## _wp11131 - Section VI concludes the paper, fleshing out some policy recommendations and proposing

## Source details

**Canonical URL:** [_wp11131 - Section VI concludes the paper, fleshing out some policy recommendations and proposing](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11131.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11131.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11131.pdf.json)

---

### Literature review and theoretical background
- Empirical evidence on effects of land-titling cited: Jimenez (1985), Alston, Libecap, and Schneider (1996), Lanjouw and Levy (2002); Besley (1995), Brasselle, Gaspart, and Platteau (2002), Jacoby, Li, and Rozelle (2002), Do and Iyer (2003); Place and Migot-Adholla (1998), Carter and Olinto (2003), Field and Torero (2002), Galiani and Schargrodsky (2011).
- Key empirical findings summarized:
  - Jimenez (1984): equilibrium price differentials between formal and informal sector unit dwelling prices in the Philippines in the range of 58.0%.
  - Field and Torero (2002): positive effects of the Peruvian titling program on labor supply, credit access, and housing investments.
  - Galiani and Schargrodsky (2011): entitled families in Argentina substantially increased housing investment, reduced household size, and enhanced children's education; land-titling reduces poverty via increased physical and human capital investment rather than credit access or entrepreneurial income.
  - Andrade (2006): positive effect of land title on income in Comunidade do Caju, Rio de Janeiro (cross-sectional sample of 200 families).
- Measurement challenges and mechanisms:
  - Potential endogeneity of ownership rights (Demsetz 1967; Alchian and Demsetz 1973); Miceli, Sirmans, and Kieyah (2001) provide direct evidence on endogeneity of formal agricultural property rights in Kenya.
  - Untitled households bear time costs for informal policing, reducing labor supply outside the home (Field and Torero 2002). Exogenous increases in formal property protection should lower opportunity cost of outside labor and increase hours worked outside the house.

### Theoretical model, comparative statics, and propositions
- Modeling framework:
  - Extension of Field (2003) with two adults (A and B) and adult and child labor; individual A has higher wage than individual B (AB ww >), initially A works more hours than B.
  - Utility: U(xA, xB; lA, lB, s(L, θ, τ)) with s(·) inputs L (hours on protection), formal property rights θ, and de facto rights τ; Uγ and sγ twice continuously differentiable, concave, and increasing.
- Key assumptions for tractability:
  - Household maximizes per capita leisure; no outside market for home security provision; security inputs (L, θ, τ) are substitutes.
- Comparative statics and empirical implication:
  - First-order conditions (equations (5) and (6)) characterize optimal lA and lB; ∂lB/∂θ < 0 unambiguous (increase in formal property rights reduces leisure of the initially lower-hours worker).
  - Proposition 1: If wage differential AB ww is sufficiently high (AB ww >>), effect of higher property rights on labor hours is larger for the individual initially working fewer hours (adult B) than for the individual initially working longer hours (adult A): AB ll θθ ∂∂ > ∂∂ (Equation (7)). Implication: land-titling programs should increase hours worked more for individuals in lower quantiles of the initial hours-worked distribution than for those in higher quantiles.

### Methodological issues and sample design
- Context and program timing:
  - Papel Passado titling program in Osasco: Osasco has 30,000 people (about 6,000 families) living under informal conditions, representing almost 4.5% of its total population (ANOREG 2007).
  - Program period for Osasco: 2007–2014. First locality titled in 2007: Jardim Canaã (500 families). Closest neighborhood DR contains 450 families and scheduled for titling in 2012.
- Minimizing selection and contamination bias:
  - Osasco City Hall claims locality priority follows a random criterion; 95 percent of first survey participants from both Jardim Canaã and DR did not expect to receive any land titles and were unaware of Papel Passado.
  - Control group cannot benefit from the program outside treatment locality; no other land title programs in the region. Program does not provide a dropout option; selling property after receiving title still counts as affected.
  - Jardim Canaã and DR contiguous, located 2.5 miles from downtown Osasco with same access to main economic center. Note: unofficial local sources allege possible "political" agenda influence.
- Sample selection procedure:
  - Two-stage survey; 326 randomly sampled families: 185 from Jardim Canaã and 141 from DR.
  - Random selection: first 150 households closest in birth dates to each field researcher, then the third and fifth households to the right.
  - Survey instrument: 39 questions mirroring PNAD (IBGE) items; included six questions on benefits associated with property formalization.

### Data collection, timing, and attrition
- Survey timing:
  - First stage: March 2007 (before titles issued to Jardim Canaã households).
  - Second stage: August 2008 (with 98 percent recall), about a year and a half after first titles were issued; ensures all households had title for at least 1 year.
  - Exact dates of title receipt obtained from the 2nd Osasco Office of Registration (2º Cartório de Osasco) and official authorization from Osasco City Hall.
- Attrition:
  - Treatment group: 8 percent of households that received the land title moved away from Canaã.
  - Control group: only one household (out of 140) moved out during the same period.

### Descriptive statistics and baseline comparability
- Baseline (2007) comparability and sample composition:
  - T-tests (2007) indicate treated and control group observations comparable in number.
  - Treated group relative to control: child labor, monthly income per capita, informal labor, and number of residents significantly lower; education of family head significantly higher.
  - Treated sample more educated and less informal but has lower income than control group.
  - Wealth index via principal component analysis summarizes durable goods ownership (TVs, radios, cars, washing machines, refrigerators, freezers).
- Selected descriptive statistics (from tables):
  - Weekly hours of adult work: 2007 mean 10.19; 2007 Std. Dev. 16.18; 2008 mean 14.33.
  - Monthly income (BRL): 2007 mean 1,126.25; Std. Dev. 1,491.92; 2008 mean 1,138.76; Std. Dev. 1,473.35.
  - Observations: 304 (both years).
  - Table 2 (2007 t-tests): Mean age Control 42.60 vs Treatment 39.40, p-value 0.06*; Weekly hours of child labor (>16 years old) Control 8.35 vs Treatment 3.30, p-value 0.00***; Years of education (family head) Control 5.00 vs Treatment 9.00, p-value 0.00***; Monthly income per capita Control 553.10 vs Treatment 255.80, p-value 0.00***; Wealth index Control 1.12 vs Treatment -0.94, p-value 0.00***; Informal worker Control 0.94 vs Treatment 0.65, p-value 0.00***; Observations (households): Control 168, Treatment 137.
  - Table 3 (Spearman correlations, 2007): Years of education with Weekly Hours Worked 0.14**; Informality with Monthly income per capita 0.23***; Monthly income per capita with Weekly Hours Worked 0.11*.

