Report No: AUS0000624 . Timor-Leste Timor-Leste Poverty Developing Timor-Leste Gender-Disaggregated Poverty Small Area Estimates Technical Report . May 2019 . POV . © 2019 The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved This work is a product of the staff of The World Bank. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Attribution—Please cite the work as follows: “World Bank. 2019. DEVELOPING TIMOR-LESTE GENDER- DISAGGREGATED POVERTY SMALL AREA ESTIMATES – TECHNICAL REPORT. © World Bank.” All queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. ii TABLE OF CONTENTS List of Figures ........................................................................................................................... ii List of Tables ........................................................................................................................... iv ABBREVIATIONS AND ACRONYMS ................................................................................ vi EXECUTIVE SUMMARY .................................................................................................... vii 1. Introduction ....................................................................................................................... 1 2. Overview of the methodology .......................................................................................... 2 3. Data .................................................................................................................................... 5 3.1 The Census ................................................................................................................. 5 3.2 The Timor Leste Survey of Living Standards ............................................................ 6 3.2 The Poverty Line............................................................................................................ 7 4. Empirical Analysis ............................................................................................................ 7 4.1 Comparing the Questionnaires ................................................................................... 7 4.2 Comparing the Variables ............................................................................................ 9 4.3 Variable Selection for Poverty Mapping Models ..................................................... 10 5. Simulation and Poverty Mapping Results ....................................................................... 11 5.1 Cross-validation with Other Welfare Indicators ...................................................... 11 5.2 Suco-Level Poverty Estimates ................................................................................. 13 6. Poverty and Gender ......................................................................................................... 20 7. Disaggregated Evidence for other Gender-Related Indicators ........................................ 24 Education ......................................................................................................................... 26 Health............................................................................................................................... 29 Labour Force.................................................................................................................... 33 8. Conclusions ..................................................................................................................... 40 References ............................................................................................................................... 43 Annex - Suco Level Maps (Full Page Version) ...................................................................... 87 i List of Figures Figure 1: Illuminated Areas in Timor Leste, 2013 DMSP Satellite F18 (using 5% Luminosity Threshold to restrict to non-ephemeral lit areas) ............................... 13 Figure 2: National-level Poverty Mapping Model Estimation ............................................... 14 Figure 3: National Model Estimation ..................................................................................... 16 Figure 4: Rural Sector Poverty Mapping Model estimation ................................................... 18 Figure 5: Urban Sector Poverty Mapping Model estimation .................................................. 19 Figure 6: Suco-Level Mean Differences of People in Male-Headed vs Female-Headed Households ............................................................................................................ 22 Figure 7: Suco-Level Relative Differences of People in Male-Headed vs Female-Headed Households ............................................................................................................ 23 Figure 8: Proportion of the Population in Households Where the Index of Male-Female Education Gaps Indicates Female Disadvantage .................................................. 27 Figure 9: Relationship Between Gender Education Disadvantage and Suco-Level Mean Consumption ......................................................................................................... 27 Figure 10: Comparison Between Direct Estimates from the Census and Survey-to-Census Imputed Values, in terms of the Proportion of the Population in Households, Where the Index of Male-Female Education Gaps Indicates Female Disadvantage ......................................................................................................... 29 Figure 11: Proportion of the Population in Households, Where the Index of Male-Female Health Gaps Indicates Female Disadvantage ........................................................ 31 Figure 12: Relationship Between Gender Health and Suco-Level Mean Consumption ........ 31 Figure 13: Proportion of the Population in Households Where the Index of Male-Female Labour Force Gaps Indicates Female Disadvantage ............................................. 35 Figure 14: Relationship Between Labour Force Disadvantage and Suco-Level Mean Consumption ......................................................................................................... 35 Figure 15: Proportion of the Population in Households, Where the Index of Female Decision Making (DM) Autonomy Indicates Female Disadvantage .................... 38 Figure 16: Proportion of the Population in Households Where the Index of Female Experience of Types of Domestic Violence (DV) Indicates Female Disadvantage ......................................................................................................... 39 ii Figure 17: Relationship Between Share of Population Living in Households With High Domestic Violence Index and Average Predicted Headcount Poverty Rate by Suco....................................................................................................................... 39 Figure 18: Suco Average of the Predicted Number of Types of Domestic Violence (DV) Reported by Females............................................................................................. 40 Figure 19: Suco-Level Estimates of Poverty Rate and 95% CI .............................................. 84 Figure 20: Suco-Level Estimates of Labor Force Index and 95% CI ..................................... 84 Figure 21. Suco-Level Estimates of Health Index and 95% CI .............................................. 85 Figure 22. Suco-Level Estimates of Education Index and 95% CI ........................................ 85 Figure 23. Suco-Level Estimats of Decision-Making Autonomy (DM) Index and 95% CI .. 86 Figure 24. Suco-Level Estimates of Domestic Violence (DV) Index and 95% CI ................ 86 iii List of Tables Table 1: Comparison between Census and TL SLS (Wall Material and Drinking Water Examples) ...............................................................................................................8 Table 2: Summary Details for Beta and Alpha Models .........................................................11 Table 3: Summary Details for Beta and Alpha Models of Gender-Related Indicators in TLSLS .....................................................................................................................33 Table 4: Summary Details for Beta and Alpha Models of Power and Agency Indicators in 2016 DHS................................................................................................................37 Table 5: Comparison of Census and TLSLS Variables for Household-Level Characteristics .........................................................................................................45 Table 6: Comparison of Census and TLSLS Variables for Person-Level Characteristics ....48 Table 7: Beta Models for Predicting Per Capita Consumption, National Sample (N=5916) 50 Table 8: Alpha Model for Heteroscedasticity, National Sample (N=5916) ..........................51 Table 9: Beta Models for Predicting Per Capita Consumption, Rural Sub-Sample (N=3898) .................................................................................................................52 Table 10: Alpha Model for Heteroscedasticity, Rural Sub-Sample (N=3898) ......................53 Table 11: Beta Models for Predicting Per Capita Consumption, Urban Sub-Sample (N=2018) ...............................................................................................................54 Table 12: Alpha Model for Heteroscedasticity, Urban Sub-Sample (N=2018).....................54 Table 13: Beta Models for Predicting Labour Force Indicator ..............................................54 Table 14: Alpha Model for Heteroscedasticity in work_pca .................................................55 Table 15: Beta Model for Education Principal Components Index .......................................56 Table 16: Alpha Model for Heteroscedasticity for the Education Index Survey-to-Census Imputation .............................................................................................................56 Table 17: Beta Model for Health Principal Components Index ............................................57 Table 18: Alpha Model for Heteroscedasticity for the Health Index Survey-to-Census Imputation .............................................................................................................57 Table 19: Comparison of Census and DHS Variables for Household-Level Characteristics .......................................................................................................58 Table 20: Comparison of Census and DHS Variables for Person-Level Characteristics ......60 Table 21: Beta Models for Predicting PCA Index of Types of Female Decision-Making Autonomy (N=7013) .............................................................................................61 Table 22: Alpha Model for Heteroscedasticity, DM Index (N=7013)...................................61 iv Table 23: Beta Models for Predicting PCA Index of Types of Domestic Violence Indicators (N=3674) ..............................................................................................62 Table 24: Alpha Model for Heteroscedasticity, DV Index (N=3674) ...................................62 Table 25: Beta Models for Predicting Count of Types of Domestic Violence Indicators (N=3674) ...............................................................................................................63 Table 26: Alpha Model for Heteroscedasticity, DV Count (N=3674)...................................63 Table 27: Suco-level Predicted Poverty and Inequality .........................................................64 Table 28: Suco-level Predicted Gender Indicators from 2014 TLSLS ..................................74 v ABBREVIATIONS AND ACRONYMS DHS Demographic and Health Survey DM Decision Making DN Digital Number DV Domestic Violence EA enumeration area ELL Elbers, Lanjouw and Lanjouw GLS Generalized Least Squares LFS Labour Force Survey OLS Ordinary Least Squares SD Standard Deviation SE Standard Error TLSLS Timor Leste Survey of Living Standards vi EXECUTIVE SUMMARY In recent years Timor-Leste has made significant progress in reducing poverty. Based on the most recent Standards of Living Survey, the number of people living below the national poverty line declined from 50.4 percent in 2007 to 41.8 percent in 2014. Measured using the US$1.90 per person per day (2011 PPP) poverty line, the decline was even more rapid, from 47.2 percent in 2007 to 30.3 percent in 2014. These impressive reductions, however, were not experienced equally across the country. In some areas people still live with high levels of extreme poverty. Although 80 percent of the poor are concentrated in rural areas, Dili, the capital, is home to the largest cluster of people living in poverty. While female-headed households are less likely to be poor than male-headed households, gender disparity is still apparent in broader socioeconomic dimensions. For example, in 2015, 61 percent of men were employed, compared with only 44 percent of women. Meanwhile, the maternal mortality rate in Timor-Leste is still far higher than that of its regional peers. Domestic violence is also pervasive. The Government of Timor-Leste has set a goal to eradicate extreme poverty by introducing more socially inclusive and gender-sensitive policies and programs; their success will depend on how effectively these policies and programs are developed and targeted. For example, if the poor are concentrated in certain areas, spatial targeting of poverty reduction programs and public services to those areas is likely to be more effective than trying to target the poor individually. The challenge in Timor-Leste, however, is that the existing consumption-based poverty estimates from the 2014 Survey of Living Standards (2014 TLSLS) are representative only at the district level and therefore do not capture the detail of the heterogeneity of living standards and access to services, and how these affect men and women differently, within districts. Earlier poverty analysis for Timor Leste only used survey data, and provided estimates only for the 13 municipalities. As such, it overlooked substantial differences within municipalities in suco-level poverty rates, as well as intra-municipality variation in living standards more generally. vii Therefore village- (suco) level, gender-disaggregated poverty statistics are needed to reveal which suco within each district have particularly high rates of poverty and gender disparity. Responding to this need, the World Bank – in close collaboration with the General Directorate of Statistics Timor-Leste – used small area estimation (SAE) techniques to develop suco-level gender-sensitive poverty maps. The current study combined data primarily from (i) the 2015 Population and Housing Census, (ii) the 2014 TLSLS, and (3) the 2016 Demographic and Health Survey (DHS) to estimate various indicators for each of the country’s 442 sucos. The analysis proceeds in two parts. The first part of the analysis employs a ‘traditional’ poverty mapping approach, which uses monetary measures of poverty, and the findings contrast with those of prior analysis. The survey-to-census imputations carried out here reveal differences within municipalities in suco level poverty rates and the intra-municipalities variation in living standards more generally, and can provide useful information for developing spatially targeted interventions, such as local development programs. These results can also inform future analyses of the driving forces behind the spatial variation in poverty in Timor Leste. The traditional poverty maps reveal an already known pattern, that the headcount poverty rate is much higher in western areas of Timor Leste than in eastern areas. However, the new generated maps also show something that was not previously known, which is that there is much more variation in poverty rates within municipalities than between municipalities. Thus, the municipality-level headcount poverty rate is of limited use for understanding the incidence of poverty of sucos within a particular municipality. For example, the Dili municipality has suco-level headcount poverty rates that range from 8–80%; in Manatuto the rates are 10–71%. One observable gender-related consumption poverty indicator is the gender of the household head, which can be determined using standard household-level data. Evidence from the 2014 TLSLS shows that female-headed households in Timor Leste are less likely to be poor than male-headed households (World Bank, 2016). While these patterns should hold, on average, across sucos, there may be considerable variation among them in the difference in poverty rates between people in male- vs. female-headed households. This spatial variation could arise from underlying variation in gender-related economic opportunities or location-specific constraints that may be stronger for female-headed households than for their male-headed counterparts. viii The SAE method can provide spatially disaggregated indicators of the difference in poverty rates according to the gender of the household head. Almost 16% of census households are headed by women, an average of just over 70 female-headed households per suco, which is a sufficient number to estimate the suco-level poverty rate for female- vs. male-headed households. Comparing these two poverty rates reveals where female-headed households face the greatest relative disadvantage. The maps on the difference in headcount poverty rates and in the poverty gap index for male- vs. female-headed households show that only in Cotabot, in Bobonaro district, do female-headed households have a higher poverty rate than male-headed ones. If the gender gap is measured using the poverty gap index, poverty rates are higher in female- than in male-headed households in five sucos – Cotabot again, Edi in Ainaro, Luculai in Liquiçá, and Muapitine and Pairara in Lautém. Considering that only 16 percent of households in Timor-Leste are female headed, the results of the analysis based on Female-Headed Households (FHHs) represent the situation of a small minority of women and girls. The second part of the analysis is non-traditional used of SAE techniques to spatially disaggregate gender-related indicators from the 2014 TLSLS and the 2016 DHS. Compared to poverty mapping models based on household consumption, the models for gender-related indicators for labour force activity, education, health, decision- making autonomy, and abuse and domestic violence have a much lower predictive power, likely due to the idiosyncratic nature of some of these indicators. The gender gap in education was constructed based on 2014 TLSLS data for two types of gaps: (i) the difference in the household-level