Effects of Land Titling

Sebastian Galiani

Universidad de San Andres

Ernesto Schargrodsky* Universidad Torcuato Di Tella

April 24, 2005

Abstract The empirical evaluation of the causal effects of land titling typically suffers from selectivity bias. The allocation of property rights across households in the population is usually not random but based on wealth, family characteristics, individual effort, or other mechanisms built on differences between the groups that acquire property rights and the groups that do not. We exploit a natural experiment in the allocation of land titles to overcome this problem. More than twenty years ago, a group of squatters occupied a piece of land in a suburban area of Buenos Aires, Argentina. When the Congress passed an expropriation law transferring the land from the former owners to the squatters, some of the former owners surrendered the land (and received a compensation), while others decide to sue in the slow Argentine courts. These different decisions by the former owners generated an allocation of property rights that is exogenous to the characteristics of the squatters. Our results show significant effects of urban land property rights on housing investment, household size, and school performance. Instead, we find only weak effects on access to credit markets and no effect on labor income.

* Sebastian Galiani, Universidad de San Andres, Vito Dumas 284, (B1644BID) Victoria, Provincia de Buenos Aires, Argentina, Tel: (54-11) 4746-2608, sgaliani@udesa.edu.ar. Ernesto Schargrodsky, Universidad Torcuato Di Tella, Saénz Valiente 1010, (C1428ATG) Buenos Aires, Argentina, Tel: (54-11) 4783-3112, eschargr@utdt.edu. We are grateful to Douglas Allen, Orazio Attanasio, Abhijit Banerjee, Jere Berhman, Tim Besley, Maitreesh Gathak, Marc Muendler, Douglas North, Kei Hirano, Debraj Ray, Emmanuel Skoufias, Chris Woodruff, and seminar participants at the AEA meetings, Econometric Society, Yale, Stanford, UCL, Washington at St. Louis, UC San Diego, ISNIE meetings, Boston University, University of Miami, Universidad de Chile, Universidad Getulio Vargas, Universidad de Montevideo, LACEA, NIP, UTDT, Warwick University and World Bank for helpful comments. Alberto Farias and Daniel Galizzi provided crucial help throughout this study. We also thank Gestion Urbana -the NGO that performed the survey-, Eduardo Amadeo, Julio Aramayo, Rosalia Cortes, Pedro Díaz, Alejandro Lastra, Héctor Lucas, Graciela Montañez, Juan Sourrouille, and Ricardo Szelagowski for their cooperation, and Matias Cattaneo and Sebastian Calonico for excellent research assistance. We also thank the World Bank and Inter-American Development Bank for financial support.

I. Introduction

Property rights are the social institutions that define or delimit the range of privileges

granted to individuals to specific assets, such as parcels of land or water. Private

ownership of these assets may involve a variety of rights, including the right to exclude

non-owners from access, the right to appropriate the stream of rents from use of and

investments in the resource, and the right to sell or otherwise transfer the resource to

others. Property right institutions critically affect decision-making regarding resource use

and, hence, affect economic behavior and performance. By allocating decision-making

authority, they also determine who are the economic actors in a system and define the

distribution of wealth in a society.

Because of its importance in society, property rights institutions have been the subject of

attention by economists, economic historians and political scientists. Secure property

rights are seen as an important determinant of economic development. The main

argument put forward is that secure property rights enhance investment incentives (see,

among others, Demsetz (1967), Libecap (1989)). North and Thomas (1973) provide a

case in point. They argued that the observed difference in growth rates in the sixteenth

and seventeenth centuries between France and Spain, on the one hand, and England

and the Netherlands, on the other, were due to a large extent to differences in property

rights arrangements.

Property rights are not necessarily delineated by the State in the form of legal rights.

They can exist de facto. Thus, the literature makes the distinction between economic

and legal rights as well as formal and informal (including usufructuary) rights. Four

characteristics of property rights are especially relevant to economic behavior: (i) clarity

of allocation, (ii) cost of alienation, (iii) security from trespass, and (iv) credibility of

persistence (Barzel, 2002). In general, de jure property rights improve on de facto

property rights in all these dimensions. Thus, the substitution of de facto by de jure

property rights is expected to affect the behavior of economic agents.

The lack of formal land titles tends to be particularly pervasive among poor households

and imposes them considerable costs. First, individuals (rich and poor) might

underinvest if others could seize the fruits of their investments. Second, houses without

1

legal titles cannot be used as collateral, therefore preventing access to credit markets

(Feder et al., 1988). Thus, the fragility of property rights may impede the use as capital

of the small amounts of wealth that the poor have reducing their consumption and

entrepreneurial opportunities. Third, the lack of well-delineated rights can affect the size

and structure of poor households. Formal titles provide poor family a savings tool in the

form of housing improvements. Moreover, the allocation of property rights within

households is also made clear by formal titles. Fourth, the non-entitled may not be able

to gain from trade (Besley, 1995). The lack of property rights restricts the possibility of

exchanging houses when their characteristics are inadequate for family needs. Fifth, in

the absence of formal property rights the poor may possibly need to spend time and

resources to protect their properties (Lanjouw and Levy (2002), Field (2003)). This self-

protection of their houses may constitute a large burden for poor families.

The recent popularity of land titling as a social intervention is a direct consequence of

the growing recognition of the fact that the absence of formal property rights imposes

considerable costs to poor households (see, among others, Binswanger et al., 1995).

The Peruvian government, for example, issued property titles to 1.2 million urban

households during the 1990s. At a smaller scale, Ecuador and Paraguay have also

developed titling programs. During his first week in office, Brazilian President Lula da

Silva announced a massive plan to award property titles to millions of people living in the

favelas of the major cities of Brazil. Some transitions economies also have large titling

programs (e.g., in Vietnam 11 million people are affected). In most land titling

interventions, the policy consists of issuing formal titles to the people that are already

settled on the land.

Effects of land titling have been documented by several studies: Jimenez (1984) for the

Philippines; Feder et al. (1988) for Thailand; Place and Migot-Adholla (1998) and Miceli,

Sirmans, and Kieyah (2001) for Kenya; Besley (1995) for Ghana; Alston, Libecap, and

Schneider (1996) and Alston, Libecap and Mueller (1999) for Brazil; Lanjouw and Levy

(2002) for Ecuador; Carter and Olinto (2002) for Paraguay; Brasselle et al. (2002) for

Burkina Faso; Do and Iyer (2002) for Vietnam; Field (2003) and Field and Torero (2003)

for Peru; inter alia. The evaluation of land titling effects typically faces the problem that

formal property rights tend to be endogenous for several reasons. The allocation of

property rights across households in the population is usually not random but based on

2

wealth, family characteristics, individual effort, and other mechanisms built on

differences between the groups that acquire those rights and the groups that do not.

This introduces an omitted variable bias on estimates that do not exploit truly exogenous

variability on the allocation of property rights. Similarly, simultaneity is another potential

source of bias. This nuisance is salient in studies of the effect of land rights on

investment, since investment tends to secure informal land rights. Finally, measurement

error is also a pervasive source of bias of estimates of the effects of formal land titles.

The distinctive characteristic of our study is that we exploit a natural experiment in the

allocation of land titles that provides us with a credible strategy to identify the effect of

land titling on the outcomes of interest.

More than 20 years ago, a large number of squatter families occupied an area of

wasteland in the outskirts of Buenos Aires, Argentina. The area was made up of different

tracts of land, each with a different legal owner. An expropriation law was subsequently

passed, ordering the transfer of land to the state in exchange for a monetary

compensation. The purpose of the law was to allow the state to later transfer legal titles

to the squatters. However, only some of the original legal owners surrendered the land,

which was then titled to the squatters. Other owners are still contesting the

compensation payment in the slow Argentine courts, resulting in weak property rights to

the squatters who happened to settle on a parcel of land on these tracts. As a result, a

group of squatters obtained formal land rights, while others are currently living on similar

parcels without paying rent, but without legal titles. Since the decision of the original

owners to challenge the expropriation was orthogonal to the characteristics of the

squatters, the allocation of property rights is exogenous in equations describing the

economic behavior of the squatters.

Moreover, our study involves the comparison of individuals possessing land titles against

squatters occupying similar tracts of land but having only usufructuary rights. Treated

squatters are moved from having only informal rights to the parcels of land they occupy

to posses formal titles on the same parcels of land. The control group remains enjoying

only informal rights to the tract of land they occupy. Thus, treatment only entails the

transfer of land titles, which coincides with the treatment of interest in policy analysis.

3

Another virtue of our study is that the causing variable is measured without error

avoiding a serious thread to the identification of the parameters of interest. We followed

the evolution of the expropriation process through all its institutional instances, including

the land registry, and obtaining precise knowledge of the current legal status of each

parcel. This eliminates any concern about the measurement of the treatment variable in

our analysis. Finally, we study the long-term effects of land titling. This is quite important

since it is likely that several effects take time to materialize.

Exploiting this natural experiment in the allocation of land titles across squatters in a

poor area, we find significant effects on housing investment, household size, and

children’s school achievement. The quality of the houses is substantially higher in the

titled parcels. Moreover, households in the titled parcels have smaller size (although

their houses are larger) and seem to invest more in the education of their children.

However, we do not find much effect on access to credit markets as a result of entitling,

and the effects on employment and labor income are not significant.

The rest of the paper is organized as follows. In the next section we describe the natural

experiment. In section III we present our data, and in section IV we discuss the empirical

strategy. Section V presents our results on housing and household variables, while

section VI concludes.

II. A Natural Experiment

The empirical evaluation of the effects of land title registration programs typically poses a

major methodological challenge. In most historical experiences, the allocation of property

rights across families is not random but based on wealth, family characteristics,

individual effort, or other selective mechanisms. Thus, the personal characteristics that

determine the likelihood of receiving land titles are likely to be correlated with the

outcomes under study. Since some of these individual characteristics are

unobservables, this correlation creates a selection problem that impedes the proper

evaluation of the effects of property right acquisition.

In this paper, we address this selectivity problem by exploiting a natural experiment in

the allocation of property rights. In 1981, about 1,800 families occupied approximately

4

two squared kilometers of wasteland in the locality of San Francisco Solano, County of

Quilmes, Province of Buenos Aires, Argentina. The occupants were groups of landless

citizens organized through the Catholic Church, who explicitly wanted to avoid creating a

shantytown and therefore partitioned the occupied land into small urban-shaped parcels.

At the beginning of the occupation, the squatters believed that the land belonged to the

state, but they later found out that it was private property.1

The squatters resisted several attempts of eviction during the military government. After

Argentina's return to democracy, the Congress of the Province of Buenos Aires passed

Law Nº 10.239 in October of 1984 expropriating these lands from the former owners to

allocate them to the new occupants. According to the law, the government would pay a

monetary compensation to the former owners and, then, it would allocate the land to the

squatters. In order to qualify for receiving the titles, the squatters should have arrived to

the parcels at least one year before the sanctioning of the law, shall not possess any

other property, and should use the parcel as their family home. The law also established

that the squatters could not transfer the property of the parcels for the first ten years after

titling.

The process of expropriation resulted to be asynchronous and incomplete. The occupied

area turned out to be composed of thirteen tracts of land belonging to different owners.

In 1986, the government offered each owner or group of owners (as several tracts of

land had more than one owner) a payment proportional to the official valuation of each

tract of land, indexed by inflation. These official valuations, assessed by the tax authority

to calculate property taxes, had been set before the land occupation. After the

government made the expropriation offers, the owner/s of each tract had to decide

whether to surrender the land (accepting the expropriation compensation) or to start a

legal dispute. Eight former owners accepted the compensation offered by the 1 On the details of the land occupation process see the documentary movie “Por una tierra nuestra” by Cespedes (1984), and CEUR (1984), Izaguirre and Aristizabal (1988), and Fara (1989). Institutional information for this paper was gathered through a series of interviews with key informants, including the Secretary of Land of the Province of Buenos Aires (Maria de la Paz Dessy), Undersecretary of Land of the Province of Buenos Aires (Alberto Farias), Directors of Land of Quilmes County (Daniel Galizzi and Alejandro Lastra), Secretary of Public Works and Land Registry of Quilmes County (Hector Lucas), General Attorney of the Province of Buenos Aires (Ricardo Szelagowski), attorney in expropiation offers’ office (Claudio Alonso), lawyer on expropiation lawsuit (Horacio Castillo), former land owners (Hugo Spivak and Alejandro Bloise -

5

government. Their tracts of land were then titled to the occupants in 1989-91, together

with formal land titles that secured the property of the parcels.2,3 Five former owners,

instead, did not accept the compensation offered by the government and sued to obtain

a higher compensation in the slow Argentine courts.

Importantly, the people who occupied parcels located on the tracts of land belonging to

the former owners that accepted the expropriation compensation were ex-ante similar

and arrived at the same time than the people who settled in the parcels of the former

owners that did not surrender the land. There was simply no way for the occupants to

know ex-ante, at the time of the occupation, which parcels of land had owners who

would accept the compensation and which parcels had owners who would dispute it. In

fact, at the time of the occupation the squatters believed that all the land was state-

owned and they could not know that an expropriation law was going to be passed, nor

what was going to be the future response of the owner of each specific parcel.

A potential problem, however, is that the different former owners’ decisions could reflect

differences in land quality. In turn, these differences could be correlated with squatters’

heterogeneity. For example, more powerful squatters could have settled in the best

parcels. An advantage of our experimental design is that the parcels of land in the

treatment (titled) and control (untitled) groups are almost identical and basically next to

each other. Indeed, after the data are described, we show in Section IV that there are no

differences in parcel observable characteristics (distance to a polluted creek, distance to

the closest non-squattered area, parcel surface, location in a corner of a block) between

heir-), squatters (Jorge Valle and Estela Gutiérrez, inter alia), and President of NGO Gestion Urbana (Claudio Aguirre). 2 The “new” urban design traced by the squatters differed from the previous land tract divisions. Thus, some “new” parcels overlapped over tracts of land that belonged to different owners. Parcels could not be titled if one portion of them was still under dispute. 3 According to the expropriation law, each squatter had to pay the government the (proportionally prorated) value of the official valuation of the occupied tract of land. The law, however, established that the payments should be made in monthly installments that could never surpass 10% of the (observable) household income and there was no indexation for inflation. Given the several hyperinflationary periods experienced by the Argentine economy during the period of analysis and the very high labor informality of this population, the payments made by the squatters were certainly inferior to the market value of the land. In practice, there are no records of the amounts paid by each squatter.

6

the treatment and control groups.4 We also show no differences in pre-treatment

observable household characteristics (gender, nationality and years of education of the

household head at the time of the occupation, and nationality and years of education of

her/his parents). Moreover, the government offers were very similar (in per-square-meter

terms) for the accepting and contesting owners, in accordance with the proximity and

alikeness of the land tracts.5 Given the similarity in land quality and compensation offers,

the different responses could instead reflect heterogeneity of the former owners

regarding subjective land value, litigation costs, and decision making, not of the quality

of the land.6 Importantly, the squatters had no direct contact with the former owners to

influence their decisions and the dwellings constructed by the squatters were explicitly

ignored for the calculation of the expropriation compensation. Finally, note that if, in spite

of this discussion, one may fear that the challenging owners did so because the quality

of their land was higher, that would imply that the squatters with no titles are standing on

land of better quality.