### Empirical strategy and estimators
- Two complementary procedures to address differences in observables:
  - Include control variables and use OLS and difference-in-difference (DD).
  - Apply same methods with propensity score technique to select a more balanced sample (common support / balanced sample).
- Main innovation: test distributional effects via quantile regression (QR) and quantile treatment effects (QTE), motivated by Proposition 1.
- Methods discussed: semi-parametric difference-in-differences matching (Heckman, Ichimura, and Todd 1997), difference-in-differences-matching (DDM), fixed effects estimator, QR, unconditional QTE (Firpo 2007), and weighting estimators using propensity scores.
- OLS specification (equation (8)): H' i = α + δ title i + β X' i + ε i; dependent variable H i : number of hours worked weekly by the household. Control vector X' i lists 12 socio-economic variables, including gender, ethnicity, marital status, age, age squared, household size, weekly hours of child labor, years of education, monthly income per capita, credit dummy, wealth index, informal work dummy.
- DD specification (equation (9)) and identification assumptions: treat s, year t, interaction treat*s * year t; assumptions include no time-varying unobserved selection, common trends absent program, no spillovers.
- QTE/QR approach: conditional quantile regression (equation (13)) and unconditional QTE (Firpo 2007) implemented via two-step weighting estimator with propensity score estimated parametrically via logit due to limited sample size.

### Empirical results — mean effects and matching
- Mean effects (OLS and Tobit, 2008):
  - OLS Naïve: Land title = 9.37*** (standard error (1.55))
  - OLS (with controls): Land title = 6.01** (2.50)
  - OLS Balanced Sample: Land title = 6.25** (2.50)
  - Tobit Balanced Sample: Land title = 9.66*** (3.55); Sigma = 18.45*** (1.01)
  - Control variables: Years of education increases weekly hours (examples: 0.49**, 0.50**, 0.68**); Access to credit increases weekly hours (examples: 4.07**, 3.68**, 5.44**); Age marginally reduces hours (some significant).
- Difference-in-Differences (2007–2008):
  - Land*year (DD interaction) = 9.01*** (1.16) DD Naïve; 8.02*** (1.24) DD Unbalanced; 8.13*** (1.26) DD Balanced.
  - Year effect = 1.00*** (0.28) DD Naïve; 1.20*** (0.35) DD Unbalanced; 1.25*** (0.38) DD Balanced.
  - DDM — kernel estimator (three bandwidths):
    - Bandwidth (0.01): Land title = 9.00*** (3.50)
    - Bandwidth (0.05): Land title = 7.98 (2.70)
    - Bandwidth (0.10): Land title = 7.8*** (2.50)
- Authors conclude average treatment effect on the treated group is around 9 hours/week.

### Empirical results — distributional effects (QR and QTE)
- First-degree stochastic dominance:
  - 2007: treatment group's distribution of adults’ weekly hours worked dominates control group.
  - 2008: dominance increases significantly, indicating positive effects across the cumulative distribution.
- Conditional Quantile Regression (Table 8, common support, balanced sample, 288 observations):
  - Land title effects by quantile:
    - 0.25 quantile: 7.77* (4.18)
    - 0.50 quantile (median): 9.66*** (3.27)
    - 0.75 quantile: 1.41 (3.26)
    - 0.90 quantile: 0.43 (4.77)
  - Highest estimated effects concentrated in first and second quartiles: 7.8 hours/week (first quartile) and 9.7 hours/week (median) per text interpretation.
  - Years of education positive and significant at middle quantiles (examples: 0.64** at median, 0.69** at 0.75).
- Interquantile comparisons (Table 9):
  - Difference 0.5-0.25 = 1.89 (4.95) — not statistically significant.
  - Difference 0.5-0.75 = 8.25** (3.62) — statistically significant.
  - Difference 0.5-0.9 = 9.23* (5.02) — statistically significant.
  - Interpretation: titling effects concentrated in the middle/lower part of the hours-worked distribution; outliers are not driving ATT.
- Unconditional QTE estimates (Table 10):
  - Whole sample (305 obs):
    - 0.10: Land Title = -12.00** (5.30)
    - 0.25: Land Title = 25.00*** (1.90)
    - 0.50: Land Title = 7.50*** (2.80)
    - 0.75: Land Title = 7.60 (4.80)
  - Common support (288 obs):
    - 0.10: Land Title = -12.20** (5.30)
    - 0.25: Land Title = 14.96*** (2.84)
    - 0.50: Land Title = 4.70 (3.16)
    - 0.75: Land Title = 6.24 (4.88)
  - Patterns:
    - Impact tends to be higher in the first half of the outcome distribution than in the second.
    - In the common support, median effect declines from 25 to 15 hours/week (text reports decline from 25 to 15 and significance at 1 percent), indicating sensitivity to sample selection.
    - Both conditional QR and unconditional QTE indicate heterogeneity and concentration of effects around the median.

### Synthesis and robustness
- Consistent evidence across methods that Papel Passado land title program had a positive effect on hours worked and labor supply.
- Average ATT estimates around 9 hours/week (DD, DDM, OLS/Tobit).
- Distributive analysis shows heterogeneous effects concentrated among households with initially lower hours worked (poorest households).
- Robustness considerations:
  - Results sensitive to sample selection/common support.
  - Limited sample size noted as caveat for QTE; excluding workers with 0 initial hours affects higher-quantile coefficients.
  - Tobit sigma significance suggests nonrandom selection into labor market; omission could yield omitted variable bias.

### Policy implications and recommendations
- Land titling can increase labor supply by reducing households’ need to stay home protecting property, particularly benefiting the poorest urban households.
- Authorities should account for heterogeneous (distributive) effects when designing public policies: impacts may be restricted to ranges within the outcome distribution, so targeting can improve efficiency.
- Program evaluation design recommendations:
  - Emphasize sampling on the common support (households with similar pre-treatment observables).
  - Use propensity score matching to minimize selection bias.
- Further research recommended on:
  - Effects on household income and utility from increased hours worked induced by titling.
  - Distributional impacts of titling on other outcomes (access to credit, fertility).
  - Broader applicability of titling programs as policy tools to address urban informality, poverty, and macroeconomic considerations.