sum of an indicator for whether a person is illiterate; and (ii) the difference in the household-level sum of an indicator for whether a person never attended school, for both males and females aged 5 and above. The results of the survey-to- census imputation of the education index show that the prevalence of female disadvantage in the education index is higher in poorer areas. Across country, the lowest gaps are observed in and around Dili. This pattern is confirmed by the finding that the poverty headcount rate for the proportion of the population living in households with negative values of the education index (which indicates female disadvantage) is negatively correlated with the suco-level mean of predicted per capita consumption for census households. ix Combined indicators in the labor force are presented in a composite index. The index of male/female gaps (defined such that negative values denote female disadvantage) in the labor force is constructed from the difference in the household-level total of males/females who engaged in no economic activity in the past week and the hours of wage labor supplied in the previous seven days across all jobs. The result shows an inverse pattern between gender disadvantage in the labor market and poverty rates. Gender-related labor force gaps are wider in suco where households are richer on average, and where poverty rates are lower. This is possibly because women from poorer households are more likely to participate in the labor market in order to support their family, and accordingly, their nonparticipation might be seen as welfare improving. This pattern is confirmed by scatter plots for the relationship between suco-level mean of predicted per capita consumption, which shows that it is in richer areas of Timor-Leste that gender gaps in labor force indicators are likely to be most apparent. While the 2015 Census contains very limited variables on health, the 2014 TLSLS contains several. Employing the SAE method, an index of health gender disparity was constructed from TLSLS data on two types of male/female gap: (i) the difference in the household-level sum of the number of days in the past 30 days that males/females of all ages were affected by ill health; and (ii) the difference in the household-level sum of the number of hospitalizations for females/males in the past 12 months. The findings suggest that females do not appear to be disadvantaged in terms of these health measures. For example, the household-level average of the number of days of female illness was 94 percent of the average number of days of male illness. Females also had fewer spells of hospitalization (although these were rare for both males and females). The map suggests that a higher proportion of the population lives in households with a female health disadvantage in Oecusse, and there are also concentrations in Baucau and Viqueque, but the patterns are more scattered than for the education index. However, there is a concern that the findings might be attributed to potential weaknesses of self-reported health status. For example, people have different levels of tolerance of illness. It has also been argued that disadvantaged populations tend to fail to perceive and report the presence of illness (Sen, A., 2002). Perhaps because of this, the gender-related health index shows the weakest relationship with predicted consumption, suggesting there is almost no correlation between a suco’s welfare and gender gaps on this health index. x The 2016 DHS contains information relating to aspects of power and agency, which are important gender indicators that can usefully be disaggregated using the SAE approach. The first indicator of power and agency used here is an index related to female autonomy in decision-making. Adult females who were married or living with a man at the time of the survey were asked whether they make decisions regarding their own health care, major purchases, and visits to their family and relatives. The suco-level results of this index suggest that the locations with the highest proportion of the population living in households with the lowest levels of female decision-making autonomy are scattered in some inland parts of the country. There are no apparent patterns with respect to average consumption levels or poverty headcount rates. The second indicator of power and agency is related to the experience of any 19 types of physical abuse, verbal threats, being afraid, or domestic violence from the current or former male partner. Only 26 percent of the respondents had no experience of any of these types of abuse or domestic violence; the mean was to have experienced 2.1 (out of 19) of these indicators of abuse and domestic violence. The index of female experience of types of domestic violence was constructed based on these 19 variables. This index is most highly correlated with elements of sexual violence (being forced to have sexual intercourse and forced with threats or in any other way to perform sexual acts) and least with being afraid of the partner. Comparing to the autonomy in decision-making, there is much clearer evidence regarding the prevalence of abuse and domestic violence. The western areas of the country, especially Oecusse, have a higher prevalence of households with high values on the domestic violence index. This geographic pattern is similar to that of poverty headcount rates, which are also higher in the west. Indeed, a scatterplot between the share of the population living in households with high scores on the domestic violence index and the average predicted poverty headcount rate shows a significant positive correlation. These gender-sensitive poverty maps at the suco level, which contain more finely grained detail on poverty variations, provide new insights on poverty in Timor-Leste. They also highlight hotspots of gender-disaggregated deprivation along dimensions such as access to economic opportunities, education, health, and power and agency. The maps can be used to inform the design of policies and programs targeting poverty at the suco level, and could help xi improve resource allocation aimed at raising the living standards of the poor, and balancing the targeting of poor areas and poor people, while also closing gender gaps in these dimensions. A further use of the results is for future analytics studies that aim to explore some of the driving forces behind the spatial variation in poverty and gender disparity in Timor-Leste. Furthermore, they provide a cost-effective way of adding value to existing census and survey data collections, and can serve as a substitute for fielding expensive new censuses or surveys. However, it is important to improve the consistency and harmonization of survey and census instruments to get much more benefit of very costly data collection. xii 1. Introduction Small area estimates of poverty and inequality statistics, through survey-to-census imputation that lets consumption be estimated for each and every household in a census, are useful for at least three reasons. First, they can help improve the effectiveness of public spending, by targeting to prevent the leakage of benefits to the non-poor (and prevent the under-coverage of the poor). If poor people are concentrated in certain areas, spatial targeting by directing extra development projects and public services to those areas, may be more feasible than trying to individually target the poor. Geographic targeting is highly relevant in countries like Timor Leste, where mountainous topography contributes to high levels of heterogeneity. In similar environments, such as Papua New Guinea, the enclave nature of some modern economic development has created high levels of spatial inequality (Gibson et al, 2005). A practical problem is that geographic targeting is most effective for smaller targeted areas.1 Yet the detailed household surveys used to measure poverty lack sufficient sample size to yield statistically precise estimates for small areas. For example, in the 2014/15 Timor Leste Survey of Living Standards (TLSLS) the sample of 5916 is almost three percent of all households in the country. The sampled households came from 400 enumeration areas (EAs), which is almost one-fifth of all of the EAs in the country. By the standards of most countries this is a relatively large sample and existing statistical capacity is unlikely to support a larger sample size without endangering data quality. Even with this sample size, only poverty rates at the municipality level (n=13) are reported, with standard errors for the head-count index about one-eighth of the value of the index and for distributional-sensitive measures, such as the poverty severity index, about one-quarter of the value of the index. At this aggregation level, any variation in living standards within municipalities is obscured and the relatively wide standard errors limit the confidence with which claims can be made about inter-area differences in poverty. One increasingly popular method to provide the information to support finer spatial targeting is small-area estimation. These estimates also provide a database of welfare measures (such as 1 In an early example, Bigman and Srinivasan (2002) show how a given budget for poverty alleviation targeted at the level of districts in India (n=340 in their sample) rather than at the broader state level (n=15) would allow an extra 4.3 million poor people to benefit from the program with no extra cost. 1 mean per capita consumption and poverty and inequality rates) for small geographic areas that can be used for further research studies. This creation of ‘new’ data is especially valuable in countries where access to the unit record data from the primary sources (censuses and surveys) is limited. The third benefit of these estimates is that they provide a way to add value to existing census and survey data collections at relatively low cost, and in some cases could be a substitute for fielding expensive new censuses or surveys. The basic details are that household survey data are used to estimate a model of consumption, with explanatory variables restricted to those that have overlapping distributions from a census. The coefficients from this model are then combined with the variables from the census, and consumption is predicted for each household in the census. With these predictions available for all households, inequality and poverty statistics can be estimated for small geographic areas (Elbers et al, 2003).2 In the results below, the poverty statistics that are calculated by using the predicted consumption data for each census household are reported at the suco level (n=442). For the headcount poverty rate, the standard errors at the suco level (relative to the poverty index) average one-quarter and so this is a comparable degree of precision to what the survey offered at the municipality level (n=13) for a variable like the poverty severity index. 2. Overview of the methodology The methodology is based on the Elbers, Lanjouw and Lanjouw (2003) approach (hereafter, referred to as ELL) and is implemented using the recently written sae set of Stata procedures (World Bank, 2017). In the first stage, a model of (log) consumption per capita for people living in household h in location c is estimated. In what follows, the location c will correspond to an enumeration area (or ‘cluster’), of which there are n=2281 in the 2015 census. There were n=400 EAs included in the TLSLS sample (noting that it was based on the census geography used in 2010 and that a concordance between the two censuses is only at the suco level): ln y ch = x 'ch β + u ch (1) 2 Bedi et al (2007) provide several examples of use of this method. A validation using census data from Brazil is provided in Elbers et al (2008) and extensions to survey-to-survey imputations for situations where survey methods have not maintained comparability over time are provided by Christiaensen et al (2012). 2 The vector of explanatory variables, x ch for household h in location c is restricted to those survey variables that can also be found in the census and that have an overlap with the distribution of the same variable in the census. The parameter vector β is not given any causal interpretation in the model because equation (1) is a prediction equation, not a model of what causes consumption. The error term has two independent components: a cluster specific effect c and a household specific effect  ch . The cluster specific effect is for aspects of the environment common to households who live in the same location. If one worked just with the survey data, these unseen elements could be controlled for with cluster fixed effects. However, because the survey sampled just a subset of all enumeration areas, there is no way to extrapolate from the fixed effects estimated on the included EAs to the remaining EAs in the census that are not in the TLSLS sample. Thus, another way has to be found to incorporate location information, otherwise it will end up in the residuals of equation (1), where it is potentially disruptive by making predicted consumption less precise (so derived poverty maps will tend to blur differences between areas). The reason for the lowered precision is that when the predictions for each household are summed or averaged, even if there are hundreds of census households in a locality, if much of the error is common to groups of households rather than being idiosyncratic and random, the gains in precision that normally come from averaging over larger numbers are muted. In order to reduce the contribution from location effects, the poverty mapping literature tends to use cluster means of household level variables, which are calculated from the census data so that they are available for all census and survey clusters. That approach is followed here as well, to reduce the contribution of the location component in the error.3 The residuals from the equation (1) regression are then decomposed into two parts; the uncorrelated household idiosyncratic components and the correlated location components: 3 In contrast to poverty mapping studies in many countries, the census averages are at the suco level (n=442) rather than the EA level (n=2281) because the survey data is based on the administrative geography from the 2010 census, while the means of the census data are from the 2015 census. The concordance between the two censuses has only been formed at the suco level rather than at the EA level. 3 ˆch = u ˆc + ˆch (2) The estimated location components given by  ˆc are the within-cluster means of the overall residuals, while the household component estimates ˆch , are the overall household-level residuals net of the location components. Additional parameters needed by the ELL method are: ˆ 2 , the variance of  c and Vˆ ( 2 ) , the variance of  2 . To allow for heteroskedasticity in  the household idiosyncratic component, a logistic model of the variance of  ch conditional on a set of explanatory variables, x ch is estimated:   ch 2  ' ln  2  = x ch ˆ + rch (3)  A −  ch  where x ch is a set of variables that are selected from a larger candidate pool by using a stepwise approach to find the model that most parsimoniously explains variation in  ch 2 . The candidate variables are not only those from equation (1) but also interactions between those variables and the predictions and squared predictions from equation (1), and A is set equal to 1.05  max { ch 2 }. The model used to estimate equation (3) is referred to as the “alpha model” and that used to estimate equation (1) is the “beta model”. The results from the alpha model feed in to the calculation of a household specific variance estimator for  ch which is:  AB  1  AB(1 − B )  ˆ 2,ch =   + Var ( r )  3  (4) 1 + B  2  (1 + B )  where exp {x 'ch ˆ}= B . These error calculations are used to produce two n  n square matrices, where n is the number of surveyed households. The first is a block matrix, where each block corresponds to a cluster, and the cell entries within each block are ˆ  . The second is a diagonal 2 matrix, with household-specific entries given by ˆ 2,ch . The sum of these two matrices is Σ ˆ , the estimated variance-covariance matrix for the consumption model. Once this matrix has been formed, the original model in equation (1) can be re-estimated by the Generalized Least Squares (GLS) method that can account for the regression disturbances not being identically and independently distributed. 4 In the simulation stage, the estimated regression coefficients from equation (1) are applied to x ch from the census to obtain predicted consumption for each household. A series of 100 simulations are conducted, and for each simulation, r, a set of beta and alpha coefficients, ̃ and ̃ are drawn, from the multivariate normal distributions described by the first stage point ~ 2 ) r a simulated estimates and their associated variance-covariance matrices. Additionally, (  value of the variance of the location error component is drawn. Combining the coefficients from the alpha model with the census data, for each census household the household-specific ~ 2 ) r is estimated. Then for each household, variance of the household error component, (  ,ch ~ r are drawn from their corresponding distributions. A ~ r and  simulated disturbance terms,  c ch ˆ ch value of consumption expenditure for each census household, y r is then simulated, which is ~ based on the combined effect of the predicted log expenditure, x 'ch; β r and the disturbance terms: ~ ~r ~r ˆ ch y r = exp (x 'ch β r +  c +  ch ) (5) Finally, the full set of simulated ˆ ch y r values are used to calculate expected values and standard errors of distributional statistics, including poverty measures, for small areas (sucos, in this case). Specifically, the simulations are repeated 100 times, drawing a new set of coefficients and disturbance terms for each simulation. The mean of a given statistic, such as the headcount poverty rate or the Gini index, can be calculated across these 100 simulated datasets for any level of geography. The mean provides the point estimate of that statistic for that location, and the standard deviation serves as an estimate of the standard error. 3. Data 3.1 The Census The Population and Housing Census was conducted in July 2015 and consisted of 54 questions for individuals (some of which were age- and gender-specific), 21 questions at the household level, which included questions about the dwelling, plus further questions about recent deaths and about former household members living abroad. There were 204,582 private households that were enumerated; these are the focus here because the approximately 1000 individuals in 5 institutions have no dwelling information available.4 The census households are located in 2281 EAs, that each have almost 90 households, on average. While the simulated cluster effects are based on the EAs, the linkage of census characteristics to survey households at the estimation stage is based on sucos (n=442) because of the lack of an EA-level concordance between the 2010 census (which the sample for TLSLS is based on) and the 2015 census. 