As explained, five former owners did not accept the compensation offered by the

government and went to trial. In these lawsuits, all the legal discussion hinges around

the determination of the monetary compensation. The law was constitutionally approved

by the Congress and, thus, the expropriation itself could not be challenged, but the

monetary offers made by the government were disputed. The squatters had no

4 There are also no differences in altitude. The Buenos Aires area is totally plane and all these parcels are within the same 5-meter topographical range. This is urban land, so agricultural productivity is not an issue. 5 In Argentine pesos (of January 1986) per square meter, the accepted offers had a mean of 0.424 and a median of 0.391. The contested offers had a mean of 0.453 and a median of 0.397. Indeed, the similitude of the offers is repeatedly used as an argument by the government attorneys in the lawsuits to show that the offers were fair, as they were similar to the ones accepted by the other owners. The same argument is utilized in a low-court judge’s verdict in one of these cases citing Supreme Court jurisprudence (Kraayenbrink de Beurts et al v. Province of Buenos Aires). 6 Although thirteen is a very small number for an statistical analysis of the former owners’ decisions, we can provide some anecdotal evidence. The average number of proprietors in the groups of accepting owners is 1.25, while the average number of proprietors for the contested tracts is 2.2. Moreover, when we defined a dummy equal to 1 if there are more than one owner sharing the same surname (i.e. same family), and 0 otherwise, the average for this dummy for the accepting owners is 0.125 while the average for the challenging owners is 0.6. Thus, it appears that having many owners and several in the same family made it more difficult for the owners to agree in surrendering the land. Within the challenging owners, we also found one case in which an owner was a lawyer and he was representing himself in the case (which may indicate lower litigation costs), while in another case, one of the original owners had deceased before the sanctioning of the law (in 1983) and her inheritance process was still under way at the time the family had to make a decision.

7

participation in these legal processes (the lawsuits were between the former owners and

the provincial government), and the value of the dwellings they constructed was

excluded from the compensation calculation.7 One of these five lawsuits finally ended

with a judge’s verdict, and the squatters on this tract of land received titles in 1997-98.

The other four lawsuits are still pending. If one is still worried about the possibility that

the former owners’ decisions of surrendering or suing was correlated with land quality,

then an additional feature of this experience is that it allows us to separately compare

the squatters in this lastly titled tract of land relative to the control group. Although these

two groups of squatters settled in eventually contested tracts of land and, therefore, are

homogenous regarding their respective former owners’ decisions of going to trial, one

group already received titles while the other is still waiting for the end of the legal

process.8

The final outcome of this expropriation process is that a group of families now has formal

property rights, while another is still living in the occupied parcels without paying rents

and enjoying rights of usufruct but without possessing formal land titles. This allocation

of land titles was the result of an expropriation process that did not depend on any

particular characteristic of the squatters. Thus, by comparing the groups that received

and did not receive land titles, we can act as if we had a randomized experiment.

Because randomization resolves the selection problem, this natural experiment allows us

to identify the effects of land titling using cross-sectional information.

III. Data Description and Collection

We followed the evolution of the expropriation process in the Land Secretary of the

Province of Buenos Aires, the office of the General Attorney of the Province of Buenos

7 This is explicit, for example, in the low-court judge’s verdict of Cordar SRL v. Province of Buenos Aires. 8 We can still wonder, within this group of challengers, why some are still in trial while one concluded. There are some clearly exogenous reasons why some of these trials are still going on. In one of these cases, the legal process was delayed by a mistake made in the description of the parcels in a low-court judge’s verdict. In another two cases, the expropriation lawsuit was delayed by the death of one of the former owners, which required an inheritance process. In the fourth case (mentioned in footnote 6) one of the original owners had died just before the sanctioning of the law and her inheritance process was still under way at the time the family had to make a decision.

8

Aires, the Quilmes County Government, the land registry, the judicial cases, and the tax

authority, obtaining precise knowledge of the legal status of each parcel.

The area affected by Expropriation Law Nº 10.239 covers a total of 1,839 parcels. 1,082

of these parcels are located in a contiguous set of blocks. However, the law also

included another non-contiguous (but close) piece of land currently called San Martin

neighborhood, which comprises 757 parcels. As this area is physically separated from

the rest, we focus on the 1,082 contiguous parcels to improve comparability, and then

pool the San Martin parcels as a robustness check of our findings.

Land titles were awarded in two phases. Property titles were awarded to the occupants

of 419 parcels in 1989-91, and to the occupants of 173 parcels in 1997-98. Land titles

were not available to the families living in 410 parcels located on tracts of land that have

not been surrendered to the government in the expropriation process. Finally, there are

80 parcels that were not titled to the squatters because the occupants had moved or

died at the time of the title offers, or had not fulfilled some of the required registration

steps, although the original owners had surrendered these pieces of land to the

government. For these potentially endogenous reasons, the inhabitants of these 80

parcels (out of the 672 parcels offered for titling) missed the opportunity to receive a title,

i.e. missed the opportunity to receive the treatment.9 Borrowing the terminology from

clinical trials, this subgroup constitutes the “non-compliers” in our study, since they were

“offered” the treatment (land title) but they did not “receive” it.

Table 1 summarizes the process of allocation of land titles for the main area. The

variable Property Right Offer equals 1 for the families occupying parcels that were

surrendered by the original owners, and 0 otherwise; while the variable Property Right

equals 1 for the squatters that received property titles, and 0 otherwise.10

9 These parcels are currently government property. 10 The 757 parcels of San Martin, which belonged to only one group of owners who accepted the expropriation compensation without suing, were offered for titling in 1989-91. 712 were titled, while 45 correspond to non-compliers.

9

Table 1 – Allocation of Land Titles

Property Right Offer = 1

Year Property Right = 1

Property Right = 0 Total

Property Right Offer = 0 &

Property Right = 0

Total

1989-91 419 23 442 1997-98 173 57 230

Total 592 80 672 410 1082

Between the months of January and March of 2003, we performed a survey interviewing

the inhabitants of 590 randomly selected parcels (out of the total of 1,839). We surveyed

617 households living in these 590 parcels (27 parcels host more than one family).

Excluding the San Martin neighborhood, we interviewed 467 households living in 448

parcels. At the same time, we sent a team of architects to perform an outside evaluation

of the housing characteristics of the parcels.11

IV. Econometric Methods

We seek to evaluate the effect of the allocation of property rights on several outcome

variables. To answer such questions, statisticians (e.g., Kish, 1987) have recommended

a formal two-stage statistical model. In the first stage, a random sample of participants is

selected from a defined population. In the second stage, this sample of participants is

randomly assigned to treatment and comparison (control) conditions. Although a thing of

beauty, the formal two-stage model has almost never been implemented in practice by

researchers in the behavioral sciences. While researchers in health frequently implement

randomized experiments testing treatments of real policy interest, these randomized

trials are nearly always implemented on a convenience sample of volunteers (one-stage

randomized trials).

11 Gestion Urbana, an NGO that works in this area, carried out the household survey and the housing evaluation. The interviewers and the architects were not informed of the hypotheses under study and were blind to the treatment status of each parcel. We distributed a food stamp of $10 (about 3 US dollars at the time of the survey) for each answered survey as a gratitude to the families willing to participate in our study. In 10 percent of the cases, the survey could not be performed because there was nobody at home in the three visit attempts, the parcel was not used as a house, rejection, or other reasons. These parcels were randomly replaced. Non-response rates were similar for titled and untitled parcels.

10

In economics even one-stage randomized experiments are rare, and in many cases

impossible to implement. Nevertheless, economists can, sometimes, preserve the

central feature of a randomized experiment by exploiting natural experiments. In a

natural experiment, like in a randomized trial, there is a control group that estimates what

would have happened to the treated group in the absence of the intervention, but nature

or other exogenous forces determine treatment status instead.12 The validity of the

control group is evaluated by examining the exogeneity of treatment status with respect

to the potential outcomes, and by testing that the pre-intervention characteristics of the

treatment and control groups are reasonably similar (see Card and Krueger (1995)).

In section II we discussed at length the exogeneity of the allocation of property rights

offers among the squatters and argued that this process was not triggered by some

phenomenon that differentially affected the treatment and control groups in our

experiment. We now test the similarity of pre-treatment characteristics between these

two groups.

In Table 2, we compare parcel characteristics for the intention-to-treat and non-intention-

to-treat groups (i.e., Property Right Offer = 0 and Property Right Offer=1, respectively) to

analyze the presence of potential differences. The four available variables are distance

to a nearby (polluted and floodable) creek (in blocks), distance to the closest non-

squattered area (in blocks), parcel surface (in squared meters), and a dummy for

whether the parcel is located in a corner of a block. We only reject the hypotheses of

equality for parcel surface (at the 8.9% level of significance). Nevertheless, the

difference in average parcel surfaces between these two groups is relatively small –

parcels are only 3% larger in the non-intention to treat group – and if something, it is the

control group the one that inhabits slightly larger parcels.

12 Examples of natural experiments are Campbell and Ross (1970), Angrist (1990), Imbens et al (2001), and Di Tella and Schargrodsky (2004).

11

Table 2 – Pre-Treatment Parcel Characteristics

Parcel Characteristics Property Right Offer=0

Property Right Offer=1 Difference

Distance to Creek (in blocks)

1.995 (0.061)

1.906 (0.034)

0.088 (0.070)

Distance to Non-Squattered Area (in blocks)

1.731 (0.058)

1.767 (0.033)

-0.036 (0.067)

Parcel Surface (in squared meters)

287.219 (4.855)

277.662 (2.799)

9.556* (5.605)

Block Corner 0.190 (0.019)

0.156 (0.014)

0.033 (0.023)

Notes: Standard errors are in parentheses. * Significant at 10%.

In Table 3, we compare pre-treatment characteristics of the “original squatter” between

the intention-to-treat and non-intention-to-treat groups (i.e., Property Right Offer = 0 and

Property Right Offer=1, respectively) for the families that arrived before treatment. We

define the “original squatter” as the household head at the time the family arrived to the

parcel they are currently occupying. We cannot reject the hypotheses of equality in

gender, nationality and years of education of the original squatter, suggesting a strong

similarity between these groups at the time of their arrival to this area. Moreover, we do

not reject the hypotheses of equality in nationality and years of education of the mother

and father of the original squatter across the groups, suggesting that these groups had

been showing similar trends in their socio-economic development before their arrival to

this area.13 The similarity across pre-treatment characteristics is consistent with the

exogeneity in the allocation of property rights described above.

13 In 23 percent of the cases, the current household head does not coincide with the original squatter, either because he/she arrived later than the first member of the family that occupied the parcel, or because he/she arrived at the same time but was not the household head at the arrival time. Nevertheless, we obtain similar results when we compare characteristics of the current household head across the groups.

12

Table 3 – Pre-Treatment Characteristics of the Original Squatter

Characteristics of the Original Squatter

Property Right Offer=0

Property Right Offer=1 Difference

Age 48.875 (0.938)

50.406 (0.761)

-1.531 (1.208)

Female 0.407 (0.046)

0.352 (0.035)

0.054 (0.058)

Argentine 0.902 (0.028)

0.903 (0.021)

-0.001 (0.035)

Years of Education 6.070 (0.187)

5.994 (0.141)

0.076 (0.234)

Argentine Father 0.794 (0.038)

0.866 (0.024)

-0.071 (0.045)

Years of Education of the Father

4.654 (0.146)

4.417 (0.076)

0.237 (0.165)

Argentine Mother 0.803 (0.037)

0.855 (0.025)

-0.052 (0.045)

Years of Education of the Mother

4.509 (0.122)

4.548 (0.085)

-0.039 (0.149)

Notes: Standard errors are in parentheses.

Once treatment status has been shown to be ignorable, estimation of average treatment

effects is straightforward. Let Y1i denote the outcome of interest of unit i if this unit has

land title, and let Y0i denote the outcome of interest otherwise. Let D be an indicator of

land titling. The average effect of land titling on outcome Y is:

E[Y1| D = 1] - E[Y0| D = 1] (1)

Note that the first term of (1) is observed, but the second term is an unobserved average

counterfactual. Simple comparisons of average outcomes by treatment status generally

fail to identify causal effects, unless land titling is determined in a manner independent of

the unit’s potential outcomes, as it is the case in a randomized trial or in a natural

experiment like the one we exploit in this paper. Exogeneity in the allocation of treatment

allows us to use E[Y0| D = 0], the average outcome of interest in the control units, as the

counterfactual realization of what would have happened to the treated group without

treatment, that is, to estimate E[Y0| D = 1].

Operationally, we analyze the effects of land titling on variable Y by estimating the

following regression model:

13

iiii εβγα +++= X Right Property Y (2)

where Yi is any of the outcomes under study, and γ is the parameter of interest, which

captures the causal effect of Property Righti (a dummy variable indicating the possession

of land title) on the outcome under consideration. Xi is a vector of pre-treatment parcel

and original squatter characteristics and εi is the error term.14

Some of the variables under study are Limited Dependent Variables (LDV). Angrist

(2001) argues that the problem of causal inference with LDV is not fundamentally

different from the problem of causal inference with continuous outcomes. If there are no

covariates or the covariates are sparse and discrete, linear models (and associated

estimation techniques like two-stage least squares) are no less appropriate for LDV than

for other types of dependent variables. Certainly, this is the case in a natural experiment

where controls are only included to improve the efficiency of the estimates, but their

omission would not bias the estimate of the parameter of interest.

A typical concern when conducting statistical inference after estimating the parameters

of equation (2) is that the errors in that equation might not be independent across

households. For example, treatment might interact with parcel characteristics and hence,

it might affect similarly households residing nearby. Moreover, the decision by each

former owner of surrender or sue simultaneously affected all the parcels located in the

tract of land she/he possessed. In order to control for these potential nuisances, we also

compute standard errors both clustering the parcels located in the same block and the

parcels belonging to the same former owner.15

To this point, we have assumed that all the squatters actually received the treatment to

which they were assigned. In many experiments, however, a portion of the participants

fail to follow the treatment protocol, a problem termed treatment noncompliance. In our 14 Our estimates show no change if we include as controls the personal characteristics of the current household head instead of those of the original squatter, when they differ. 15 For the former owner clustering, if a parcel overlaps on the borders of the previous tract divisions, occupying a piece of land that belonged to one owner and another piece that corresponded to another owner, the former owner is defined as the combination of the two former owners. For the block clustering, a block is defined as both sides of the segment of a street between two corners. These procedures define 18 former owner clusters and 83 block clusters.

14

case, this might be of potential concern since a number of families that were offered the

possibility of obtaining land titles did not receive them for reasons that may also affect

their outcomes. In order to address this problem of non-compliance, we also report

estimates of the treatment effects by the method of Two-Stages Least Squares (2SLS)

where we instrument the Property Right dummy variably using the “intention to treat”

variable Property Right Offer, a dummy variable indicating the availability of land title

offers (see Angrist et al, 1996).