### Appendix — selected auxiliary results
- Propensity score logit estimates (Table A):
  - Observations: 305 (Unmatched Sample); 288 (Matched Sample).
  - Pseudo-R2: 0.62 (Unmatched Sample); 0.63 (Matched Sample).
  - Selected coefficients (Unmatched Sample, standard errors in parentheses):
    - Years of education (head): 0.14*** (0.05)
    - Monthly income per capita: -0.01** (0.00)
    - TV (= 1 if have): -1.48** (0.69)
    - Radio (= 1 if have): -1.68*** (0.50)
    - Washing machine (= 1 if have): 2.19*** (0.65)
    - Refrigerator (= 1 if have): -6.07*** (1.07)
    - Informal worker: -1.73*** (0.62)
- Figure A: Distribution of Wealth Index, 2008 (X-axis ticks -4 -2 0 2; Density ticks 0 .2 .4 .6 .8).

*Italic source: IMF Working Paper content unit from _wp11131 - Appendix shows the distribution of this index.*

### Section VI concludes the paper, fleshing out some policy recommendations and proposing

### _wp11131 - Section VI concludes the paper, fleshing out some policy recommendations and proposing

### Literature review and theoretical background
- Empirical evidence on effects of land-titling cited:
  - Real estate values: Jimenez (1985), Alston, Libecap, and Schneider (1996), Lanjouw and Levy (2002).
  - Agricultural investment: Besley (1995), Brasselle, Gaspart, and Platteau (2002), Jacoby, Li, and Rozelle (2002), Do and Iyer (2003).
  - Credit access, labor supply, housing investment, and income: Place and Migot-Adholla (1998), Carter and Olinto (2003), Field and Torero (2002), Galiani and Schargrodsky (2011).
- Key empirical findings summarized:
  - Jimenez (1984): equilibrium price differentials between formal and informal sector unit dwelling prices in the Philippines in the range of 58.0%.
  - Field and Torero (2002): positive effects of the Peruvian titling program on labor supply, credit access, and housing investments.
  - Galiani and Schargrodsky (2011): entitled families in Argentina substantially increased housing investment, reduced household size, and enhanced children's education; land-titling reduces poverty via increased physical and human capital investment rather than credit access or entrepreneurial income.
  - Andrade (2006): positive effect of land title on income in Comunidade do Caju, Rio de Janeiro (cross-sectional sample of 200 families).
- Measurement challenges:
  - Potential endogeneity of ownership rights (Demsetz 1967; Alchian and Demsetz 1973).
  - Miceli, Sirmans, and Kieyah (2001) provide direct evidence on endogeneity of formal agricultural property rights in Kenya.
- Mechanism emphasized:
  - Untitled households bear time costs for informal policing, reducing labor supply outside the home (Field and Torero 2002).
  - Exogenous increases in formal property protection should lower opportunity cost of outside labor and increase hours worked outside the house.

### Theoretical model, comparative statics, and propositions
- Modeling framework:
  - Extension of Field (2003) with two different adults in the household (A and B) and adult and child labor.
  - Individual A has higher wage than individual B (AB ww >), and initially A works more hours than B.
  - Utility includes per capita leisure, consumption, and a tenure security function s(·) with inputs: hours spent on property protection, formal property rights θ, and de facto rights τ.
  - Utility specified as U(xA, xB; lA, lB, s(L, θ, τ)) with assumptions: Uγ and sγ twice continuously differentiable, concave, and increasing.
- Key assumptions for tractability:
  - Household maximizes per capita leisure (not individual leisure).
  - No outside labor market for home security provision.
  - Security inputs (L, θ, τ) are substitutes in production.
- First-order conditions and comparative statics:
  - First-order conditions (equations (5) and (6)) characterize optimal leisure choices lA and lB.
  - Under weakly positive first derivatives and weakly negative second derivatives of U and s, sign of ∂lA/∂θ is ambiguous.
  - Increase in formal property rights unambiguously reduces leisure of the initially lower-hours worker: ∂lB/∂θ < 0.
- Proposition 1 (empirical implication):
  - If the wage differential between A and B is sufficiently high (AB ww >>), the effect of higher property rights on labor hours is larger for the individual initially working fewer hours (adult B) than for the individual initially working longer hours (adult A):
    - AB ll θθ ∂∂ > ∂∂ (Equation (7)).
  - Implication: land-titling programs should increase hours worked more for individuals in lower quantiles of the initial hours-worked distribution than for those in higher quantiles.

### Methodological issues and sample design
- Context of evaluation:
  - Papel Passado titling program in Osasco; Osasco has 30,000 people (about 6,000 families) living under informal conditions, representing almost 4.5% of its total population (ANOREG 2007).
  - Program period for Osasco: 2007–2014.
  - First locality titled in 2007: Jardim Canaã (500 families). Closest neighborhood DR contains 450 families and scheduled for titling in 2012.
- Minimizing selection bias:
  - Osasco City Hall claims locality priority follows a random criterion; this randomization minimizes selection and self-selection concerns.
  - 95 percent of first survey participants from both Jardim Canaã and DR did not expect to receive any land titles and were unaware of Papel Passado, reducing randomization bias.
  - Contamination bias avoided: control group cannot benefit from the program outside treatment locality and there are no other land title programs in the region.
  - Program does not provide a dropout option; selling property after receiving title still counts as affected, reducing attrition bias.
  - Jardim Canaã and DR are geographically contiguous (no physical border), located 2.5 miles from downtown Osasco, with the same access to the main economic center, supporting similarity in observable and unobservable characteristics between treatment and control.
  - Note: unofficial local sources allege a possible "political" agenda may have influenced the decision.
- Sample selection procedure:
  - Two-stage survey; 326 randomly sampled families: 185 from Jardim Canaã and 141 from DR.
  - Random selection method based on Bolfarine and Bussab (2005): researchers chose first 150 households closest in birth dates to each field researcher, then selected the third and fifth households to the right.
  - Survey instrument: 39 questions, mirroring PNAD (IBGE) items; included six questions on benefits associated with property formalization.

### Data collection, timing, and attrition
- Survey timing:
  - First stage: March 2007 (before titles issued to Jardim Canaã households), conducted by researchers not originally from Osasco.
  - Second stage: August 2008 (with 98 percent recall), about a year and a half after first titles were issued; ensures all households had title for at least 1 year.
  - Exact dates of title receipt obtained from the 2nd Osasco Office of Registration (2º Cartório de Osasco) and official authorization from Osasco City Hall.
- Attrition:
  - Treatment group: 8 percent of households that received the land title moved away from Canaã.
  - Control group: only one household (out of 140) moved out during the same period.