3.2 The Timor Leste Survey of Living Standards The data on consumption expenditures come from the TLSLS, fielded from April 2014 to March 2015. The survey is based on a sample of 400 EAs, that was stratified by urban and rural sector within each municipality.5 A target of 15 households were to be surveyed per EA (and 5916 of the target 6000 households have data available). The sampling weights (expansion factors) reflect the unequal probabilities of selection inherent in a sample that is representative at municipality level. While there is a six-fold difference in population between the most (Dili) and the least (Manatuto) populous municipality, the sample has just a three-fold difference for the municipalities with the largest and smallest samples (because small municipalities have to be over-sampled to get a large enough minimum sample size – set as 255 households – to give representative estimates). Thus, the sampling weights have a wide range, from 3 to 300 (with a mean of 31). The weights are applied to the survey data at all stages of the poverty mapping process, and the simulations also take account of variation in household size (larger households tend to be poorer, so the poverty rate amongst people living in poor households is higher than the poverty rate calculated at the household level). The consumption estimates are in terms of US$ per person per month. These rely on a 7-day recall of consumption for 135 food groups, plus a mixed period (month, 3-month, or annual) recall for expenditure on 53 non-food groups. The consumption aggregate also includes an estimate of imputed rent, which is based on a hedonic housing equation. 4 The number of households is slightly fewer than in prior reports because duplicate records are removed. 5 Specifically, the 400 EAs were a random selection of the 472 EAs used for the 2012 Labour Force Survey (LFS), with the LFS sample of EAs based on the 2010 census. 6 3.2 The Poverty Line The poverty lines are based on the Poverty in Timor Leste: 2014 report and only basic details are provided here. A cost of basic needs poverty line was calculated for each municipality (but with no urban-rural sectoral differences within municipalities). Food baskets were anchored to a nutritional norm of 2100 calories per person per day, and were meant to “correspond to the average food consumption pattern of the poor in that domain” (World Bank, 2016, p.14). In other words, there was a separate food basket used to anchor the poverty line for each of the 13 municipalities. This is problematic if the food basket of one municipality is preferred to that of another (Arndt and Simler, 2010) and likely yields inconsistent poverty comparisons.6 In the interests of maintaining comparability with the already-reported poverty estimates, the same poverty line is used here, even with these doubts about its spatial consistency. On top of the food poverty line, a rent poverty line and a non-food poverty line were added, where these were based on a hedonic rent model and on the average spending on non-food consumption of households whose combined food and rent consumption was within 5% of the food plus rent poverty lines. The value of the total poverty lines ranged from a high of $56.16 per person per month in Dili to a low of $37.97 in Ermera and Liquiçá. Just over one-half of the cost of the total poverty lines was due to the food poverty lines. 4. Empirical Analysis 4.1 Comparing the Questionnaires The first step in the empirical analysis was to compare the questions and response options that were available in the Population and Housing Census with those in the TLSLS. These were divided into two types: 28 personal characteristics or attributes collected at person level (and calculated either for the household head or for a group such as all adults), and 72 dwelling and household-level attributes. The number of possible matching variables was quite limited, considering how many questions are available in the census, because the sets of response options in the census were often different to those in the survey. The following provides two 6 A further concern with the description of the poverty line formation is that defining a food basket based on the consumption patterns of poor households in a municipality pre-supposes that one knows who the poor are. Yet it is precisely because we do not know who the poor are that we need to estimate a poverty line. An iterative solution to this issue of circular logic is suggested by Pradhan et al (2001). 7 examples, for materials of the dwelling wall and for sources of drinking water (with various responses to these questions typically good predictors of household consumption levels). Good practice is to use the same wording and set of options offered in the census for a survey question on the same topic. There is no guarantee that the distributions will overlap, if for example, the survey sample is not fully representative but consistent question wording and answer options at least makes overlap possible. In contrast, if a survey has options that are unavailable in the census (e.g. “unbaked brick” or “tap water”) or uses different wording (e.g. “tin” rather than “corrugated iron/zinc”) then there is no possibility of getting consistent overlap. It seems likely that Timor Leste is not getting as much value out of census and survey data as would be possible if a more coherent and consistent set of questions and responses were used when asking about the same topic in the census and in surveys. Table 1: Comparison between Census and TL SLS (Wall Material and Drinking Water Examples) Comparison 2015 Population and Housing 2014 Timor Leste Survey of Census Living Standards Wall Material Question H2. What is the main (2.01) What is the major construction material for your construction material of the external walls? external walls? Answer * Concrete/brick * Brick Category * Wood * Concrete * Bamboo * Unbaked brick * Corrugated iron/zinc * Wood * Clay/Soil * Bamboo * Palm Trunk (Bebak) * Bebak (Piko) * Rock * Tin * Other * Mud * Other Drinking Water Question H13. What is the main source (2.10) What is the main source of drinking water used by of water for drinking for your household members? household? Answer * Piped or pump indoors * Bottled water Category * Piped or pump outdoors * Tap water * Public piped/tap * Pump * Tubewell/borehole * Protected well * Protected well/protected * Unprotected well spring * Rainwater collection * Protected spring * Bottle water * Unprotected spring 8 Comparison 2015 Population and Housing 2014 Timor Leste Survey of Census Living Standards * Not protected well or spring * River, stream, lake, pond * Water vendors/tank * Rainwater * River, lake, stream, channel * Other * Other 4.2 Comparing the Variables Questions may seem similar in the census and the survey but there is no guarantee that variables derived from these questions will have an overlapping distribution. Therefore, the next step in the analysis was to empirically compare the distributions of these variables coming from the survey (using the sampling weights to expand up to national totals) with the distributions for what should ostensibly be the same variable in the census. The comparisons take account of the clustered nature of the survey (which reduces the precision of the survey estimates). The details from this comparison of census and survey variables are provided in Table 5 and Table 6 (on Appendix). If there is a statistically significant (at p0.10 level) difference in the mean of a variable from the survey compared to the mean from the census, it is not considered further. For the variables that survive this test, the comparison also considers their standard deviations. There are 22 household level variables in Table 5 and 7 person-level variables in Table 6 that are candidates for using to project survey data (specifically log per capita consumption) onto the census households. These variables (shown in yellow highlight) include the gender, migrant status and higher education of the household head, four types of wall and one type of floor variables, several livestock variables, two durables ownership variables, and some variables related to cooking, lighting and sanitation. With so few variables that have overlapping distributions, the lack of consistency in ways that census and survey questions were phrased and were populated with answer options emerges as a constraint on the analysis because it would be typical to have a longer list of overlapping variables when the census is as detailed (90 questions, 9 pages) as is the Timor Leste census. A particular constraint is that almost all of the overlapping variables are categorical, so it is not possible to use squares and cubes of these variables to increase the number of possible candidates for including in the poverty mapping models (while with continuous variables the squares and cubes could be used). 9 4.3 Variable Selection for Poverty Mapping Models A databank of 58 variables was created from (a) the variables in Table 5 and Table 6 that were not excluded on the grounds of having non-overlapping distributions, and (b) the suco-level means of the same variables, calculated over the census households. All of these variables were candidates for initial beta models of log per capita consumption from the survey data. A backwards stepwise procedure was used to select variables for the beta model, with eligibility for removal set at p0.05. These models were estimated over three domains: nationally, rural households, and urban households. The OLS estimates of the beta model are shown in Table 7 for the national model, in Table 9 for the rural model and in Table 11 for the urban model. The residuals from these initial beta models were then decomposed, following equation (2) and the alpha model for the variance of the household idiosyncratic component was estimated. Specifically, the covariates for the beta models were interacted with the predicted values and the squares of the predicted values of equation (1), and these formed the candidate variables for the alpha models. The finally included variables were also selected using the backwards stepwise with p0.05 eligibility for removal. The alpha models are reported in Table 8, Table 10, and Table 12. Once the alpha models were complete, the GLS models that account for heteroscedasticity were estimated, and these are reported in the last three columns of Table 7, Table 9, and Table 11. Some summary details on these beta and alpha models, and their success in dealing with the location component of the residuals are reported below. The adjusted R2 values for the beta models are reasonably low compared to those found in poverty mapping exercises in other countries (for example, in the Solomon Islands the values ranged from 0.46 to 0.60). There are at least two contributing factors. First, the lack of EA-level concordance between the 2010 and 2015 census meant that the incorporation of census means into the models had to be at the suco level (so the census means capture conditions from a broader area than just the EA that the household is located in). Second, a lack of consistency in question wording and answer options between the census and the survey limited the number of variables that were available to consider for testing for overlapping distributions. Consequently, a majority of the variables that are selected in the stepwise models are census-means, rather than household level attributes. 10 Notwithstanding these lower values for the adjusted R2, one key diagnostic suggests that the prediction equations should be reasonably successful. Specifically, the ratio of the variance of the location error to that of the total error,  ˆ2  ˆ u2 , was between 0.09 and 0.18, so less than one-fifth of the error is due to factors that are correlated within locations.7 The ratio for the national level model (0.16) was lower than for the recent poverty maps for the Solomon Islands (0.19) and this is a good sign. Moreover, the pattern of a larger relative location component in rural areas is plausible because having unobserved common factors affecting economic livelihoods of the households in the same EA is more likely in the countryside, where people typically work where they live. In contrast, in urban areas there may be a geographic separation between the location of employment, the location of places of human capital investment (e.g. schools) and the location of places of residence. Table 2: Summary Details for Beta and Alpha Models Domain National Rural Urban Beta Model Number of predictor variables used 27 29 18 Adjusted R-squared 0.337 0.239 0.346 Relative variance of location error,  ˆ2  ˆ u2 0.161 0.178 0.085 Alpha Model Number of predictor variables used 19 15 8 Adjusted R-squared 0.026 0.026 0.019 Note: Full details on the models are reported in Table 6 up to Table 11. 5. Simulation and Poverty Mapping Results 5.1 Cross-validation with Other Welfare Indicators The headcount poverty rate, calculated as the number of people living in census households whose imputed per capita consumption is below the poverty line, is just over 40 percent, with a standard error of 1.1 percent. The equivalent figure from the survey was a headcount poverty 7 With most of the error variance being due to the idiosyncratic household component, rather than due to the correlated location component, the small-area estimates based on the imputed consumption for each census household should be more precise. 11 rate of 41.8 percent (with a standard error of 1.4 percent). For the poverty gap and poverty severity index, the census-based estimates are 11.1% (SE=0.5%) and 4.3% (SE=0.2%) while from the survey they are 10.4% (SE=0.5%) and 3.7% (SE=0.2%).8 There is even closer match for the Gini coefficient for inequality; using the imputed census consumption data the Gini is 0.29, just the same as the survey estimate. Likewise, the census-based estimates of the rural (Gini=0.27) and urban (Gini=0.29) also exactly match the survey estimates. Moving down a level, to the survey estimates of poverty rates for each municipality, there is a correlation with the estimates from the imputed census data of 0.75, 0.73, and 0.70 for the headcount index, the poverty gap index, and the poverty severity index. Another source of cross-validation comes from night-time lights, which are increasingly used to proxy for sub-national economic activity (Donaldson and Storeygard, 2016). The pattern of night-time lights in Timor Leste in 2013 is shown in Figure 1; this is the most recent year with data available and is just one year before the TLSLS was fielded so spatial patterns should be similar in both years. A common measure of night-lights is the “digital number” (DN) which ranges from 0-63, where 63 is for the brightest, most-saturated light and 0 is total darkness.9 If the average DN value for each of the 442 sucos is correlated with log average per capita consumption, which is based on the survey-to-census imputed data, one obtains a correlation coefficient of 0.76 (and 0.72 if using the mean dollar value of per capita consumption for the suco). Thus, a completely independent source of sub-national income data has spatial patterns that are highly correlated with the spatial patterns in the imputed consumption for census households. The two exercises in this section add to the confidence in the imputed measures. 8 The poverty gap index gives the mean distance below the poverty line expressed as a proportion of that line, where the mean is formed over the entire population (counting the non- poor as having zero poverty gap). The poverty severity index uses the squared poverty gaps, to put more weight on the poorest. SE=standard error. 9 This digital number is because the sensors on the DMSP satellites providing the night-lights data can only store six bits of data and 26=64. 12 Figure 1: Illuminated Areas in Timor Leste, 2013 DMSP Satellite F18 (using 5% Luminosity Threshold to restrict to non-ephemeral lit areas) 5.2 Suco-Level Poverty Estimates While the standard errors for poverty rates at the national level are similar for the survey-based and census-based estimates, at the municipality level the standard errors for poverty measures based on the census data are only one-half of the size of the survey-based estimates. No further disaggregation of the survey poverty estimates is provided below municipalities, but with the census-based estimates it is possible to get fairly precise values even at suco level. The suco level results from the simulations are shown in two ways. The first way of reporting these results is in maps that show the headcount poverty rate and also the number of people in the households that are predicted to be poor in each suco (Figure 2). The reason for having both types of maps is that a focus on poverty rates may be misleading when the population is unequally distributed over space; there may be more poor people in an area where the poverty rate is not as high as in a higher rate area with low population density area. Thus, it is helpful to know about both rates and numbers when designing geographically targeted interventions. 13 Figure 2: National-level Poverty Mapping Model Estimation (a) Predicted Headcount Poverty Rate (b) Number in Poor Households 14 Figure 2 reveals an already known pattern, that the headcount poverty rate is much higher in western areas of Timor Leste than in eastern areas. However, the figure also shows something that was not previously known, which is that there is much more variation in poverty rates within municipalities than between municipalities. Specifically, the standard deviation of the headcount poverty rate within municipalities (0.111) is 54-times higher than the standard deviation between municipalities (0.002). In other words, knowing the headcount poverty rate at the municipality level, which is what the survey has already reported, is not very informative about the poverty rate of sucos within that municipality. A good example of this effect is for the Dili municipality, which has suco-level headcount poverty rates that range from 8% to 80%. Likewise, in Manatuto, the suco-level headcount poverty rates range from 10% to 71%. Panel (b) of Figure 2 displays the number of people living in poor households in each suco. Even though the headcount poverty rate is low in most sucos within Dili, because of the high population density, there are a large number of people who live in poor households in these sucos (especially in western areas of Dili). A dense belt with high numbers of poor people per suco goes from Dili through Liquiçá and Ermera, and also along the western boundary of Ainaro. There are also high numbers of people in poor households in Oecusse. Figure 3 maps the poverty rates for the poverty gap index and the poverty severity index. The spatial patterns are largely the same as what Figure 2 shows with the headcount index. If the variation is decomposed into between-municipality and within-municipality effects, only two percent of the total variation in the suco-level poverty gap index is between municipalities while the vast majority of it is within-municipality variation. Once again, only knowing the value of the poverty gap index or of the poverty severity index at a municipality level will not be a very good basis for spatial targeting, due to the substantial degree of variation in poverty amongst the sucos within a municipality. 