get a ration of subsidized

milk or bread.

Recent assessments of geographic targeting in different regions

have provided contradictory results. On one hand, Morris et al. (1999)

mention that an analysis carried out by the World Bank (Baker and

Grosh, 1994) "has demonstrated that, at a national level, in Latin

America, geographic targeting reduces the leakage of program benefits

to the non-needy as compared to untargeted programs". On the other

hand, their own assessment of nutrition interventions in Adidjan and

Accra reveals that "neighborhood targeting of poverty-alleviation or

nutrition intervention in these and similar cities could lead to under-

coverage of the truly needy". Baker and Grosh (1994) drew the same

conclusion, which found that the under-coverage of the truly needy

increases in geographic target based programs in Latin America.

13

The second principle is that of self-targeting. The strategy is to

define a set of incentives and costs of participation (adhesion criteria)

and let potential beneficiaries themselves decide whether to join a

program. The underlying goal is that, though the adhesion criteria are

common to all potential beneficiaries (rural poor, for instance), it would

actually provide differentiated incentives for the adhesion of a particular

sub-group of rural poor. It is assumed that self-selected beneficiaries

would be endowed with non-observable (and highly desirable) personal

(family) characteristics that would increase the probability of success of

a program. Though such characteristics shall vary according to each

program objective and design, they should be easily identified by the

analysis of the structure of governance and overall program strategy.11

One common operational problem faced by self-selection based

interventions is how to exclude the non-poor from self-selecting

themselves to participate in a program. There are many ways to prevent

significant leakages without intervening in the self-selection principle

nor resorting to ex-ante bureaucratic selection mechanisms. One

alternative is to increase the participation "burden" or "tax".

This is a key element in the self-selection process: to function as

an effective deterrent of nor-poor adhesion, the weight of the burden

should be obviously heavier to the better off than to the actual poor and

needy. The size of the benefit is another key variable to avoid leakages.

If self-selection mechanisms were properly designed, non-poor

population would not bother to incur the adhesion burden to grasp

what they perceive as a small benefit. As mentioned above, setting up

these criteria and mechanisms is not a trivial task.

Another common problem faced by programs based on self-

selection targeting is how to avoid the beneficiaries' dependency on

program benefits. High dependency on benefits can undermine the

sustainability and success of programs. There are also many ways to

11 Heckman et al. (1999) provide a deeper analysis of targeting issues, in particular the problems related to self-selection bias in the evaluation of program results and socioeconomic effects.

14

tackle this issue, and the most common is to require a feasible and

desirable counterpart contribution from the beneficiaries. The

contribution would function as an adhesion cost that is expected to

exclude those potential beneficiaries that are not willing to truly commit

themselves to a program. A good example is to require beneficiaries of a

program that distributes basic food staples to keep their children in

school, for the school’s premises can be used as the distribution locus.

Both types of narrow targeting are helpful to reduce the main

problems which greatly undermine the use of broad targeting by most

poverty combat programs: the leakage of benefits to non-poor, the

imperfect coverage and the high cost associated with universal benefits.

On the other hand, the narrow targeting approach is not free from

problems. In addition to those mentioned above, it is worthwhile to

point out the following: (i) the administrative costs generated by poor

information concerning the potential beneficiary population. Lack of

information can lead to high identification costs; (ii) the problems

associated with the changes in behavior of the beneficiaries in response

to programs' embedded incentives;12 (iii) the costs arising from the

participation/interference of interest groups. While the first cost can be

usually measured, the other two are rather obscure and difficult to

quantify.

In addition to the reduction of leakages and program costs, self-

selection based programs are expected to perform better than either

broad selection or bureaucratic selection mechanisms. On one hand, if

properly designed, self-selection mechanisms are expected to attract the

share of potential beneficiaries that are both best prepared to benefit

from the program and truly committed to making it work. On the other

hand, this approach has the advantage of reducing the power of local

members of elite groups to use public intervention as a political

12 A number of such problems have been identified in the literature, ranging from family disruption due to the assignment of property rights to women rather than to men, drops in production due to the withdrawal of child labor from the crop fields, beneficiary withdrawal from traditional income raising occupations to join the program and so on.

15

instrument. This advantage should not be underestimated, especially in

those countries where actual participative democracy is still largely a

desirable goal and corruption is a common practice in public-private

relations.

As will be detailed in section 2.3, the Cédula da Terra selection

mechanisms are based on a combination of both principles of the

narrow targeting approach: indicator targeting and self-targeting13. Our

general concern is to assess, if even on a preliminary basis, how well

the program structure of governance is working with respect to the

selection of beneficiaries.

The specific questions we are raising are: (i) has decentralized

community-based land reform intervention actually been reaching the

rural poor as expected, or has government withdrawal from land

acquisition and selection of beneficiaries allowed for leakages of benefits

in favor of non-poor families? (ii) Which poor are the program selection

mechanisms actually reaching? Is the program attracting beneficiaries

who are suited to run a family farm business or are the program’s

rather simple selection criteria also attracting families which are truly

poor but do not actually fit into agrarian reform programs? The

performance and sustainability of the Cédula da Terra program are

understood to be greatly dependent on how well its selection

mechanism is working.

2.3 CÉDULA DA TERRA PROGRAM AND EFFICIENT STRUCTURES OF

GOVERNANCE

The Cédula da Terra pilot program was launched at the end of

1998 and is expected to benefit approximately 15,000 families at a total

cost of US$150 million. The program’s target population is comprised of

landless rural workers and poor rural farmers with insufficient land to

ensure their subsistence and the sustainable process of capital

13 The structure of governance of the program imposes certain limitations to access (indicator targeting), and, for the resulting subgroup, there is room for the voluntary adhesion to the program. See Buainain et al. (1998) for a formal scheme of the self-selection process.

16

accumulation. Five states in the Northeast14 (Bahia, Ceará, Maranhão,

Minas Gerais, and Pernambuco) have been contemplated in this pilot

phase.

Program implementation has been considerably disrupted by both

the financial crisis, which has required restrictive fiscal policies, and

the occurrence of the most severe drought ever during the last 50 years.

The question is whether or not the Cédula may become a valid

alternative to INCRA’s traditional land redistribution scheme. It is

particularly interesting to analyze how the program performs in those

key issues (of cost, selection and sustainability), which have hindered

agrarian reform and reduced its positive socioeconomic results (see

Buainain et al. (1999a) for a preliminary evaluation of the Cédula da

Terra).15

Several authors are critical of traditional forms of distributing

land (these forms being one of the basic components of rural wealth).16

Such criticism is largely associated with factors that reduce effort

allocation by beneficiaries, in particular the paternalistic relationship

between government and potential beneficiaries and the ubiquity of

property rights over the distributed land. It is argued that the land

expropriation-free distribution approach does not produce an adequate

structure of governance to lead and push beneficiaries towards and

alongside a sustainable accumulation track.