### Descriptive statistics and initial findings
- Sample composition and baseline comparability:
  - Table 2 T-tests (2007) comparing control and treatment: numbers of observations in treated and control groups are comparable.
  - Treated group characteristics (relative to control):
    - Child labor, monthly income per capita, informal labor, and number of residents are significantly lower in the treated group.
    - Education of the family head is significantly higher in the treated group.
    - Treated sample is more educated and less informal but has a lower income than the control group.
  - Wealth index computed via principal component analysis summarizes durable goods ownership (TVs, radios, cars, washing machines, refrigerators, freezers).
- Changes in weekly adult work hours:
  - For combined sample (treated and control), weekly hours of adult work increased between 2007 and 2008.
  - Treatment group increased its number of hours worked after the program (Figure 3).
  - Control group weekly hours remained practically constant over time (Figure 4).
- Perceived welfare change:
  - Figure 2 shows most households who received the land title felt their lives had improved by the time of the second-stage survey, even though most had not previously expected the land title.

*Italic source: _wp11131 - Section VI concludes the paper, fleshing out some policy recommendations and proposing*

### Appendix shows the distribution of this index.

### _wp11131 - Appendix shows the distribution of this index.

### Education, formal vs informal employment, and income-hours correlation
- Households with higher levels of education tend to have more access to formal jobs in Osasco and "tend to receive additional perks that are not reflected in the cash payroll" as is customary for formal employees in Brazil.25
- Informal workers do not have those benefits and "rely, essentially, on cash income to compensate for the lack of perks."26
- In the data sample, 233 households are informal workers. They represent 92 percent of the workers of the control group and 64 percent of the treated group (see Table 2).
- Table 3 shows a positive correlation among monthly income per capita and hours worked weekly: 0.11.27
- Note: the sample does not contain information on wages, which is the actual variable the underlying assumption is based upon.27

*Footnotes preserved from source:*
- 25: "For example, a formal Brazilian employee usually receives a health care plan for the whole family, subsidized transportation, and meal plans."
- 26: "In our data sample, 233 households are informal workers. They represent 92 percent of the workers of the control group and 64 percent of the treated group (see Table 2)."
- 27: "Unfortunately the sample does not contain information on wages, which is the actual variable that our assumption is based upon."

### Empirical strategy — overview
- Two complementary econometric procedures are adopted to address differences in observables:
  - Include control variables and use ordinary least squares (OLS) and difference-in-difference (DD).
  - Apply the same methods but use a propensity score technique to select a more balanced sample (common support / balanced sample).
- Main innovation: test potential distributional effects of the titling program via quantile regression (QR) and quantile treatment effects (QTE), motivated by Proposition 1.
- Methods and estimators discussed include semi-parametric difference-in-differences matching (Heckman, Ichimura, and Todd 1997), difference-in-differences-matching (DDM), fixed effects estimator, QR, unconditional QTE (Firpo 2007 approach), and weighting estimators using propensity scores.

### A. The OLS regression analysis
- Mean effect of land-titling estimated via standard OLS with equation (8):
  - H' i = α + δ title i + β X' i + ε i  (equation (8) as in source)
- Dependent variable:
  - H i : number of hours worked weekly by the household.
- Control vector X' i includes the following social-economic variables:
  - (i) gender;
  - (ii) ethnicity;
  - (iii) marital status;
  - (iv) age;
  - (v) age squared, in order to test non-linear effects of age;
  - (vi) number of members of the household;
  - (vii) weekly hours of child labor;
  - (viii) years of education;
  - (ix) monthly income per capita;
  - (x) a dummy for access to credit;
  - (xi) household wealth, measured by the wealth index;
  - (xii) a dummy for informal work.
- Treatment indicator:
  - title i = 1 if the household participated in the titling program, and 0 otherwise.
  - Since all households in Canaã received the title, the parameter of interest is the average treatment effect on the treated (ATT) instead of the intent to treat (ITT).
- Semi-parametric ATT estimated via a variant of Heckman, Ichimura, and Todd (1997) difference-in-differences matching; propensity score estimated via logit for this paper given limited sample size.
- OLS estimates also computed on the subset of households pertaining to the common support (balanced sample).

### B. The difference-in-difference methodology
- DD estimator compares before-after differences for treated and control groups; implemented with regression (9):
  - H ist = α + β1 treat s + β2 year t + β3 (treat s * year t) + β' X ist + u ist  (represented in source as equation (9))
- Notation and definitions:
  - H ist : number of hours worked weekly by household i in community s at time t.
  - treat st : dummy = 1 if individual resides in the treated community (1 s = ), and 0 otherwise.
  - year t : dummy = 0 in 2007 (baseline period), and = 1 in 2008.
  - X ist : vector of observable characteristics of household i in community s, changing through time.
  - u ist : error term.
- Parameter of interest:
  - DD α (coefficient on the interaction) identifies the effect of the treatment on the treated group.
- Identification assumptions for DD:
  - (i) selection for treatment does not depend on unobservable individual and community characteristics changing over time;
  - (ii) the difference between treated and comparison groups would be the same in the absence of the program (time-invariant common effect);
  - (iii) the treatment does not affect access to credit of households living in neighboring areas (no spillover effects).
- These assumptions imply expressions (10) and (11) in source (see Meyer 1995; and Imbens and Wooldridge 2008).
- Main issue: self-selection / anticipation problem (households may change hours in expectation of receiving title).
- Robustness checks:
  - Use control variables to account for baseline differences (2007).
  - Apply fixed effect estimator to account for group-specific time-invariant differences.
  - Estimate difference-in-differences-matching (DDM) per Heckman, Ichimura, and Todd (1997) using a kernel weighting function to construct counterfactuals based on propensity score distances; DDM estimator expressed in source (summation formula) computed on common support p S.
- DDM identification condition stated: E[ΔH | P(Land)=1] = E[ΔH | P(Land)=0] under common support.