Finally, in any investigation where the impact takes time to materialize (like the

investment and household size variables considered in this paper), some participants will

inevitably drop out from the analysis. As one illustration, Biglan et al. (1991) reviewed

longitudinal studies of substance abuse prevention programs and found that attrition

rates ranged from 5% to 66%. Participation attrition, hence, is another potential problem

that might bias the estimates of causal effects in long-term studies.

In our survey, we found that some families arrived to the parcel they are currently

occupying after the time the former owners made, during 1986, the decision of surrender

the land or sue. From the sample of 467 interviewed households, we found that the

original squatter had arrived to the parcel before the end of 1985 in 313 families, while

154 families had arrived after 1985.16 As it is plausible to argue that the families that

arrived after the former owners’ decisions could have known the different expropriation

status (i.e., the different probabilities of receiving the land) associated to each parcel, to

guarantee exogeneity we need to exclude from the analysis the families that arrived to

the parcel they are currently occupying after 1985. This raises a problem of attrition. If

some families arrived after 1985, they could have replaced some original squatters in our

treatment and control parcels that had left the area before we ran our survey in 2003.17

Similar results are obtained using other clustering units, such as each side of a block or the “square” block. 16 To identify with accuracy whether the time of arrival of each family to the parcel they are currently occupying was prior to 1986, in our first survey we asked each household head where she/he was living when Diego Maradona scored the Hand of God goal in the 1986 World Cup game against England. It is impossible for an Argentine not to remember where she/he was on that day (see Amis (2004)). 17 For the families that arrived after 1985, our questionnaire attempted to collect information on the names and destination of the previous occupants of the parcels. In both treatment and control parcels, the current occupants could provide a name and/or destination of the previous occupant only for less than 20 percent of the cases. Although the information obtained is very poor, it does

15

Moreover, the availability of titles could have affected household migration decisions.

Indeed, the first column of Table 4 shows that 62.4 percent of the parcels in the non-

intention-to-treat group are inhabited by families that arrived before 1986, while the

percentage is 70.0 percent for the intention-to-treat group.

Table 4 – Household Attrition

Variables Property

Right Offer=0

Property Right

Offer=1

Property Right Offer 1989-91=1

Property Right Offer 1997-98=1

Household arrived before 1986=1

0.624 (0.036)

0.700 (0.028)

0.729 (0.051)

0.689 (0.033)

Difference relative to Property Right Offer=0 -0.076*

(0.045) -0.105* (0.063)

-0.064 (0.049)

Notes: Standard errors are in parentheses. * Significant at 10%.

Of course, the migration decision could be potentially correlated with the outcomes

under study. However, the distortions generated by this selection problem may not be

that severe. We exploit two alternative strategies to address this potential nuisance.

First, we will exploit the asynchronous timing in the titling process. The third column of

Table 4 shows a significant difference in attrition for the parcels titled in 1989-91 (early

treatment) relative to the control group. Instead, the last column shows no statistically

significant differences in attrition for the parcels more recently titled in 1997-98 (late

treatment) relative to the control group. Thus, once we incorporate the fact that the

analysis must be done on the survivors of the experiment, the non-intention-to-treat

group appears a priori as a better control group for the late-treated group than for the

early-treated squatters. First, there is no differential attrition between these two groups.

Second, the unobservable variables that might have affected migration decisions are, a

priori, more likely to be ignorable when comparing these two groups than when

comparing the control and early treatment groups. This is so because these two groups

not only faced similar shocks since they arrived to this neighborhood –i.e. are in the

same labor markets, etc.- but being untitled for most of the period, also had similar

incentives to respond to them. Thus, the estimated effects of land titling for the group of

recently titled parcels are unlikely to be biased by attrition. Additionally, the comparison

of these coefficients with those corresponding to the estimated effects of land titling for

not suggest that the households that left the untitled parcels moved to richer areas than the families that left the titled parcels.

16

the early treated group leads to an indirect test of whether attrition in the latter group is

also ignorable.

A more standard strategy to address the problem of attrition is to model the selection

mechanism. A traditional parametric approach was developed by Heckman (1979). A

number of less restrictive semi-parametric approaches have been recently introduced in

the literature. A common thread in all these formulations, however, is that they need to

characterize the mean of the regression error term conditional on the regression

covariates and the sample selection rule. An alternate approach is discussed in Ahn and

Powell (1993). They propose to eliminate the selection bias by differencing observations

with similar probabilities of selection sidestepping the problem of estimating the unknown

conditional mean function.18 The identification of the effect of land titling on a certain

outcome necessitates that at least one of the pre-treatment characteristics predicts

attrition. The idea is then to compare the outcomes for treated and control survivors with

similar pre-treatment characteristics. This is equivalent to match observations based on

the propensity score of sample selection (see Rosenbaum and Rubin (1983)).

The only pre-treatment characteristics available for the whole set of squatters (attrited

and non-attrited) are the parcel characteristics reported in Table 2. We estimate a Logit

model of the likelihood of survival since 1985 on these parcel characteristics, and find

that a further distance to the nearby polluted and floodable creek significantly increases

this likelihood.19 We exploit the variability in attrition induced by this pre-treatment

characteristic to correct for sample selection.

We implement the matching selection correction by means of the method of stratification

matching. First, we eliminate observations outside the common support of the estimated

propensity score for the distributions of treated and control groups. Second, we divide

the range of variation of the propensity score in intervals such that within each interval,

treated and control units have on average the same propensity score. Third, within each

interval, the difference between the average outcomes of the treated and the controls is

computed. The parameter of interest is finally obtained as an average of the estimates of

18 The strategy relies on the fact that in latent index models, the selected mean of the regression error is an invertible function of the probability of selection given covariates. 19 If we add the titling offer dummy variable to the model, this variable also affects the likelihood of survival as suggested in Table 4. Distance to the nearby creek remains statistically significant.

17

each block with weights given by the distribution of treated units across blocks.

Bootstrap standard errors are computed.

V. Results In this section we investigate the average causal effect on housing investment,

household structure, schooling outcomes, access to credit, and labor earnings of moving

a squatter household from usufructuary rights to formal land titling. This is exactly the

treatment of interest in order to test the effects of secure property rights on individual

behavior. It is also the exact treatment of interest to policy analysis where the question is

what happens to poor squatter households when they receive the formal title to the piece

of land they usufruct.

V.1. Effects on Housing Investment

Given the importance of investment to long-term poverty alleviation, it is critical to

understand whether land titles affect investment. The possession of land titles may affect

the incentives to invest in housing improvements. The traditional view emphasizes

security from trespass. Individuals do not invest if others may seize the fruits of their

investments.20 Investment and land titles could also be linked via enhanced possibilities

from gains from trade. Land titles reduce the cost of alienation of the assets encouraging

investment. A third link emphasized in the literature is through the credit market. Well-

defined property rights –including clarity of allocation and credibility of persistence- might

make easier to use the land as collateral, diminishing the constraints on funding

investments. Finally, a fourth link might appear through savings strategies. Poor

households, especially in unstable macroeconomic environments, are likely to lack good

savings instruments. Land titles could provide them with such an instrument: the

accumulation of housing value. If households are also constrained in the labor market

opportunities, land titles might allow them to substitute present leisure for future

20 In our survey, we investigated about the subjective perceptions on eviction threats among squatters. In the untitled parcels, 7.8 percent of the respondents fear eviction by the former owners, while that figure is negligible for the entitled households. However, the main fear of these families is not to be evicted by the former owners, but to have their parcels occupied by new squatters. 17.8 percent of the families without legal titling report suffering this fear, while the percentage is half of that number for the entitled households.

18

consumption via self-made investments in their houses.21 We now investigate empirically

the impact of legal land titles on housing investment.

Previous research has documented inconclusive results on the impact of land rights on

investment. For example, in a well-known study, Besley (1995) reports that land rights

have a positive effect on investment in the region of Angola but a less noticeable impact

on the region of Wassa. Other studies also find no impact of land rights on investment

(see, among others, Brasselle et al. (2002)).

In Table 5 we summarize the analysis of the effect of property rights on housing

investments. An important clarification is that before the occupation this was a wasteland

area without any construction. Thus, the treatment and control areas had a similar (i.e.,

zero) baseline investment level before the occupation. In each column, we present the

coefficient of the treatment dummy Property Right on a different housing characteristic.

All the estimates reported in Table 5 are from regressions including controls for pre-

treatment characteristics of the parcel and the original squatter.

Table 5 - Housing Investment

Good Walls

(1)

Good Roof

(2)

Constructed Surface

(3)

Concrete Sidewalk

(4)

Points in City of

Quilmes (5)

Property Right 0.20*** 0.15** 8.26** 0.11** 8.42*** (3.47) (2.48) (2.33) (2.17) (3.64) Control Group Mean 0.50 0.31 67.63 0.67 22.71 %∆ 40.00% 48.39% 12.21% 16.42% 37.08% Notes: Constructed surface in squared meters. Good Roof, Good Walls and Concrete Sidewalk are dummy variables that equal 1 if the house has a roof of good quality, walls of good quality and a sidewalk made of concrete, respectively, and 0 otherwise. Points in City of Quilmes measures overall quality of houses relative to houses in downtown Quilmes (the main locality of the county) from 0 to 100 points. All the regressions control for parcel and original squatter pre-treatment characteristics: surface of the parcel (in square meters); distance to creek (in blocks); distance to nearest non-squattered area (in blocks); a corner-of-the-block dummy; age, gender, nationality, and years of education of the original squatter; and nationality and years of education of father and mother of the original squatter. The robustness of each of these regressions is presented in Appendix Tables A.1 through A.5. Absolute value of t statistics are in parentheses * significant at 10%; ** significant at 5%; *** significant at 1%. 21 Houses in these socioeconomic groups are mainly self-constructed (“The rich first builds and then inhabits, the poor first inhabits and then builds”, Fundación ProVivienda Social, institutional

19

The first two columns present large effects of land titling on the probability of having

walls (first column) and roof (second column) of good quality. The proportion of houses

with good quality walls rises by 40 percent under land titling, while the increase reaches

48 percent for good quality roof. The third column presents the effect of land titling on

the total surface constructed in the parcel (measured in squared meters). Our results

suggest a statistically significant increase of about 12 percent in constructed surface

under the presence of land titles. The fourth column shows a statistically significant

increase of 16 percent in the proportion of houses with sidewalks made of concrete. In

the last column, the dependent variable Points in Quilmes measures the overall quality

of each house relative to houses in downtown Quilmes (the main locality of the county)

using an index from 0 to 100 points assigned by the team of architects following a set of

guidelines. The coefficient shows a large and significant effect of land titling on housing

quality. Relative to the baseline average sample value, the estimated effect represents

an overall housing improvement of 37 percent associated to titling.22 Thus, we conclude

that moving a poor household from usufructuary rights to land titles substantially

increases housing investment. Consequently, our results show that securing property

rights significantly enhances investment levels.

For each one of these investment variables, in Appendix Tables A.1 to A.5 we show the

robustness of the results regarding the methodological concerns discussed in section IV.

In each Appendix table, Column (1) reports again the model in Table 5 but displaying all

the coefficients (and t-statistics) for the control variables. In column (2), we start with a

simple model without including any control variables. For the five outcome variables

considered in Table 5, the point estimates are slightly lower than the ones obtained from

the models that includes the full set of control variables, but the differences are very

small and never statistically significant. In Column (3) we add back the control variables

for parcel characteristics –i.e. Surface, Distance to Creek, Block Corner and Distance to

Non-Squattered Area-. Again, in all cases, the point estimates are very similar to those in

Column (1). In Column (4) we add the observations for the San Martin neighborhood that

were excluded from the baseline analysis in order to enhance geographical

presentation). Thus, the labor necessary to construct these houses mainly emanates from leisure substitution. 22 Similar results are obtained using an alternative index of housing quality that measures the overall quality of each house relative to the houses within this same neighborhood.

20

comparability between treatment and control groups (see Section III). Once more, the

point estimates are similar to those in the baseline model in Column (1) and the

differences are never statistically significant.

Columns (5) and (6) address the potential presence of clusters in the errors of the

models. There, we report t-statistics computed with clustered standard errors after

clustering the parcels located in the same block and the parcels belonging to the same

former owner. In most cases, t-statistics change little and the level of significance of the

test remains unaltered. Only for concrete sidewalk, clustering the standard errors

noticeable reduces the significance of the treatment variable.

Column (7) deals with the potential problem of non-compliance. We report Two-Stages

Least Squares (2SLS) estimates instrumenting the Property Right dummy variable with

the intention-to-treat variable Property Right Offer. The estimates are very similar to

those obtained from Ordinary Least Squares (OLS) in the baseline specification and the

differences are not statistically significant at conventional levels of significance.

In Columns (8) and (9) we address the concern that these results might be caused by

the attrition in the original populations and are not the cause of treatment. In Column (8)

we separately report average effects for early and late treatment groups. Both early and

late land titling have positive significant effects on all these investment variables (but

Concrete Sidewalk for the late treatment group). For all the variables, the point estimates

for the late treatment dummy variables are very similar to the ones in the baseline

specification in Column (1). Moreover, the F-statistics show that we cannot reject the null

hypotheses that the effects for the early-treated group and late-treated group are similar

at conventional levels of significance.23

Column (9) reports the stratified matching estimates discussed in the previous section.

Again, for all the variables considered, the point estimates are quite similar to those in

the baseline specification and the differences are never statistically significant. Thus, the

23 If one was still worried about the possibility that the former owners’ decisions of accepting or contesting the government offer was correlated with land quality, the equality of these estimates should be reassuring. In both the late treated and control areas, the squatters settled on eventually contested tracts of land and, therefore, are homogenous regarding their respective former owners’ decisions (see section II).

21

evidence suggests that the estimates in Table 5 identify the causal effect of land titling

on investment and not a statistical artifact due to attrition.

Finally, in Column (10) we consider the whole sample of 446 parcels where households

were interviewed, instead of considering only the parcels occupied by households that

arrived before the time the former owners decided to surrender the land or sue. This

analysis investigates a different parameter than the one considered so far. The

estimated coefficient measures the causal effect of securing property rights on

investment in a given parcel regardless of whether the same family has been occupying

it or not, i.e. regardless of whether the selection of the family that is currently occupying

depended on titling status.24 The estimated coefficients are of very similar magnitude.

This evidence directly contributes to a large literature on the effects of secure property

rights on growth by providing micro evidence supportive of the hypothesis that secure

property rights institutions are conducive to development.

A final question relates to the interpretation of the identified causal effect of land titling on

investment. Is this an incentive effect induced by owning formal property rights or it is

mainly a wealth effect from titled households that became richer? We argue that our

research design suggests the treatment operates by affecting the incentives to invest.

First, several factors lessen the size of the wealth transfer. The market value of land

parcels in the neighboring non-squattered area of similar size to the ones occupied by

the squatters represented approximately 3 times the average monthly total household

income in the official household survey (EPH) of October 1986.25 For the households in

the first quintile of the EPH income distribution, the market value of these parcels

represented 7.4 times the monthly total household income. This figure, however,

constitutes only an upper bound of the wealth transfer. Our study involves the

comparison of squatters receiving formal land titles against squatters occupying similar

tracts of land, enjoying usufructuary rights without paying rent, and who have a high

24 In these regressions that ignore household attrition, the estimated coefficient can be interpreted as “what grows in a parcel when it is entitled” regardless of whether the same family has been occupying it or has been replaced by another one. Instead, the estimates obtained exclusively on the non-attrited households measure “what a given family builds in a parcel when receives a land title”. 25 Market value of parcels obtained from evidence presented in Kraayenbrink de Beurts et al v. Province of Buenos Aires.