15 Figure 3: National Model Estimation (a) Predicted Poverty Gap Index (b) Poverty Severity Index 16 The second way that the suco-level small-area estimation results are reported is shown in Table 27. For each of the 442 sucos in the 2015 census, this file reports the mean of per capita consumption for census households, the headcount poverty rate, the poverty gap index, the poverty severity index, and the Gini index of inequality. For each of these five indicators there are standard errors reported, where these standard errors are based on the variation in the predictions amongst the 100 replications of the simulations. Relative to their respective poverty indexes, the standard errors average 0.21, 0.30, and 0.37 for the headcount, poverty gap, and poverty severity measures. The relative precision is even higher for predicted consumption and for the Gini, at 0.09 and 0.07. The spreadsheet file also includes the number of households and the number of individuals in each suco, because for many purposes it is population-weighted averages or sums of the suco-level welfare indicators that would be required. Applications of the ELL poverty mapping approach typically consider sub-national domains rather than just relying on a national-level model. Within a domain, the parametric relationship between characteristics (regressors) and consumption is expected to be the same across areas, while the relationship may differ between domains. In this case, the sub-national models will provide a better basis for imputing consumption of census households and for deriving the small-area welfare statistics. Given that Timor Leste is a much smaller country than most countries where poverty mapping techniques have been used, there may be less need to use sub-national models, although it is still possible that the rate at which personal and household characteristics are transformed into consumption will differ between urban and rural areas. The predicted poverty rates for rural households, at suco-level for the headcount index and for the number living in poor households, is shown in Figure 4.10 Most sucos have rural households and the majority (71%) of the population is rural, so the map looks fairly similar to the maps in Figure 2 that are based on all households. The corresponding maps for urban households (in Figure 5) show much more grey, for the sucos with no urban households. The sucos with the highest urban poverty rates are found in Dili, Aileu, Ermera, and Bobonaro. 10 These estimates use the rural sector beta and alpha models in Table 6 and Table 7. 17 Figure 4: Rural Sector Poverty Mapping Model estimation (a) Predicted Headcount Poverty Rate (b) Number in Poor Households 18 Figure 5: Urban Sector Poverty Mapping Model estimation (a) Predicted Headcount Poverty Rate (b) Number in Poor Households 19 There is a high correlation between the suco-level headcount poverty rates estimated using the national model (as mapped in Figure 2) and the poverty rates that come from combining the separate predictions for rural and urban households (in Figure 4 and Figure 5). Specifically, the correlation coefficient for the poverty rates is 0.96, and for the number of people in households that are predicted to be poor in each suco it is even higher, at 0.99. Given these high correlations, the results from the national-level model (based on the regressions in Tables 6 and 7) should be a sufficient guide for policy and research. Therefore, it is just the results from the national-level model that are provided in the accompanying spreadsheet that has details for each suco. 6. Poverty and Gender The impetus for the development of the small area estimates reported above was a desire for spatially disaggregated gender indicators of standards of living for Timor Leste. In particular, it was expected that knowledge gaps identified in the Timor Leste Country Gender Action Plan may benefit from having suco-level indicators. Generally speaking, it is impossible to measure individual-level consumption and poverty because most people consume at least some items that they share with others, such as household-level public goods (e.g. heat, light, shelter) and also because the attempt to measure individual level consumption is likely to distort actual behavior. Most analysis therefore measures consumption at the household level and assumes some sharing rule (such as the equal sharing implied by using per capita consumption) in order to get estimates on an individual basis. One observable gender-related consumption indicator is the gender of the household head, and poverty analysis according to this characteristic can use the standard household-level data. Consequently, the poverty profile by gender of the household head is typically reported from surveys around the world. Along these lines, the existing evidence from the 2014 TLSLS shows that female-headed households in Timor Leste are less likely to be poor than are male-headed households (World Bank, 2016). Furthermore, a comparison with the previous TLSLS results (for 2007) showed that female-headed households enjoyed a faster rate of poverty reduction than male-headed households. Although female-headed households tend to have 1.9 fewer household members than male-headed households, the poverty advantage for female-headed households in 2014 appeared to be less independent of household composition given female- headed households tend to have higher female household members but smaller young-age- 20 dependency ratio than male-headed households. While these patterns should hold, on average across sucos, there may be considerable variation over space in the difference in poverty rates between people in male-headed households and people in female-headed households. This spatial variation could arise from underlying variation in gender-related economic opportunities and in location-specific constraints that may bind more on female-headed households than on male-headed households. The small-area estimation method can provide spatially disaggregated indicators of the difference in poverty rates according to the gender of the household head. At the simulation stage, a value of consumption is imputed for each census household, and those data can then be linked back to the characteristics of the household in the original census unit-record data. Almost 16% of census households (n=32,400) are female-headed, and that is equivalent to just over 70 female-headed households per suco, on average, which is a sufficient number to estimate the suco-level poverty rate for female-headed households. Of course, because the majority of households are male-headed, the corresponding male-headed household poverty rate at suco level can also be estimated. It is the difference between these two poverty rates that is likely to be of greatest interest. Figure 6 maps the difference in head-count poverty rates (in panel a) and in the poverty gap index (in panel b) for male-headed households compared to female-headed households. Any suco with a negative value (shown in red on the map) is a place where female-headed households have poverty rates that exceed those of male-headed ones. The only place this occurs for the headcount rate is for Cotabot suco, in Bobonaro municipality. If the gender- related gap is measured, instead, using the poverty gap index, there are five sucos where poverty rates amongst female-headed households are higher than amongst male-headed households; Cotabot again, Edi in Ainaro, Luculai in Liquiçá, and Muapitine and Pairara in Lautém. 21 Figure 6: Suco-Level Mean Differences of People in Male-Headed vs Female-Headed Households (a) Predicted Poverty Headcount Rate (b) Poverty Gap Index 22 Figure 7: Suco-Level Relative Differences of People in Male-Headed vs Female-Headed Households (a) Predicted Poverty Headcount Rate (b) Poverty Gap Index 23 One factor that affects the interpretation of the maps in Figure 6 is that poverty rates are much higher in western areas of Timor Leste. The gender gap that is mapped is the difference in poverty rates, and there is more scope for this difference to be bigger where poverty rates are higher. Consequently, most western areas showed high positive values in Figure 6 because poverty is higher there. An alternative way to visualize the data is to consider the percentage difference in the poverty rates of male-headed versus female-headed households because the percentage differences are then independent of the underlying average poverty rate. In other words, a relative difference may be more informative than the absolute differences shown in Figure 6. The maps that present these relative differences are shown in Figure 7. With this indicator, the locations where poverty rates in male-headed households are much higher than in households with a female head are spread more widely, away from the concentration in the west that is apparent with the absolute gaps in Figure 6. 7. Disaggregated Evidence for other Gender-Related Indicators Considering that only 16 percent of households in Timor-Leste are female headed, the results of the analysis based on female-headed households represent the situation of a small minority of women and girls. This suggests the need to go beyond household headship to individual- level characteristics of education, health, labor force, and power and agency to better capture the much more meaningful standard-of-living and gender disparities of women and girls. The results in Section 6 rely on the traditional poverty mapping method, of survey-to-census imputation of consumption, and just use the gender of the household head as a way to regroup the predictions and then contrast them between male-headed and female-headed households. Another way to get small area estimates of gender-related indicators would be to use the estimation framework in Section 2 to project an index of female disadvantage – calculated for each surveyed household – onto each and every census household. In other words, rather than having a model of (log) per capita consumption in equation (1) the dependent variable would be a (continuous) index of female disadvantage. The right-hand side variables used to predict this index would be ones with an overlapping distribution in the census so that the disadvantage index can then be predicted for every census household, and then aggregated and mapped. As noted in Section 2, a key advantage of using the ELL method for this is that it can account for the location effects and provide more accurate standard errors of the small area estimates. 24 There are a few examples of adapting the ELL method to non-consumption indicators in this way, although not with a specific focus on gender. The main example is Fujii (2010), who mapped the prevalence of child stunting and underweight (based on z-scores for height for age and weight for age) in Cambodia. The original Demographic and Health Survey (to be referred as DHS) data that were reported for 17 strata in Cambodia were then able to be disaggregated to the commune level (the third sub-national level). The precision of the commune-level estimates, reported for about 1600 communes, was comparable to that of the survey-only estimates at the stratum level. A similar application of the ELL method is by Sohnesen et al (2017) for DHS data on child anthropometric indicators (z-scores for height for age and weight for age) in Ethiopia which were projected onto census data and then mapped at the district (Woreda) level. In both of these examples, there is a continuous variable (z-scores) and a threshold that can be treated like a poverty line (for example, a z-score below -2 for height for age is a common indicator of a child being stunted). It is notable that the ELL estimation approach is not designed for binary indicator variables, so examples in the literature of small area estimation of such indicators (e.g., for contraceptive use from DHS data) use rather more complex estimation methods (Amoako et al, 2012). In order to carry out this approach with the TLSLS data, it is necessary to convert indicators of gender-related disadvantage that are mainly binary into continuous variables. Also, because it takes considerable time to go through the steps in the ELL method, it is infeasible to do the survey-to-census imputation for more than a few variables. The principal components method can help with each of these two requirements, because the first principal component extracts the largest share of variability in the data from a set of variables. In other words, a ‘family’ of related indicators can be reduced to a single indicator (the first principal component), which is also a continuous variable (with mean zero), even if the underlying variables are binary. The three families considered with the TLSLS data are the labour force, education, and health, and these can be thought of as different domains in which gender disparities might be revealed. For each of these three domains, two underlying indicators are reduced to a single index. For the education index, the same variables that are used to construct it with the TLSLS data are also available in the census. This duplication allows the possibility of cross-validation by comparing the survey-to-census imputed values of the education index with the values of the index that is directly estimated from the census data (along the lines of what is reported in Section 5.1, 25 above, where the cross-checks were based on using the survey estimates of poverty at the municipality level and using the night time lights at the suco level as a proxy for local economic activity to compare with the survey-to-census imputed values). Education A principal components index (educ_pca) was constructed, for the first principal component formed from two types of male-female gaps (defined so that negative values denote female disadvantage): • The difference in the household level sum of an indicator for whether the person is illiterate (based on the self-reported criteria of not being able to read a letter), for females aged 5 and up, and for males aged 5 and up, expressed as a proportion of all household members aged 5 and up (where proportions allow for variation in household size and for the households lacking either eligible males or eligible females which prevents use of a gender-specific denominator) • The difference in the household level sum of an indicator for whether the person never attended school (either formal or informal), for females aged 5 and up, and for males aged 5 and up, expressed as a proportion of all household members aged 5 and up The correlation of the index with each of the two components is 0.99, and in terms of the component variables, the average number of unschooled or illiterate females per TLSLS household is almost 30% higher than the corresponding number of unschooled or illiterate males. The results of the survey-to-census imputation of the education index are mapped in Figure 8. The prevalence of female disadvantage in the education index is higher in poorer areas, while it is lowest in and around Dili. This pattern is also seen in Figure 9, where the headcount rate for the proportion of the population living in households that have negative values of the education index (which indicates female disadvantage) is negatively correlated (r=-0.76) with the suco-level mean of predicted per capita consumption for census households. In other words, the highest rate of education-related female disadvantage is found in poorer areas. 26 Figure 8: Proportion of the Population in Households Where the Index of Male-Female Education Gaps Indicates Female Disadvantage Figure 9: Relationship Between Gender Education Disadvantage and Suco-Level Mean Consumption .6 r = -0.76 .5 .4 .3 Circles are proportional to suco population .2 40 60 80 100 120 140 Suco Mean of Predicted Per Capita Consumption ($ per month) 27 The TLSLS variables used to construct the education index have counterparts in the census, although the question on literacy is asked in a different way (separately for Tetun, Portuguese, Bahasa Indonesia, and for English, and distinguishing speaking, reading, and writing, but not specifying ‘reading a letter’ as in the TLSLS). With these variables, it is therefore possible to calculate the education index directly with census data, without going through the steps in the survey-to-census imputation. In the census data the correlation of the principal components index with each of the two components is 0.96, which is slightly lower than the correlation of 0.99 in the TLSLS data. In terms of the component variables, the number of unschooled females in census households averages 26% higher than for unschooled males, which is almost the same as the 27.5% higher average number of unschooled females than unschooled males per household in the TLSLS data. However the census shows less of a gender difference in illiteracy, with the number of illiterate females per household averaging 18% above the number of illiterate males while in the TLSLS the margin is 26%; the difference between the census and the TLSLS in the gender gap in illiteracy may reflect the different way that the two questionnaires were structured, with rather more detail (four questions and more options for each) in the census than in the survey. Notwithstanding these differences, there is a reasonably close relationship between the directly estimated prevalence of female educational disadvantage based on the census data, and the estimates that come from the survey-to-census imputation. The scatterplot in Figure 10 shows that there is a correlation of 0.73 between the two sets of suco-level estimates of the proportion of the population living in households where the education index indicates female disadvantage, where one set of estimates are directly from the census and the other set of estimates are from the survey-to-census imputed values. This degree of similarity for the direct and indirect estimates is about the same as the correlations reported in Section 5.1 for poverty rates using the direct estimates from the survey compared to the poverty rates estimated indirectly with the survey-to-census imputation. Thus, even though the beta model for the education index (and for the labour force and health indexes) has a rather low degree of predictive fit, as shown below, the overall performance in creating a spatially disaggregated indicator compared to the benchmark of the same indicator from census data, is about the same as for the ‘traditional’ survey-to-census imputation approach for head count poverty rates estimated from a consumption model. 