### C. The quantile treatment effects (QTE) and quantile regression (QR)
- Rationale:
  - Distribution of outcome may change in ways not revealed by averages (Frölich and Melly 2008); QTE captures heterogeneous effects across distributional quantiles.
  - QTE corresponds to horizontal differences between two cumulative distribution functions (CDFs).
- First-order stochastic dominance analysis compares CDFs for hours worked in 2007 and 2008.
  - Definition: F1(h) first-degree stochastically dominates F2(h) iff F1(h) ≤ F2(h) for all h in [0,1], and F1(h) < F2(h) for some h in [0,1] (Gravelle and Rees 2004).
- Quantile regression advantages (listed in source):
  - (i) allows modeling full conditional distribution of response given regressors;
  - (ii) point slope estimates for each quantile use the complete data set;
  - (iii) applicable when distribution is not Gaussian;
  - (iv) provides outlier robustness;
  - (v) can be more efficient than OLS if error terms are non-normal;
  - (vi) parameter confidence intervals can be estimated directly from conditional quantiles;
  - (vii) QR can be represented as a linear programming model.
- Conditional quantile function as starting point (equation (12) in source):
  - Q H i (τ | X i ) = F^{-1}_{H|X i} (τ)  (represented in source as equation (12))
- Quantile regression specification (equation (13) in source):
  - H i = α τ + δ τ title i + X' i α τ + ε i τ  (represented in source as equation (13))
- Interpretation caution:
  - QR results do not directly tell the effect on specific individuals; they indicate how effects vary across the outcome distribution. Under rank-invariance (rank-preserving), one may interpret larger effects at lower quantiles as larger impacts on individuals originally in lower tail.
- Unconditional QTE:
  - Define potential outcomes H i (1) and H i (0); observed H i = treat i * H i (1) + (1 - treat i) * H i (0).
  - Unconditional QTE at quantile τ: Q H(1) (τ) - Q H(0) (τ) (equation (14) in source).
- Advantages of unconditional QTE (Frölich and Melly 2008):
  - Unconditional QTE definition does not change when covariates change (does not depend on X).
  - Unconditional effects can be estimated consistently at rate n without parametric restrictions.
  - For policy makers, effects on entire population often more relevant than many conditional effects.
- Estimation approach used in paper:
  - Osasco titling program treated as unconditional QTE with exogenous treatment.
  - Follow Firpo (2007) two-step weighting estimator of check-functions under (i) ignorability of treatment, (ii) common support, and (iii) quantile monotonicity existence.
  - Weight function depends on propensity score estimated in first step. Given limited sample size, propensity score estimated parametrically via logit in this paper (rather than non-parametric local logit).
- QTE estimator used (as in source):
  - Δ α τ = arg min over α τ of Σ i ω̂ i ρ τ (H i - Δ treat i - X i α τ) subject to Δ̂ = treat i - ... (formula and notation preserved from source)
  - Weight function ω̂ i defined as (treat i / p̂(X i)) + ((1 - treat i) / (1 - p̂(X i))) where p̂(X i) is propensity score estimated parametrically.
- Two sets of QTE estimates reported:
  - First set not restricted to common support.
  - Second set conditioned to the common support.

### Empirical results — structure announced
- Section V begins empirical analysis by:
  - Discussing findings for the mean (OLS / DD / DDM / ATT).
  - Considering distributive effects estimated via QTE methods.
- Table 4 is indicated to present estimation results of model (8) for the baseline year, with subsequent sections reporting the detailed empirical outcomes.

*Italic source attribution: IMF Working Paper content unit from _wp11131 - Appendix shows the distribution of this index.*

### 2007. In that year, the program had not yet been implemented and no household had received

### _wp11131 - 2007. In that year, the program had not yet been implemented and no household had received

### Background and identification
- Pre-program year: 2007 — program not yet implemented and no household had received the land title (pre-program period) from Papel Passado.
- Pre-program checks find no significant impact of land title on weekly hours worked in 2007; ownership prior to program did not lead, on average, to a statistically significant difference in hours worked.
- Prior weak incentives to increase labor supply due to land title reduce the possibility of self-selection (anticipation) bias among titling program participants.

### Mean effects (OLS, Tobit)
- Table 5 (OLS, 2008 data) provides initial support that land title increases labor supply and weekly hours worked.
- Key point estimates (post-program, 2008):
  - OLS Naïve: Land title = 9.37*** (standard error (1.55))
  - OLS (with controls): Land title = 6.01** (2.50)
  - OLS Balanced Sample: Land title = 6.25** (2.50)
  - Tobit Balanced Sample: Land title = 9.66*** (3.55); Sigma = 18.45*** (1.01)
- Control variables with expected signs:
  - Years of education (family head) increases weekly hours: coefficient examples 0.49**, 0.50**, 0.68** (see Table 5).
  - Access to credit increases weekly hours: examples 4.07**, 3.68**, 5.44** (see Table 5).
  - Age marginally reduces hours (negative coefficients; some significant at 10 or 5 percent).
- Tobit model: high significance of sigma suggests nonrandom selection into the labor market; omission could yield omitted variable bias. Wald test: Prob > F(1, 275) = 0.34 indicating no statistically significant difference between Models 3 and 4 coefficients.

### Difference-in-Differences and Matching (DD, DDM)
- DD estimates (2007–2008) corroborate that land title increases weekly hours:
  - Land*year (DD interaction) = 9.01*** (1.16) in DD Naïve; 8.02*** (1.24) DD Unbalanced; 8.13*** (1.26) DD Balanced.
  - Year effect = 1.00*** (0.28) DD Naïve; 1.20*** (0.35) DD Unbalanced; 1.25*** (0.38) DD Balanced.
- Difference-in-Difference Matching (DDM) — kernel estimator with three bandwidths:
  - Bandwidth (0.01): Land title = 9.00*** (3.50)
  - Bandwidth (0.05): Land title = 7.98 (2.70)
  - Bandwidth (0.10): Land title = 7.8*** (2.50)
- Authors conclude average treatment effect on the treated group is around 9 hours/week.

### Distributive effects — Quantile Regression (QR)
- First-degree stochastic dominance:
  - 2007: treatment group's distribution of adults’ weekly hours worked dominates control group.
  - 2008: dominance increases significantly, indicating positive effects across the cumulative distribution, not only the mean.
- Conditional Quantile Regression (Table 8, common support, balanced sample, 288 observations):
  - Land title effects by quantile:
    - 0.25 quantile: 7.77* (4.18)
    - 0.50 quantile (median): 9.66*** (3.27)
    - 0.75 quantile: 1.41 (3.26)
    - 0.90 quantile: 0.43 (4.77)
  - Highest estimated effects concentrated in first and second quartiles: 7.8 hours/week (first quartile) and 9.7 hours/week (median) per text interpretation.
  - Years of education positive and significant at middle quantiles (examples: 0.64** at median, 0.69** at 0.75).
- Interquantile comparisons (Table 9):
  - Difference 0.5-0.25 = 1.89 (4.95) — not statistically significant.
  - Difference 0.5-0.75 = 8.25** (3.62) — statistically significant.
  - Difference 0.5-0.9 = 9.23* (5.02) — statistically significant.
- Interpretation: titling effects concentrated in the middle/lower part of the hours-worked distribution; outliers are not driving ATT.