22

probability of receiving the land titles at some point in the future. Moreover, the titled

families had to make some payments to receive the titles (see footnote 3) and they were

restricted from transferring the property of the parcels for the first ten years after titling.

Thus, in order to estimate the net wealth transfer, the value of these concepts should be

deducted from the market value of a similar parcel.

Second, the families could not have financed the investments with the wealth transfer.

Again, they were restricted from transferring the property of the parcels for the first ten

years after titling; their access to credit only marginally improved (see section V.4); and

the families that could have sold the land are excluded by design from our estimates

(because they attrited from our sample).

Third, Appendix Table 6 shows no differences in the consumption of several durable

goods (refrigerators, freezers, washing machine, TV sets and cellular phones). This

suggests that the large increase in household investment presented in this section is a

result of a change in the economic returns to housing investment induced by the land

titles and not just a response to a wealth effect that should have affected the

consumption of these goods.

V.2. Effects on Household Size Land titling might also affect the way households structure. There are passive and active

reasons for that to happen. Regarding the former ones, lack of land titles might reduce

the ability of household heads to restrict other members of their extended family from

residing in their houses. This is likely to affect welfare directly. And it also generates a

situation where even the usufructuary rights on the land are divided, inducing an

inefficient allocation of resources. In particular, the lack of clarity of right allocations on

the house that occur when the head of the household does not have land titles may

generate, on average, larger households among the untitled group.

Among the active reasons, insurance motives seem to be most important. The poor lack

access to well-functioning insurance markets. With little possibility for risk diversification,

the need for insurance has to be satisfied by other means. An important strategy for

insurance is to use traditional systems of support. These include the village, the

23

commons, and the extended family.26 Another possibility is to use children as a substitute

for insurance. In particular, especially in urban areas, there might be old-age security

motives for higher fertility (see, among others, Cain, 1985, Nugent, 1985 and Portner,

2001). Clearly, land titling, directly and indirectly –i.e., through housing investment or

through the ability to access credit-- might provide some of the insurance needed by

households, also inducing smaller households, on average, among the titled group.27

In Table 6, we find large differences in household size between titled and untitled

groups. Untitled families have an average of 6 members, while titled households have

one member less. The difference is robust (see Appendix Tables A.7 to A.12). Table 6

also shows that the difference in household size does not originate in a more frequent

presence in the control group of a spouse of the household head (column 2), or of

offspring of the household head older than 14 years old, i.e. born before the first land

titles were issued (column 3).28 This last result is important because it suggests that

attrition is not causing the reported results in this section. If we were going to find

differences in the number of children of the household head born before treatment, given

the claim that intention-to-treat is ignorable, that difference could only be an statistical

artifact induced by attrition.

26 “[A]n important question is whether having many children and/or a large extended household is an optimizing strategy allowing households to derive benefits otherwise lost due to poorly functioning markets” (Birdsall 1988, pp. 502). 27 David and Sundstrom (1984) explain the fertility changes in US history using an argument similar to this one. Suppose, they argue, that large families were designed to be old-age insurance for the parents. At the time of independence, the superabundance of arable land meant that the price of land would not rise over time sufficiently to be a nest egg for old age, and children would, or could, be induced to care for the aged parents. When, late in the nineteenth century, the best lands were growing scarce, then the rent, and therefore the price, of land already owned and settled would increase becoming a nest egg due to its capital gain. Thus, investment in land operated as a substitute for more children. The scarcer the land, the higher the economic rent and capital gain, and the fewer children needed to provide for the declining years of the parents. 28 Remember that the first wave of titles started in 1989 and the survey was performed in 2003.

24

Table 6 - Household Size

Number of Household Members

(1)

Household Head Spouse

(2)

Children of the

Household Head (>14 years old)

(3)

Others Relatives (no

Spouse or Children of Household

Head) (4)

Property Right -0.95*** -0.01 0.03 -0.68*** (2.83) (0.27) (0.16) (3.53) Control Group Mean 6.06 0.73 1.52 1.25 %∆ -15.68% -1.37% 1.97% -54.40%

Children of the Household Head

(6-14 years old)

Children of the Household Head

(0-5 years old) (5) (6) (7) (8) Property Right -0.21 -0.08 (1.45) (0.92) Property Right -0.43** -0.14 1989-91 (2.13) (1.18) Property Right -0.09 -0.04 1997-98 (0.57) (0.47) Control Group Mean 1.12 1.12 0.43 0.43 %∆ Property Right -18.75% -18.60% %∆ Property Right 1989-91 -38.39% -32.56% %∆ Property Right 1997-98 -8.04% -9.30% Notes: En each column, the dependent variable is the number of household members of each group. All the regressions control for parcel and original squatter pre-treatment characteristics: surface of the parcel (in square meters); distance to creek (in blocks); distance to nearest non-squattered area (in blocks); a corner-of-the-block dummy; age, gender, nationality, and years of education of the original squatter; nationality and years of education of the father of the original squatter; and nationality and years of education of the mother of the original squatter. The robustness of each of these regressions is presented in Appendix Tables A.7 through A.12. Absolute value of t statistics are in parentheses * significant at 10%; ** significant at 5%; *** significant at 1%.

The difference in household size seems to originate in two factors. First, column 4 of

Table 6 shows a higher presence (0.68 members) of non-nuclear relatives in untitled

households. Untitled households report a much larger number of further relatives of the

25

household head (brothers, sisters, parents, in-laws, grandchildren, etc.) living in the

house than titled household.29

Second, the entitled households show a smaller number of offspring of the household

head born after the title allocation. To analyze this result better we split the household

head offspring into those born between the first and the second title allocation (children

between 6 and 14 years old), and those born after the second title allocation (children

between 0 and 5 years old). For the 6-14 age group, column 6 of Table 6 shows that the

reduction in the number of household head’s children is significant for the early-treated

household, but there is no reduction in the number of children born before land titling for

the late-treated group. This result is also reassuring since treatment could not have

affected fertility for the late-treated group in the 6-14 age bracket because these children

were born before titling for this group. The effect, however, is not significant for the

number of household head’s children in the 0-5 age group for both the late and early

treated households.30 Thus, we find that the larger household size in the untitled parcels

is due to both a larger number of offspring of the household head and a more frequent

presence of non-nuclear relatives.

V.3. School Performance

Following the seminal work of Becker and Lewis (1973) where the level of household

expenditure on children, or child “quality”, is an explicit choice variable, one would expect

that if land titling cause a reduction in fertility, it would also induce households to choose

higher quality offspring. In this section we explore this hypothesis by looking at schooling

outcomes.

29 Only two households in our sample include members who are not relatives of the household head. 30 A plausible explanation for the lack of significant effects for the offspring of the household head of 0-5 years of age is that our household heads are now fairly old and, therefore, their fertility rate is low. Remember that three quarters of them were already the heads of their households at the time of the occupation (see footnote 13). In our sample, the average household head age is 46 years old, and the average age of the female head (the household head if female or the age of his spouse if male) is 43.7 years old.

26

In Table 7, we analyze the performance of children at school. These regressions are

estimated at the child level.31 We collapse differences in school attrition, in grade

repetition, and in age of school initiation in the School Achievement variable, the

difference between the school grade the child is currently attending or the maximum

grade attained (if she/he is not currently attending school) minus the grade

corresponding to her/his age. For the offspring of the household head in the 6-14 age

group in the early treated group, the group of children for which in column 6 of Table 6

we found a reduction in the number of household members, column (2) of Table 7 shows

a large effect in School Achievement. Relative to children in the control group, the

children in titled parcels reach an additional 0.4 year of schooling for their age. Instead,

the effect is not significant for children in the late-treated parcels, which had not shown a

reduction in the number of members. Finally, columns (3) and (4) of Table 7 show a

reduction of 0.4 days associated to titling status in the number of days the child missed

school out of the last five days of classes.

Table 7 – Education

Children of the Household Head (6-14 years old)

School Achievement School Absenteeism (1) (2) (3) (4) Property Right 0.10 -0.37** (0.85) (2.39) Property Right 1989-91 0.39** -0.55** (2.02) (2.14) Property Right 1997-98 0.00 -0.31* (0.00) (1.81) Control Group Mean -1.12 -1.12 0.64 0.64 Notes: In columns (1) and (2), the dependent variable is the difference between the school grade each child is currently attending or the maximum grade attained (if not attending school) minus the grade corresponding to the child age. In columns (3) and (4), the dependent variable is the number of days the child missed school out of the last five days of classes. All the regressions control for child, parcel and original squatter pre-treatment characteristics: child age, child gender, surface of the parcel (in square meters); distance to creek (in blocks); distance to nearest non-squattered area (in blocks); a corner-of-the-block dummy; age, gender, nationality, and years of education of the original squatter; nationality and years of education of the father of the original squatter; and nationality and years of education of the mother of the original squatter. The robustness of each of these regressions is presented in Appendix Tables A.13 and A.14. Absolute value of t statistics are in parentheses * significant at 10%; ** significant at 5%; *** significant at 1%.

31 In addition to clustering the standard errors at the block and former owner levels, Appendix Tables A.13 and A.14 also report standard errors clustered at the household level.

27

V.4. Access to Credit

Markets in developing countries are highly imperfect and those who do not have enough

wealth tend to underinvest. This reduces the overall productivity of the economy

affecting the process of economic development. The imperfection in the credit market

largely affects the poor. The fact that the rate of interest on deposits is much lower than

the rate on loans means that the opportunity cost of capital for those who just want to

invest their own money is much lower than the opportunity cost for those who have to

borrow. The richer people also enjoy better direct access to credit. Thus, the poor then

underinvest relative to the rich, but also relative to what would happen if markets

functioned properly (Banerjee, 2004).

Land titles might serve as collateral improving the access of the untitled poor to capital

markets. To be functional, however, they should be appropriable by the lender at

reasonable cost in the event of default, thus bounding the lender’s losses and

encouraging the borrower to take care to avoid default (Besley, 1995).

The promise of land titling as a powerful policy instrument to attack poverty has been

recently reinvigorated in policy circles by the work of De Soto (2000). His work argues

vehemently that titled property creates capital because formal landholders could use

these assets as collateral for loans. In turn, this credit could be invested as capital

increasing their labor productivity and hence, the income of the poor.32 However,

rigorous evidence backing up these effects is scant and ambiguous (see, among others,

Feder et al., 1988; Alston et al., 1996; Carter and Olinto, 2002; Place and Migot-Adholla,

1998, and Field and Torero, 2003). In this section we investigate whether the possession

of a land title affects access to credit while in the next section we investigate the effect of

land titles on labor market outcomes.

Table 8a shows no differences across groups in the access to credit cards and banking

services; to credit from a bank, the government, a union or a cooperative; to non-

mortgage banking credit; or to mortgage credit. Indeed, these families show very low

access to any type of formal credit. The access to credit is higher for informal credit from

relatives, colleagues, neighbours, and friends, and for on-trust credit that families may

28

receive from the stores in which they perform their daily purchases (fiado), but titling

status show no effect on access to these informal sources of credit.33

Table 8b redoes the exercise separating the effect for the early and late treatment

groups. This separation is important because the late treatment group is not yet in a

legal situation to access the mortgage markets, which is the market in which the land

titles might be more effective.34 In this case, we do find a small but statistically significant

effect of land titling on the access to mortgage markets. Entitled households are 4%

more likely to access to this market.

Table 8a - Access To Credit

Credit Card and

Bank Account

(1)

Loan Received

(Bank, Govt, Union, Coop)

(2)

Banking Non-

Mortgage Loan

Received

(3)

Banking Mortgage

Loan Received

(4)

Informal Credit

(Relatives, Colleagues, Neighbors,

Friends) (5)

Grocery Store Loan

(6) Property Right -0.01 0.02 -0.01 0.02 -0.06 0.01 (0.71) (0.60) (0.64) (1.57) (1.00) (0.16) Control Group Mean 0.05 0.09 0.08 0.00 0.40 0.27

32 However, see Woodruff (2001) for a critical review of De Soto’s book. 33 The effects of land titling on access to credit remain non-significant when we perform our robustness checks. For the sake of space, we do not include an Appendix Table for each of these non-significant results, but they are available from the authors. 34 Remember from section II that the law established that the squatters could not transfer the property of the parcels for the first ten years after titling. Ten years from the 1997-98 titling had not elapsed at the time of the survey.

29

Table 8b - Access To Credit

Credit Card and

Bank Account

(1)

Loan Received

(Bank, Govt, Union, Coop)

(2)

Banking Non-

Mortgage Loan

Received

(3)

Banking Mortgage

Loan Received

(4)

Informal Credit

(Relatives, Colleagues, Neighbors,

Friends) (5)

Grocery Store Loan

(6) Property Right 1989-91 -0.01 0.05 0.01 0.04*** -0.04 0.02 (0.41) (1.05) (0.44) (3.19) (0.50) (0.31) Property Right 1997-98 -0.01 0.00 -0.04 0.00 -0.06 0.00 (0.70) (0.12) (1.16) (0.06) (1.03) (0.01) Control Group Mean 0.05 0.09 0.08 0.00 0.40 0.27 F-stat

0.02

0.79

1.64

8.61***

0.10

0.08

Notes: Credit Card and Bank Account is a dummy variable that equals 1 if the household head has a credit card or bank account, and 0 otherwise. Loan Received, Banking Non-Mortgage Loan Received, Banking Mortgage Loan Received, Informal Credit, and Grocery Store Loan are dummy variables that equal 1 if the household has received a credit from a bank, government, union, or cooperative; a non-mortgage credit from a bank; a mortgage credit from a bank; informal credit from relatives, colleagues, neighbors or friends; and credit from a grocery store, respectively, and 0 otherwise. All the regressions control for parcel and original squatter pre-treatment characteristics: surface of the parcel (in square meters); distance to creek (in blocks); distance to nearest non-squattered area (in blocks); a corner-of-the-block dummy; age, gender, nationality, and years of education of the original squatter; nationality and years of education of the father of the original squatter; and nationality and years of education of the mother of the original squatter. Absolute value of t statistics are in parentheses. The F-stat test the null hypothesis: Property Right 1989-91= Property Right 1997-98. * significant at 10%; ** significant at 5%; *** significant at 1%.

V.5. Performance in the Labor Market

Finally, we investigate the effect of land titling on labor market outcomes. One advantage

of our experiment in order to study the effects of land titles on labor market outcomes is

that both treated and control units are all in the same labor market. Heckman et al.

(1997) show that comparing individuals from different labor markets is the most

important source of bias when evaluating the impact of training programs on labor

market outcomes.