28 Figure 10: Comparison Between Direct Estimates from the Census and Survey-to-Census Imputed Values, in terms of the Proportion of the Population in Households, Where the Index .6 of Male-Female Education Gaps Indicates Female Disadvantage r = 0.73 .5 .4 .3 Circles are proportional to suco population .2 .1 .2 .3 .4 .5 .6 Prevalence of Gender Bias in Census Health Due to limited data, a principal components index (health_pca) was constructed using self- reported variables in TLSLS, for the first principal component formed from two types of male- female gaps (defined so that negative values denote female disadvantage): • The difference in the household level sum of the number of days in the last 30 days that females of all ages were affected by ill-health, compared to the sum of the number of days where males were affected by ill-health, with the difference normalized by the total number of person-days available for that household within the survey reference period • The difference in the household level sum of the number of episodes that females were hospitalized in the last 12 months, compared to the number of episodes for males, where the difference is normalized by household size 29 The correlation of the index with each of the two components is 0.74. In contrast to the other indicators, females did not appear to be disadvantaged in terms of these health measures. For example, the household-level average of the number of days of female illness was 94% of the average number of days of male illness. Females also had fewer spells of hospitalization (although these were rare for both males and females). Perhaps because of this, the gender- related health index in Figure 12 shows the weakest relationship with predicted consumption, with a correlation of just -0.25. The map in Figure 11 suggests that there is a higher proportion of the population living in households with a female health disadvantage in Oecusse, and there are also concentrations in Baucau and Viqueque but the patterns are more scattered than for the education index. The findings, however, might be attributed to potential weaknesses of self-reported health status. For example, people have different levels of tolerance of illness. It has also been argued that disadvantaged populations tend to fail to perceive and report the presence of illness (Sen, A., 2002). Moreover, given the same health problems, women are less likely to use health services than men. 30 Figure 11: Proportion of the Population in Households, Where the Index of Male-Female Health Gaps Indicates Female Disadvantage Figure 12: Relationship Between Gender Health and Suco-Level Mean Consumption .6 r = -0.25 .55 .5 .45 Circles are proportional to suco population .4 40 60 80 100 120 140 Suco Mean of Predicted Per Capita Consumption ($ per month) 31 In addition to the maps (Figure 13, Figure 8 and Figure 11), the point estimates and standard errors of the small area estimates of the headcount indexes for the labour force, education and health indicators are reported for each suco in Table 28 (in appendix). The table also includes the number of households and the number of individuals in each suco, because for many purposes it is population-weighted averages or sums of the suco-level welfare indicators that would be required. Recall that for each indicator, the standard errors are based on variation in the predictions amongst the 100 replications of the simulations. Relative to their respective headcount indexes, the standard errors average 0.13, 0.11, and 0.10 for the labour force, education and health indicators. In comparison, for the ‘traditional’ approach of modelling log per capita consumption and then predicting a headcount poverty index the standard errors averaged 0.25 of the index (and if these comparisons are repeated by weighting each suco by its population then the average of the relative standard errors are 0.10, 0.10 and 0.07 while for the headcount poverty index they were 0.21). It is surprising that the relative precision of the small area estimates for the gender-related indicators exceeds that of the poverty indicators because the underlying beta and alpha models have much lower predictive power than for the corresponding models of (log) per capita consumption. This lower predictive power is partly because the models are smaller, with fewer variables selected by the stepwise procedure (as noted in the table below, and this flows through into quite low values of the adjusted R2). This selection of only a few predictors presumably reflects the fact that household-level gender gaps reflect somewhat idiosyncratic factors that are less observable with survey data than are some of the predictors of per capita consumption. Moreover, the modelling for Timor Leste already starts in a weaker position than it would in some other countries because only a few variables in the TLSLS have overlapping distributions with variables in the census, because of the different ways that questions were phrased and the different sets of answer options. The relatively sparse models that were estimated may also be a cause of the location effect in the error having zero variance, for the education and health models, and this feature persisted across many different specifications that were attempted. 32 Table 3: Summary Details for Beta and Alpha Models of Gender-Related Indicators in TLSLS Domain National Rural Urban Beta Model Number of predictor variables used 8 9 8 Adjusted R-squared 0.046 0.128 0.011 Relative variance of location error,  ˆ2  ˆ u2 0.041 0 0 Alpha Model Number of predictor variables used 3 7 9 Adjusted R-squared 0.035 0.143 0.039 Note: Full details on the models are reported in Table 13 up to Table 18. Labour Force A principal components index (work_pca) was constructed, comprised of the first principal component for two series: the male-female gap (defined so that negative values denote female disadvantage) in terms of: • The household level sum of an indicator for the person having no economic activity in the past week (including wage and salary work, income in-kind and own-account activity), for females aged 10 and up, and for males aged 10 and up, expressed as a proportion of all household members aged 10 and up (where using a proportion is to allow for variation in household size and for households lacking either eligible males or eligible females which would make a gender-specific denominator zero and produce an undefined index) • The difference in the household level sum of female hours of wage labour compared to male hours of wage labour supplied in the previous seven days across all jobs, for all persons aged 10 and up The correlation of the index with each of the two components is 0.74. The index has a mean of zero, by construction. Negative values are used as the indicator of female disadvantage and the mapping is in terms of the proportion of the population living in households that exhibit this net female labour force disadvantage (that is, the results are only presented for the head-count index because the poverty gap index is undefined for an underlying series that has both positive and negative values). The results mapped in Figure 13 show an inverse pattern between gender disadvantage in the labour market and poverty rates. In other words, the gender-related labour force gaps are bigger 33 in sucos where households, on average, are richer and where poverty rates are lower. This same pattern is highlighted in Figure 14, which has scatter plots for the relationship between the suco-level mean of predicted per capita consumption (based on the beta and alpha models in Table 7 and Table 8) and the headcount index for the gender disadvantage indicators. For the labour force indicator of gender disadvantage there is a correlation with predicted per capita consumption of 0.79, which shows that it is in richer areas of Timor Leste that gender gaps in labour force indicators are likely to be most apparent. It should be noted that, due to data constraints, the labor force index does not include key labor indicators reflecting the quality of employment, such as returns to labor force or employment segregation. 34 Figure 13: Proportion of the Population in Households Where the Index of Male-Female Labour Force Gaps Indicates Female Disadvantage Figure 14: Relationship Between Labour Force Disadvantage and Suco-Level Mean Consumption .7 r = 0.79 .6 .5 .4 Circles are proportional to suco population .3 40 60 80 100 120 140 Suco Mean of Predicted Per Capita Consumption ($ per month) 35 Power and Agency The TLSLS did not collect qualitative information that relates to aspects of agency and power, which are important gender indicators that could usefully be disaggregated with small-area estimation. However, this type of information is available in the 2016 Demographic and Health Survey (DHS), and quantitative indexes can be derived from these data along the lines of what is done with the education, health, and labour force indicators discussed above. 11 The first indicator of power and agency used here is a principal components index related to female autonomy in decision-making. Specifically, it is for adult females who were married or living with a man at the time of the survey, where the underlying dummy variables indicate whether the respondent makes decisions about her own health-care, about major purchases, and about visits to her family and relatives. The index is most highly correlated with autonomy on health-care decisions (r=0.83) and least with major purchase decisions (r=0.77). Two other indexes are created from a smaller sample of women, selected from amongst the adult females in the households in the DHS sample, who answered (in privacy) a module on domestic violence. For this sample, their answers could refer to the current male partner or a former male partner, and they were asked about: • any experience of 10 types of physical abuse or domestic violence from the male partner • any experience of five types of the male partner being jealous, angry, limiting access to friends and family, and otherwise limiting autonomy • any experience of three types of verbal threats and abuse from the male partner • and whether the respondent was afraid of the current or former male partner most of the time Only 26% of the respondents to this module had no experience of any of these types of abuse or domestic violence, while the mean was for respondents to have experienced 2.1 (out of 19) of these various indicators of abuse and domestic violence. Based on these 19 dummy variables, one indicator used for the small area estimation is the first principal component (mean 0, SD=2.2), while the other indicator is a simple count of the types of abuse and violence 11 The DHS was fielded in 455 EAs between September and December 2016. The sample size varies with the type of indicator (e.g. a sub-set of all adult female respondents are given the domestic violence module). 36 experienced. The principal component index is most highly correlated with elements of sexual violence (being forced to have sexual intercourse, r=0.62, and force with threats or in any other way to perform sexual acts, r=0.59) and least with being afraid of the partner (r=0.23). In order to use these data from the DHS for survey-to-census imputation, the same steps that are described in Section 4.1 and Section 4.2 were carried out with the DHS data. There were even fewer variables in the DHS that had overlapping distributions with variables in the census than was the case for the comparison of the TLSLS with the census. The results of the comparisons are in Table 19 and Table 20, and these show that there were just 12 household level variables and four person level variables where one would not reject the hypothesis of equality of means in the DHS and the census (and taking account of the clustered nature of the DHS sample which makes it even easier to not reject the null hypothesis) and where the standard deviations were of similar magnitude in the two data sources. In addition to these variables, the suco-level means, from the full set of census households, are obtained for these 12 and four variables, and so in total that provided 32 candidate variables for the stepwise model selection. Table 4: Summary Details for Beta and Alpha Models of Power and Agency Indicators in 2016 DHS Domain National Rural Urban Beta Model Number of predictor variables used 12 13 15 Adjusted R-squared 0.026 0.049 0.06 Relative variance of location error,  ˆ2  ˆ u2 0.104 0.07 0.085 Alpha Model Number of predictor variables used 14 10 13 Adjusted R-squared 0.03 0.086 0.059 Note: Full details on the models are reported in Table 21 up to Table 26. The OLS and GLS estimates of the beta models are reported in Table 21, Table 23, and Table 25, while the results of the alpha models are reported in Table 22, Table 24, and Table 26. A summary of the models is presented below, and it is apparent that there is only a low degree of predictive power for these indicators of power and agency. A more positive feature of these models, in terms of the precision of the small-area estimates, is that the location component is only a small share of the total error variance (ranging from 0.070 to 0.104). 37 The results from mapping the index of female autonomy in decision-making at suco level are shown in Figure 15. The locations with the highest proportion of the population living in households where female autonomy for decision making is lowest are scattered through some inland parts of the country. There are no apparent patterns with respect to average consumption levels (r=-0.03) or to headcount poverty rates (r=0.03). Figure 15: Proportion of the Population in Households, Where the Index of Female Decision Making (DM) Autonomy Indicates Female Disadvantage There is much clearer evidence with respect to the prevalence of abuse and domestic violence. Whether in terms of the proportion of the population living in households with high values of the domestic violence index Figure 16), or where the average number of types of abuse and domestic violence experienced is higher (Figure 18), the western areas, and especially Oecusse, appear to have higher prevalence of these problems. This geographic pattern is similar to the pattern of headcount poverty rates, which are also higher in the west, and so a scatterplot ( Figure 17) between the share of the population living in households with high domestic violence index and the average predicted headcount poverty rate shows a significant positive correlation (r=0.58). Thus, interventions designed to deal with partner abuse and domestic violence may usefully be targeted at poorer areas. 38 Figure 16: Proportion of the Population in Households Where the Index of Female Experience of Types of Domestic Violence (DV) Indicates Female Disadvantage Figure 17: Relationship Between Share of Population Living in Households With High Domestic Violence Index and Average Predicted Headcount Poverty Rate by Suco 39 Figure 18: Suco Average of the Predicted Number of Types of Domestic Violence (DV) Reported by Females 8. Conclusions In this study, data primarily from the 2015 Population and Housing Census and the 2014/15 Timor Leste Survey of Living Standards, along with more limited data from the 2016 Demographic and Health Survey were combined to estimate various indicators for each of the 442 sucos in Timor Leste. The first part of the analysis is a ‘traditional’ poverty mapping approach, where the focus is on monetary measures of poverty, and the findings contrast with prior poverty analysis for Timor Leste that uses only the survey data and that provides estimates just for the 13 municipalities. As such the prior analysis ignores the substantial differences within municipalities in suco-level poverty rates and the intra-municipality variation in living standards more generally. The survey-to-census imputations carried out here enable this variation to be revealed, and can provide useful information for developing spatially targeted interventions, such as local development programs. A further use of the results reported here is for future analytic studies that aim to explore some of the driving forces behind the spatial variation in poverty in Timor Leste. While some Timorese researchers may be only limited access to unit record data from the surveys and censuses, useful analyses may also be carried 40 out by using the type of constructed poverty and inequality variables that are here reported at suco level.12 The second part of the analysis (in Sections 6 and 7) is non-traditional, in using the small area estimation techniques to spatially disaggregated gender-related indicators from the Living Standards Survey and from the 2016 Demographic and Health Survey. Compared to the predictive fit of the poverty mapping models based on household consumption, the models for gender-related indicators for labour force activity, education, health, decision-making autonomy, and abuse and domestic violence have much lower predictive power, which is likely to reflect the idiosyncratic nature of some of these indicators. Nevertheless, the precision of the derived suco-level indicators was at least as high as for the traditional poverty indicators, in part because the common location component in the errors for the gender-related indicator prediction equations was generally smaller than it was in the errors for the consumption (again, reflecting the smaller role of area characteristics and the larger role of idiosyncratic factors in influencing these gender-related indicators). In terms of substantive findings, this attempt to spatially disaggregate gender-related indicators revealed two key patterns. First, it is in poorer areas of Timor Leste where more people are affected by abuse of females and domestic violence against females, and it is also in poor areas where there is more educationally-related female disadvantage. In contrast, if the focus is on female disadvantages in the labour force, these show up the most in the more economically developed sucos. A secondary finding from the analysis is that this type of research would be enhanced if future surveys in Timor Leste more closely followed the wording of questions and answer options used in the census so that better predictive models of gender-related indicators could be produced from the available data. 12 For example, small-area welfare statistics from poverty mapping can be linked to measures of environmental change, such as deforestation, that is available from remote sensing. Using the same techniques as in the current study, Gibson (2018) shows how prior deforestation in the Solomon Islands is associated with higher inequality and poverty. These types of relationship should be of interest to policymakers and researchers. 