### Quantile Treatment Effects (QTE, unconditional, Firpo 2007)
- Unconditional QTE estimates (Table 10):
  - Whole sample (305 obs):
    - 0.10: Land Title = -12.00** (5.30)
    - 0.25: Land Title = 25.00*** (1.90)
    - 0.50: Land Title = 7.50*** (2.80)
    - 0.75: Land Title = 7.60 (4.80)
  - Common support (288 obs):
    - 0.10: Land Title = -12.20** (5.30)
    - 0.25: Land Title = 14.96*** (2.84)
    - 0.50: Land Title = 4.70 (3.16)
    - 0.75: Land Title = 6.24 (4.88)
- Key patterns from QTE:
  - Impact tends to be higher in the first half of the outcome distribution than in the second.
  - In the common support, median effect declines from 25 to 15 hours/week (text reports decline from 25 to 15 and significance at 1 percent), suggesting sensitivity to sample selection.
  - Both conditional QR and unconditional QTE indicate heterogeneity and concentration of effects around the median.

### Synthesis of empirical findings
- Consistent evidence across methods that the Papel Passado land title program had a positive effect on hours worked and labor supply.
- Average ATT estimates around 9 hours/week (DD, DDM, OLS/Tobit).
- Distributive analysis shows heterogeneous effects: benefits concentrated among households with initially lower hours worked (poorest households).
- Robustness considerations:
  - Results sensitive to sample selection/common support.
  - Limited sample size noted as caveat for QTE; excluding workers with 0 initial hours affects higher-quantile coefficients.

### Policy implications and recommendations
- Land titling can increase labor supply by reducing households’ need to stay home protecting property, particularly benefiting the poorest urban households.
- Authorities should account for heterogeneous (distributive) effects when designing public policies: impacts may be restricted to ranges within the outcome distribution, so targeting can improve efficiency.
- Program evaluation should emphasize sampling on the common support (households with similar pre-treatment observables) and consider propensity score matching to minimize selection bias.
- Further research recommended on:
  - Effects on household income and utility from increased hours worked induced by titling.
  - Distributional impacts of titling on other outcomes (access to credit, fertility).
  - Broader applicability of titling programs as policy tools to address urban informality, poverty, and macroeconomic considerations.

### Selected descriptive and test statistics (from tables)
- Table 1 (Pre-Program 2007 vs Post-Program 2008, means):
  - Weekly hours of adult work: 10.19 (2007 mean), 16.18 (2007 Std. Dev.), 14.33 (2008 mean)
  - Monthly income (BRL): 1,126.25 (2007 mean), Std. Dev. 1,491.92; 1,138.76 (2008 mean), Std. Dev. 1,473.35
  - Observations: 304 (both years)
- Table 2 (2007 t-tests; control mean vs treatment mean, p-values):
  - Mean age: Control 42.60 vs Treatment 39.40, p-value 0.06*
  - Weekly hours of child labor (>16 years old): Control 8.35 vs Treatment 3.30, p-value 0.00***
  - Child labor2: Control 0.34 vs Treatment 0.14, p-value 0.00***
  - Years of education (family head): Control 5.00 vs Treatment 9.00, p-value 0.00***
  - Monthly income per capita: Control 553.10 vs Treatment 255.80, p-value 0.00***
  - Wealth index: Control 1.12 vs Treatment -0.94, p-value 0.00***
  - Informal worker: Control 0.94 vs Treatment 0.65, p-value 0.00***
  - Observations (households): Control 168, Treatment 137
- Table 3 (Spearman correlations, 2007):
  - Years of education correlated with Weekly Hours Worked: 0.14**
  - Informality correlated with Monthly income per capita: 0.23***
  - Monthly income per capita correlated with Weekly Hours Worked: 0.11*
- Table 7 (DDM kernel standard errors via bootstrap, 100 replications): see bandwidth-specific estimates above.

*Italic source attribution: Canonical PDF: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11131.pdf*