The Cédula da Terra takes a rather different approach to promote

the rural poor’s access to land. The main characteristics of the Cédula

da Terra program are:

(i) Contrary to traditional interventions, which are marked by

14 Brazil’s Northeastern region is situated almost entirely in an arid zone. The evolution of its socioeconomic structure, with its depleted environment and poor natural endowments, lies at the root of bad records of human development, similar to the poorest countries. 15 Such evaluation yields enough information to sustain an optimistic view about its potentials. It also suggests a number of weaknesses that, if not properly addressed, could jeopardize its performance. 16 For a survey, see Buainain, Silveira and Teófilo (2000).

17

strong authoritarianism, centralization, and bureaucratic

paternalism, the Cédula da Terra is a decentralized program.

It establishes general criteria for the asset redistribution

process in a determined region and provides special loans for

beneficiaries' own initiatives. There is a limit on the price of

land acquisition and on the total loan. The decision regarding

land selection, the negotiation of land acquisition, and the

definition of agricultural activities to be implemented are left

to the beneficiaries themselves.

(ii) The program is founded on the idea of self-selection of

beneficiaries. In other words, the program does not select

participants, but merely defines the basic characteristics of

potential beneficiaries and the access conditions. Then, those

interested in getting into the program are expected to seek it

out, being served on a first-come first-served basis.

(iii) Participation in the program is on an associative, not

individual, basis, as only associations of farmers can get

credit.

(iv) Land is not distributed, but sold through an operation of

agrarian credit (complemented by other lines of credit)

contracted by an association and the program’s financial

agent. The cost of the loans is subject to special discounts but

follows market evolution. Following a 4-year grace period, the

loan must be paid yearly, under the penalty of losing the land.

It is expected that the obligation of payment of the land will

create incentives for production while associative initiatives

will reduce the financial institutions’ monitoring costs and

credit restrictions.

(v) The associations of beneficiaries have the autonomy to make

decisions on both how to use the financial resources and on

the specific productive strategy they will follow. They also have

to decide how the land will be distributed among the families

18

and how the common land and individual parcels will be used.

The association assumes the financial obligations, which are,

in fact, a mutual responsibility of its members.

The aforementioned characteristics of the program are expected to

generate incentives and conditions for an efficient and sustainable

allocation of resources. Specifically, the selection of land and of

beneficiaries could be improved with respect to the traditional

approach.

The works of Buainain et al (1988;1999) and Silveira et al (1999)

sustain that the performance of the program is strongly conditioned by

its structure of governance, whose mechanisms affect the selection of

beneficiaries, choice of the property, investments, etc. The authors

sustain that, in conceptual terms, the governance structure of the

program defines mechanisms and incentives which are adequate to

avoid some of the traditional problems faced by family producers , as

well as to make it possible to fully use the productive potential as well

as any eventual competitive advantages associated with the family

production. For example, the hypothesis that the beneficiaries will

choose adequate properties for sustainable exploitation can be derived

directly from the program’s governance structure. As the beneficiaries

must pay for the land, it is reasonable to suppose that they will look for

the best available option, and that they will only buy properties whose

productive potential is compatible with the generation of enough income

to pay the loan. Waiving the obligation of payment would change the

governance structure, and consequently, many of the hypotheses about

the agents' presumed behavior and about the dynamics of the projects

and of the program17. If the beneficiaries do not believe that the

payment of land will be demanded and the beneficiaries are betting that

their default won't have any consequences, the ex-ante consistency of

the program and of its governance structure will fall apart. The 17 One of the advisors of the “Revista Economia” has brought up the hypothesis that the public financing of land does, in practice, relieve the beneficiaries from future payment. This is, no doubt, a real possibility, and it is based on the vast experience of the re-negotiation of debts by large and small scale producers.

19

preliminary research done by Buainain et al. (1999) didn't reveal any

signs that the associations were betting on the hypothesis of

default.18Nevertheless, the beneficiaries would hardly reveal the

intention of not paying, in case it existed.

In this sense, Buainain, Silveira and Teófilo (2000) call attention

to the fact that the governance structure is sensitive to the real

conditions of implementation of the program, which, at their limits,

could produce governance structures that draw themselves away from

its concept.

In this article we are concerned only with the selection of

beneficiaries.

The question is: do decentralized mechanisms of self-selection of

beneficiaries in rural poverty alleviation programs work? Is the set of

characteristics of the beneficiaries of the Cédula da Terra program

seemingly consistent with its goals?

Under the traditional model, it is difficult to know the true

capacity, interest, and disposition of a candidate beforehand. The

inefficiency of administration selection processes has been pointed out,

even in projects that outline rigorous and detailed criteria for the choice

of beneficiaries, such as those of colonization and irrigation.

Selection of the target population in programs of this nature

presents a clear trade-off: rigid criteria for the identification raise costs,

but would, in theory, later reduce monitoring costs. In conditions in

which the empowerment process of local communities is not well

developed, this process of targeting is subject to bureaucratic

interference and the politics of central and local power. A distortion in

the selection of the target public can occur, since the people who really

need to raise their initial endowment, are precisely the least capable of

getting into the program.

18 This hypothesis (of associations betting on default) should be studied carefully during the pilot phase of the program but, as of now, it is possible to affirm that the program’s credibility will only be put to the test after the period of grace has expired.

20

These difficulties indicate that the self-selection approach adopted

by the Cédula program may offer advantages. It is possible to reduce

both identification and post-selection monitoring costs; eventual errors

in selection are also likely to be reduced, as the community must

approve candidates before they can join the association. It is therefore

reasonable to expect that the self-selection mechanism of the Cédula da

Terra program will make it possible to overcome some of the difficulties

associated with the government selection of beneficiaries in agrarian

reform.

While the program is open to all families that fit into the selection

criteria, actual participation is subject to a set of rules and restrictions,

such as being a member of a legally formed association, the purchase of

the land instead of their being settled by the government on

expropriated land, the payment of the loan used to purchase the land

and so on.

These criteria are bound to attract, in the majority of cases,

candidates whose profiles are more suited to taking advantage of the

opportunities created by the program and are willing and able to fulfill

the obligations.

On the other hand, the selection mechanism has been designed to

reduce eventual selection distortions. The definition of a rather

restrictive financing ceiling that establishes the credit that each family

can get is expected to prevent non-poor participation. Adhesion through

an association should also be a powerful instrument to avoid both

leakages and selection errors associated with insufficient information.