Our results show no differences between control and treatment households in their

performance in the labor market. Table 9 shows no significant differences in income of

the household head, total household income, total household income per capita, and

total household income per adult. There are also no differences in the employment rates

of the household heads, or in their pension status. These effects of land titling on labor

30

market performance also remain non-significant when we perform our robustness

checks.35

Table 9 - Labor Market

Household Head

Income

(1)

Total Household

Income

(2)

Total Household

Income per Capita

(3)

Total Household

Income per Adult

(4)

Employed Household

Head

(5)

Retired Household

Head

(6) Property Right -27.35 -43.56 1.04 -4.45 0.03 -0.02 (1.10) (1.27) (0.13) (0.38) (0.63) (0.68) Control Group Mean 272.54 374.59 73.71 118.73 0.72 0.08 Notes: Household Head Income is the total income earned by the household head in the previous month. Total Household Income is the total income earned by the household members in the previous month. Total Household Income per Capita is Total Household Income divide by the number of household members. Total Household Income per Adult is Total Household Income divide by the number of household members older than 16 years old. Retired Household Head is a dummy variable that equals 1 if the household head receives a pension from the social security system, and 0 otherwise. Employed Household Head is a dummy variable that equals 1 if the household head was employed the week before the survey, and 0 otherwise. All the regressions control for parcel and original squatter pre-treatment characteristics: surface of the parcel (in square meters); distance to creek (in blocks); distance to nearest non-squattered area (in blocks); a corner-of-the-block dummy; age, gender, nationality, and years of education of the original squatter; nationality and years of education of the father of the original squatter; and nationality and years of education of the mother of the original squatter. Absolute value of t statistics are in parentheses * significant at 10%; ** significant at 5%; *** significant at 1%.

VI. Conclusions

Eliminating poverty in the world is the most important goal of the international

community. But, what are the best policy instruments to achieve this goal? Academics,

policymakers and social actors are all involved in a process of learning the answer to this

important question. Behavioral hypothesis are formulated, interventions proposed and

their effects evaluated in a continuous search for better approaches to reduce poverty.

Recently, the potential positive effects of entitling the urban poor have gained attention

in policy circles and many countries adopted or are in the process of adopting

interventions to entitle urban squatters. Are these interventions going to be effective or

will end up being another delusion? What exactly could be expected from them? In other

words, what are the causal effects of urban land titling?

35 Field (2003) reports that land titles reduce the child labor. She also finds a substitution effect between working at home and in the market. Titled males appear to work more in the market while untitled males work more at home.

31

The empirical evaluation of the effects of land title registration programs typically poses a

major methodological challenge. In most historical experiences, the allocation of property

rights across families is not random but based on wealth, family characteristics,

individual effort, or other selective mechanisms. Thus, the personal characteristics that

determine the likelihood of receiving land titles are likely to be correlated with the

outcomes under study. Since some of these individual characteristics are

unobservables, this correlation creates a selection problem that impedes the proper

evaluation of the effects of property right acquisition. Previous work exploited standard

exclusion restrictions or the geographically variability of policy interventions to deal with

this nuisance. In this paper, instead, we exploited a natural experiment in the allocation

of land titles that we believe provides us with a credible strategy to identify the effect of

land titling on the outcomes of interest. Naturally, this assertion does not neglect the

unavoidable fact that in observational studies behind every causal conclusion there must

lay some causal assumption that is not directly testable. Much to the contrary, it is

precisely the recognition of this circumstance that leads us to believe that this paper

contributes greatly to the existent literature.

A natural experiment provides an instrument to understand the source of variability of the

data exploited in order to defend that the control group estimates what would have

happened to the treated group in the absence of the intervention. In section II we

provided considerable evidence on the exogeneity of the allocation of land titles. Section

IV scrutinized implications of this evidence. We showed that the observable

characteristics of the parcels were on average very similar between treatment and

control groups. We also showed that the pre-treatment characteristics of the squatters

were also very similar between these two groups.

Exploiting this natural experiment in the allocation of land titles across squatters in a

poor suburban area of Buenos Aires, Argentina, we find significant effects on housing

investment, household size, and school achievement. The quality of the houses is

substantially higher in the titled parcels. Moreover, households in the titled parcels have

smaller size (although their houses are larger) and seem to invest more in the education

of their children. However, we do not find much effect on the access to credit markets as

32

a result of titling, and hence, we do not find significant effects on employment and labor

income.

33

References Ahn, Hyungtaik, and James L. Powell (1993). Semiparametric Estimation of Censored Selection

Models with a Nonparametric Selection Mechanism. Journal of Econometrics 58, pp. 3-29. Alston L, G. Libecap, and R. Schneider (1996). The Determinants and Impact of Property Rights:

Land Titles on the Brazilian Frontier. Journal of Law, Economics & Organization, 12: pp. 25-61.

Alston L, G. Libecap, and B. Mueller (1999). Titles and Land Use: The Development of Property Rights on the Brazilian Amazon, University of Michigan Press.

Amis, Martin (2004) “In Search of Dieguito”, The Guardian, October 1. Angrist, J. (1990). Lifetime Earnings and the Vietnam Era Draft Lottery: Evidence from Social

Security Administrative Records. American Economic Review, Vol. 80, No. 3, pp. 313-336. Angrist, J. (2001). Estimation of Limited Dependent Variable Models with Dummy Endogenous

Regressors: Simple Strategies for Empirical Practice, Journal of Business and Economic Statistics 19, pp. 2-16.

Angrist J.D., G. Imbens and D. Rubin (1996). Identification of Causal Effects Using Instrumental Variables (with discussion), Journal of the American Statistical Association 91: 444-455.

Banerjee, A. (2004): Inequality and Investment, mimeo MIT. Barzel, Yoram (2002): A Theory of the State: Economic Rights, Legal Rights, and the Scope of

the State, Cambridge University Press. Becker, Gary and H. Lewis (1973). On the Interaction between the Quantity and Quality of

Children. Journal of Political Economy 81(2), pp. S279-88. Besley, T. (1995). Property Rights and Investments Incentives: Theory and Evidence from

Ghana. Journal of Political Economy 103: pp. 903-937. Biglan, A, D. Hood, P. Borozovsky, L. Ochs, D. Ary, and C. Black (1991). Subject Attrition in

Prevention Research, in Drug abuse prevention intervention research: Methodological issues, C. G. Luekfeld and W. Bukowski (eds.), National Institute on Drug Abuse.

Binswanger, H., K. Deninger and G. Feder (1995): “Power, Distortions, Revolt, and Reform in Agricultural Land Relations”, Handbook of Development Economics, Volume IIIB, Amsterdam, Elsevier.

Birdsall, N. (1988). “Economic Approaches to Population Growth.” Chapter 12 in the Handbook of Development Economics, Volume 1, edited by H. Chenery and T.N. Srinivasan. North Holland: Elsevier Sciences Publishers.

Brasselle, A., F. Gaspart and J.P. Platteau (2002): Land Tenure Security and Investment Incentives: Puzzling Evidence from Burkina Faso. Journal of Development Economics 67 (2), pp. 373-418.

Cain, M. (1985): On the Relationship between landholding and Fertility, Population Studies 39, pp. 5-15.

Campbell, Donald T. and H. Lawrence Ross (1970). The Connecticut Crackdown on Speeding: Time Series Data in Quasi-Experimental Analysis, In Edward R. Tuffs, ed. The Quantitative Analysis of Social Problems. Reading, Mass: Addison-Wesley, pp. 110-118.

Card, D. and A. Kruger (1995): Myth and Measurement, Princeton University Press. Carter M. and P. Olinto (2002). Getting Institutions Right for Whom? Credit Constraints and the

Impact of Property Rights on the Quantity and Composition of Investment. American Journal of Agricultural Economics 85 (1), pp. 173-86.

Céspedes, M. (1984). Por una Tierra Nuestra, documentary movie. CEUR (1984). Condiciones de Hábitat y Salud de los Sectores Populares. Un Estudio Piloto en el

Asentamiento San Martín de Quilmes. Buenos Aires: Centro de Estudios Urbanos y Regionales.

David, P. And W. Sundstrom (1984): “Bargains, Bequests, and Births: An Essay on Intergenerational Conflict, Reciprocity, and the Demand for Children in Agrarian Societies”, Stanford Project on the History of Fertility Control, Working Paper 12.

Demsetz, H. (1967): “Toward a theory of property rights”, American Economic Review 57 (2), 347-359.

De Soto, H. (2000). The Mystery of Capital: Why Capitalism Triumphs in the West and Fails Everywhere Else. New York: Basic Books.

34

Di Tella, R. and E. Schargrodsky (2004). Do Police Reduce Crime? Estimates Using the Allocation of Police Forces after a Terrorist Attack. American Economic Review 94 (1), pp. 115-133.

Do QT, L. Iyer (2002). Land Rights and Economic Development: Evidence from Vietnam, mimeo, MIT.

Fara, L. (1989). Luchas Reivindicativas Urbanas en un Contexto Autoritario. Los Asentamientos de San Francisco Solano, en Elizabeth Jelin (comp.), Los Nuevos Movimientos Sociales. Buenos Aires: Centro Editor de América Latina.

Feder, G., T. Onchan, Y. Chalamwong and C. Hongladarom (1988). Land Policies and Farm Productivity in Thailand, Johns Hopkins University Press, Baltimore.

Field, E. and M. Torero (2003). Do Property Titles Increase Access to Credit? Evidence from Peru, mimeo, Harvard University.

Field, E. (2003). Entitled to Work: Urban Property Rights and Labor Supply in Peru, Research Program in Development Studies Working Paper #220, Princeton University.

Heckman, J. (1979): “Sample Selection Bias as a Specification Error”, Econometrica, 47, pp. 153-61.

Heckman, James, H. Ichimura, and P. Todd (1997). Matching as an Econometric Evaluation Estimator: Evidence from Evaluating a Job Training Programme. Review of Economic Studies 64(4), pp. 605-54.

Imbens, G., D. Rubin and B. Sacerdote (2001). Estimating the Effect of Unearned Income on Labor Supply, Earnings, Savings and Consumption: Evidence from a Survey of Lottery Players. American Economic Review 91 (4), pp.778-794.

Izaguirre, I., and A. Aristizabal (1988). Las Tomas de Tierras en la Zona Sur del Gran Buenos Aires. Un Ejercicio de Formación de Poder en el Campo Popular. Buenos Aires: Centro Editor de América Latina.

Jimenez, E. (1984). Tenure Security and Urban Squatting. REStat 66: pp. 556-67. Kish, L. (1987): Statistical Designs for Research. New York: Wiley. Lanjouw, J., and P. Levy (2002). Untitled: A Study of Formal and Informal Property Rights in

Urban Ecuador. Economic Journal 112: pp.986-1019. Libecap, G. (1989): Contracting for Property Rights, Cambridge University Press. Miceli T, C. Sirmans and J. Kieyah (2001). The Demand for Land Title Registration: Theory with

Evidence from Kenya. American Law and Economics Review, 3: pp. 275-287. North, D. and P. Thomas (1973): The Rise of the Western World: A New Economic History,

Cambridge University Press. Nugent (1985): “The Old-Age Security Motive for Fertility”, Population and Development Review,

11, pp. 75-97. Place, F. and S.E. Migot-Adholla (1998). The Economic Effects of Land Registration for

Smallholder Farms in Kenya: Evidence from Nyeri and Kakamega Districts. Land Economics, 74 (3), pp. 360-73.

Portner, Clauss (2001): “Children as Insurance”, Journal of Population Economics, 14, pp. 119-136.

Rosenbaum, P. and D. Rubin (1983): “The Central Role of the Propensity Score in Observational Studies for Causal Effects”, Biometrika 70, pp. 41-55.

Woodruff, Christopher (2001). Review of De Soto’s The Mistery of Capital. Journal of Economic Literature 39: pp.1215-1223.

35

Appendix Table A.1 - GOOD WALLS (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Property Right 0.20*** 0.19*** 0.19*** 0.14*** 0.20*** 0.20*** 0.18*** 0.19*** 0.10** (3.47) (3.32) (3.37) (2.64) (3.19) (4.26) (2.63) (3.22) (2.35)Property Right 1989-91 0.23*** 2.78) (Property Right 1997-98 0.19*** 2.90) (Surface -0.00** -0.00*** -0.00** -0.00** -0.00*** -0.00*** -0.00*** -0.00**

(2.59) (2.65) (2.25) (2.14) (4.16) (2.63) (2.60) (2.44)Distance to Creek 0.07** 0.07** 0.02 0.07** 0.07** 0.07** 0.07** 0.08*** (2.21) (2.29) (1.01) (2.33) (2.80) (2.11) (2.00) (3.27)Block Corner -0.05 -0.09 -0.00 -0.05 -0.05 -0.05 -0.05 -0.08

(0.53) (1.15) (0.05) (0.55) (0.67) (0.57) (0.53) (1.28)Distance to Non-Squattered 0.03 0.04 0.05* 0.03 0.03 0.03 0.03 0.02Area (0.94) (1.49) (1.89) (0.81) (1.61) (0.93) (0.94) (0.89)Age of Original Squatter <49 0.28 0.31 0.28 0.28* 0.27 0.28 (1.25) (1.53) (1.15) (1.80) (1.20) (1.26)Age of Original Squatter > 50 0.26 0.33 0.26 0.26 0.25 0.26 (1.18) (1.59) (1.06) (1.58) (1.14) (1.19)Female Original Squatter 0.05 -0.04 0.05 0.05 0.05 0.05 (0.83) (0.86) (0.84) (0.82) (0.82) (0.82)Argentine Original Squatter -0.16 -0.12 -0.16 -0.16 -0.16 -0.16 (1.11) (0.95) (1.23) (1.24) (1.15) (1.13)Years of Education of the -0.02 -0.01 -0.02 -0.02 -0.02 -0.02Original Squatter (0.96) (0.73) (0.92) (1.34) (0.96) (0.97)Argentine Father of the -0.23** -0.16 -0.23** -0.23*** -0.23** -0.23**Original Squatter (2.04) (1.53) (2.62) (3.57) (2.01) (2.02)Years of Education of 0.02 -0.01 0.02 0.02 0.01 0.02Original Squatter’s Father (0.58) (0.38) (0.64) (0.78) (0.53) (0.59)Argentine Mother of the 0.27** 0.15 0.27** 0.27 0.28** 0.27**Original Squatter (2.40) (1.42) (2.19) (1.49) (2.42) (2.37)Years of Education of 0.00 0.03 0.00 0.00 0.00 0.00Original Squatter’s Mother (0.05) (1.00) (0.05) (0.08) (0.08) (0.04)Constant 0.44 0.50*** 0.58*** 0.34 0.44 0.44 0.47 0.45 0.57***

(1.34) (11.92) (3.81) (1.21) (1.11) (1.68) (1.43) (1.37) (4.78)F-stat 0.17Observations 291 295 295 393 291 291 291 291 275 441Notes: The dependent variable is a dummy that equals 1 if the house has walls of good quality (Ladrillo, piedra, bloque u hormigón con revoque o revestimiento externo), and 0 otherwise. Column (1) is the regression summarized in Column 1 of Table 5. Column (2) uses no controls, and Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. Column (9) presents the matching estimator on the propensity score of the probability of attrition. The regression in Column (10) is run on all the interviewed households (for any time of household arrival). Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