### References

### _wp11131 - References

### References
- Abadie, Alberto, 2002, “Bootstrap Tests for Distributional Treatment Effects in Instrumental Variable Models,” Journal of the American Statistical Association, Vol. 97, No. 457, pp. 284–292.
- Abadie, Alberto, 2005, "Semiparametric Difference-in-Differences Estimators," Review of Economic Studies, Vol. 72(1), pp. 1–19, 01.
- Alchian, Armen, and Harold Demsetz, 1973, “The Property Rights Paradigm,” The Journal of Economic History, Vol. 33, No. 1, pp. 16–27.
- Alston, Lee, Gary Libecap, and Robert Schneider, 1996, “The Determinants and Impact of Property Rights: Land Titles on the Brazilian Frontier,” Journal of Law, Economics & Organization, Vol. 12, pp. 25–61.
- Andrade, Maria T., 2006, “Direitos de Propriedade e Renda Pessoal: Um Estudo de Caso das Comunidades do Caju,” Revista do BNDES, Vol. 13, No. 26, pp. 261–274, Rio de Janeiro.
- Angelucci, Manuela, and Orazio Attanasio, 2009, “Oportunidades: Program Effect on Consumption, Low Participation, and Methodological Issues,” Economic Development and Cultural Change, Vol. 57, No.3, pp. 479–506.
- Angrist, Joshua D., and Jörn-Steffen Pischke, 2009, “Mostly Harmless Econometrics: An Empiricist’s Companion.” Princeton University Press, 2009.
- Associação dos Notários e Registradores do Estado de São Paulo—ANOREG, 2007, “Sistema de Biblioteca. Cartilha dos Registros Públicos,” São Paulo: Versão III, pp. 3–5. Available at: http://www.anoreg.org.br/.
- Baharoglu, Deniz, 2002, “World Bank Experience in Land Management and the Debate on Tenure Security,” World Bank Housing Research Background—Land Management Paper, July.
- Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan, 2004, “How Much Should We Trust Differences-in-Differences Estimates?” The Quarterly Journal of Economics, Vol. 119, No.1, pp. 249–275.
- Berhman, Jere R., and Petra Todd, 1999, “Randomness in the Experimental Samples of PROGRESA: Report Submitted to PROGRESA,” International Food Policy Research, Washington DC, mimeo.
- Besley, Timothy, 1995, “Property Rights and Investment Incentives: Theory and Evidence from Ghana,” The Journal of Political Economy, Vol. 103, No. 5, pp. 903–937.
- Binswanger, Hans, Klaus Deninger, and Gershon Feder, 1995, “Power, Distortions, Revolt, and Reform in Agricultural Land Relations,” Handbook of Development Economics, ed. 1, Vol. 3, Chapter 42, pp. 2659–2772, Elsevier.
- Bitler, Marianne P., Jonah B. Gelbach, and Hilary W. Hoynes, 2006, “What Mean Impact Miss: Distributional Effects of Welfare Reform Experiments,” The American Economic Review, Vol. 96, No. 4, pp. 988-1012.
- Blundell, Richard, and Monica C. Dias, 2002, “Alternative Approaches to Evaluation in Empirical Microeconomics.” CeMMAP Working Paper, CWP10/02, October.
- Blundell, Richard, and Thomas MaCurdy, 1999, “Labor Supply: A Review of Alternative Approaches,” in Handbook of Labor Economics, Vol. 3, Chapter 27, ed. by Orley C. Ashenfelter and David Card, pp. 1559–1695 (Amsterdam: Elsevier).
- Bolfarine, Heleno, and Wilton Bussab, 2005, “Elementos de Amostragem,” São Paulo: Edgar Blucher, First edition, Vol. 1.
- Brasselle, Anne-Sophie; Frederic Gaspart, and Jean-Philippe Platteau, 2002, “Land Tenure Security and Investment Incentives: Puzzling Evidence from Burkina Faso,” Journal of Development Economics, Vol. 67, Issue 2, pp. 373–418.
- Carter, Michael, and Pedro Olinto, 2003, “Getting Institutions Right for Whom? Credit Constraints and the Impact of Property Rights on the Quantify and Composition of Investment,” American Journal of Agricultural Economics, Vol. 85, pp. 173–186.
- Crump, Richard K., V. Joseph Hotz, and Guido W. Imbens, 2006, “Moving the Goalposts: Addressing the Limited Overlap in Estimation of Average Treatment Effects by Changing the Estimand,” IZA Discussion Paper No. 2347. Institute for the Study of Labor.
- Dammert, Ana C, 2009, “Heterogeneous Impacts of Conditional Cash Transfers: Evidence from Nicaragua,” Economic Development and Cultural Change, Vol. 58(1), pp. 53–83, University of Chicago Press.
- Demsetz, Harold, 1967, “Toward a Theory of Property Rights,” American Economic Review, Vol. 57, Issue 2, pp. 347–359.
- De Soto, Hernando, 2000, “O Mistério do Capital.” Rio de Janeiro: Editora Record.
- Do, Quy-Toan, and Lakshmi Iyer, 2003, “Land Rights and Economic Development: Evidence from Vietnam,” Policy Research Working Paper Series No. 3120, The World Bank, August.
- Duflo, Esther, Rema Hanna, and Stephen Ryan, 2005, “Monitoring Works: Getting Teachers to Come to School,” NBER Working Paper No. 11880, December.
- Field, Erica, 2003, “Entitled to Work: Urban Property Rights and Labor Supply in Peru,” Research Program in Development Studies Working Paper No. 223, Princeton University.
- Field, Erica, 2007, “Entitled to Work: Urban Property Rights and Labor Supply in Peru,” The Quarterly Journal of Economics, Vol. 122, No. 4, pp. 1561–1602.
- Field, Erica, and Maximo Torero, 2002, “Do Property Titles Increase Credit Access Among the Urban Poor? Evidence from Peru,” Research Program in Development Studies Working Paper No. 223, Princeton University.
- Firpo, Sérgio, 2007, “Efficient Semiparametric Estimation of Quantile Treatment Effects,” Econometrica, Vol. 75(1), pp. 259–276.
- Frölich, Markus, 2007, “Propensity Score Matching Without Conditional Independence Assumption With an Application to the Gender Wage in the UK,” Econometrics Journal, Vol. 139, pp. 35–75.
- Frölich, Markus, and Blaise Melly, 2008, “Unconditional Quantile Treatment Effects under Endogeneity,” IZA Discussion Paper No. 3288. Institute for the Study of Labor, January.
- Galiani, Sebastian, and Ernesto Schargrodsky, 2011, “Property Rights for the Poor: Effects of Land Titling,” Journal of Public Economics, forthcoming.
- Goldsmith, Arthur A, 1995, “Democracy, Property Rights and Economic Growth,” Journal of Development Studies, Vol. 32(2), pp. 157–174.
- Graham, John W., and Carole A Green, 1984, “Estimating the Parameters of a Household Production Function with Joint Products,” Review of Economics and Statistics, Vol. 66, Issue 2, pp. 277–82.
- Gravelle, Hugh, and Ray Rees, 2004, Microeconomics, 3rd ed. (London: Prentice Hall).
- Gronau, Rubin, 1977, “Leisure, Home Production, and Work―The Theory of Allocation of Time Revisited,” Journal of Political Economy 85, pp. 1099–1123.
- Heckman, James, and Joseph Hotz, 1989, “Choosing Among Alternative NX Methods for Estimating the Impact of Social Programs: The Case of Manpower Training,” Journal of the American Statistical Association, Vol. 84, pp. 862–874.
- Heckman, James, Hidehiko Ichimura, and Petra E. Todd, 1997, “Matching As An Econometric Evaluation Estimator: Evidence from Evaluating a Job Training Program,” Review of Economic Studies, Vol. 64, pp. 605–654.
- IBGE, 2007, “PNAD—Síntese dos Indicadores 2007, 2008 - Microdados,” Rio de Janeiro. Available at: <http://www.ibge.gov.br/home/estatistica/populacao/trabalhoerendimento/pnad2007/sintesepnad2007.pdf>. Access on September 28, 2008.
- Imbens, Guido W., and Jeff M. Wooldridge, 2008, “Recent Developments in the Econometrics of Program Evaluation,” NBER Working Paper No. W14251, August.
- Jacoby, Hanan G., Guo Li, and Scott Rozzelle, 2002, “Hazards of Expropriation: Tenure Insecurity and Investment in Rural China,” American Economic Review, Vol. 92, Issue 5, pp. 1420–1447.
- Jimenez, Emmanuel, 1984, “Tenure Security and Urban Squatting,” Review of Economics and Statistics, Vol. 66, Issue 4, pp. 556–567.
- Jimenez, Emmanuel, 1985, “Urban Squatting and Community Organization in Developing Countries,” Journal of Public Economics, Vol. 27, pp. 69–92.
- Koenker, Roger, 2005, “Quantile Regression in R: A Vignette,” Cambridge Books, Cambridge University Press, Number 9780521608275, pp. 5.
- Lanjouw, Jean O., and Philip Levy, 2002, Untitled: A Study of Formal and Informal Property Rights in Urban Ecuador. The Economic Journal, Vol. 112 (482), pp. 986–1019.
- Melly, Blaise, 2006, “Estimation of Counterfactual Distributions Using Quantile Regression,” Swiss Institute for International Economics and Applied Economic Research (SIAW), mimeo.
- Meyer, Bruce D., 1995, “Natural and Quasi-Experiments in Economics,” Journal of Business & Economic Statistics, Vol.13, No. 2, pp. 151–61.
- Micelli, Thomas J., C.F. Sirmans, and Joseph Kieyah, 2001, “The Demand for Land Title Registration: Theory with Evidence from Kenya,” American Law and Economics Review, Vol. 3, No. 2, pp. 275–287.
- Mishra, Prachi, Peter Montiel, and Antonio Spilimbergo, 2010, “Monetary Transmission in Low Income Countries,” IMF Working Paper 10/223 (Washington: International Monetary Fund).
- North, Douglass C, 1990, “Institutions, Institutional Change and Economic Performance,” Cambridge: Cambridge University Press.
- North, Douglass C, and Richard P. Thomas, 1973, “The Rise of the Western World: A New Economic History,” Cambridge: Cambridge University Press.
- Place, Frank, and Shem Migot-Adholla, 1998, “The Economic Effects of Land Registration for Smallholder Farms in Kenya: Evidence from Nyeri and Kakamega Districts,” Land Economics, Vol. 74, No. 3, pp. 360–373.
- Rivera, Berta, and Luis Currais, 2005, “Individual Returns to Health in Brazil: a Quantile Regression Analysis,” in Health and Economic Growth: Findings and Policy Implications ed. by Gullem López-Casasnovas, Berta Rivera, and Luis Currais, Cambridge, MIT Press.
- Rubin, Donald, and Thomas N. Herzog, 2000, “Combining Propensity Score Matching With Additional Adjustments for Prognostic Covariates,” Journal of the American Statistical Association, Vol. 95, pp. 573–585.
- Silva, Everton, and Sabino Porto Junior, 2004, “Sistema Financeiro e Crescimento Econômico: uma Aplicação da Regressão Quantílica,” Proceedings of the 32nd Brazilian Economics Meeting, ANPEC.
- Skoufias, Emmanuel, 2001, “PROGRESA and its Impacts on the Human Capital and Welfare of Households in Rural Mexico: A Synthesis of the Results of an Evaluation by IFPRI International Food Policy Research Institute, International Food Policy Research Institute: Washington, DC.
- Smith, Jeffrey A., and Petra E. Todd, 2005, “Does Matching Overcome LaLonde’s Critique of Nonexperimental Estimators?” Journal of Econometrics, Vol. 125, pp. 305–354.
- The Economist, 2007, “Property Rights: China's Next Revolution?” March 12th.
- Torstensson, Johan, 1994, “Property Rights and Economic Growth: An Empirical Study,” Kiklos, Vol. 47, Issue 2, pp. 231–247.
- United Nations, 2005, “Habitat Report,” United Nations: New York.
- Wooldridge, Jeffrey M., 2002, Econometric Analysis of Cross Section and Panel Data, (Cambridge, MA: MIT Press Books).
- World Bank, 2000, “Urban Poor Gain Access to Property Market,” World Bank Development New Archives, Peru, February 2nd.
- Zylberstajn, Helio, and Giacomo Neto, 1999, “As Teorias de Desemprego e as Políticas Públicas de Emprego,” Estudos Econômicos, Vol. 29, No. 1, pp. 129–149.