A preliminary evaluation carried out by Buainain et al. (1999a)

has revealed that true self-selection does not work at the very first

phase of the implementation of the program. In spite of state

government efforts to inform the target population about the program,

in many cases potential beneficiaries were actually encouraged by

government officials to join the program. Their decision to self-select

themselves was, at least to some extent, induced by program exogenous

21

stimuli, stimuli these which may not be necessarily in line with their

own assessment of the opportunities the program is likely to raise nor

with their potential to take advantage of such opportunities and to fulfill

preset conditions.

In other cases, especially when demand has exceeded short run

"vacancies", due to unexpected adhesion to the program, state agencies

have intervened in the selection criteria and arbitrated in favor of the

poorest. In one state program, operations were concentrated in

predefined priority districts. Both the extent of intervention and the

specific selection criteria have varied from state to state. Buainain et

alii (2000) strongly emphasize that such intervention was both desirable

and positive at the initial stage of the program implementation. But they

also remark that the self-selection mechanisms have not been fully

operational under the conditions that have shaped the implementation

of the program during its first year.

One must return, at this point, to the aim of the program: does

the self-selection based on the associations of beneficiaries really work?

Some points that follow can help us to focus adequately on the

program’s intentions, from an empirical point of view:

a) As has already been stated, the selection process of the

beneficiaries is conditioned to a set of governance mechanisms

that make up the governance structure that is implicit in the

design of the program. From this frame of reference, it is

important to note that a misguided (self) selection of the

communities and of the beneficiaries may jeopardize the

program’s performance, mainly with respect to its aim of

reducing poverty;

b) The selection process best fits into the category of “narrow

targeting”. Using the population of potential beneficiaries in

eligible communities as a point of reference, the process tries

to determine to what extent the true beneficiaries have unique

characteristics which set them apart from the others, and to

22

evaluate how these characteristics match the expectations and

can contribute to the program’s success. As will be seen in the

following pages, the results obtained in section 3 show that the

choice of methodology (which was obviously made ex-ante) was

correct in how it analyzed the selection process in light of the

existing tug among the program’s aims, which are: on one

hand, to fulfill the generic objective of being an instrument in

the reduction of poverty (among other objectives) and, on the

other hand, to maximize growth in the communities, through

the adequate distribution of land property rights to self-

selected poor families.19

c) It is difficult to separate the effects of each of the governance

mechanisms on the selection process and on the dynamics of

the program. The analysis taken up was based on a

representative sampling of the beneficiaries who have just

adhered (initial phase). This means that self-selection is based

on “pure” expectations with regards to the possibilities that

the program will create, possibilities these which have not yet

passed through reality and its filtering effects nor have been

modified by the program’s dynamics during its implementation

and the evolution of the projects.

d) Finally, it also indicates that the analysis has not yet taken

into account that the selection process of beneficiaries has

passed through the selection of communities and that, as has

been show by Buainain et al. (1999), in this initial phase, some

communities have in fact been pre-selected and stimulated to 19 In this respect, one has avoided hastily putting together theories about the individuals’ behavior in light of the proposed governance mechanisms. In this phase of research there would be no way to check if an agent who had demonstrated interest in the program would not have the human resources needed to become a producer, let alone to imagine that a beneficiary, acting in a community-based manner, would adhere to a program merely to gain advantages in terms of land ownership, an act which could be defined as opportunistic behavior. [Of course, if the characteristics of the behavior show that they are rural poor with some level of capital... FALTA ALGO PARA COMPLETAR A IDÉIA! O texto original, contém uma condição, mas a idéia ficou incompleta...]

23

participate in the program.

3. EMPIRICAL EVALUATION OF THE CÉDULA DA TERRA SELECTION PROCESS

Data concerning the beneficiaries and non-beneficiaries has

allowed for an empirical analysis of the selection. Descriptive statistics

of socioeconomic variables and a logit model are provided. The results

obtained are useful for indicating to what extension the process of

defining a target public (targeting) has been compatible to the program’s

goals. The results also indicate necessary corrections.

The data of this study was obtained from two sources: a special

survey, which was conducted among beneficiaries during February,

1999, and the 1997 National Household Sample Survey (PNAD).

3.1 DESCRIPTIVE STATISTICS OF BENEFICIARIES

Descriptive statistics of beneficiaries have been built using data

from a special survey, which was conducted among beneficiaries during

February 1999. This survey has yielded information on 222

representative beneficiaries from the 5 States where the program has

been carried out.

The age structure of the beneficiaries is consistent with the

expected profile for a program like the Cédula da Terra, which requires

time and dedication for its full maturation (Table A). The average age of

the beneficiaries was 39.4 years old (between 37.6 and 41.2, a result

having a 90% factor of reliability). Most of them (56.3%) were young,

ranging from 18 to 40 years old (2.7% being up to 21 years old and

53.6% being between 22 and 40). Only 3.8% were over 61. The families

were also young; this fact is confirmed by the high percentage of

children under 14 (37.8%) in the total number of the family members.

The percentage rises to 51.8% if 15- to 20-year-old members are

included.20

20 Navarro (1998) considers that the feasibility of the program requires the intense participation of young farmers. He raised the criticism that the program was not attracting young families. The evidence produced in this report does not back up such criticism.

24

Table A

Beneficiaries' Demographic Indicators

Demographic AspectsAverage Age (years) 39.4 1.1 37.6 41.2Men (%) 88.2 2.9 83.3 93.0Women (%) 11.8 2.9 7.0 16.7

AgeUp to 21 years old (%) 2.7 (2) (2) (2)

From 22 to 40 years old (%) 53.6 4.5 46.3 60.9From 41 to 60 years old (%) 39.9 4.3 32.8 47.061 years old or older (%) 3.8 (2) (2) (2)

(1) Interval with a reliability factor of 90%.(2) The number of observations did not reach the interval established to guarantee reliability

tototoguguaranteetotoreliabilitythe confidence interval determination.

Range ofReliability (1)

Indicators Value StandardError Inferior

LimitSuperior

Limit

The level of schooling of beneficiaries is relatively low: 31.7% of

the sampled population was illiterate (Table B). If we add those who can

only read and write (which represent 4.5%) to the figure, we have a

group of 36.2% of the total. The most numerous group (47.1%) attended

the first years of school (1st to 4th grade). Few beneficiaries (3 percent)

had higher education.

Table B

Beneficiaries' Educational Level

Educational LevelIlliterate (%) 31.7 4.0 25.2 38.2Read and Write (%) 4.5 (2) (2) (2)

Preschool (%) 0.7 (2) (2) (2)

1o to 4 o year (%) 47.1 4.5 39.7 54.55o to 8o year (%) 13.1 3.4 7.5 18.7College or Superior Education (%) 3.0 (2) (2) (2)

(1) Interval with a reliability factor of 90%.(2) The number of observations did not reach the interval established to guarantee reliability

Range ofReliability (1)

Indicators Value StandardError Inferior

LimitSuperior

Limit

The survey collected information regarding the occurrence of any

chronic diseases and their consequent damage to the beneficiary’s

25

capacity to work. An insignificant number of beneficiaries indicated the

presence of chronic diseases that had affected their capacity to work in

the previous year. This answer must be analyzed with great care, since

some beneficiaries could have omitted the existence of some disease.