41 The overall findings suggest the importance of using gender-disaggregated individual level analysis, beyond the male/female household headship, to better assess poverty of women and men and gender disparity. However, the individual-level indicators are generally still underexplored and therefore data constraints are big. Moreover, since this work is a new field, the method and indicators are still in a developmental stage and there is still fairly limited literature, especially on large-scale empirical work at village-level, to use as reference. This study hopefully will lead to more and further study so as to contribute towards enriching the existing literature in this area. 42 References Amoako J., S. Padmadas, H. Chandra, Z. Matthews, and N. Madise (2012), ‘Estimating unmet need for contraception by district within Ghana: an application of small-area estimation techniques’, Population Studies 66(2): 105-122. Arndt, C., and Simler, K. (2010). ‘Estimating utility-consistent poverty lines with applications to Egypt and Mozambique’, Economic Development and Cultural Change 58(3): 449-474. Bedi, T., A. Coudouel, and K. Simler (2007), More Than a Pretty Picture: Using Poverty Maps to Design Better Policies and Interventions. World Bank Publications. Bigman, D., and Srinivasan, P. (2002), ‘Geographical targeting of poverty alleviation programs: methodology and applications in rural India’, Journal of Policy Modelling 24(3): 237-255. Christiaensen, L., P. Lanjouw, J. Luoto, and D Stifel. (2012), ‘Small area estimation-based prediction methods to track poverty: validation and applications’, The Journal of Economic Inequality 10(2): 267-297. Donaldson, D., and Storeygard, A. (2016), ‘The view from above: Applications of satellite data in economics’, Journal of Economic Perspectives 30(4): 171-98. Elbers, C., J. Lanjouw, and P. Lanjouw (2003), ‘Micro-level estimation of poverty and inequality’, Econometrica 71(1): 355 – 364. Elbers, C., P. Lanjouw, and P. Leite (2008), ‘Brazil within Brazil: Testing the poverty map methodology in Minas Gerais’, Policy Research Working Paper Series No. 4513, The World Bank. Fujii, T. (2010), ‘Micro-level estimation of child undernutrition indicators in Cambodia’, The World Bank Economic Review, 24(3): 520-553. Gibson, J. (2018), ‘Forest loss and economic inequality in the Solomon Islands: Using small - area estimation to link environmental change to welfare outcomes ’, Ecological Economics 148(1): 66-76. Gibson, J., G. Datt, B. Allen, V. Hwang, M. Bourke and D. Parajuli (2005), ‘Mapping poverty in rural Papua New Guinea’, Pacific Economic Bulletin 20(1): 27 – 43. Pradhan, M., Suryahadi, A., Sumarto, S., & Pritchett, L. (2001), ‘Eating like which “Joneses?” An iterative solution to the choice of a poverty line “reference group’, Review of Income and Wealth 47(4): 473-487. 43 Sohnesen, T., A. Ambel, P. Fisker, C. Andrews, and Q. Khan (2017), ‘Small area estimation of child undernutrition in Ethiopian woredas’, PloS One, 12(4), e0175445. World Bank (2016), Poverty in Timor-Leste 2014 (English). Washington, D.C. : World Bank Group. http://documents.worldbank.org/curated/en/577521475573958572/Poverty-in- Timor-Leste-2014 World Bank. (2017). “Small Area Estimation: An extended ELL approach.” World Bank. 44 Table 5: Comparison of Census and TLSLS Variables for Household-Level Characteristics Dwelling attributes and household variables Census TLSLS t-test p-value Mean SD Mean SD Main material of dwelling walls is concrete or brick 0.38 0.49 0.38 0.49 0.08 0.933 Main material of dwelling walls is wood 0.04 0.20 0.02 0.13 9.80 0.000 Main material of dwelling walls is bamboo 0.25 0.43 0.23 0.42 1.23 0.220 Main material of dwelling walls is corrugated iron/zinc/tin 0.05 0.22 0.04 0.19 1.90 0.057 Main material of dwelling walls is palm trunk (bebak) 0.25 0.43 0.31 0.46 3.84 0.000 Main material of dwelling walls is any plant matter (wood, bamboo, or bebak) 0.54 0.50 0.56 0.50 0.91 0.362 Main material of dwelling roof is concrete or brick 0.04 0.21 0.00 0.05 27.15 0.000 Main material of dwelling roof is tin or corrogated iron 0.75 0.43 0.80 0.40 3.49 0.000 Main material of dwelling roof is palm leaves, thatch, or grass 0.19 0.39 0.12 0.32 5.99 0.000 Main material of dwelling floor is concrete or brick 0.39 0.49 0.30 0.46 6.55 0.000 Main material of dwelling floor is tile 0.09 0.29 0.13 0.34 3.88 0.000 Main material of dwelling floor is wood or bamboo 0.04 0.19 0.04 0.20 0.73 0.468 Main material of dwelling floor is earth, clay or dirt 0.46 0.50 0.51 0.50 2.23 0.026 Dwelling is owned by a member or members of the household 0.95 0.21 0.96 0.20 0.68 0.496 Dwelling is in good condition 0.19 0.39 0.17 0.37 2.03 0.043 Dwelling is in mediocre condition 0.53 0.50 0.46 0.50 6.47 0.000 Dwelling is a little damaged 0.20 0.40 0.27 0.44 5.59 0.000 Dwelling is severely damaged 0.08 0.27 0.11 0.31 3.73 0.000 Number of rooms in the dwelling (excl bathrooms, kitchens etc) 2.89 1.33 3.06 1.32 4.71 0.000 Dwelling has indoor bath or shower 0.14 0.35 0.10 0.31 3.50 0.000 Household members bath or shower outdoors 0.63 0.48 0.62 0.49 0.27 0.789 Household members bathe in river, pond etc 0.20 0.40 0.22 0.41 1.18 0.240 Household members use flush toilet 0.36 0.48 0.09 0.29 23.14 0.000 Household members use ventilated, improved pit latrine 0.12 0.33 0.20 0.40 6.51 0.000 Household members use pit latrine with slab 0.09 0.28 0.19 0.39 8.98 0.000 45 Dwelling attributes and household variables Census TLSLS t-test p-value Mean SD Mean SD Household members use open pit latrine 0.08 0.27 0.16 0.36 7.49 0.000 Household members have no toilet facility or use the bush 0.21 0.40 0.25 0.43 2.18 0.029 Sewerage disposal into septic tank 0.29 0.46 0.29 0.45 0.11 0.912 Sewerage disposal into a hole 0.30 0.46 0.30 0.46 0.28 0.778 Main energy source for cooking is electricity 0.10 0.30 0.08 0.28 2.01 0.045 Main energy source for cooking is kerosene 0.05 0.21 0.03 0.16 3.91 0.000 Main energy source for cooking is wood 0.81 0.39 0.80 0.40 0.91 0.365 Main energy source for lighting is electricity 0.67 0.47 0.69 0.46 0.64 0.525 Main energy source for lighting is kerosene 0.11 0.31 0.02 0.14 17.91 0.000 Main energy source for lighting is solar panel 0.15 0.35 0.16 0.37 0.83 0.407 Main energy source for lighting is candle, battery or torch 0.03 0.18 0.13 0.34 7.03 0.000 Drinking water is mainly from protected well or spring 0.08 0.26 0.15 0.36 6.34 0.000 Drinking water is mainly from unprotected well or spring 0.09 0.28 0.16 0.37 4.81 0.000 Drinking water is mainly from river, lake or stream 0.14 0.35 0.06 0.23 9.05 0.000 Household has at least one radio 0.27 0.44 0.08 0.28 30.33 0.000 Household has at least one television 0.37 0.48 0.34 0.47 1.58 0.115 Household has at least one mobile telephone 0.81 0.39 0.61 0.49 15.65 0.000 Household has at least one computer 0.16 0.37 0.09 0.29 6.96 0.000 Household has at least one refrigerator 0.16 0.36 0.01 0.11 23.75 0.000 Household has at least one sewing machine 0.04 0.19 0.02 0.13 7.26 0.000 Household has at least one bicycle 0.14 0.35 0.07 0.25 9.91 0.000 Household has at least one motor cycle 0.24 0.43 0.20 0.40 2.43 0.015 Household has at least one car, van or truck 0.07 0.25 0.03 0.18 5.96 0.000 Household has at least one boat 0.02 0.14 0.02 0.12 1.11 0.266 Household uses a tractor 0.16 0.36 0.09 0.28 6.00 0.000 Household rears livestock 0.87 0.33 0.88 0.33 0.38 0.701 46 Dwelling attributes and household variables Census TLSLS t-test p-value Mean SD Mean SD Household grows crops 0.80 0.40 0.81 0.39 0.64 0.520 Number of chickens owned by household 4.54 7.73 5.12 6.13 3.01 0.003 Number of pigs owned by household 2.05 3.61 2.41 2.91 4.30 0.000 Number of sheep owned by household 0.20 1.94 0.13 1.57 1.74 0.082 Number of goats owned by household 0.78 2.61 0.90 3.13 1.51 0.133 Number of cattle owned by household 1.08 3.57 1.27 3.68 2.07 0.039 Number of buffalo owned by household 0.63 3.01 0.74 3.40 1.55 0.121 Household grows rice 0.35 0.48 0.19 0.39 8.37 0.000 Household grows maize 0.70 0.46 0.67 0.47 1.10 0.272 Household grows cassava 0.64 0.48 0.56 0.50 3.44 0.001 Household grows sweet potato 0.55 0.50 0.30 0.46 12.63 0.000 Household grows vegetables 0.52 0.50 0.25 0.43 15.12 0.000 Household grows coffee 0.38 0.48 0.19 0.39 8.84 0.000 Household grows coconuts 0.51 0.50 0.17 0.38 23.09 0.000 Household uses organic fertiliser 0.11 0.32 0.01 0.11 20.64 0.000 Household uses inorganic fertiliser 0.08 0.27 0.01 0.09 18.08 0.000 Household uses pesticide 0.08 0.27 0.05 0.21 3.86 0.000 Household uses herbicide 0.06 0.24 0.04 0.20 2.10 0.036 Household cultivates less than 1 hectare 0.51 0.50 0.55 0.50 2.10 0.036 Household cultivates 1 to 5 hectares 0.25 0.43 0.26 0.44 0.80 0.422 Household cultivates more than 5 hectares 0.02 0.13 0.00 0.06 8.31 0.000 47 Table 6: Comparison of Census and TLSLS Variables for Person-Level Characteristics Person-level characteristics Census TLSLS t-test p-value Mean SD Mean SD Number of persons (from census front cover) 5.76 2.98 5.42 2.63 5.55 0.000 Number of males 2.92 1.86 2.71 1.67 5.85 0.000 Number of females 2.84 1.77 2.70 1.59 4.26 0.000 Number aged less than 6 years 0.89 1.02 0.73 0.95 8.98 0.000 Number aged 6 to 15 years 1.52 1.50 1.59 1.48 2.70 0.007 Number aged 60 and above 0.47 0.72 0.47 0.72 0.08 0.939 Number aged 16 to 59 years 2.88 1.92 2.62 1.70 5.14 0.000 Number of unregistered births amongst 0-5 year olds 0.09 0.37 0.09 0.38 0.03 0.973 Number of 6-15 year olds who have never attended school 0.18 0.55 0.17 0.46 1.48 0.139 Number of 16-59 year olds who were wage employees in the prior week 0.52 0.86 0.41 0.71 4.11 0.000 Number of 16-59 year olds who are were economically active in the prior week 1.33 1.58 1.70 1.56 7.92 0.000 Household head is male 0.84 0.37 0.85 0.36 0.96 0.336 Age of household head 48.06 15.08 49.68 14.02 5.09 0.000 Household head is married 0.85 0.36 0.79 0.41 8.56 0.000 Household head's mother tongue is tetun 0.31 0.46 0.15 0.36 8.30 0.000 Household head is a migrant (born outside the current sub-district) 0.21 0.41 0.19 0.39 0.87 0.385 Household head never attended school 0.41 0.49 0.50 0.50 5.58 0.000 Household head's highest education level is primary 0.20 0.40 0.17 0.37 4.24 0.000 Household head's highest education level is pre-secondary 0.09 0.28 0.09 0.28 0.05 0.961 Household head's highest education level is secondary 0.16 0.37 0.15 0.36 0.93 0.352 Household head's highest education level is tertiary (polytech/university) 0.10 0.30 0.05 0.22 7.62 0.000 Household head was not economically active in prior week 0.13 0.33 0.45 0.50 19.93 0.000 Share of household who are males 0.50 0.21 0.49 0.21 3.30 0.001 Share of household who are females 0.50 0.21 0.51 0.21 2.76 0.006 Share of household who are aged less than 6 years 0.14 0.16 0.12 0.15 8.43 0.000 48 Person-level characteristics Census TLSLS t-test p-value Mean SD Mean SD Share of household who are aged 6 to 15 years 0.23 0.21 0.25 0.21 6.63 0.000 Share of household who are aged 60 and above 0.13 0.24 0.15 0.28 3.72 0.000 Share of household who are aged 16 to 59 years 0.51 0.26 0.48 0.25 4.47 0.000 49 Table 7: Beta Models for Predicting Per Capita Consumption, National Sample (N=5916) Variable OLS Model GLS Model Coefficient Std Error z-stat Coefficient Std Error z-stat age60up 0.0384 0.0110 3.48 0.0252 0.0094 2.68 birth_unreg -0.0989 0.0178 5.54 -0.0791 0.0119 6.64 cook_wood -0.2211 0.0211 10.47 -0.1677 0.0207 8.09 did_crops -0.1238 0.0251 4.93 -0.1302 0.0250 5.20 did_livestock -0.0631 0.0296 2.13 -0.0702 0.0271 2.60 has_tv 0.0783 0.0200 3.92 0.0861 0.0177 4.86 head_male -0.1463 0.0249 5.87 -0.1494 0.0249 6.00 head_secondary 0.0655 0.0223 2.93 0.0433 0.0189 2.30 mean_age60up 0.2108 0.0836 2.52 0.2526 0.1354 1.86 mean_bath_outdoor -0.1371 0.0494 2.77 -0.1413 0.0816 1.73 mean_birth_unreg 0.6203 0.1357 4.57 0.5987 0.2322 2.58 mean_cult_1_5_ha 0.1464 0.0616 2.38 0.1470 0.1016 1.45 mean_did_crops -0.5770 0.1546 3.73 -0.5660 0.2538 2.23 mean_did_livestock 0.4022 0.1737 2.32 0.4120 0.2780 1.48 mean_grow_maize 0.4443 0.1074 4.14 0.4250 0.1786 2.38 mean_has_boat 1.2882 0.2147 6.00 1.3194 0.3510 3.76 mean_has_tv 0.2908 0.1054 2.76 0.3093 0.1744 1.77 mean_head_mig 0.2903 0.0786 3.69 0.3041 0.1320 2.30 mean_hhsize -0.0450 0.0123 3.66 -0.0399 0.0205 1.95 mean_light_elect -0.1241 0.0422 2.94 -0.1270 0.0703 1.81 mean_n_buffalo 0.0477 0.0107 4.47 0.0501 0.0179 2.80 mean_n_goat -0.0394 0.0161 2.45 -0.0479 0.0262 1.83 mean_n_sheep 0.0761 0.0214 3.56 0.0843 0.0357 2.36 mean_sewer_tank 0.1218 0.0535 2.28 0.1176 0.0877 1.34 noschool615 -0.1266 0.0158 8.00 -0.1153 0.0115 9.99 wall_concretebrick 0.2175 0.0181 12.02 0.2165 0.0161 13.42 wall_tin 0.1376 0.0400 3.44 0.1060 0.0352 3.01 _cons 4.1104 0.1166 35.26 4.0341 0.1892 21.32 Notes: Dependent variable is log per capita consumption. The OLS model has an adjusted-R2 of 0.337 and the F-statistic is 112.2. Variable names that are "mean_*" are means of the attribute for census households at the suco level. Other variables defined in Table 5. 50 Table 8: Alpha Model for Heteroscedasticity, National Sample (N=5916) Variable Coefficient Std Error z-stat cook_wood 4.2933 1.9534 2.20 did_crops -0.3208 0.1410 2.28 mean_bath_outdoor -57.9470 15.5134 3.74 birth_unreg_yhat -0.5649 0.2910 1.94 cook_wood_yhat -1.0933 0.4807 2.27 head_male_yhat -0.0751 0.0326 2.30 head_secondary_yhat -0.8279 0.4071 2.03 mean_age60up_yhat -0.2745 0.0934 2.94 mean_bath_outdoor_yhat 29.2599 7.7710 3.77 mean_birth_unreg_yhat -5.1578 2.1280 2.42 mean_cult_1_5_ha_yhat -3.3650 1.1370 2.96 mean_has_tv_yhat -0.4024 0.1024 3.93 mean_n_buffalo_yhat 0.0451 0.0130 3.46 birth_unreg_yhat2 0.1500 0.0771 1.95 head_secondary_yhat2 0.1999 0.0988 2.02 mean_bath_outdoor_yhat2 -3.7110 0.9776 3.80 mean_birth_unreg_yhat2 1.3542 0.5317 2.55 mean_cult_1_5_ha_yhat2 0.8286 0.2918 2.84 mean_light_elect_yhat2 0.0387 0.0137 2.83 _cons -3.4800 0.3780 9.21 Notes: The alpha model has an adjusted-R2 of 0.026 and the F-statistic is 9.2. Variable names with suffix "_yhat" are interacted with the predicted value from the beta model for log per capita consumption, and with suffix "_yhat2" are interacted with the square of the predicted value from the beta model. Other notes, see Table 7. 51 Table 9: Beta Models for Predicting Per Capita Consumption, Rural Sub-Sample (N=3898) Variable OLS Model GLS Model Coefficient Std Error z-stat Coefficient Std Error z-stat age60up 0.0422 0.0118 3.56 0.0291 0.0101 2.89 birth_unreg -0.0850 0.0204 4.17 -0.0837 0.0142 5.89 cook_wood -0.1894 0.0251 7.54 -0.1154 0.0250 4.61 did_crops -0.1530 0.0360 4.25 -0.1920 0.0361 5.32 has_tv 0.0945 0.0235 4.01 0.1057 0.0207 5.12 head_male -0.1626 0.0283 5.75 -0.1507 0.0297 5.08 head_mig 0.0745 0.0329 2.26 0.0211 0.0286 0.74 head_secondary 0.0664 0.0279 2.38 0.0434 0.0239 1.81 mean_age60up 0.2446 0.0894 2.73 0.2899 0.1505 1.93 mean_bath_outdoor -0.1490 0.0525 2.84 -0.1581 0.0895 1.77 mean_birth_unreg 0.5165 0.1391 3.71 0.5525 0.2386 2.32 mean_cult_1_5_ha 0.1640 0.0642 2.56 0.1959 0.1102 1.78 mean_did_crops -0.3863 0.1956 1.97 -0.3231 0.3329 0.97 mean_did_livestock 0.5466 0.2227 2.45 0.5118 0.3744 1.37 mean_grow_maize 0.2129 0.1155 1.84 0.1675 0.1990 0.84 mean_has_boat 1.3380 0.2195 6.10 1.4320 0.3933 3.64 mean_head_mig 0.3378 0.0846 3.99 0.3876 0.1407 2.75 mean_hhsize -0.0445 0.0128 3.48 -0.0417 0.0218 1.91 mean_light_elect -0.0764 0.0303 2.52 -0.0805 0.0513 1.57 mean_n_buffalo 0.0466 0.0116 4.03 0.0446 0.0201 2.22 mean_n_goat -0.0330 0.0167 1.97 -0.0410 0.0285 1.44 mean_n_sheep 0.0954 0.0272 3.51 0.1035 0.0484 2.14 mean_sewer_tank 0.1777 0.0645 2.76 0.1895 0.1104 1.72 mean_wall_anyplant -0.6356 0.2139 2.97 -0.6912 0.3699 1.87 mean_wall_concretebrick -0.4590 0.2351 1.95 -0.5177 0.4064 1.27 mean_wall_tin -0.6728 0.2696 2.50 -0.6604 0.4661 1.42 noschool615 -0.1301 0.0168 7.74 -0.1190 0.0118 10.05 wall_concretebrick 0.1931 0.0217 8.89 0.2036 0.0189 10.75 wall_tin 0.1469 0.0457 3.21 0.1288 0.0397 3.24 _cons 4.5254 0.2653 17.06 4.5126 0.4536 9.95 Notes: Dependent variable is log per capita consumption of rural households. The OLS model has an adjusted-R2 of 0.239 and the F-statistic is 43.1. Variable names that are "mean_*" are means of the attribute for rural households in the census, averaged at the suco level. Other variables defined in Table 5 and Table 6. 52 Table 10: Alpha Model for Heteroscedasticity, Rural Sub-Sample (N=3898) Variable Coefficient Std Error z-stat birth_unreg 32.5027 13.6039 2.39 did_crops -5.5847 1.6203 3.45 mean_head_mig 128.2914 56.8943 2.25 mean_n_buffalo 0.1543 0.0563 2.74 noschool615 -0.2880 0.1005 2.87 birth_unreg_yhat -18.2206 7.3273 2.49 did_crops_yhat 1.2802 0.3987 3.21 head_male_yhat -0.1350 0.0386 3.50 mean_birth_unreg_yhat 8.5475 3.4153 2.50 mean_head_mig_yhat -63.5143 28.1849 2.25 wall_concretebrick_yhat 0.9300 0.5005 1.86 birth_unreg_yhat2 2.5257 0.9843 2.57 mean_birth_unreg_yhat2 -2.1628 0.8700 2.49 mean_head_mig_yhat2 7.7993 3.4837 2.24 wall_concretebrick_yhat2 -0.2343 0.1250 1.87 _cons -3.5148 0.2477 14.19 Notes: The alpha model has an adjusted-R2 of 0.026 and the F-statistic is 7.9. Variable names with suffix "_yhat" are interacted with the predicted value from the beta model on per capita consumption, and with suffix "_yhat2" are interacted with the square of the predicted value from the beta model. Other notes, see Table 9. 