Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
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, [email protected]. Ernesto Schargrodsky, Universidad Torcuato Di Tella, Saénz Valiente 1010, (C1428ATG) Buenos Aires, Argentina, Tel: (54-11) 4783-3112, [email protected]. 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)