Most occurrences mentioned were related to family members, especially

the wives and young children. The main diseases mentioned were

associated with circulatory problems, backaches, allergies and

bronchitis.

We have made an attempt to trace a detailed profile of the

beneficiaries’ occupations before getting into the program, as well as

their occupational background and history of migrations. Beneficiaries'

previous occupations and migration history are relevant characteristics

to differentiate beneficiaries from non-beneficiaries, as will be seen in

Section 4.2. The main results are summarized in Table C.

Table C

Previous Occupations of the Cédula da Terra Beneficiaries

Owner (%) 7.1 2.2 3.5 10.8Non owner (%) 54.2 3.6 48.3 60.0Extraction of natural resources (%) 7.1 1.8 4.2 10.0Qualified Activities (%) 6.7 2.0 3.3 10.0Day-worker (%) 55.5 3.9 49.0 62.0Emergency Front (%) 5.9 1.5 3.5 8.4Non-Agricultural Occupations (%) 11.9 3.4 6.4 17.5Merchant (%) 3.9 (2) (2) (2)

House keeper (%) 2.0 (2) (2) (2)

Special Occupations (%) 2.5 (2) (2) (2)

Without Occupation (%) 0.5 (2) (2) (2)

Beneficiary’s workplaceRural Zone (%) 88.9 2.7 84.5 93.2Urban Zone (%) 8.5 2.3 4.7 12.3Rural and urban Zone (%) 2.6 (2) (2) (2)

(1) Interval with a reliability factor of 90%.(2) The number of observations did not reach the interval established to guarantee reliability.

Beneficiary’s occupation beforeadhering to the program

Range ofreliability (1)Standard

Error SuperiorLimit

Indicators ValueInferior

Limit

Almost 90% of the beneficiaries had had rural work, 8.5% urban

26

work and 2.6% had taken up both occupations. Since this information

refers to the last occupation before getting into the program, it is not

possible to deduce that those who had declared urban work had not

had experience with agriculture. However, it is possible to indicate that

many beneficiaries were forced to look for occupations in urban areas

due to the long period of drought that devastated the region throughout

1998.

The majority of beneficiaries had had more than one occupation.

The most frequent ones were day-workers, rural producers who worked

on small parcels of their own land, leased land, or some kind of

partnership. Many producers, who were working on their own land,

were actually working on their father’s or father-in-law’s land, or on

inherited land without the proper legal ownership. The most common

tenure arrangements on dry areas were of cession of land in exchange

for planted grazing or palm cactus; it was common to share output in

more humid areas. A few cases of fixed payment, in cash, were

observed.

It is estimated that 7.1% of the program’s beneficiaries were

landowners. Most of them were smallholders and minifundistas who

worked as day-workers, partners and tenants on someone else’s land,

and even in emergency fronts. Only 2.7% had not had any other

occupation. The sample registered two retailers: one had owned a bar

and the other had sold cereals.

The group of non-owners comprised 54.2% of the sample, while

day-workers were 55.5%. Earnings of the former group ranged from

R$4.50 to R$5.50 per day. It is important to note that nobody had

declared to have work all days of the week. The modal answer was 3

days a week in “good” months.

Approximately 12% of the sampled population had had non-

agricultural activities, which included the activities of bricklayer, tailor,

cabinet-maker, barber, vendor, brick maker, watchman and tinker.

Some individuals of this group were working in rural areas, while others

27

were working in urban ones.

The group of "highly qualified workers" (tractor driver, driver,

herdsman, sawyer) was around 6.7% of the sample. They had worked as

non-owner farmers and daily rural workers. A few of them had migrated

to other areas, especially to urban areas. Very few had ever had any

experience with any kind of rural enterprise (or even urban) before

getting into the program. Most of them had come from rural milieu,

their parents were farmers in the same locality or nearby areas; few

mentioned any other professional experience apart from rural

worker/producer before getting into the program. In these cases, the

following occupations were mentioned: bricklayer, watchman, public

official, mechanic help, truck driver and handyman.

We have made an attempt to examine the source of income of the

beneficiaries before getting into the program. The difficulties are

innumerous: the beneficiaries did not maintain any bookkeeping record;

most families were in transition from the prior situation to that of

beneficiary of the program; there were many cases of engagement in

occasional activities; and the year of 1998 could be characterized as

atypical, as production and income had been reduced due to harvest

losses or inadequate weather conditions to sow. Also, in 1998, workers

had difficulties in finding occupation, having to rely on emergency

fronts.

Only monetary gross income could be calculated, as data on input

expenditures was not available. Subsistence production and non-

declared or non-evident marketed production were not taken into

account in the estimation of income. The standard food staples provided

by government aid have been converted into monetary value and

accounted for in their income. This procedure allowed us to identify the

role of donations in the survival of this population during years of

drought.

28

Table D

Beneficiaries' Income in 1998

A. Beneficiary's Income (R$) 958.64 118.14 764.89 1152.39

A1. Off-farm Labor Income (R$) 478.10 74.65 355.67 600.52

A2. Agric./Livestock Activities' Total Income (R$) 340.17 78.03 212.20 468.13Back yard Production (R$) 34.28 11.91 14.76 53.81Animal Production (R$) 9.94 4.81 2.06 17.83Animal hiring (R$) ((n.a.) (n.a.) ((n.a. ((n.a.

Vegetal Production (R$) 221.82 66.82 112.24 331.40

Other Properties and Leased Land (R$) 74.12 39.78 8.88 139.36

A3. Financial Income (R$) 16.81 17.96 0.00 46.26

A4. Other Incomes (R$) 93.75 34.42 37.30 150.19

A5. Beneficiary Pension (R$) 29.82 15.96 3.64 55.99

B. Consort Income (R$) 207.38 48.91 127.17 287.59

C. Other family members’ income(R$)

891.80 106.63 716.93 1066.67

C1. Resident members’ off-farm income (R$) 245.34 68.83 132.47 358.22

C2. Family Members' Pensions (R$) 145.40 56.93 52.04 238.76

C3. Other aid received by family (R$) 501.06 61.67 399.92 602.19Remittance from non resident familiar (R$) 79.21 33.56 24.17 134.26Donations received by family (R$) 62.68 12.93 41.48 83.88Social Program Aid (R$) 359.16 47.26 281.65 436.68

D. Family Income I [A + B] (R$) 1166.02 125.03 960.96 1371.08

E. Family Income II [A + B + C1 + C2] (R$) 1556.76 156.33 1300.38 1813.14

F. Family Income III [A + B + C] (R$) 2057.82 159.75 1795.82 2319.81(1) Interval with a reliability factor of 90%.(v.a.) Value not available

Range ofreliability (1)Average

ValueStandard

Error InferiorLimit

SuperiorLimit

Indicators

The beneficiary’s average income was of R$ 958.64 for the year of

1998, comprising: income generated by work outside the holding and

agricultural activities; financial income; pensions; and other income

obtained through the sale of non-agricultural products and services,

commercial activities and machines and equipment hiring (Table D). If

we take the upper limit of the range of reliability in this average, we

have an average monthly income of R$ 92.00, which is equivalent to

73% of the national minimum wage.