53 Table 11: Beta Models for Predicting Per Capita Consumption, Urban Sub-Sample (N=2018) Variable OLS Model GLS Model Coefficient Std Error z-stat Coefficient Std Error z-stat bath_outdoor -0.0982 0.0311 3.15 -0.0918 0.0278 3.31 birth_unreg -0.1520 0.0356 4.27 -0.1459 0.0313 4.66 cook_wood -0.2233 0.0345 6.47 -0.1937 0.0318 6.10 did_crops -0.1125 0.0334 3.36 -0.0865 0.0309 2.80 did_livestock -0.1428 0.0414 3.45 -0.1385 0.0391 3.54 head_male -0.1048 0.0490 2.14 -0.1156 0.0425 2.72 mean_birth_unreg 0.8087 0.3608 2.24 0.7367 0.5103 1.44 mean_cook_wood -0.3883 0.1483 2.62 -0.3935 0.2027 1.94 mean_cult_1_5_ha -0.4408 0.1428 3.09 -0.4141 0.1901 2.18 mean_dwell_own 1.0598 0.3772 2.81 1.0603 0.5069 2.09 mean_head_secondary 1.1101 0.3368 3.30 1.0275 0.4411 2.33 mean_n_buffalo 0.0463 0.0240 1.93 0.0271 0.0316 0.86 mean_noschool615 0.6015 0.1557 3.86 0.5938 0.2042 2.91 mean_wall_anyplant 0.3109 0.1344 2.31 0.2662 0.1856 1.43 mean_wall_bamboo -0.3566 0.1510 2.36 -0.4649 0.2024 2.30 n_buffalo 0.0076 0.0030 2.49 0.0105 0.0039 2.67 noschool615 -0.1318 0.0429 3.07 -0.1444 0.0331 4.36 wall_concretebrick 0.2498 0.0338 7.39 0.2458 0.0285 8.63 _cons 3.3580 0.3846 8.73 3.3877 0.5043 6.72 Notes: Dependent variable is log per capita consumption of urban households. The OLS model has an adjusted-R2 of 0.346 and F-statistic is 60.2. Variable names that are "mean_*" are means of the attribute for urban households in the census, averaged at the suco level. Other variables defined in Table 5 and Table 6. Table 12: Alpha Model for Heteroscedasticity, Urban Sub-Sample (N=2018) Variable Coefficient Std Error z-stat mean_wall_anyplant -9.4989 4.8846 1.94 n_buffalo 0.0219 0.0168 1.30 wall_concretebrick -4.8885 2.4631 1.98 mean_birth_unreg_yhat 1.0193 0.3943 2.59 mean_cook_wood_yhat 3.5218 1.5888 2.22 wall_concretebrick_yhat 1.2505 0.6078 2.06 mean_cook_wood_yhat2 -0.8524 0.3850 2.21 mean_wall_anyplant_yhat2 0.5989 0.2919 2.05 _cons -5.8791 0.4403 13.35 Notes: The alpha model has an adjusted-R2 of 0.019 and the F-statistic is 5.9. Variable names with suffix "_yhat" are interacted with the predicted value from the beta model on per capita consumption, and with suffix "_yhat2" are interacted with the square of the predicted value from the beta model. Other notes, see Table 11. Table 13: Beta Models for Predicting Labour Force Indicator Variable OLS Model GLS Model Coefficient Std z-stat Coefficient Std z-stat Error Error 54 birth_unreg -0.0886 0.0441 2.01 -0.0749 0.0419 1.79 cult_1_5_ha 0.142 0.0434 3.27 0.1441 0.0379 3.80 did_crops 0.2499 0.0594 4.21 0.2762 0.0628 4.40 has_tv -0.2108 0.0457 4.61 -0.2089 0.0426 4.91 mean_head_secondary -1.1167 0.2934 3.81 -0.9402 0.3239 2.90 mean_sewer_tank 0.217 0.0985 2.20 0.1627 0.1147 1.42 n_sheep 0.0177 0.0113 1.56 0.0077 0.0072 1.08 wall_tin -0.2705 0.0975 2.78 -0.1892 0.0939 2.01 _cons -0.0587 0.0741 0.79 -0.099 0.0775 1.28 Notes: Dependent variable is the principal component for labour force indicators. Variable names that are "mean_*" are means of the attribute for census households at the suco level. Other variables defined in Table 5. Table 14: Alpha Model for Heteroscedasticity in work_pca Variable Coefficient Std Error z-stat mean_sewer_tank 0.6653 0.2077 3.20 mean_sewer_tank_yhat -3.3443 0.6686 5.00 mean_sewer_tank_yhat2 -3.1688 1.6610 1.91 _cons -5.0208 0.0592 84.79 Notes: Variable names with suffix "_yhat" are interacted with the predicted value from the beta model for log per capita consumption, and with suffix "_yhat2" are interacted with the square of the predicted value from the beta model. Other notes, see Table 13. 55 Table 15: Beta Model for Education Principal Components Index Variable Coefficient Std Error t-stat |Prob | > t Label _intercept_ -0.9425 0.0742 -12.71 0.000 Intercept AGE60UP -0.0924 0.0233 -3.96 0.000 age60up HEAD_MALE_1 1.0887 0.0540 20.15 0.000 Dummy for HEAD_MALE=1 HEAD_PRESEC_1 -0.1612 0.0540 -2.98 0.003 Dummy for HEAD_PRESEC=1 LIGHT_SOLAR_1 -0.0761 0.0438 -1.74 0.083 Dummy for LIGHT_SOLAR=1 MEAN_NOSCHOOL615 -0.1958 0.1251 -1.57 0.118 mean_noschool615 MEAN_SEWER_HOLE 0.2133 0.0889 2.40 0.017 mean_sewer_hole MEAN_SEWER_TANK 0.2444 0.0756 3.23 0.001 mean_sewer_tank NOSCHOOL615 0.0769 0.0343 2.24 0.025 noschool615 N_SHEEP 0.0140 0.0099 1.41 0.157 n_sheep Note: Dependent variable is the principal components education index described in text. Variables named "MEAN_*" are means of the attribute for census households at the suco level. Other variables defined in Table 5. Table 16: Alpha Model for Heteroscedasticity for the Education Index Survey-to-Census Imputation Coeffic Variable Std Error t-stat |Prob | > t Label ient _intercept_ -5.7583 0.1780 -32.35 0.000 Intercept MEAN_BATH_OUTDOOR 1.5753 0.2763 5.70 0.000 mean_bath_outdoor MEAN_SEWER_HOLE -1.0213 0.2638 -3.87 0.000 mean_sewer_hole MEAN_WALL_ANYPLAN -1.9357 0.2746 -7.05 0.000 mean_wall_anyplant* T*_yhat_ _yhat_ MEAN_WALL_CONCRET -1.6798 0.2296 -7.32 0.000 mean_wall_concreteb EBRICK rick MEAN_WALL_CONCRET -2.8443 0.3953 -7.20 0.000 mean_wall_concreteb EBRICK*_yhat_ rick*_yhat_ NOSCHOOL615 0.8561 0.1125 7.61 0.000 noschool615 NOSCHOOL615*_yhat_ 1.0945 0.3019 3.63 0.000 noschool615*_yhat_ Note: Variables named with suffix "_yhat" are interacted with the predicted value from the beta model. Other notes, see Table 15. 56 Table 17: Beta Model for Health Principal Components Index Std Variable Coefficient t-stat |Prob | > t Label Error _intercept_ -0.0872 0.1201 -0.73 0.468 Intercept HEAD_MALE_1 0.2300 0.0478 4.81 0.000 Dummy for HEAD_MALE=1 HEAD_MIG_1 -0.0981 0.0452 -2.17 0.030 Dummy for HEAD_MIG=1 HEAD_PRESEC_1 -0.0838 0.0475 -1.77 0.078 Dummy for HEAD_PRESEC=1 LIGHT_ELECT_1 -0.0618 0.0426 -1.45 0.147 Dummy for LIGHT_ELECT=1 MEAN_COOK_WOOD -0.2201 0.0962 -2.29 0.022 mean_cook_wood MEAN_LIGHT_ELECT 0.1922 0.0787 2.44 0.015 mean_light_elect MEAN_LIGHT_SOLAR 0.1558 0.0964 1.62 0.106 mean_light_solar MEAN_WALL_TIN -0.5532 0.2921 -1.89 0.058 mean_wall_tin Note: Dependent variable is the principal components health index described in text. Variables named "MEAN_*" are means of the attribute for census households at the suco level. Other variables defined in Table 5. Table 18: Alpha Model for Heteroscedasticity for the Health Index Survey-to-Census Imputation Coeffic Std Variable t-stat |Prob | > t Label ient Error _intercept_ -9.4685 0.3468 -27.30 0.000 Intercept _yhat_ -6.5471 1.1022 -5.94 0.000 _yhat_ HEAD_MALE_1 0.6741 0.3486 1.93 0.053 Dummy for (HEAD_MALE)=1 HEAD_MALE_1*_yhat_* 43.043 8.5009 5.06 0.000 Dummy for _yhat_ 3 (HEAD_MALE)=1* _yhat_*_yhat_ HEAD_MIG_1 -0.9261 0.1754 -5.28 0.000 Dummy for (HEAD_MIG)=1 HEAD_PRESEC_1 -0.9059 0.2185 -4.15 0.000 Dummy for (HEAD_PRESEC)=1 HEAD_PRESEC_1*_yhat_ -3.0047 2.0968 -1.43 0.152 Dummy for (HEAD_PRESEC)=1 *_yhat_ LIGHT_ELECT_1*_yhat_ - 5.0409 -2.92 0.004 Dummy for *_yhat_ 14.720 (LIGHT_ELECT)=1 8 *_yhat_*_yhat_ MEAN_COOK_WOOD -1.9675 0.3910 -5.03 0.000 mean_cook_wood MEAN_COOK_WOOD*_ 13.264 6.6879 1.98 0.047 mean_cook_wood*_yh yhat_*_yhat_ 7 at_*_yhat_ Note: Variables named with suffix "_yhat" are interacted with the predicted value from the beta model. Other notes, see Table 17. 57 Table 19: Comparison of Census and DHS Variables for Household-Level Characteristics Dwelling attributes and household variables 2015 Census 2016 DHS t-test p- Mean SD Mean SD value Number of rooms in the dwelling (excl bathrooms, kitchens etc) 2.891 1.326 2.886 1.365 0.14 0.887 Dwelling has indoor bath or shower 0.143 0.350 0.150 0.357 0.59 0.554 Household members use flush toilet 0.362 0.481 0.465 0.499 5.11 0.000 Household members use ventilated, improved pit latrine 0.122 0.328 0.026 0.161 20.26 0.000 Household members use pit latrine with slab 0.087 0.282 0.149 0.356 5.79 0.000 Household members use open pit latrine 0.078 0.268 0.014 0.119 17.90 0.000 Household members use hanging toilet/latrine 0.109 0.312 0.057 0.232 7.38 0.000 Household members have no toilet facility or use the bush 0.206 0.404 0.271 0.445 4.07 0.000 Household members use improved toilet (flush or VIP) 0.484 0.500 0.491 0.500 0.33 0.744 Household members use latrine with pit slab or hanging 0.196 0.397 0.206 0.405 0.78 0.435 Household members use open pit, bush or have no toilet 0.284 0.451 0.286 0.452 0.13 0.901 Dwelling has a kitchen within the dwelling (shared or exclusive) 0.251 0.433 0.118 0.322 14.65 0.000 Main energy source for cooking is electricity 0.103 0.304 0.082 0.275 2.89 0.004 Main energy source for cooking is gas 0.026 0.159 0.006 0.080 9.51 0.000 Main energy source for cooking is biogas 0.007 0.083 0.002 0.047 5.37 0.000 Main energy source for cooking is kerosene 0.046 0.209 0.044 0.206 0.20 0.843 Main energy source for cooking is coal 0.004 0.060 0.000 0.022 9.34 0.000 Main energy source for cooking is wood 0.813 0.390 0.864 0.343 3.76 0.000 Drinking water is mainly from piped or pumped indoors 0.050 0.218 0.228 0.419 12.52 0.000 Drinking water is mainly from piped or pumped outdoors 0.123 0.329 0.097 0.296 3.37 0.001 Drinking water is mainly from public tap/public pipe 0.412 0.492 0.236 0.424 12.82 0.000 Drinking water is mainly from tubewell/borehole 0.066 0.249 0.038 0.190 4.05 0.000 Drinking water is mainly from protected well or spring 0.075 0.263 0.111 0.314 4.43 0.000 Drinking water is mainly from rainwater collection 0.003 0.052 0.001 0.026 4.45 0.000 Drinking water is mainly from water bottles 0.018 0.133 0.053 0.224 4.20 0.000 Drinking water is mainly from unprotected well or spring 0.089 0.284 0.159 0.366 5.94 0.000 58 Dwelling attributes and household variables 2015 Census 2016 DHS t-test p- Mean SD Mean SD value Drinking water is mainly from water vendor/tanker 0.009 0.097 0.009 0.096 0.03 0.977 Drinking water is mainly from river, lake or stream 0.142 0.349 0.037 0.188 14.52 0.000 Drinking water is piped or from tap (public or private) 0.586 0.493 0.560 0.496 1.45 0.148 Household has at least one radio 0.270 0.444 0.245 0.430 3.02 0.003 Household has at least one television 0.368 0.482 0.402 0.490 1.72 0.086 Household has at least one telephone/mobile 0.813 0.390 1.000 0.000 65.61 0.000 Household has at least one computer 0.164 0.370 0.109 0.312 5.06 0.000 Household has at least one refrigerator 0.155 0.362 0.196 0.397 2.66 0.008 Household has at least one sewing machine 0.037 0.189 0.030 0.171 2.56 0.011 Household has at least one bicycle 0.139 0.346 0.146 0.353 0.69 0.492 Household has at least one motor cycle 0.236 0.425 0.318 0.466 5.97 0.000 Household has at least one car, van or truck 0.065 0.246 0.049 0.216 3.10 0.002 Household has at least one boat 0.019 0.137 0.006 0.078 7.21 0.000 Household has at least one radio or one TV 0.487 0.500 0.496 0.500 0.52 0.604 Household is using banking facility 0.422 0.494 0.491 0.500 6.61 0.000 Household rears livestock 0.872 0.334 0.829 0.377 3.51 0.000 Number of chickens owned by household 4.540 7.730 6.279 13.147 7.39 0.000 Number of pigs owned by household 2.049 3.613 2.337 4.378 3.48 0.001 Number of sheep owned by household 0.198 1.936 0.189 2.818 0.21 0.832 Number of goats owned by household 0.775 2.608 1.095 4.338 4.26 0.000 Number of cattle owned by household 1.084 3.573 1.182 4.729 1.25 0.212 Number of buffalo owned by household 0.627 3.010 0.904 5.349 3.33 0.001 Number of horses owned by household 0.248 0.804 0.286 2.183 1.27 0.204 59 Table 20: Comparison of Census and DHS Variables for Person-Level Characteristics Person-level characteristics 2015 Census 2016 DHS t-test p- Mean SD Mean SD value Number of persons (from census front cover) 5.759 2.979 5.313 2.723 7.70 0.000 Number of males 2.924 1.862 2.672 1.704 7.60 0.000 Number of females 2.835 1.770 2.641 1.640 6.65 0.000 Number aged less than 6 years 0.887 1.019 0.787 0.967 6.90 0.000 Number aged 6 to 15 years 1.519 1.498 1.490 1.436 1.42 0.156 Number aged 60 and above 0.473 0.718 0.508 0.733 2.57 0.010 Number aged 16 to 59 years 2.881 1.918 2.528 1.700 7.43 0.000 Number of unregistered births amongst 0-5 year olds 0.076 0.338 0.259 0.591 19.71 0.000 Number of 6-15 year olds who have never attended school 0.180 0.551 0.135 0.426 5.31 0.000 Household head is male 0.842 0.365 0.825 0.380 3.12 0.002 Age of household head 48.063 15.077 50.053 14.965 5.93 0.000 Household head is married 0.851 0.356 0.794 0.405 10.64 0.000 Household head never attended school 0.413 0.492 0.440 0.496 1.94 0.052 Household head's highest education level is primary 0.196 0.397 0.226 0.418 5.10 0.000 Household head's highest education level is pre-secondary 0.087 0.281 0.082 0.274 1.34 0.180 Household head's highest education level is secondary 0.163 0.369 0.165 0.371 0.34 0.733 Household head's highest education level is tertiary (polytech/university) 0.097 0.296 0.085 0.279 1.47 0.141 Share of household who are males 0.504 0.213 0.497 0.217 2.45 0.014 Share of household who are females 0.496 0.213 0.503 0.217 2.45 0.014 Share of household who are aged less than 6 years 0.139 0.159 0.128 0.156 4.88 0.000 Share of household who are aged 6 to 15 years 0.227 0.205 0.242 0.211 4.78 0.000 Share of household who are aged 60 and above 0.127 0.242 0.153 0.271 4.92 0.000 Share of household who are aged 16 to 59 years 0.508 0.256 0.478 0.253 5.34 0.000 60 Table 21: Beta Models for Predicting PCA Index of Types of Female Decision-Making Autonomy (N=7013) Variable OLS Model GLS Model Coefficient Std Error z-stat Coefficient Std Error z-stat cook_kero -0.2483 0.1103 2.25 -0.2083 0.1073 1.94 mean_cook_kero 2.1923 0.5718 3.83 2.2863 0.9552 2.39 mean_dwell_rooms 0.1489 0.0631 2.36 0.131 0.0996 1.32 mean_has_bike -0.6863 0.3035 2.26 -0.5569 0.5004 1.11 mean_has_radiotv 0.4624 0.2205 2.1 0.386 0.3557 1.09 mean_head_tertiary -2.4147 0.5304 4.55 -2.3705 0.865 2.74 mean_toilet_improved 0.6791 0.1507 4.51 0.6468 0.2313 2.8 mean_toilet_noopen 0.8869 0.1271 6.98 0.8087 0.1855 4.36 mean_water_carttank 1.7617 0.3728 4.73 1.7428 0.6414 2.72 mean_water_improved -0.3067 0.0897 3.42 -0.3249 0.1413 2.3 toilet_improved 0.1519 0.0499 3.04 0.096 0.0529 1.82 toilet_slabhang -0.1192 0.0584 2.04 -0.086 0.0573 1.5 _cons -0.8986 0.1892 4.75 -0.7721 0.2871 2.69 Notes: Dependent variable is PCA Index for three dummy variables for decision-making autonomy in different domains. The OLS model has an adjusted-R 2 of 0.026 and the F -statistic is 16.3. Variable names that are "mean_*" are means of the attribute for census households at the suco level. Other variables defined in Table 19 and Table 20. Table 22: Alpha Model for Heteroscedasticity, DM Index (N=7013) Variable Coefficient Std Error z-stat mean_dwell_rooms 0.4197 0.1007 4.17 mean_has_bike 0.2495 0.3772 0.66 mean_toilet_noopen 0.745 0.2474 3.01 mean_water_carttank 3.6338 2.037 1.78 toilet_improved 0.3156 0.0871 3.62 mean_cook_kero_yhat 11.835 2.781 4.26 mean_dwell_rooms_yhat 0.8592 0.1619 5.31 mean_toilet_improved_yhat -4.0357 0.7873 5.13 mean_toilet_noopen_yhat -0.4105 0.8016 0.51 mean_water_carttank_yhat -0.4292 3.9263 0.11 toilet_slabhang_yhat -1.8986 0.405 4.69 mean_has_radiotv_yhat2 -2.5407 1.8976 1.34 mean_toilet_noopen_yhat2 -4.5575 1.8053 2.52 mean_water_improved_yhat2 -1.4971 1.6349 0.92 _cons -4.6346 0.3478 13.32 Notes: The alpha model has an adjusted-R 2 of 0.030 and the F -statistic is 16.3. Variable names with suffix "_yhat" are interacted with the predicted value from the beta model for DV Index, and with suffix "_yhat2" are interacted with the square of the predicted value from the beta model. 61 Table 23: Beta Models for Predicting PCA Index of Types of Domestic Violence Indicators (N=3674) Variable OLS Model GLS Model Coefficient Std Error z-stat Coefficient Std Error z-stat cook_kero -0.5165 0.2406 2.15 -0.4427 0.1221 3.63 has_bike 0.2063 0.1194 1.73 0.2286 0.0756 3.02 has_radiotv -0.2436 0.0944 2.58 -0.2393 0.0656 3.65 head_secondary -0.3569 0.1097 3.25 -0.3171 0.0658 4.82 head_tertiary -0.4588 0.1688 2.72 -0.2698 0.0973 2.77 mean_bath_indoor 3.7054 0.8468 4.38 2.3647 0.8731 2.71 mean_dwell_rooms -0.4143 0.1098 3.77 -0.4401 0.1156 3.81 mean_head_secondary 2.432 0.8799 2.76 2.1968 0.9293 2.36 mean_head_tertiary -7.9966 1.4238 5.62 -6.6549 1.417 4.7 mean_n_cattle -0.1531 0.0429 3.57 -0.1561 0.0444 3.51 mean_n_sheep -0.4274 0.1032 4.14 -0.4213 0.0883 4.77 n_sheep 0.0248 0.016 1.55 0.0377 0.0158 2.38 water_improved -0.2545 0.0852 2.99 -0.1284 0.0611 2.1 _cons 1.6882 0.3069 5.5 1.7361 0.3297 5.27 Notes: Dependent variable is PCA Index for 19 dummy variables for experience of actual or threatened abuse, being afraid of partner, and having limits placed on autonomy (see text for details). The OLS model has an adjusted-R 2 of 0.049 and the F -statistic is 15.6. Variable names that are "mean_*" are means of the attribute for census households at the suco level. Other variables defined in Table 19 and Table 20. Table 24: Alpha Model for Heteroscedasticity, DV Index (N=3674) Variable Coefficient Std Error z-stat head_secondary -0.3546 0.1365 2.6 head_tertiary -0.5665 0.2056 2.76 mean_head_secondary 1.1638 0.6253 1.86 head_secondary_yhat 1.3941 0.282 4.94 mean_bath_indoor_yhat 1.717 1.1117 1.54 mean_head_secondary_yhat 1.3599 1.4452 0.94 mean_n_sheep_yhat 1.591 0.381 4.18 n_sheep_yhat -0.1104 0.0378 2.92 mean_n_sheep_yhat2 0.8183 0.2194 3.73 n_sheep_yhat2 0.033 0.0161 2.05 _cons -6.1647 0.0989 62.33 Notes: The alpha model has an adjusted-R 2 of 0.086 and the F -statistic is 32.4. Variable names with suffix "_yhat" are interacted with the predicted value from the beta model for DV Index, and with suffix "_yhat2" are interacted with the square of the predicted value from the beta model. 62 Table 25: Beta Models for Predicting Count of Types of Domestic Violence Indicators (N=3674) Variable OLS Model GLS Model Coefficient Std Error z-stat Coefficient Std Error z-stat cook_kero -0.7155 0.292 2.45 -0.7331 0.182 4.03 has_radiotv -0.2284 0.1159 1.97 -0.2632 0.096 2.74 head_secondary -0.422 0.1334 3.16 -0.3943 0.0986 4 head_tertiary -0.4407 0.2053 2.15 -0.3447 0.1395 2.47 mean_bath_indoor 4.6393 1.1374 4.08 3.3271 1.3033 2.55 mean_dwell_rooms -0.3837 0.1371 2.8 -0.4277 0.1566 2.73 mean_has_bike 1.8206 0.7302 2.49 2.2535 0.9253 2.44 mean_head_tertiary -10.1201 1.655 6.11 -9.151 1.8552 4.93 mean_n_cattle -0.2013 0.0551 3.65 -0.2206 0.0647 3.41 mean_n_sheep -0.6544 0.1284 5.1 -0.6759 0.1333 5.07 mean_toilet_slabhang -0.6448 0.2793 2.31 -0.4576 0.3213 1.42 n_sheep 0.0359 0.0194 1.84 0.0531 0.0338 1.57 toilet_noopen 0.2145 0.135 1.59 0.1098 0.1185 0.93 toilet_slabhang 0.2712 0.1441 1.88 0.1547 0.1233 1.25 water_improved -0.2803 0.1058 2.65 -0.1209 0.089 1.36 _cons 4.1014 0.4016 10.21 4.2007 0.4528 9.28 Notes: Dependent variable is count of 19 dummy variables for experience of actual or threatened abuse, being afraid of partner, and having limits placed on autonomy (see text for details). The OLS model has an adjusted-R 2 of 0.060 and the F -statistic is 16.7. Variable names that are "mean_*" are means of the attribute for census households at the suco level. Other variables defined in Table 19 and Table 20. Table 26: Alpha Model for Heteroscedasticity, DV Count (N=3674) Variable Coefficient Std Error z-stat head_secondary -2.1484 0.3445 6.24 mean_n_cattle 0.1001 0.0406 2.46 mean_toilet_slabhang 0.5915 0.2388 2.48 n_sheep 0.1007 0.087 1.16 toilet_slabhang -2.3968 1.1022 2.17 head_secondary_yhat 0.9004 0.179 5.03 head_tertiary_yhat -0.3795 0.0985 3.85 n_sheep_yhat -0.0367 0.0537 0.68 toilet_slabhang_yhat 1.8578 0.9464 1.96 mean_has_bike_yhat2 0.3586 0.0616 5.82 n_sheep_yhat2 0.0033 0.008 0.42 toilet_slabhang_yhat2 -0.3187 0.2024 1.57 water_improved_yhat2 0.043 0.0157 2.75 _cons -5.9708 -0.1176 50.78 Notes: The alpha model has an adjusted-R 2 of 0.059 and the F -statistic is 18.6. Variable names with suffix "_yhat" are interacted with the predicted value from the beta model for DV Count, and with suffix "_yhat2" are interacted with the square of the predicted value from the beta model. 