### Appendix

- Table A. Propensity Score: Logit Estimates for the Selection of the Treatment Group, 2007
  - Context: Dummy = 1 if a household lives in the treated area (Canaã)
  - Column headings: (Unmatched Sample) and (Matched Sample)
  - Variables and estimated coefficients (standard errors in parentheses):
    - Gender(= 1 if have): 0.32 (0.48) ; 0.11 (0.51)
    - Ethinicity (non-white): 0.04 (0.45) ; 0.02 (0.45)
    - Marital status(= 1 if have): 0.58 (0.47) ; 0.35 (0.49)
    - Age: -0.03* (0.01) ; -0.01 (0.01)
    - Weekly hours of adult work: 0.02 (0.01) ; 0.01 (0.01)
    - Weekly hours of child labor: -0.03 (0.02) ; -0.01 (0.02)
    - Years of education (head): 0.14*** (0.05) ; 0.05 (0.08)
    - Monthly income per capita: -0.01** (0.00) ; -0.01 (0.00)
    - TV ( = 1 if have): -1.48** (0.69) ; -0.68 (0.85)
    - DVD (= 1 if have): -0.64 (0.53) ; -0.29 (0.58)
    - Radio (= 1 if have): -1.68*** (0.50) ; -0.60 (0.84)
    - Car (= 1 if have): -0.28 (0.45) ; -0.09 (0.48)
    - Washing machine (= 1 if have): 2.19*** (0.65) ; 1.06 (0.92)
    - Refrigerator (= 1 if have): -6.07*** (1.07) ; -2.76 (2.15)
    - Informal worker: -1.73*** (0.62) ; -0.75 (0.85)
    - Credit: -0.17 (0.43) ; -0.03 (0.45)
    - Constant: 8.18*** (1.62) ; 1.87 (4.09)
  - Model fit and sample:
    - Pseudo-R2: 0.62 (Unmatched Sample) ; 0.63 (Matched Sample)
    - Prob > Chi2(16): 0.00 (Unmatched Sample) ; 1.00 (Matched Sample)
    - Observations: 305 (Unmatched Sample) ; 288 (Matched Sample)
  - Note: ***, **,* Statistically significant at 1%, 5% and 10%, respectively.

- Figure A. Distribution of Wealth Index, 2008
  - Y-axis label: Density
  - X-axis label: Wealth Index
  - X-axis tick labels shown as: -4 -2 0 2
  - Density axis tick labels shown as: 0 .2 .4 .6 .8

*Source: _wp11131 - References*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11131.pdf_