Approximately, half (between 43% and 52%) of the beneficiary’s

income came from off-farm work. Income from agricultural activities

was the second component in order of importance, making up from 27%

to 40%. Taking the upper limit as a basis for calculation, we would have

29

a monthly income of only R$ 39.00, or less than 1/3 of the minimum

wage. This contribution does not represent the typical income from

agricultural activities for these people, especially in areas of the states

of Ceará, Pernambuco and Bahia, which were strongly devastated by a

long period of drought.

Considering all sources of income indicated in Table D, we

estimate an average family income of R$ 2,057.00, in 1998. The average

monthly family income was R$171.40. Considering that the average

number of resident persons in each family is 5.2, a per capita income of

R$ 32.90 is obtained. Taking the upper limit, the average annual

income would be R$ 2,319.81, with a monthly income of R$ 193.00 per

family and a monthly per capita income of R$ 39.00. Considering these

values, there is no doubt that the beneficiary population is poor, finding

themselves below the poverty line for these states; even if we take into

consideration the likely underestimation of income.

However, it is important to evaluate whether poverty is a

characteristic of the beneficiaries or a prevailing characteristic of the

regions where the program has been implemented. One of the criteria

for eligibility is low income. To what extent is the concern to attend to

the poor, by defining eligibility criteria, not just a trivial fact that arises

from regional conditions? This question will be further assessed below

through the comparative analysis of the socioeconomic indicators of the

Cédula beneficiaries and of the potential beneficiaries of the program.

3.2 ASSESSING THE SELECTION PROCESS THROUGH A LOGIT MODEL

A logit model was used to examine the differences between

beneficiaries and non-beneficiaries, thus the targeting process could be

evaluated through the socioeconomic characteristics of both categories.

The dependent variable is a dichotomous one, which indicates the

household group.21 It assumes a value of 1 if the household is a

beneficiary of the Program and 0 if it is not. In this binary logit model,

the dependent variable, say y, is expressed by a set of independent 21 See Maddala (1986) for a review on limited dependent variable models.

30

variables, which are used as proxies to test for economic and social

determinants (see definitions in Table E).

Table E

Definitions of independent variables

GENDER Dummy variable indicating gender. It assumes a value of 1 if the head ofthe household is a woman and 0 otherwise.

MIGRATION1 Dummy variable indicating how long the head of the household hascontinuously lived in the municipality. It assumes a value of 1 if he/shehas continuously lived in the municipality for up to 4 years, 2 for up to 9years, 3 for up to 10 years or more, and 4 if he/she has lived in themunicipality since he/she was born.

MIGRATION2 Dummy variable indicating how long the head of the household hascontinuously lived in the state. It assumes a value of 1 if he/she hascontinuously lived in the state for up to 4 years, 2 for up to 9 years, 3 forup to10 years or more, and 4 if it has been since he/she was born.

INCOME Household total income (PCT income benefits were taken out).

ADULTS Number of members of the household aged between 14 and 60 years old.

CHILDREN Variable indicating the proportion of school-age children in the totalnumber of household members. It is the number of children up to 14 yearsold divided by the total number of household members.

GOODS Dummy variable indicating stock of durable household goods. It assumes avalue of 1 if the household was supplied with a set of durable goods(television, fridge and oven) and 0 otherwise.

ASSETS Dummy variable indicating stock of assets. It assumes a value of 1 if thehousehold possessed a house, real estate, or land; and 0 otherwise.

EDUCATION Dummy variable indicating schooling. It assumes a value of 1 if the head ofthe household is illiterate; 2 if he/she has 1 to 3 years of schooling; 3 ifhe/she has 4 to 7 years of schooling; 4 if he/she has 8 to 10 years ofschooling; and 5 if he/she has 11 years or more of schooling.

VILLAGE Dummy variable indicating whether a household is located in a ruralvillage. It assumes a value of 1 if it is located in a rural village and 0otherwise.

The data set used in this analysis comprises information

regarding the beneficiaries (the special survey sub-sample) and non-

beneficiaries (PNAD sub-sample). From the PNAD data set, we have

selected 36,337 households in the 5 States where the special survey on

beneficiaries took place. From these, we have selected 3,413 households

matching the following characteristics: (a) the household head is ≥ 18

years old and ≤ 60 years old; (b) he/she has an agricultural occupation;

and (c) the household income was ≤ R$240,00 per month in September,

1997. These are characteristics of the target population previously

31

defined by the program. Missing information has also been cut-off,

resulting in a full set of 3,583 valid observations (200 beneficiaries and

3,383 observations from PNAD).

As the two sets of data (regarding non-beneficiaries of the Cédula

da Terra and beneficiaries) were obtained from different and

independent surveys, they cannot be used without some precautions.

Data obtained from different sources is not entirely comparable, mainly

due to differences in the definition of some variables. Any statistical

analysis that does not take these differences into account would most

likely produce biased estimators and, therefore, distorted inferences. To

minimize this problem, the analysis carried out in this study has only

used data available in one sub-sample that matches the correspondent

data available in the other sub-sample. Even more, data definitions and

data collection instructions are alike in both PNAD and the Cédula da

Terra surveys.22 For example, for the variable ASSETS, information

about the ownership of a house, real estate and land is available in both

sub-samples. For this reason, only these kinds of assets were

considered, as is clearly made known in the definition of the variable.

See Appendix for the data transformation procedures used).

The variable INCOME was one of major concern. PNAD income is

exclusively based on wage revenues.23 For this reason, the variable

INCOME was constrained by this PNAD shortcoming and revenues from

government transfers, private transfers, as well as capital and property

incomes were not considered in the estimation of Cédula da Terra

beneficiaries’ income. Earnings from non-family members of the

household were also excluded from the total income of PCT households,

as they were not included in the PNAD sub-sample.