36

Appendix Table A.2 - GOOD ROOF (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Property Right 0.15** 0.14** 0.15*** 0.14** 0.15** 0.15** 0.16** 0.13** 0.11*** (2.48) (2.41) (2.65) (2.56) (2.27) (2.69) (2.22) (2.15) (2.55)Property Right 1989-91 0.22*** 2.63) (Property Right 1997-98 0.11* 1.66) (Surface 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 (0.42) (0.46) (1.19) (0.36) (0.25) (0.43) (0.39) (0.98)Distance to Creek 0.03 0.03 0.02 0.03 0.03 0.03 0.02 0.02 (0.93) (1.00) (0.91) (0.83) (1.32) (0.94) (0.54) (1.07)Block Corner 0.08 0.08 0.13* 0.08 0.08 0.08 0.08 0.01 (0.90) (1.02) (1.67) (0.94) (0.68) (0.91) (0.91) (0.26)Distance to Non-Squattered Area -0.01 -0.02 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 (0.31) (0.60) (0.30) (0.31) (0.54) (0.31) (0.31) (0.58)Age of Original Squatter <49 -0.09 -0.01 -0.09 -0.09 -0.09 -0.09 (0.41) (0.03) (0.33)(0.40) (0.40) (0.40)Age of Original Squatter > 50 -0.09 0.00 -0.09 -0.09 -0.08 -0.08 (0.38) (0.00) (0.30)(0.36) (0.37) (0.35)Female Original Squatter -0.04 -0.04 -0.04 -0.04 -0.04 -0.04 (0.64) (0.87) (1.21)(0.65) (0.64) (0.65)Argentine Original Squatter 0.18 0.17 0.18 0.18 0.18 0.17 (1.20) (1.31) (1.39)(1.14) (1.21) (1.15)Years of Education of the Original 0.00 0.01 0.00 0.00 0.00 0.00Squatter (0.22) (0.42) (0.21) (0.13) (0.22) (0.20)Argentine Father of the Original -0.02 0.03 -0.02 -0.02 -0.02 -0.01Squatter (0.15) (0.25) (0.14) (0.21) (0.16) (0.11)Years of Education of Original -0.00 -0.01 -0.00 -0.00 -0.00 -0.00Squatter’s Father (0.03) (0.52) (0.03) (0.05) (0.02) (0.00)Argentine Mother of the Original -0.09 -0.14 -0.09 -0.09 -0.09 -0.09Squatter (0.73) (1.26) (0.62) (0.62) (0.74) (0.78)Years of Education of Original 0.00 0.01 0.00 0.00 0.00 0.00Squatter’s Mother (0.17) (0.19) (0.18) (0.26) (0.16) (0.14)Constant 0.21 0.32*** 0.22 0.14 0.21 0.21 0.20 0.24 0.22* (0.62) (7.53) (1.41) (0.49) (0.58) (0.51) (0.59) (0.72) (1.85)F-stat 1.48Observations 293 297 297 395 293 293 293 293 278 445Notes: The dependent variable is a dummy that equals 1 if the house has a roof of good quality (Cubierta asfáltica o membrana, Baldosa o losa, and Pizarra o teja), and 0 otherwise. Column (1) is the regression summarized in Column 2 of Table 5. Column (2) uses no controls, and Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. Column (9) presents the matching estimator on the propensity score of the probability of attrition. The regression in Column (10) is run on all the interviewed households (for any time of household arrival). Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

37

Appendix Table A.3 - CONSTRUCTED SURFACE (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Property Right 8.26** 7.99** 9.89*** 4.98 8.26** 8.26 9.87** 8.10** 8.61*** (2.33) (2.33) (2.87) (1.56) (2.16) (1.45) (2.40) (2.41) (3.02)Property Right 1989-91 10.36** 2.09) (Property Right 1997-98 7.17* (1.80) Surface -0.01 -0.00 0.01 -0.01 -0.01 -0.01 -0.01 0.01

(0.33) (0.02) (0.66) (0.35) (0.33) (0.28) (0.35) (0.68)Distance to Creek 5.85*** 6.42*** 2.62** 5.85*** 5.85*** 6.03*** 5.50*** 4.79*** (3.00) (3.42) (2.05) (3.08) (3.01) (3.07) (2.69) (3.15)Block Corner 4.53 3.98 3.64 4.53 4.53 4.72 4.54 8.37**

(0.89) (0.82) (0.80) (0.78) (1.62) (0.93) (0.89) (2.10)Distance to Non-Squattered 3.65** 4.51*** 3.10* 3.65 3.65 3.67** 3.66** 2.02Area (2.05) (2.60) (1.93) (1.60) (1.42) (2.06) (2.05) (1.36)Age of Original Squatter <49 -3.94 -6.36 -3.94 -3.94 -3.31 -3.83 (0.29) (0.52) (0.50) (0.39) (0.25) (0.28)Age of Original Squatter > 50 -2.32 -3.87 -2.32 -2.32 -1.79 -2.10 (0.17) (0.31) (0.28) (0.21) (0.13) (0.16)Female Original Squatter -0.69 -1.97 -0.69 -0.69 -0.68 -0.72 (0.19) (0.65) (0.20) (0.23) (0.19) (0.20)Argentine Original Squatter -7.53 -6.71 -7.53 -7.53 -7.16 -7.74 (0.87) (0.86) (0.85) (1.49) (0.83) (0.89)Years of Education of the -0.06 -0.11 -0.06 -0.06 -0.07 -0.08Original Squatter (0.06) (0.13) (0.06) (0.05) (0.07) (0.08)Argentine Father of the -5.47 -3.29 -5.47 -5.47 -5.70 -5.30Original Squatter (0.80) (0.51) (0.83) (1.14) (0.83) (0.77)Years of Education of Original -3.32** -3.25** -3.32** -3.32* -3.24** -3.30**Squatter’s Father (2.06) (2.39) (2.22) (1.91) (2.01) (2.05)Argentine Mother of the 10.74 7.41 10.74 10.74*** 10.56 10.52Original Squatter (1.55) (1.14) (1.28) (3.06) (1.52) (1.52)Years of Education of Original 3.58** 3.54** 3.58* 3.58* 3.52** 3.55**Squatter’s Mother (2.08) (2.24) (1.69) (1.78) (2.05) (2.07)Constant 56.53*** 67.63*** 46.28*** 62.45*** 56.53*** 56.53** 54.33*** 57.53*** 49.96***

(2.87) (26.49) (5.00) (3.73) (3.06) (2.66) (2.73) (2.91) (6.78)F-stat 0.37Observations 295 299 299 397 295 295 295 295 279 447Notes: The dependent variable is the constructed surface in squared meters. Column (1) is the regression summarized in Column 3 of Table 5. Column (2) uses no controls, and Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. Column (9) presents the matching estimator on the propensity score of the probability of attrition. The regression in Column (10) is run on all the interviewed households (for any time of household arrival). Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

38

Appendix Table A.4 - CONCRETE SIDEWALK

(1) (2) (3) (4) (5) (6) (8) (9) (10)Property Right 0.11** 0.08 0.12** 0.10** 0.11 0.10 0.09 0.16*** (2.17) (1.43) (2.24) (2.30) (1.55) (1.63) (1.61) (3.85)Property Right 1989-91 0.16** (2.16)Property Right 1997-98 0.09 (1.54)

-0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00(0.36) (0.69) (0.50) (0.33) (0.20) (0.39) (0.39)

Distance to Creek 0.09*** 0.09*** 0.08*** 0.09** 0.09** 0.09*** 0.08*** 0.10*** (3.20) (3.19) (4.55) (2.35) (2.23) (3.13) (4.89)Block Corner -0.11

(7)0.11

(1.61)

Surface

(1.15)

(2.81)-0.14** -0.12* -0.11 -0.11* -0.11 -0.11 -0.07

(1.46) (1.98) (1.95) (1.53) (1.85) (1.48) (1.45) (1.29)Distance to Non-Squattered -0.07** -0.07*** -0.05** -0.07* -0.07 -0.07** -0.07** -0.08***Area (2.52) (2.66) (2.20) (1.70) (1.72) (2.52) (2.51) (4.14)Age of Original Squatter <49 -0.29 -0.27* -0.29*** -0.29*** -0.29 -0.29 (1.59) (1.78) (3.60) (4.34) (1.60) (1.57)Age of Original Squatter > 50 -0.19 -0.20 -0.19*** -0.19** -0.19 -0.19 (1.05) (1.31) (2.75) (2.22) (1.06) (1.02)Female Original Squatter -0.05 -0.06 -0.05 -0.05 -0.05 -0.05 (0.94) (1.60) (0.90) (1.45) (0.95) (0.96)Argentine Original Squatter 0.07 0.07 0.07 0.07 0.06 0.06 (0.51) (0.66) (0.42) (0.76) (0.48) (0.47)Years of Education of the -0.02 -0.02 -0.02* -0.02** -0.02 -0.02Original Squatter (1.38) (1.44) (1.74) (2.73) (1.38) (1.40)Argentine Father of the Original -0.04 -0.03 -0.04 -0.04 -0.03 -0.03Squatter (0.35) (0.34) (0.42) (0.33) (0.33) (0.31)Years of Education of Original 0.03 0.03 0.03 0.03* 0.03 0.03Squatter’s Father (1.35) (1.45) (1.55) (1.74) (1.33) (1.37)Argentine Mother of the Original -0.02 -0.04 -0.02 -0.02 -0.02 -0.03Squatter (0.20) (0.46) (0.14) (0.22) (0.18) (0.24)Years of Education of Original -0.03 -0.03 -0.03 -0.03 -0.03 -0.03Squatter’s Mother (1.18) (1.37) (1.14) (1.02) (1.16) (1.20)Constant 1.02*** 0.67*** 0.71*** 1.01*** 1.02*** 1.02** 1.04*** 1.04*** 0.67*** (3.68) (17.19) (5.17) (4.78) (3.60) (2.33) (3.71) (3.74) (6.34)F-stat 0.75Observations 296 300 300 398 296 296 296 296 280 448Notes: The dependent variable is a dummy that equals 1 if the house has a sidewalk made of concrete, and 0 otherwise. Column (1) is the regression summarized in Column 4 of Table 5. Column (2) uses no controls, and Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. Column (9) presents the matching estimator on the propensity score of the probability of attrition. The regression in Column (10) is run on all the interviewed households (for any time of household arrival). Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

39

Appendix Table A.5 - POINTS IN CITY OF QUILMES (1) (2) (3) (4) (5) (6 (7) (8) (9) (10)

Property Right 8.42*** 7.45*** 8.07*** 5.26*** 8.42*** 8.42*** 10.18*** 7.51*** 6.97*** (3.64) (3.38) (3.58) (2.60) (3.93) (2.97) (3.80) (3.47) (3.87)Property Right 1989-91 6.30* 1.95) (Property Right 1997-98 9.53*** 3.67) (Surface -0.02 -0.02 -0.01 -0.02 -0.02* -0.02 -0.02 -0.02*

(1.22) (1.46) (1.42)(0.68) (1.92) (1.13) (1.19) (1.87)Distance to Creek 2.40* 2.32* -0.15 2.40* 2.40** 2.60** 2.76** 3.00***

(1.88) (1.90) (1.80)(0.18) (2.85) (2.02) (2.07) (3.13)Block Corner 0.38 -0.70 2.25 0.38 0.38 0.58 0.37 0.02

(0.11) (0.22) (0.11)(0.78) (0.08) (0.18) (0.11) (0.01)Distance to Non-Squattered Area -0.03 -0.14 0.06 -0.03 -0.03 -0.00 -0.03 -0.48 (0.02) (0.13) (0.02)(0.06) (0.03) (0.00) (0.03) (0.52)Age of Original Squatter <49 -2.08 -2.03 -2.08 -2.08 -1.39 -2.18 (0.24) (0.26) (0.15) (0.16) (0.16) (0.25)Age of Original Squatter > 50 -2.78 -1.87 -2.78 -2.78 -2.19 -3.01 (0.32) (0.24) (0.20) (0.21) (0.25) (0.34)Female Original Squatter -3.19 -4.29** -3.19 -3.19 -3.18 -3.16 (1.38) (2.22) (1.45) (1.71) (1.37) (1.37)Argentine Original Squatter 6.85 3.97 6.85 6.85** 7.25 7.06 (1.21) (0.80) (0.96) (2.35) (1.28) (1.25)Years of Education of the Original 0.73 0.46 0.73 0.73 0.71 0.74Squatter (1.12) (0.88) (0.91) (0.77) (1.10) (1.14)Argentine Father of the Original -9.85** -6.54 -9.85** -9.85** -10.10** -10.02**Squatter (2.20) (1.61) (2.21) (2.41) (2.25) (2.23)Years of Education of Original -1.15 -0.66 -1.15 -1.15** -1.06 -1.17Squatter’s Father (1.09) (0.77) (1.39) (2.24) (1.01) (1.11)Argentine Mother of the Original -1.92 -1.93 -1.92 -1.92 -2.12 -1.69Squatter (0.42) (0.47) (0.33) (0.45) (0.47) (0.37)Years of Education of Original -1.24 -0.82 -1.24 -1.24 -1.30 -1.21Squatter’s Mother (1.11) (0.82) (1.02) (0.98) (1.16) (1.08)Constant 36.75*** 22.71*** 24.63*** 35.68*** 36.75** 36.75* 34.33*** 35.74*** 24.73***

(2.86) (4.08)(13.82) (3.36) (2.22) (1.98) (2.64) (2.77) (5.31)F-stat 0.88Observations 295 299 299 397 295 295 295 295 279 446Notes: The dependent variable measures the overall quality of each house relative to houses in downtown Quilmes (the main locality of the county) from 0 to 100 points. Column (1) is the regression summarized in Column 5 of Table 5. Column (2) uses no controls, and Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. Column (9) presents the matching estimator on the propensity score of the probability of attrition. The regression in Column (10) is run on all the interviewed households (for any time of household arrival). Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

40

Appendix Table A.6 - DURABLE CONSUMPTION Refrigerator with

Freezer Refrigerator without

Freezer

(1)

(2)

Washing Machine

(3)

TV

(4)

Cellular Phone

(5) Property Right 0.05 0.04 0.04 -0.01 -0.01

(0.92) (0.61) (0.67) (0.40) (0.32)Surface -0.00 0.00 0.00 -0.00 0.00

(1.29) (0.83) (0.52) (0.45) (0.97)Distance to Creek 0.09*** -0.03 0.06** 0.06*** 0.03*