Another source of potential bias was the method we used to arrive

at the annual income of PNAD households. PNAD data on income refers 22 This is not a mere coincidence. The questionnaire of the Cédula da Terra evaluation study was designed to ensure as much compatibility as possible with the PNAD and IBGE Agricultural Census. 23 For a discussion on the use of PNAD data in income see FERREIRA, LANJOUW E NERI (2000).

32

to the income earned in September 1997, while PCT data refers to the

income earned in the whole year of 1998. Thus, PNAD annual income

was estimated on the assumption that the September earnings were the

same throughout the year. We are aware of the problems that can arise

from these transformations. However, as data availability is a major

constraint, we found that this exercise could provide some inferences,

even if we could not rely too much on the results provided by the

variable INCOME.

The set of variables has enabled us to evaluate how economic and

social characteristics are related to the program’s selection process, and

how these characteristics can affect the program’s success.

As we are interested in evaluating the selection process, all data

concerning beneficiaries refers to their condition before getting into the

program. For example, the variable ASSETS refers to their possessions

before the acquisition of land through the program’s line of credit, as all

of them have come to own land after getting into the program.

The results obtained from the fitting of the logit model are

presented in Table F.24 Independent variables were controlled to avoid

bivariate correlation. The likelihood ratio (LR) was used to test the

hypothesis that all coefficients are zero. The restricted log likelihood

value is –771.4389 and the unrestricted one is –650.0815. The LR test

statistic is therefore 241.7148. With 10 degrees of freedom, the critical

value at 5% significance level is 18.31, and so the joint hypothesis that

the coefficients in the full set of variables are all zero is rejected.

24 LIMDEP Software Package was used for estimation.

33

Table F:

Logit Model — Coefficients and Marginal Effects —

Variables (1) Coefficients P[ Z >= z] MarginalEffects P[ Z >= z]

CONSTANT

-3.8989 0.00000 -0.12795 0.00000MIGRATION1

-0.51246 0.00000 -0.16818E-01

0.00000MIGRATION2

0.18019 0.08884 0.59135E-02

0.08870VILLAGE

1.3850 0.00000 0.45454E-01

0.00000INCOME

-0.93341E-04

0.25616 -0.30633E-05

0.25732CHILDREN

1.3871 0.00022 0.45523E-01

0.00016ADULTS

0.47304 0.00000 0.15524E-01

0.00000GOODS

0.88225 0.00003 0.28954E-01

0.00004ASSETS

-1.0393 0.00000 -0.34107E-01

0.00000EDUCATION

0.36885 0.00005 0.12105E-01

0.00005GENDER

0.88639E-01

0.75665 0.29090E-02

0.75697

Log-Likelihood Restr. (slopes=0) Log-L -650.0815 -771.4389

(1) Marginal effects are computed at the means of thevariables.

The model assumes and tests some propositions derived from the

declared intentions of the program’s formulators, which were presented

in section 3.

Let us start with the aspects related to the beneficiaries’ past

history. The governance structure of the program was set to select local

people, who would supposedly have closer relations with landowners,

better access to networks of social relations and information on the

local market of land.25 Other desirable characteristics of beneficiaries

are "life experience" and leadership, which make them more capable of

handling challenging circumstances.

Three variables were used to check this: MIGRATION1,

MIGRATION2 and VILLAGE. It is assumed that an individual obtains

life experience through migration.26 MIGRATION1 indicates whether

they are ‘locals’ (local residents), according to the length of time they

25 It should be stressed that the region is characterized by unbalanced land distribution and marked co-existence of large farms and smallholdings. 26 Menezes (1997) has shown that Northeastern rural people, who have migrated to the Southeastern region, have acquired experience in a unionized labor market. Some of them have returned after some years and have become rural worker leaders in the Northeastern region.

34

have lived in one municipality. It has also allowed us to examine

whether they have migrated around (within the state). MIGRATION2

indicates whether they are ‘locals’ at a state level and have migrated

nationally. The sign of the coefficient of MIGRATION1 is negative, which

means that, an increase in the period lived in a municipality decreases

the probability of the individual being in the group of beneficiaries. A

positive sign was found for the coefficient of MIGRATION2, which means

that, an increase in the period lived in one state increases the

probability of the individual being in the group of beneficiaries. These

two variables indicate that the program is selecting people who are

‘locals’ at a state level, but have migrated around (within the state).

Both characteristics are desirable since they are assumed to have better

access to information regarding the regional land market, and are more

integrated to the local social networks than outsiders. The fact that they

have migrated, if even over short distances, may be taken as an

indication of "life experience".

The variable VILLAGE portrays two possible inferences. First,

inhabitants of rural villages are presumed to have better access to

information and more "life experience" than their counterparts who live

on farms; at least within the income level constraint imposed in the

sample.27 The positive sign and high significance of VILLAGE confirms

that the program tends to select the most experienced people among

those who are constrained at the income level of up to R$240.00 per

month.28 Second, inhabitants of rural villages, at this income level

constraint, are generally rural workers and landless producers; that is,

they are the target population of the program. The positive sign of the

coefficient of VILLAGE confirms that the selection mechanism is

working in this direction.

27 It should be stressed that the sub-sample of non-beneficiaries does not comprise individuals whose income is greater than R$240.00 per month, which is the maximum allowed for candidates who wish to apply for the program’s benefits. 28 Thus, the result does not mean they are the most experienced among the whole rural population. Individuals whose income level is greater than that limit could have more experience, but they would not be qualified to apply. This hypothesis has not been tested in this model.

35

The next step is to examine whether a beneficiary’s previous

income has affected his/her decision to get into the program. The

coefficient of the variable INCOME presented a low level of significance,

which means that the variable cannot be used to distinguish

beneficiaries from non-beneficiaries in this model. In fact, one can say

that the sample as a whole is truncated at a low level of income, given

that an income limit (R$240.00 per month) has been imposed on the

sub-sample of non-beneficiaries, which is the same imposed by the

selection criterion of the program. Thus, one could expect that INCOME

would not be a good predictor, since the range that was established for

variation is narrow and is at the bottom of the social pyramid. If the

selected beneficiaries had greater income than the one set up by the

program’s selection rules, a possible result in this model would be a

positive and highly significant coefficient. As this is not the model

result, one can say that the income selection rule of the program has

been satisfactorily attained.

One question that can be brought into the analysis is whether the

beneficiaries and associates have perceived a desirable family profile.

The number of members and composition of the household must be

coherent with the double objective of poverty alleviation and project

sustainability. Local governments were adopting ‘number of family

members’ as a criterion for selection. The reason for this is to benefit a

large number of people. Apart from social goals, this selection criterion

has some effects on future production achievements in the household.

On one hand, a large number of adults could mean the household has

enough supply of workforce to succeed. On the other hand, families

with a high percentage of school-age children would have a low supply

of workforce and a high level of expenditure with non-productive

individuals.29

Two variables were used in the model to consider the matter:

29 It should be stressed that a president of a local producers association mentioned that a high number of school-age children has been used as a positive criterion in the selection process, given the social aim of the program.