(2.99) (0.83) (1.98) (3.11) (1.88)Block Corner -0.04 0.09 0.06 0.03 0.02

(0.50) (1.06) (0.74) (0.65) (0.54)Distance to Non-Squattered Area -0.01 0.02 0.06** 0.02 0.00 (0.27) (0.80) (2.00) (0.95) (0.32)Age of Original Squatter <49 -0.06 -0.16 0.34 -0.08 0.03 (0.29) (0.77) (1.58) (0.68) (0.34)Age of Original Squatter > 50 -0.09 -0.15 0.32 -0.09 0.04 (0.45) (0.71) (1.47) (0.79) (0.43)Female Original Squatter 0.02 -0.04 -0.05 0.03 0.00 (0.35) (0.74) (0.85) (1.04) (0.20)Argentine Original Squatter 0.05 -0.06 -0.02 -0.22*** -0.01 (0.36) (0.37) (0.13) (2.67) (0.22)Years of Education of the Original 0.02 -0.02 0.02 0.01 -0.00Squatter (1.00) (1.28) (1.39) (1.54) (0.65)Argentine Father of the Original -0.10 0.05 -0.05 0.06 0.04Squatter (0.86) (0.43) (0.41) (0.90) (0.90)Years of Education of Original 0.01 0.00 -0.01 -0.00 0.02* Squatter’s Father (0.21) (0.18) (0.57) (0.14) (1.91)Argentine Mother of the Original -0.03 0.06 -0.07 0.08 -0.02Squatter (0.30) (0.50) (0.61) (1.18) (0.49)Years of Education of Original 0.01 -0.01 -0.05* -0.01 -0.02**Squatter’s Mother (0.36) (0.39) (1.83) (0.81) (2.05)Constant 0.28 0.70** 0.33 0.95*** -0.05

(0.93) (2.16) (1.07) (5.35) (0.37)Observations 307 307 307 308 308Notes: The dependent variable of each column is a dummy that equals 1 if the household possess the good (or service), and 0 otherwise. Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

41

Appendix Table A.7 – NUMBER OF HOUSEHOLD MEMBERS

(1) (2) (3) (4) (5) (6) (7) (8) (9)Property Right -0.95*** -0.87*** -0.86*** -0.97*** -0.95** -0.95** -1.18*** -1.00*** (2.83) (2.68) (2.58) (3.24) (2.60) (2.85) (3.00) (2.88)Property Right 1989-91 -1.18** (2.50)Property Right 1997-98 -0.83** (2.19)Surface 0.00 0.00 0.00 0.00 0.00 0.00 0.00

(0.66) (0.65) (0.90) (0.71) (1.26) (0.59) (0.67)Distance to Creek 0.02 -0.03 0.07 0.02 0.02 -0.01 0.05

(0.08) (0.20) (0.59) (0.08) (0.09) (0.04) (0.28)Block Corner 0.01 0.04 0.11 0.01 0.01 -0.01 0.01

(0.03) (0.09) (0.26) (0.03) (0.03) (0.03) (0.02)Distance to Non-Squattered 0.02 -0.01 0.09 0.02 0.02 0.02 0.02Area (0.13) (0.11) (0.62) (0.13) (0.15) (0.12) (0.13)Age of Original Squatter <49 -0.89 -0.95 -0.89 -0.89 -0.94 -0.90 (0.75) (0.87) (0.68) (0.65) (0.78) (0.76)Age of Original Squatter > 50 -1.96 -1.79 -1.96 -1.96 -1.99* -1.99* (1.64) (1.62) (1.49) (1.53) (1.66) (1.66)Female Original Squatter -0.07 -0.13 -0.07 -0.07 -0.08 -0.07 (0.21) (0.45) (0.20) (0.16) (0.23) (0.21)Argentine Original Squatter -0.90 -0.53 -0.90 -0.90 -0.95 -0.87 (1.07) (0.71) (1.15) (1.17) (1.14) (1.04)Years of Education of the -0.07 -0.10 -0.07 -0.07 -0.07 -0.06Original Squatter (0.70) (1.25) (0.70) (0.48) (0.69) (0.67)Argentine Father of the Original 1.23* 0.98 1.23* 1.23* 1.26* 1.22*Squatter (1.84) (1.58) (1.79) (1.91) (1.88) (1.81)Years of Education of Original -0.18 -0.18 -0.18 -0.18 -0.19 -0.18Squatter’s Father (1.18) (1.41) (1.27) (1.34) (1.24) (1.19)Argentine Mother of the Original -0.74 -0.69 -0.74 -0.74 -0.72 -0.72Squatter (1.10) (1.10) (1.05) (1.45) (1.07) (1.06)Years of Education of Original 0.06 0.03 0.06 0.06 0.07 0.07Squatter’s Mother (0.39) (0.22) (0.43) (0.41) (0.42) (0.41)Constant 8.35*** 6.06*** 5.74*** 8.29*** 8.35*** 8.35*** 8.62*** 8.23***

(4.64) (24.99) (6.36) (5.43) (4.77) (4.93) (4.74) (4.55)F-stat 0.47Observations 309 313 313 415 309 309 309 309 292Notes: The dependent variable is the total number of household members. Column (1) is the regression summarized in Column 1 of Table 6. Column (2) uses no controls. Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. Column (9) presents the matching estimator on the propensity score of the probability of attrition. Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

42

Appendix Table A.8 – HOUSEHOLD HEAD SPOUSE

(1) (2) (3) (4) (5) (6) (7) (8) (9)Property Right -0.01 -0.02 -0.01 -0.02 -0.01 -0.01 0.01 -0.03

(0.27) (0.37) (0.37) (0.50) (0.26) (0.41) (0.20) (0.58)Property Right 1989-91 -0.03 (0.36) Property Right 1997-98 -0.01 (0.13) Surface -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00

(0.93) (1.45) (0.96) (0.97) (1.53) (0.88) (0.92)Distance to Creek 0.02 0.01 0.01 0.02 0.02 0.02 0.02 (0.65) (0.47) (0.73) (0.69) (0.57) (0.74) (0.69)Block Corner 0.03 0.06 0.02 0.03 0.03 0.03 0.03

(0.42) (0.84) (0.38) (0.45) (0.46) (0.46) (0.42)Distance to Non- 0.01 0.00 0.02 0.01 0.01 0.01 0.01Squattered Area (0.37) (0.03) (1.08) (0.36) (0.39) (0.37) (0.36)Age of Original 0.15 0.06 0.15 0.15 0.15 0.15Squatter <49 (0.81) (0.33) (0.68) (0.63) (0.84) (0.80)Age of Original 0.12 0.02 0.12 0.12 0.12 0.12Squatter > 50 (0.65) (0.12) (0.53) (0.44) (0.66) (0.64)Female Original -0.28*** -0.31*** -0.28*** -0.28*** -0.28*** -0.28***Squatter (5.40) (7.13) (4.84) (4.68) (5.38) (5.39)Argentine Original -0.01 0.01 -0.01 -0.01 0.00 -0.00Squatter (0.05) (0.08) (0.05) (0.04) (0.00) (0.03)Years of Education of 0.02* 0.01 0.02* 0.02** 0.02* 0.03*the Original Squatter (1.71) (0.99) (1.78) (2.26) (1.71) (1.72)Argentine Father of the -0.05 -0.02 -0.05 -0.05 -0.05 -0.05Original Squatter (0.45)

(0.19) (0.55) (0.78) (0.48) (0.45)Years of Education of 0.00 0.02 0.00 0.00 0.00 0.00Original Squatter’s Father (0.02) (0.86) (0.02) (0.02) (0.07) (0.02)Argentine Mother of -0.06 -0.08 -0.06 -0.06 -0.06 -0.06the Original Squatter (0.58) (0.87) (0.64) (0.54) (0.60) (0.56)Years of Education of -0.02 0.00 -0.02 -0.02 -0.02 -0.02Original Squatter’s Mother (0.81) (0.12) (0.65) (0.86) (0.84) (0.80)Constant 0.77*** 0.74*** 0.84*** 0.73*** 0.77*** 0.77** 0.74*** 0.76***

(2.81) (19.44) (6.01) (3.15) (2.70) (2.42) (2.67) (2.77)F-stat 0.05Observations 309 313 313 415 309 309 309 309 292Notes: The dependent variable is a dummy that equals 1 if the household head lives with a spouse, and 0 otherwise. Column (1) is the regression summarized in Column 2 of Table 6. Column (2) uses no controls. Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. Column (9) presents the matching estimator on the propensity score of the probability of attrition. Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

43

Appendix Table A.9 - CHILDREN >14 YEARS OLD

(1) (2) (3) (4) (5) (6) (7) (8) (9)Property Right 0.03 0.03 0.05 -0.03 0.03 0.03 -0.12 0.02

(0.16) (0.16) (0.32) (0.18) (0.16) (0.17) (0.58) (0.15)Property Right 1989-91 -0.22 (0.89)Property Right 1997-98 0.16 (0.81)Surface 0.00 0.00 0.00 0.00 0.00 0.00 0.00

(0.58) (0.65) (1.63) (0.56) (0.90) (0.50) (0.62)Distance to Creek 0.09 0.07 0.02 0.09 0.09 0.07 0.13 (0.91) (0.81) (0.36) (0.79) (1.21) (0.75) (1.29)Block Corner -0.06 -0.03 0.14 -0.06 -0.06 -0.08 -0.06

(0.24) (0.14) (0.62) (0.26) (0.36) (0.30) (0.25)Distance to Non- 0.09 0.07 0.07 0.09 0.09 0.09 0.09Squattered Area (1.01) (0.85) (0.83) (1.02) (1.44) (1.00) (1.01)Age of Original 0.65 0.76 0.65 0.65 0.62 0.63Squatter <49 (1.02) (1.32) (1.09) (1.10) (0.97) (1.00)Age of Original 0.45 0.60 0.45 0.45 0.43 0.42Squatter > 50 (0.71) (1.04) (0.75) (0.74) (0.67) (0.66)Female Original -0.01 -0.12 -0.01 -0.01 -0.01 -0.01Squatter (0.06) (0.79) (0.06) (0.07) (0.08) (0.05)Argentine Original -0.10 0.08 -0.10 -0.10 -0.13 -0.07Squatter (0.22) (0.21) (0.24) (0.19) (0.30) (0.15)Years of Education of -0.00 -0.01 -0.00 -0.00 -0.00 -0.00the Original Squatter (0.06) (0.17) (0.06) (0.05) (0.05) (0.02)Argentine Father of the 0.46 0.21 0.46 0.46* 0.48 0.44Original Squatter (1.30) (0.65) (1.66) (1.86) (1.35) (1.25)Years of Education of 0.01 0.00 0.01 0.01 0.00 0.01Original Squatter’s Father (0.13) (0.04) (0.13) (0.11) (0.06) (0.11)Argentine Mother of the -0.14 -0.04 -0.14 -0.14 -0.13 -0.11Original Squatter (0.39) (0.11) (0.43) (0.31) (0.35) (0.32)Years of Education of 0.04 0.01 0.04 0.04 0.04 0.04Original Squatter’s Mother (0.44) (0.18) (0.46) (0.56) (0.48) (0.48)Constant 0.09 1.52*** 1.03** 0.10 0.09 0.09 0.27 -0.04

(0.10) (12.12) (2.22) (0.12) (0.12) (0.11) (0.28) (0.04)F-stat 2.06Observations 309 313 313 415 309 309 309 309Notes: The dependent variable is the number of sons or daughters of the household head older than 14 years old living in the house. Column (1) is the regression summarized in Column 3 of Table 6. Column (2) uses no controls. Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. Column (9) presents the matching estimator on the propensity score of the probability of attrition. Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

44

Appendix Table A.10 – OTHERS

(1) (2) (3) (4) (5) (6) (7) (8)Property Right -0.68*** -0.53*** -0.55*** -0.57*** -0.68*** -0.68*** -0.89*** -0.78***

(2.75) (2.75) (3.35) (3.54) (4.98) (3.96) (2.67)Property Right

(9)

(3.53) -0.36

1989-91 (1.34)Property Right -0.85***1997-98 (3.93)Surface 0.00 0.00 0.00 0.00 0.00 -0.00 -0.00

(0.02) (0.02) (0.08) (0.02) (0.05) (0.09) (0.03)Distance to Creek -0.09 -0.09 -0.02 -0.09 -0.09 -0.11 -0.14 (0.85) (0.85) (0.37) (0.82) (1.24) (1.05) (1.31)Block Corner 0.06 0.00 -0.02 0.06 0.06 0.04 0.07

(0.23) (0.01) (0.10) (0.29) (0.37) (0.15) (0.24)Distance to Non- -0.12 -0.11 -0.07 -0.12 -0.12** -0.12 -0.12Squattered Area (1.26) (1.18) (0.87) (1.37) (2.23) (1.27) (1.26)Age of Original -2.49*** -1.95*** -2.49** -2.49** -2.53*** -2.47***Squatter <49 (3.64) (3.11) (2.20) (2.20) (3.69) (3.63)Age of Original -2.16*** -1.59** -2.16* -2.16** -2.18*** -2.12***Squatter > 50 (3.14) (2.53) (1.96) (2.12) (3.18) (3.10)Female Original 0.32* 0.26 0.32* 0.32 0.32 0.32*Squatter (1.67) (1.58) (1.79) (1.59) (1.64) (1.67)Argentine Original -0.71 -0.52 -0.71 -0.71 -0.76 -0.75Squatter (1.48) (1.24) (1.49) (1.37) (1.58) (1.57)Years of Education of -0.09* -0.08* -0.09** -0.09 -0.09* -0.10*the Original Squatter (1.71) (1.92) (2.00) (1.63) (1.69) (1.77)Argentine Father of 0.97** 0.84** 0.97* 0.97* 1.00*** 0.99***the Original Squatter (2.53) (2.40) (1.91) (1.83) (2.60) (2.60)Years of Education of -0.06 -0.07 -0.06 -0.06 -0.07 -0.06Original Squatter’s Father (0.67) (0.99) (0.53) (0.65) (0.77) (0.64)Argentine Mother of -0.37 -0.32 -0.37 -0.37 -0.35 -0.40the Original Squatter (0.95) (0.91) (0.83) (1.04) (0.90) (1.04)Years of Education of 0.02 -0.02 0.02 0.02 0.03 0.02Original Squatter’s Mother (0.27) (0.28) (0.25) (0.30) (0.33) (0.21)Constant 4.70*** 1.25*** 1.63*** 4.07*** 4.70*** 4.70*** 4.94*** 4.87***

(4.56) (8.66) (3.05) (4.69) (3.56) (4.38) (4.75) (4.72)F-stat 2.89*Observations 309 313 313 415 309 309 309 309 292Notes: The dependent variable is the number of household members that live in the house excluding the household head, household head spouse and sons or daughters of the household head. Column (1) is the regression summarized in Column 4 of Table 6. Column (2) uses no controls. Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. Column (9) presents the matching estimator on the propensity score of the probability of attrition. Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

45

Appendix Table A.11 - CHILDREN 6-14 YEARS OLD

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)Property Right -0.21 -0.27* -0.28* -0.25* -0.21 -0.21 -0.15 -0.17

(1.45) (1.79) (1.84) (1.86) (1.61) (1.49) (0.90) (0.87)Property Right 1989-91 -0.43** -0.39 (2.13) (1.47)Property Right 1997-98 -0.09 -0.04 (0.57) (0.19)Surface 0.00 0.00 0.00 0.00 0.00 0.00 0.00

(0.58) (0.62) (0.28) (0.61) (0.61) (0.62) (0.63)Distance to Creek -0.02 -0.06 0.04 -0.02 -0.02 -0.02 0.01 (0.30) (0.76) (0.80) (0.34) (0.55) (0.23) (0.16)Block Corner 0.02 0.03 -0.03 0.02 0.02 0.02 0.01