36

CHILDREN and ADULTS. The coefficients of both were found positive

and significant. The conclusion drawn from this is that beneficiary

families have a larger number of adult members, which is positive for

their success, but also have a higher proportion of children. The latter

could restrain investments during the first years of existence of the

settlement, which is a period when capital must be built to increase

land productivity. Failure to do so could jeopardize the program.

However, in the future, when children become adults, households will

have a good supply of workforce.

We were interested in checking whether the program had selected

people who were well off in terms of possessions, productive or not,

among the target population. Unfortunately, the PNAD lacks supportive

information in the matter. However, the coefficient of two variables

(GOODS and ASSETS) could be tested and some inferences were taken.

The negative sign of the coefficient of ASSETS indicates that

beneficiaries were less endowed in terms of a stock of assets (a house,

real estate, or land). The positive sign of the coefficient of GOODS,

however, indicates they were not at the extreme bottom below the

poverty line. These results indicate that the program had selected

people who had not been able to accumulate assets, but had had

enough income, or donations/gifts, to possess a set of household

durable goods. The fact that they tend to live in rural villages

contributes to explain this. They have had easier access to power

supply and they have been more exposed to a modern way of life. Also,

as will be shown below, a better level of schooling influences their

standard of living and consumption, which is based on a greater range

of durable and non-durable goods. For example, a refrigerator allows for

diversification in food consumption, and a TV set allows for better

access to information and the inducement of consumption of new

goods.

A desirable characteristic of the beneficiaries is a better level of

schooling, as it is normally related to efficient agricultural exploitation.

Literature has shown that a high level of education is correlated with

37

better access to information and social integration (see Souza

Filho (1997) for a review). These are expected characteristics of the

program’s beneficiaries (being associated is a condition for participating

in the program). The high significance of the coefficient of EDUCATION

confirms this hypothesis. Also, it shows that the governance structure

of the program has been able to select people who are in better

conditions to succeed.

Finally, the model tested whether the selection process is biased

with respect to gender. The low significance of the coefficient of

GENDER shows that this is not the case.

4. CONCLUSIONS The results obtained from descriptive statistics as well as from the

logit model confirm that the governance structure of the program has

indeed shaped a distinctive group of “self-selected” beneficiaries.

Beneficiaries are poor, but not extremely poor. They have larger families

and a higher percentage of children in their families than potential

beneficiaries (from the PNAD sample) do. They live in rural villages and

work mostly as day paid rural workers and landless small farmers,

typically as both.

An analysis of the beneficiaries’ profile indicates that they are

poor, but not discriminated or excluded from local economy. They have

managed to survive in extremely inadequate living conditions. This is a

characteristic that differentiates them from typical Northeastern

emigrants who have been forced, by natural, social and economic

conditions, to search for alternative occupations in other regions. The

logit model has confirmed this. The result indicates that the

beneficiaries’ likely migration pattern is consistent with individuals that

migrate within the state searching for occupation, but do not move

definitely away from their communities.

The coherence of the selection process, confirmed by the analysis

above, can be taken as initial evidence of the relevance of the program’s

conception and its decentralized selection procedures in assigning land

38

property rights to the group. In fact, the analysis of the selection

process has shown that beneficiaries are indeed rural poor with some

experience and potential to exploit agricultural activities.

The analysis also indicates that beneficiaries have a low level of

education, although it is greater than the level of the PNAD group. This

is a consequence of selecting the poor, and a matter of great concern. In

fact, the program has selected the most qualified individuals among the

poor, which is a great achievement. However, they still have a low level

of education, which suggests that educational policies and technical

assistance should be considered in order to increase the probability of

success of the program.

As for our final words, we would like to stress that this article

provides an analysis on the selection process only. We have not

addressed a number of relevant issues that could jeopardize the Cédula

da Terra Program. The difficulty in enforcing contracts, a common

problem of rural credits, is one of these issues. Where the political cost

of foreclosing debtors is high, incentive to repayment is reduced and the

risk of default increases. This is a particularly relevant issue for the

Cédula da Terra Program and should be further investigated.

5. APPENDIX Selection of observations and data transformation to make PNAD

data comparable to PCT data

39

6. Selection/Transformation Variables (PNDA code)1 Summation of family income per household V09053162 Selection of rural households in which household

income was≤ R$240,00 per monthV0905316, V4105

3 Selection of households in which the head was 18 to 60years old

V03034, V0402

4 Selection of households in which the head hadagricultural occupation

V090071, V090072, V090062,V090073, V090074, V090063

5 Addition of non-resident sons and daughters toresident members of the household

V11051, V11052, V0401,V11041, V11042

6 Family members, except pensioners and employees V0401, V030347 Changing the range of values of dummy variables (with

recodification)8 Grouping categories to obtain a new range of values of

dummy variables (with recodification)9 Merging variables (with recodification for dummy

variables)

V0203, V0204, V0207,V0217, V0218, V0219,V0221, V0223,V0225,V0226, V0227,V0228,V0401, V0501, V0502,V0504, V05061, V05063,V05065, V0511, V05121,V05123, V05125, V0601,V0602, V0607, V06073,V0610, V1101, V4105

40

REFERENCES BAKER, J. L.; GROSH, M. Poverty Reduction through Geographical Targeting:

Does it Work? World Development. 22(7), p. 983-94, 1994.

BARDHAN, P.; UDRY, C. Development Microeconomics. 1.ed. Oxford: Oxford University Press, 1999, 242p.

BESLEY, T. Targeting Taxes and Transfer: Administrative Costs and Policy Design in Developing Countries. In: HOFF, BRAVEMAN, STIGLITZ (orgs.) The Economics of Rural Organization. 1.ed. World Bank Book, 590p., 1993.

BISWANGER, H.; DEININGER, K.; FEDER, G. Poder, Distorções, Revolta e Reforma nas Relações de Terras Agrícolas. 1999. Available from World Wide Web: <http://www.dataterra.org.br.>, 109p. (Mimeo).

BUAINAIN, A. Trajetória Recente da Política Agrícola Brasileira: da intervenção planejada à intervenção caótica. Campinas: Instituto de Economia/UNICAMP, 1999. 325p. (Tese, Doutorado em Economia). (Mimeo.).

BUAINAIN, A. et al. (1998). “Metodologia de Avaliação de Impactos Sócio-Econômicos”. Relatório Final. Campinas, IE/UNICAMP, 233p. (Mimeo.).

BUAINAIN, A. et al. Avaliação Preliminar do Programa Cédula da Terra. Brasília, NEAD/World Bank, 1999a, 320p. (Mimeo.).