(0.08) (0.15) (0.16) (0.08) (0.08) (0.11) (0.07)Distance to Non-Squattered 0.02 -0.00 0.05 0.02 0.02 0.02 0.02Area (0.31) (0.09) (0.75) (0.36) (0.29) (0.31) (0.31)Age of Original Squatter <49 1.42*** 1.08** 1.42*** 1.42*** 1.43*** 1.40*** (2.73) (2.21) (7.05) (10.01) (2.75) (2.72)Age of Original Squatter > 50 0.43 0.27 0.43** 0.43*** 0.43 0.40 (0.82) (0.55) (2.28) (5.15) (0.84) (0.77)Female Original Squatter -0.15 -0.06 -0.15 -0.15 -0.14 -0.14 (0.99) (0.46) (0.90) (1.06) (0.98) (0.98)Argentine Original Squatter 0.29 0.26 0.29 0.29 0.30 0.32 (0.80) (0.78) (0.80) (1.04) (0.84) (0.88)Years of Education of the 0.01 0.01 0.01 0.01 0.01 0.01Original Squatter (0.27) (0.29) (0.27) (0.37) (0.26) (0.31)Argentine Father of the -0.28 -0.17 -0.28 -0.28 -0.29 -0.30Original Squatter (0.96) (0.62) (0.90) (1.04) (0.99) (1.02)Years of Education of -0.12* -0.12** -0.12* -0.12** -0.12* -0.13*Original Squatter’s Father (1.86) (2.08) (1.74) (2.15) (1.83) (1.90)Argentine Mother of the -0.24 -0.32 -0.24 -0.24* -0.24 -0.21Original Squatter (0.81) (1.16) (0.64) (2.04) (0.83) (0.73)Years of Education of 0.07 0.06 0.07 0.07 0.07 0.07Original Squatter’s Mother (0.97) (0.95) (0.82) (0.65) (0.94) (1.02)Constant 0.42 1.12*** 1.07*** 0.65 0.42 0.42 0.35 0.30

(0.54) (10.16) (2.61) (0.95) (0.69) (1.27) (0.45) (0.38)F-stat 2.43Observations 309 313 313 415 309 309 309 309 292 170 236Notes: The dependent variable is the number of sons or daughters of the household head between 6 and 14 years old living in the house. Column (1) is the regression summarized in Column 5 of Table 6. Column (2) uses no controls. Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) is the regression summarized in Column 6 of Table 6, which estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. For treatment, early treatment and late treatment, respectively, Column (9) through (11) present the matching estimators on the propensity score of the probability of attrition. Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

46

Appendix Table A.12 - CHILDREN 0-5 YEARS OLD

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)Property Right -0.08 -0.08 -0.07 -0.10 -0.08 -0.08 -0.03 -0.03 (0.92) (1.04) (0.86) (1.38) (0.90) (1.07) (0.30) (0.37)Property Right 1989-91 -0.14 -0.10 (1.18) (0.80)Property Right 1997-98 -0.04 0.02 0.47) 0.29) ( (Surface 0.00 0.00 0.00 0.00 0.00 0.00 0.00

(0.90) (1.03) (0.07) (0.89) (0.98) (0.96) (0.92)Distance to Creek 0.02 0.02 0.02 0.02 0.02 0.03 0.03 (0.49) (0.64) (0.59) (0.53) (0.60) (0.60) (0.69)Block Corner -0.04 -0.02 0.00 -0.04 -0.04 -0.03 -0.04

(0.32) (0.21) (0.02) (0.36) (0.31) (0.27) (0.32)Distance to 0.02 0.03 0.03 0.02 0.02 0.02 0.02Non-Squattered Area (0.51) (0.79) (0.73) (0.54) (0.44) (0.51) (0.51)Age of Original Squatter <49 -0.61** -0.90*** -0.61 -0.61 -0.60** -0.62** (2.06) (3.33) (1.47) (1.53) (2.03) (2.07)Age of Original Squatter > 50 -0.80*** -1.09*** -0.80* -0.80* -0.79*** -0.81*** (2.69) (4.02) (1.92) (2.01) (2.66) (2.70)Female Original Squatter 0.04 0.10 0.04 0.04 0.04 0.04 (0.44) (1.40) (0.48) (0.66) (0.46) (0.45)Argentine Original Squatter -0.38* -0.35* -0.38** -0.38*** -0.37* -0.37* (1.81) (1.93) (2.08) (4.24) (1.75) (1.77)Years of Education of the -0.01 -0.03 -0.01 -0.01 -0.01 -0.01Original Squatter (0.24) (1.42) (0.28) (0.22) (0.25) (0.22)Argentine Father of the 0.13 0.12 0.13 0.13* 0.12 0.12Original Squatter (0.76) (0.76) (1.16) (1.76) (0.72) (0.73)Years of Education of Original -0.01 -0.01 -0.01 -0.01 -0.01 -0.01Squatter’s Father (0.26) (0.31) (0.38) (0.67) (0.21) (0.28)Argentine Mother of the 0.06 0.08 0.06 0.06 0.06 0.07Original Squatter (0.38) (0.50) (0.50) (0.83) (0.35) (0.42)Years of Education of Original -0.05 -0.02 -0.05 -0.05 -0.05 -0.05 Squatter’s Mother (1.18) (1.44) (1.42) (0.62) (1.22) (1.16)

** ** 0.16 1.75* 1.31*** 1.34*(3.06) (7.33) (0.75) (4.65) (2.38) (2.68) (2.91)

F-stat 0.57Observations 313 3 9 309 313 415 309 0 309 309 292 170 236Notes: The dependent variable is the numbe ughters of the ehold he ween 0 and 5 years o iving in hous n (1)summarized in Column 7 of Table 6. Column (2) uses no controls. Column (3) only controls for parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the block level in Column (5), and at the former owner level in Column (6). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (7). Column (8) is the regression summarized in Column 8 of Table 6, which estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. For treatment, early treatment and late treatment, respectively, Column (9) through (11) present the matching estimators on the propensity score of the probability of attrition. Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

r of sons or da hous ad bet ld l the e. Colum is the regression

Constant 1.37* 0.44* ** 1.37** 1.37** ** (2.97)

47

Appendix Table A.13 – SCHOOL ACHIEVEMENT - CHILDREN 6-14 YEARS OLD

(1) (6) (2) (3) (4) (5) (7) (8) (9) (10) (11) (12)Property Right 0.10 0.10 0.03 0.08 0.09 0.10 0.10 0.18 0.18 (0.85) (0.30) (0.72) (0.83) (0.86) (0.72) (0.82) (1.18) (0.56)

0.39** 0.37(2.02) (0.58)

Property Right 1997-98 0.00 -0.05 (0.00) (0.13)Surface -0.00 -0.00 -0.00 0.00 -0.00 -0.00 -0.00 -0.00 (1.09) (0.91) (0.48) (1.06) (1.13) (1.68) (1.02) (1.22)

0.03 0.04 -0.02 0.03 0.03 0.04 0.00(0.72) (0.45) (0.50) (0.45) (0.41) (0.66) (0.02)

Block Corner -0.01 0.06 0.04 -0.01 0.02 -0.01 -0.01 0.01 (0.05) .06) (0.45) (0.26) (0.05) (0 (0.06) (0.03) (0.09) Distance to Non-Squattered Area -0.09 -0.05 -0.11* -0.09 -0.09 -0.09 -0.09 -0.09 (1.31) (0.91) (1.85) (1.33) (1.41) (1.54) (1.42) (1.33) Male -0.03 -0.05 -0.03 -0.04 -0.03 -0.03 -0.03 -0.03 -0.02 (0.28) (0.52) (0.30) (0.45) (0.25) (0.21) (0.32) (0.28) (0.23)Age ** -0.13*** -0.13*** -0.12*** -0.12*** -0.13*** -0.13*** -0.13*** -0.13*** -0.13* (6.30) (6.34) (6.07) (6.27) (6.31) (6.80) (7.25) (6.29) (6.15) Age of Original Squatter <49 0.41 0.20 0.41 0.41 0.41 0.41 0.36 (0.94) (0.53) (1.10) (1.52)(1.09) (0.94) (0.82)Age of Original Squatter > 50 0.33 0.14 0.33 0.33 0.33 0.34 0.30 (0.74) (0.37) (0.87)(0.87) (1.09) (0.75) (0.67)

0.02 0.06 0.02 0.02 0.03 0.00(0.16) (0.60) (0.15)(0.15) (0.14) (0.23) (0.01)

Argentine Original Squatter -0.00 0.13 -0.00 -0.00 -0.00 0.01 -0.04 (0.01) (0.44) (0.01)(0.01) (0.01) (0.04) (0.11)

0.05 0.07** 0.05 0.05 0.05*** 0.05 0.05Squatter (1.38) .47) (1.40) (1.48) (2.34) (1 (3.36) (1.51)Argentine Father of the Original -0.33 -0.33 -0.28 -0.33 -0.33 -0.36 -0.30Squatter (1.33) (1.14) (1.43) (1.34) (1.48) (1.43) (1.23)Years of Education of Original 0.03 0.06 0.03 0.03 0.03 0.03 0.03Squatter’s Father (0.54) (1.02) (0.53) (0.57) (0.70) (0.54) (0.61)Argentine Mother of the Original 0.31 0.23 0.31 0.31 0.31** 0.31 0.27Squatter (1.47) (1.13) (1.25) (1.05) (2.21) (1.50) (1.30)Years of Education of Original 0.02 0.02 0.00 0.02 0.02 0.03 0.01Squatter’s Mother (0.31) (0.08) (0.30) (0.30) (0.28) (0.38) (0.08) Constant -0.34 0.26 0.37 -0.64 -0.34 -0.34 -0.34 -0.41 -0.14

(0.52) (1.17) (1.08) (1.16) (0.59) (0.59) (0.71) (0.61) (0.22)F-stat 3.52* Observations 282 282 282 284 284 361 282 282 282 266 153 233Notes: The dependent variable is the difference between the school grade each child is currently attending or the maximum grade attained (if not attending school) minus the grade corresponding to the child age. Column (1) is the regression summarized in Column 1 of Table 7. Column (2) only controls for child age and gender. Column (3) controls for child age, gender and parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the household level in Column (5), at the block level in Column (6), and at the former owner level in Column (7). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (8). Column (9) is the regression summarized in Column 2 of Table 7, which estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. For treatment, early treatment and late treatment, respectively, Column (10) through (12) present the matching estimators on the propensity score of the probability of attrition. Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

Property Right 1989-91

Distance to Creek 0.03 (0.53)

Female Original Squatter 0.02

Years of Education of the Original

48

49

Appendix Table A.14 - SCHOOL ABSENTEEISM - CHILDREN 6-14 YEARS OLD (1) (5) (2) (3) (4) (6) (7) (8) (9) (10) (11) (12)

Property Right -0.37** -0.32** * -0.35* -0.27** -0.37** -0.37** -0.37* -0.44** -0.34* (2.39) (2.28) (2.41) (2.03) (2.06) (2.13) (1.78) (2.19) (1.86)Property Right 1989-91 -0.55** -0.46** (2.14) (2.41)Property Right 1997-98 -0.31* -0.40* (1.73) (1.81)Surface 0. 0 -0.00 0.00*0 -0.00 -0.00 -0.00 -0.00 -0.00

(0.26) (0.14) (1.68) (0.29) (0.28) (0.22) (0.30) (0.20)Distance to Creek -0.15* -0.12 -0.05 -0.15 -0.15 -0.15 -0.16* -0.13

(1.71) (1.49) (0.84) (1.08)(1.09) (1.07) (1.78) (1.40)Block Corner 0.09 0.08 0.15 0.09 0.09 0.09 0.08 0.08

(0.42) (0.76) (0(0.30) .33) (0.33) (0.35) (0.34) Distance to Non-Squattered -0.06 -0.03 0.00 -0.06 -0.06 -0.06 -0.05 -0.06Area (0.67) (0.45) (0.00) (0.36) (0.35) (0.35) (0.58) (0.65)Male -0.01 -0.04-0.03 -0.010.01 -0.01 -0.01 -0.01 -0.02 (0.07) (0.21) (0.31) (0.04) (0.08) (0.08) (0.07) (0.07) (0.11)Age -0.04 -0.04 -0.03-0.04 -0.04-0.04 -0.04 -0.04 -0.04

(1.34) (1.40) (1.55) (1.31) (1.15) (1.34) (1.41)Age of Original Squatter <49 0.30 0.34 0.53 0.30 0.30 0.30 0.30 (0.53) (1.13) (0.67) (0.72) (1.01) (0.52) (0.59)Age of Original Squatter > 50 0.43 0.72 0.43 0.43 0.43* 0.42 0.45 (0.73) (1.50) (1.08) (1.24) (1.83) (0.72) (0.77)Female Original Squatter 0.15 0.16 0.12 0.15 0.15 0.15 0.14 (0.95) (0.92) (0.89) (1.18)(0.77) (0.90) (1.01)Argentine Original -0.33 0.00 -0.33 -0.33 -0.33 -0.34 -0.31Squatter (0.77) (0.00) (0.90) (0.92) (1.45) (0.81) (0.72)Years of Education of the -0.01 -0.02 -0.01 -0.01 -0.01 -0.01 -0.01Original Squatter (0.29) (0.30) (0.48) (0.28) (0.28) (0.29) (0.24) Argentine Father of the 0.23 -0.03 0.23 0.23 0.23 0.26 0.22Original Squatter (0.71) (0.10) (0.88) (0.83) (0.72) (0.79) (0.67)Years of Education of -0.14* -0.12* -0.14** -0.14** -0.14*** -0.14* -0.14*Original Squatter’s Father (1.90) (1.71) (2.46) (2.21) (3.06) (1.90) (1.93)Argentine Mother of the 0.42 0.43* 0.42* 0.42 0.42* 0.41 0.44Original Squatter (1.52) (1.59) (1.72) (1.76) (1.66) (2.03) (1.49)Years of Education of 0.14 0.13 0.14 0.14 0.14* 0.14 0.15*Original Squatter’s Mother (2.16) (1.56) (1.63) (1.49) (1.42) (1.51) (1.66)Constant 0.92 1.30***1.05*** -0.11 0.92 0.92 0.92 0.99 0.80

(1.08) (3.53) (2.84) (1.01)(0.16) (0.98) (1.13) (1.14) (0.92)F-stat 0.77Observations 264 151 279 281 281 357 279 279 279 279 279 231Notes: The dependent variable is the number of da each ssed sc l out of the last five days o classes. Column (1) is the re on summarized in ColuTable 7. Column (2) only controls for child age and gender. Column (3) controls for child age, gender and parcel characteristics. Column (4) includes the San Martin neighborhood. The standard errors are clustered at the household level in Column (5), at the block level in Column (6), and at the former owner level in Column (7). The 2SLS regression (instrumenting the treatment variable “Property Right” with the intention-to-treat variable “Property Right Offer”) is presented in Column (8). Column (9) is the regression summarized in Column 4 of Table 7, which estimates separately the effect of early and late treatments. The F-stat tests the null hypothesis: Property Right 1989-91= Property Right 1997-98. For treatment, early treatment and late treatment, respectively, Column (10) through (12) present the matching estimators on the propensity score of the probability of attrition. Absolute value of t statistics in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1%.

ys child mi hoo f gressi mn 3 of

(1.41) (1.29)

(0.41)