BUAINAIN, A. et al. Community-Based Land Reform Implementation in Brazil: A new way of reaching out the marginalized ? Paper presented in Bonn Seminar/ Global Development Network. Bonn, November, downloadable <htpp://www.gdnet.org.>, 106p. 1999b.

BUAINAIN, A.; SILVEIRA, J. M. F. J. DA; SOUZA, H.; MAGALHÃES, M. (2000). Programa Cédula da Terra: resultados preliminares, desafios e obstáculos. Núcleo de Estudos Agrários, 2000, 113p. (Mimeo.).

BUAINAIN, A.; SILVEIRA, J. M. F. J.; TEÓFILO, E. O Programa Cédula da Terra no contexto das novas políticas de reforma agrária, desenvolvimento e participação: uma discussão das transformações necessárias e possíveis. In: MDA/NEAD, Reforma Agrária e Desenvolvimento Sustentável. 1.ed. Brasília: Ed. Paralelo 15, 380p. pp.157-175, 2000.

DAVID, M. et al. Demandas por Políticas de Desenvolvimento Rural no Brasil. In: ENCONTRO NACIONAL DE ECONOMIA, 1999. Anais do XXVII...: ANPEC, 1999. p. 183-207.

DEININGER, K. Fazendo a Reforma Agrária Negociada funcionar: a experiência da Colômbia, Brasil e África do Sul. In: MDA/NEAD, Reforma Agrária e Desenvolvimento Sustentável. 1.ed. Brasília: Ed. Paralelo 15., 380p. pp. 157-175;213-239, 2000.

FERREIRA, F.H.G. LANJOUW, P. AND NERI, M. A New Poverty Profile for Brazil Using PPV, PNAD and Census Data, Departamento de Economia, PUC-Rio, Texto para Discussão N0. 418, marco 2000.

GUANZIROLLI, C. Land Reform and Globalization of the Economy: The Brazilian Case. Brasília, DF, 1998. (Mimeo.).

HECKMAN, J. et al. Characterizing Selection Bias Using Experimental Data. Econometrica, v. 66(5), pp.1017-1098. 1998.

HOFF, K. Wealth Distribution, Economic Efficiency and Incentives. Paper

41

apresentado ao Seminário Internacional: Distribuição de Riqueza, Pobreza e Crescimento Econômico. NEAD-World Bank, julho de 1998, p. 14-17. 31p. 1998. (Mimeogr.).

KANBUR, R. Measurement and Alleviation of Poverty. IMF Staff Papers. 34:6085.

LEITE, S.; PALMEIRA, M. Debates econômicos, processos sociais e lutas políticas: reflexão sobre a Questão Agrária. Cadernos de Debates, CPDA, 1, 1997.

MADDALA, G. S. Limited-Dependent and Qualitative Variables. In: ECONOMETRICS. Econometric Society Monographs. 1.ed. Cambridge: Cambridge University Press, 401p. 1983.

MALLETA, H.; BUAINAIN, A.; VILLALOBOS, R. Rural Poverty in Brazil. Paper to BID. Bolonha, Nomisma, SPA, 1999.

MENEZES, M. A. DE (1997.) Peasant-migrant workers: social networks and practices of resistance. Manchester: University of Manchester. (PhD thesis).

MORRIS, S. et al. Does Geographic Targeting of Nutrition Interventions Make Sense in the Cities? Washington, D.C.: World Development, 27(11), pp.2011-2019, 1999.

NAVARRO, Z. O projeto piloto “Cédula da Terra”. Data terra. 23p. Available from World Wide Web: <http://www.dataterra.org.br>, 1998.

RAVALLION, M. Poverty Alleviation Through Regional Targeting. In: HOFF, BRAVEMAN ; STIGLITZ (ORGS.) The Economics of Rural Organization. 1.ed. World Bank Book, 590p., 1993.

REYDON, B. P.; PLATA, L. A. Politicas de mercados de tierras en Brasil. In: PERSPECTIVAS SOBRE MERCADOS DE TIERRAS RURALES EN AMERICA LATINA. BID, Depto. de Desarrollo Sostenible. Washington D.C., 124, pp. 56-93, 1998.

SADOULET, E.; DE JANVRY, A. Quantitative Development Analysis. Baltimore: The Johns Hopkins University Press, 1995.

SILVEIRA, J. M. F. J DA, BUAINAIN, A.; MAGALHÃES, M. Novas Formas de Reorganização Fundiária no Brasil: O Programa Cédula da Terra. In: XXVII ENCONTRO BRASILEIRO DE ECONOMIA, Belém, 1999. Anais. v.III, p.115-141: ANPEC, 1999.

SILVEIRA, J. M. F. J DA; MAGALHÃES, M.; BUAINAIN, A. Analysis of the Sustainability of Cedula da Terra Program. In: XXXVII CONGRESSO BRASILEIRO DE ECONOMIA E SOCIOLOGIA RURAL. Rio de Janeiro, agosto de 2000, Anais..., SOBER, 2000, 15p. (Mimeo.).

SOUZA FILHO, H. M DE. The adoption of sustainable agricultural technologies: a case study in the State of Espírito Santo, Brazil. England: Ashgate Publishing, 1997.

SOUZA FILHO, H. M.; BUAINAIN, A; SILVEIRA, J.M.F.J DA; MAGALHÃES, M. Community-based Land Reform in Brazil: assessing the selection process. Texto para Discussão do Instituto de Economia da UNICAMP, 96. <www.eco.unicamp.br>, 2000.

SUBBARAO, K.; AHMED, A.; TEKLU, T. Selected Social Safety Net Programs. In: THE PHILIPPINES. Washington, D.C.: World Bank Discussion

42

Papers, 317, 72p. 1996.

TEÓFILO E. et al. Diretrizes para a Política de Desenvolvimento Agrário. Brasília: NEAD, 1998. (Mimeo.).

VAN DE WALLE, D. The Distribution of the Benefits from Social Services in Indonesia. 1978-87. Policy Research Working Paper. 871. Washington, D.C.: Development Economics Research Group, World Bank, 1992.

VAN DE WALLE, D. The Distribution of Subsidies through Public Health Services in Indonesia. The World Bank Economics Review, 8(2), p. 279-309, 1994.

VAN DE WALLE, D. Targeting Revisited, The World Bank Research Observer, v.13 (2), p. 231-48, 1998.

VILLALOBOS, R.; BUAINAIN, A.; MALETTA, H. Brazilian Agriculture and the rural sector: a framework for rural development and sustainable growth. Bologna, document prepared for NOMISMA SPA and Interamerican Development Bank (IDB), 1999.

WILLIAMSON, O. The Mechanisms of Governance. N.Y. Oxford University Press, 429p, 1996.