63 Table 27: Suco-level Predicted Poverty and Inequality Area Identification Number of Expenditure Poverty Poverty Poverty Gini Population Rate Gap Severity Ratio District Subdistrict Suco HH Indiv. Min Max Mean Mean Mean Mean Mean Aileu Aileu Vila Aisirimou 326 2,206 7.72 298.81 56.46 0.36 0.08 0.03 0.23 Aileu Aileu Vila Bandudato 204 1,138 8.87 245.98 48.07 0.50 0.13 0.05 0.22 Aileu Aileu Vila Fahiria 301 1,823 10.40 311.42 54.50 0.40 0.10 0.03 0.23 Aileu Aileu Vila Fatubosa 357 2,033 8.41 282.59 50.15 0.47 0.11 0.04 0.22 Aileu Aileu Vila Hoholau 218 1,365 5.12 287.26 40.96 0.66 0.22 0.10 0.26 Aileu Aileu Vila Lahae 119 698 9.75 618.02 58.57 0.34 0.07 0.02 0.23 Aileu Aileu Vila Lausi 212 1,420 7.45 332.55 53.94 0.40 0.10 0.03 0.22 Aileu Aileu Vila Saboria 131 781 6.91 1,068.89 56.21 0.42 0.11 0.04 0.26 Aileu Aileu Vila Seloi Craic 550 3,584 6.31 936.38 48.07 0.54 0.14 0.05 0.25 Aileu Aileu Vila Seloi Malere 740 4,813 7.84 383.52 57.43 0.36 0.08 0.03 0.24 Aileu Aileu Vila Suco Liurai 688 4,122 6.92 544.07 47.04 0.54 0.15 0.06 0.24 Aileu Laulara Cotolau 183 1,283 7.94 322.04 44.96 0.58 0.16 0.06 0.22 Aileu Laulara Fatisi 210 1,357 9.81 521.63 48.25 0.54 0.14 0.05 0.24 Aileu Laulara Madabeno 241 1,543 7.69 292.18 41.92 0.65 0.18 0.07 0.21 Aileu Laulara Talitu 336 2,220 9.22 388.66 48.34 0.52 0.14 0.05 0.24 Aileu Laulara Tohumeta 96 674 9.81 248.78 41.47 0.66 0.18 0.06 0.20 Aileu Liquidoe Acubilitoho 131 864 11.50 204.69 46.21 0.54 0.13 0.04 0.19 Aileu Liquidoe Bereleu 205 1,280 9.11 286.35 42.97 0.63 0.18 0.07 0.23 Aileu Liquidoe Betulau 99 670 11.95 245.76 49.57 0.47 0.11 0.04 0.20 Aileu Liquidoe Fahisoi 169 1,204 8.11 244.88 45.56 0.57 0.15 0.05 0.21 Aileu Liquidoe Faturilau 99 719 10.82 229.68 43.16 0.62 0.17 0.07 0.21 Aileu Liquidoe Manucasa 88 518 7.65 275.59 47.81 0.53 0.14 0.05 0.23 Aileu Liquidoe Namoleso 239 1,510 11.75 418.02 50.77 0.46 0.10 0.03 0.21 Aileu Remexio Acumau 391 2,689 6.26 345.89 52.27 0.45 0.11 0.04 0.24 Aileu Remexio Fadabloco 286 1,896 9.80 199.60 40.06 0.69 0.21 0.08 0.22 Aileu Remexio Fahisoi 193 1,290 9.34 303.71 49.31 0.49 0.12 0.04 0.22 Aileu Remexio Faturasa 165 1,125 7.14 446.36 53.94 0.42 0.11 0.04 0.24 Aileu Remexio Hautoho 133 856 5.80 269.89 41.48 0.66 0.20 0.08 0.22 Aileu Remexio Maumeta 89 532 4.85 474.75 43.27 0.62 0.19 0.08 0.24 Aileu Remexio Suco Liurai 53 375 9.42 163.77 42.42 0.62 0.19 0.08 0.21 Aileu Remexio Tulataqueo 346 2,170 9.52 374.91 48.46 0.50 0.11 0.04 0.20 Ainaro Ainaro Ainaro 937 5,263 8.77 360.30 55.60 0.32 0.07 0.02 0.23 Ainaro Ainaro Cassa 508 2,916 4.47 3,489.35 60.11 0.36 0.09 0.04 0.31 Ainaro Ainaro Manutasi 323 2,110 9.93 285.40 50.32 0.39 0.08 0.03 0.22 Ainaro Ainaro Mau-Nuno 184 1,102 7.66 533.96 58.08 0.31 0.07 0.02 0.25 Ainaro Ainaro Mau-Ulo 300 1,492 7.61 424.58 45.34 0.51 0.14 0.05 0.23 Ainaro Ainaro Soro 320 1,949 9.77 520.23 51.67 0.38 0.08 0.03 0.23 Ainaro Ainaro Suro-Craik 182 1,104 11.84 261.37 55.92 0.28 0.06 0.02 0.20 Ainaro Hatu-Udo Foho-Ai-Lico 963 4,939 6.91 763.06 52.72 0.39 0.10 0.04 0.26 Ainaro Hatu-Udo Leolima 940 5,360 7.19 391.79 57.37 0.29 0.06 0.02 0.24 Ainaro Hatu-Builico Mau-Chiga 411 2,458 11.16 294.50 48.97 0.40 0.08 0.02 0.20 Ainaro Hatu-Builico Mulo 1,104 6,333 8.11 587.22 45.94 0.49 0.13 0.05 0.24 Ainaro Hatu-Builico Nuno-Mogue 654 4,175 7.10 488.21 48.05 0.46 0.12 0.04 0.24 Ainaro Maubisse Aitutu 862 5,131 6.54 417.85 46.03 0.50 0.14 0.06 0.26 64 Area Identification Number of Expenditure Poverty Poverty Poverty Gini Population Rate Gap Severity Ratio District Subdistrict Suco HH Indiv. Min Max Mean Mean Mean Mean Mean Ainaro Maubisse Edi 353 2,459 1.07 2,983.93 59.83 0.34 0.08 0.03 0.28 Ainaro Maubisse Fatu-Besi 233 1,491 4.90 441.09 52.10 0.40 0.10 0.04 0.25 Ainaro Maubisse Horai-Quic 281 1,704 3.39 565.08 46.74 0.47 0.12 0.04 0.22 Ainaro Maubisse Suco Liurai 134 744 1.71 849.55 55.70 0.43 0.12 0.05 0.30 Ainaro Maubisse Manelobas 200 1,276 9.68 517.72 46.99 0.48 0.12 0.04 0.22 Ainaro Maubisse Manetu 351 2,238 8.44 242.15 48.24 0.41 0.09 0.03 0.21 Ainaro Maubisse Maubisse 951 6,096 6.06 397.93 46.25 0.49 0.13 0.05 0.24 Ainaro Maubisse Maulau 409 2,477 5.62 345.62 45.96 0.50 0.15 0.06 0.26 Baucau Baguia Afaloicai 219 1,106 10.66 238.27 46.98 0.49 0.12 0.04 0.20 Baucau Baguia Alaua Craic 357 1,643 5.28 425.68 50.28 0.47 0.13 0.05 0.25 Baucau Baguia Alaua Leten 216 996 2.98 483.59 49.33 0.50 0.15 0.06 0.27 Baucau Baguia Defa Uassi 205 1,012 8.22 290.51 51.21 0.46 0.12 0.05 0.25 Baucau Baguia Haeconi 451 2,319 7.44 337.49 44.98 0.56 0.15 0.06 0.23 Baucau Baguia Lari Sula 250 1,167 3.13 577.15 56.58 0.41 0.12 0.05 0.29 Baucau Baguia Lavateri 309 1,455 5.18 392.81 50.89 0.46 0.12 0.05 0.25 Baucau Baguia Ossu-Huna 166 838 6.58 470.69 47.83 0.50 0.13 0.05 0.23 Baucau Baguia Samalari 358 1,822 4.31 751.36 49.75 0.51 0.16 0.07 0.29 Baucau Baguia Uacala 111 576 4.87 440.13 56.00 0.36 0.09 0.03 0.24 Baucau Baucau Bahu 1,340 8,154 12.25 684.02 72.07 0.14 0.02 0.01 0.22 Baucau Baucau Bucoli 391 2,443 10.21 387.22 60.08 0.30 0.07 0.02 0.24 Baucau Baucau Buibau 952 5,838 10.28 645.73 67.18 0.23 0.05 0.02 0.25 Baucau Baucau Buruma 611 3,245 9.65 571.91 64.22 0.24 0.05 0.02 0.23 Baucau Baucau Caibada 318 1,984 8.95 433.20 66.24 0.21 0.04 0.01 0.22 Baucau Baucau Gariuai 899 4,962 9.98 1,452.81 72.22 0.19 0.04 0.01 0.25 Baucau Baucau Samalari 319 1,499 7.77 862.10 75.09 0.20 0.04 0.01 0.27 Baucau Baucau Seical 395 1,975 8.30 1,214.57 84.29 0.14 0.03 0.01 0.27 Baucau Baucau Trilolo 1,811 11,348 7.42 641.97 68.16 0.21 0.04 0.01 0.25 Baucau Baucau Triloca 370 2,345 11.38 316.03 50.90 0.45 0.11 0.04 0.23 Baucau Baucau Wailili 654 3,471 3.02 774.94 63.40 0.29 0.06 0.02 0.26 Baucau Laga Atelari 313 1,616 7.05 516.34 50.86 0.46 0.12 0.05 0.25 Baucau Laga Libagua 133 729 9.11 320.49 54.75 0.36 0.08 0.03 0.22 Baucau Laga Nunira 308 1,772 10.01 549.53 60.51 0.30 0.07 0.02 0.24 Baucau Laga Saelari 382 2,439 4.73 731.96 53.67 0.44 0.13 0.05 0.28 Baucau Laga Sagadati 600 2,750 5.02 436.74 53.41 0.43 0.12 0.04 0.26 Baucau Laga Samalari 534 2,677 7.13 436.26 55.12 0.40 0.11 0.04 0.26 Baucau Laga Soba 482 2,744 11.67 510.61 61.71 0.27 0.06 0.02 0.23 Baucau Laga Tequino Mata 778 3,477 4.29 948.56 53.59 0.45 0.13 0.05 0.29 Baucau Quelicai Abafala 210 927 6.36 413.86 56.17 0.37 0.09 0.03 0.24 Baucau Quelicai Abo 131 524 4.08 958.72 64.09 0.34 0.09 0.03 0.29 Baucau Quelicai Afaça 218 1,204 7.20 802.30 58.46 0.33 0.08 0.03 0.24 Baucau Quelicai Baguia 287 1,456 7.99 479.78 52.83 0.42 0.10 0.04 0.24 Baucau Quelicai Bualale 438 2,057 9.06 392.93 50.46 0.44 0.11 0.04 0.23 Baucau Quelicai Guruca 315 1,520 6.27 460.24 54.86 0.38 0.09 0.03 0.24 Baucau Quelicai Locoliu 286 1,285 9.28 433.98 47.54 0.51 0.13 0.05 0.23 Baucau Quelicai Laisorolai De 234 954 10.61 403.07 52.82 0.39 0.09 0.03 0.21 Baixo 65 Area Identification Number of Expenditure Poverty Poverty Poverty Gini Population Rate Gap Severity Ratio District Subdistrict Suco HH Indiv. Min Max Mean Mean Mean Mean Mean Baucau Quelicai Laisorolai De 357 1,306 9.16 441.07 53.09 0.40 0.10 0.03 0.23 Cima Baucau Quelicai Lelalai 185 809 8.21 384.99 51.53 0.44 0.11 0.04 0.23 Baucau Quelicai Letemuno 265 1,379 7.38 386.12 54.49 0.38 0.09 0.03 0.23 Baucau Quelicai Macalaco 237 919 6.08 586.38 52.66 0.42 0.10 0.04 0.23 Baucau Quelicai Maluro 168 763 7.92 417.70 56.14 0.37 0.09 0.03 0.24 Baucau Quelicai Namanei 218 1,095 5.37 617.10 60.86 0.34 0.09 0.03 0.27 Baucau Quelicai Waitame 257 1,241 6.68 374.49 49.96 0.47 0.13 0.05 0.25 Baucau Vemase Caicua 22 77 13.94 448.60 84.90 0.12 0.02 0.01 0.21 Baucau Vemase Loilubo 280 1,282 7.71 523.67 55.15 0.39 0.10 0.04 0.25 Baucau Vemase Ossoala 207 1,067 6.36 784.48 62.30 0.31 0.08 0.03 0.27 Baucau Vemase Ostico 221 1,206 8.53 712.57 62.79 0.29 0.07 0.02 0.25 Baucau Vemase Uaigae 155 758 6.79 409.17 55.98 0.36 0.09 0.03 0.24 Baucau Vemase Uatu-Lari 137 711 7.38 349.41 51.96 0.45 0.12 0.05 0.26 Baucau Vemase Vemase 795 4,542 10.56 568.85 64.85 0.26 0.06 0.02 0.25 Baucau Venilale Bado Ho'O 506 2,664 6.67 382.44 50.67 0.44 0.11 0.04 0.23 Baucau Venilale Baha Mori 328 1,761 8.20 410.60 53.99 0.38 0.09 0.03 0.23 Baucau Venilale Fatulia 499 2,788 7.21 463.12 51.92 0.45 0.12 0.05 0.26 Baucau Venilale Uaiolo 245 1,088 7.67 319.80 47.78 0.51 0.14 0.05 0.24 Baucau Venilale Uailaha 446 2,643 8.89 410.63 50.32 0.44 0.10 0.03 0.22 Baucau Venilale Uataco 436 2,360 8.70 620.56 51.54 0.44 0.10 0.03 0.23 Baucau Venilale Uma Ana Ico 229 1,290 6.86 373.49 50.01 0.49 0.14 0.06 0.26 Baucau Venilale Uma Ana Ulu 432 2,520 9.17 623.98 48.79 0.50 0.13 0.05 0.24 Bobonaro Atabae Aidabaleten 917 5,403 8.59 473.36 55.99 0.37 0.08 0.03 0.23 Bobonaro Atabae Atabae 302 1,679 8.11 259.19 42.34 0.64 0.18 0.07 0.22 Bobonaro Atabae Hataz 376 2,212 8.79 394.50 48.45 0.53 0.14 0.05 0.24 Bobonaro Atabae Rairobo 299 1,623 9.54 292.86 48.34 0.51 0.13 0.05 0.22 Bobonaro Balibo Balibo Vila 743 3,928 7.88 405.09 52.07 0.45 0.11 0.04 0.23 Bobonaro Balibo Batugade 504 2,678 6.94 341.94 50.57 0.47 0.12 0.04 0.23 Bobonaro Balibo Cowa 322 1,707 9.54 286.72 47.92 0.50 0.12 0.04 0.21 Bobonaro Balibo Leohito 608 3,159 5.58 451.19 45.61 0.57 0.16 0.06 0.24 Bobonaro Balibo Leolima 484 2,210 10.22 439.67 44.46 0.59 0.16 0.06 0.22 Bobonaro Balibo Sanirin 382 2,184 8.35 271.73 47.96 0.52 0.14 0.05 0.23 Bobonaro Bobonaro Ai-Assa 387 1,960 9.74 344.77 55.21 0.40 0.10 0.03 0.24 Bobonaro Bobonaro Atu-Aben 144 860 5.94 375.98 45.25 0.59 0.19 0.08 0.26 Bobonaro Bobonaro Bobonaro 348 1,924 9.01 935.30 63.33 0.29 0.06 0.02 0.24 Bobonaro Bobonaro Carabau 358 2,136 5.96 445.56 54.98 0.43 0.11 0.04 0.26 Bobonaro Bobonaro Colimau 232 1,364 8.30 269.75 51.05 0.47 0.13 0.05 0.24 Bobonaro Bobonaro Cotabot 34 228 14.57 225.27 63.34 0.23 0.04 0.01 0.18 Bobonaro Bobonaro Ilat-Laun 273 1,633 7.22 436.80 52.54 0.43 0.11 0.04 0.23 Bobonaro Bobonaro Leber 226 1,225 8.54 297.70 51.13 0.44 0.10 0.03 0.21 Bobonaro Bobonaro Lour 193 961 11.01 273.04 56.53 0.33 0.07 0.02 0.20 Bobonaro Bobonaro Lourba 270 1,421 9.65 798.81 65.42 0.29 0.07 0.02 0.26 Bobonaro Bobonaro Male-Ubu 371 2,216 7.63 367.02 46.87 0.55 0.15 0.06 0.24 Bobonaro Bobonaro Malilait 255 1,373 8.01 383.79 56.36 0.36 0.08 0.03 0.22 Bobonaro Bobonaro Molop 340 1,620 6.00 486.82 55.76 0.42 0.11 0.04 0.27 66 Area Identification Number of Expenditure Poverty Poverty Poverty Gini Population Rate Gap Severity Ratio District Subdistrict Suco HH Indiv. Min Max Mean Mean Mean Mean Mean Bobonaro Bobonaro Oe-Leu 194 1,192 11.22 322.07 54.91 0.38 0.08 0.03 0.22 Bobonaro Bobonaro Sibuni 195 1,148 8.21 422.83 51.63 0.46 0.11 0.04 0.23 Bobonaro Bobonaro Soilesu 226 1,336 8.64 432.69 47.57 0.53 0.15 0.06 0.24 Bobonaro Bobonaro Tapo 150 624 9.32 339.33 50.31 0.47 0.12 0.04 0.21 Bobonaro Bobonaro Tebabui 271 1,496 6.59 390.21 48.20 0.54 0.16 0.06 0.26 Bobonaro Cailaco Atudara 284 1,543 9.05 977.54 46.57 0.56 0.15 0.05 0.22 Bobonaro Cailaco Dau Udo 94 500 9.52 198.35 38.33 0.74 0.22 0.08 0.19 Bobonaro Cailaco Goulolo 223 1,149 5.38 482.52 51.32 0.50 0.13 0.05 0.25 Bobonaro Cailaco Guenu Lai 81 472 10.94 338.75 45.05 0.60 0.16 0.06 0.21 Bobonaro Cailaco Manapa 291 1,688 9.29 317.82 50.98 0.46 0.11 0.04 0.22 Bobonaro Cailaco Meligo 511 3,018 7.76 502.20 51.88 0.44 0.10 0.03 0.22 Bobonaro Cailaco Purugoa 168 931 6.34 971.27 47.28 0.55 0.14 0.05 0.23 Bobonaro Cailaco Raiheu 211 1,104 6.15 280.08 42.48 0.65 0.21 0.09 0.25 Bobonaro Lolotoe Deudet 91 454 8.69 223.04 55.54 0.38 0.09 0.03 0.22 Bobonaro Lolotoe Gildapil 268 1,208 10.69 277.85 56.98 0.34 0.08 0.02 0.22 Bobonaro Lolotoe Guda 186 966 14.79 252.70 53.39 0.37 0.07 0.02 0.18 Bobonaro Lolotoe Lebos 203 966 10.18 443.10 55.49 0.39 0.09 0.03 0.23 Bobonaro Lolotoe Lontas 138 612 11.81 355.74 55.52 0.39 0.09 0.03 0.23 Bobonaro Lolotoe Lupal 216 1,058 7.65 348.45 50.72 0.46 0.11 0.03 0.21 Bobonaro Lolotoe Opa 324 1,537 10.94 337.21 58.67 0.32 0.07 0.02 0.22 Bobonaro Maliana Holsa 863 4,972 9.55 587.94 59.94 0.31 0.07 0.02 0.23 Bobonaro Maliana Lahomea 762 4,523 11.16 588.20 63.25 0.26 0.05 0.01 0.23 Bobonaro Maliana Odomau 728 4,433 6.91 1,036.98 63.84 0.28 0.06 0.02 0.25 Bobonaro Maliana Raifun 256 1,623 11.65 365.52 59.66 0.30 0.06 0.02 0.21 Bobonaro Maliana Ritabou 1,048 6,318 7.13 671.04 54.93 0.40 0.09 0.03 0.24 Bobonaro Maliana Saburai 471 2,268 10.35 267.52 49.58 0.48 0.11 0.04 0.22 Bobonaro Maliana Tapo/Memo 813 4,235 5.04 401.87 50.79 0.47 0.11 0.04 0.23 Covalima Fatululic Fatululic 121 595 7.59 256.49 51.57 0.55 0.16 0.06 0.24 Covalima Fatululic Taroman 293 1,391 9.49 298.70 48.46 0.61 0.16 0.06 0.20 Covalima Fatumean Belulik Leten 346 1,698 9.24 443.30 57.47 0.45 0.11 0.04 0.23 Covalima Fatumean Fatumea 171 787 15.76 202.20 54.46 0.46 0.09 0.03 0.17 Covalima Fatumean Nanu 155 831 9.66 296.02 47.74 0.62 0.17 0.07 0.21 Covalima Forohem Dato Rua 164 803 12.34 209.89 52.17 0.51 0.12 0.04 0.17 Covalima Forohem Dato Tolu 215 1,021 14.21 251.18 52.05 0.52 0.12 0.04 0.18 Covalima Forohem Lactos 148 577 10.00 281.02 54.28 0.49 0.12 0.04 0.21 Covalima Forohem Fohoren 327 1,683 8.64 506.92 48.38 0.61 0.17 0.07 0.22 Covalima Maukatar Belecasac 378 2,368 8.37 488.67 44.43 0.69 0.21 0.08 0.22 Covalima Maukatar Holpilat 347 1,597 4.69 313.72 45.36 0.68 0.21 0.09 0.24 Covalima Maukatar Matai 549 3,013 8.81 314.39 49.14 0.60 0.17 0.06 0.23 Covalima Maukatar Ogues 376 1,915 7.54 562.64 49.08 0.61 0.17 0.07 0.23 Covalima Suai Beco 739 3,756 8.16 773.14 61.92 0.40 0.09 0.03 0.24 Covalima Suai Camenaça 693 3,668 7.71 457.91 61.86 0.36 0.08 0.02 0.22 Covalima Suai Debos 1,946 11,285 7.66 763.60 60.35 0.42 0.10 0.03 0.24 Covalima Suai Labarai 578 3,275 9.55 520.85 47.07 0.64 0.18 0.07 0.22 Covalima Suai Suai Loro 832 3,758 10.21 647.01 64.56 0.35 0.08 0.03 0.24 67 Area Identification Number of Expenditure Poverty Poverty Poverty Gini Population Rate Gap Severity Ratio District Subdistrict Suco HH Indiv. Min Max Mean Mean Mean Mean Mean Covalima Tilomar Foholulic 540 2,267 5.61 412.25 50.75 0.58 0.18 0.07 0.26 Covalima Tilomar Casabauc 376 1,655 8.26 342.97 54.40 0.50 0.14 0.05 0.23 Covalima Tilomar Lalawa 329 1,439 10.49 472.68 55.57 0.49 0.13 0.05 0.23 Covalima Tilomar Maudemo 518 2,524 8.58 349.01 57.59 0.46 0.11 0.04 0.23 Covalima Zumalai Fatuleto 155 766 6.99 610.77 62.06 0.38 0.08 0.03 0.22 Covalima Zumalai Lepo 179 1,302 11.62 385.25 58.20 0.43 0.10 0.03 0.22 Covalima Zumalai Lour 372 2,048 12.64 375.10 55.73 0.47 0.11 0.04 0.21 Covalima Zumalai Mape 59 313 8.50 267.38 48.33 0.60 0.16 0.06 0.20 Covalima Zumalai Raimea 672 3,444 7.29 400.50 46.83 0.65 0.19 0.07 0.23 Covalima Zumalai Tashilin 422 2,275 5.69 625.48 52.76 0.56 0.16 0.06 0.26 Covalima Zumalai Ucecai 40 247 10.25 480.72 68.23 0.34 0.08 0.03 0.23 Covalima Zumalai Zulo 520 2,868 4.89 700.63 51.88 0.55 0.15 0.06 0.23 Dili Atauro Beloi 325 1,678 10.24 359.49 62.49 0.48 0.12 0.04 0.22 Dili Atauro Biceli 418 2,076 16.17 331.91 89.22 0.12 0.02 0.01 0.16 Dili Atauro Macadade 343 1,632 7.39 213.54 43.59 0.80 0.28 0.12 0.21 Dili Atauro Maquili 361 2,062 9.16 261.96 55.56 0.59 0.16 0.06 0.21 Dili Atauro Atauro 301 1,826 7.74 333.82 54.74 0.61 0.18 0.07 0.22 Vila/Maumeta Dili Cristo Rei Balibar 239 1,688 11.21 360.36 54.37 0.62 0.17 0.07 0.22 Dili Cristo Rei Becora 3,160 22,121 7.66 2,129.90 82.57 0.25 0.05 0.02 0.24 Dili Cristo Rei Bidau Santana 929 6,480 10.85 769.02 90.67 0.17 0.03 0.01 0.23 Dili Cristo Rei Camea 1,920 13,481 5.22 1,916.48 70.21 0.43 0.12 0.04 0.28 Dili Cristo Rei Culu Hun 1,009 7,513 1.69 1,596.52 95.72 0.18 0.04 0.01 0.26 Dili Cristo Rei Hera 1,339 8,853 5.93 923.00 61.12 0.54 0.16 0.06 0.26 Dili Cristo Rei Meti Aut 282 2,045 13.35 439.68 86.51 0.18 0.04 0.01 0.20 Dili Dom Aleixo Bairro Pite 5,259 34,777 5.46 1,458.82 87.67 0.23 0.05 0.02 0.26 Dili Dom Aleixo Comoro 12,261 76,387 2.73 2,022.80 90.07 0.22 0.05 0.02 0.26 Dili Dom Aleixo Fatuhada 2,381 14,789 3.04 1,612.80 95.72 0.20 0.04 0.01 0.27 Dili Dom Aleixo Kampung Alor 678 3,531 1.38 16,827.66 130.46 0.21 0.07 0.03 0.40 Dili Metinaro Duyung (Sereia) 633 4,021 8.16 459.29 55.11 0.62 0.19 0.08 0.25 Dili Metinaro Sabuli 213 1,627 8.93 287.38 52.70 0.64 0.20 0.08 0.22 Dili Nain Feto Acadiru Hun 489 3,164 8.26 830.55 100.82 0.12 0.02 0.01 0.23 Dili Nain Feto Bemori 614 4,084 11.27 787.39 100.28 0.09 0.02 0.00 0.20 Dili Nain Feto Bidau Lecidere 162 1,177 19.00 429.15 96.36 0.09 0.01 0.00 0.18 Dili Nain Feto Gricenfor 172 917 17.51 662.94 108.68 0.09 0.02 0.00 0.22 Dili Nain Feto Lahane Oriental 1,982 13,606 3.79 1,395.85 82.91 0.29 0.07 0.02 0.27 Dili Nain Feto Santa Cruz 1,480 9,701 11.60 731.97 93.60 0.13 0.02 0.01 0.21 Dili Vera Cruz Caicoli 934 5,053 8.73 695.50 87.43 0.22 0.05 0.02 0.25 Dili Vera Cruz Colmera 313 1,839 5.24 1,970.74 107.07 0.14 0.03 0.01 0.26 Dili Vera Cruz Dare 437 2,994 5.90 1,938.99 65.27 0.50 0.13 0.05 0.26 Dili Vera Cruz Lahane 625 5,152 5.07 436.65 74.31 0.31 0.06 0.02 0.22 Ocidental Dili Vera Cruz Mascarenhas 896 5,827 8.86 859.03 88.20 0.21 0.04 0.01 0.24 Dili Vera Cruz Motael 766 4,962 12.69 657.33 93.61 0.15 0.03 0.01 0.22 Dili Vera Cruz Vila Verde 1,564 10,311 6.56 1,073.99 99.59 0.15 0.03 0.01 0.25 Ermera Atsabe Atara 433 2,737 6.83 1,086.39 50.68 0.38 0.10 0.04 0.27 Ermera Atsabe Baboi Craic 343 1,965 9.11 343.69 42.84 0.49 0.12 0.04 0.23 68 Area Identification Number of Expenditure Poverty Poverty Poverty Gini Population Rate Gap Severity Ratio District Subdistrict Suco HH Indiv. Min Max Mean Mean Mean Mean Mean Ermera Atsabe Beboi Leten 185 1,106 3.83 796.79 48.98 0.44 0.11 0.04 0.28 Ermera Atsabe Batumanu 167 1,034 7.42 265.76 38.24 0.60 0.18 0.07 0.24 Ermera Atsabe Lasaun 381 2,041 3.49 546.14 44.17 0.49 0.14 0.06 0.27 Ermera Atsabe Laclo 282 1,624 9.78 300.43 45.26 0.43 0.10 0.03 0.22 Ermera Atsabe Laubono 179 1,061 7.26 281.91 46.02 0.42 0.10 0.04 0.23 Ermera Atsabe Leimea Leten 392 2,234 6.28 742.85 47.11 0.42 0.11 0.04 0.26 Ermera Atsabe Atadame/Malabe 256 1,566 8.11 356.07 48.06 0.38 0.08 0.03 0.23 Ermera Atsabe Obulo 188 1,022 7.92 221.65 46.75 0.40 0.10 0.04 0.23 Ermera Atsabe Paramin 287 1,683 7.25 334.58 44.94 0.47 0.12 0.05 0.25 Ermera Atsabe Tiarlelo 87 490 10.10 233.35 46.99 0.37 0.08 0.03 0.20 Ermera Ermera Estado 521 3,022 7.29 398.38 41.70 0.52 0.14 0.05 0.24 Ermera Ermera Humboe 376 2,305 9.17 337.86 45.82 0.40 0.09 0.03 0.22 Ermera Ermera Lauala 503 3,150 7.28 271.14 40.42 0.54 0.14 0.05 0.23 Ermera Ermera Leguimea 469 2,828 6.42 294.49 41.16 0.53 0.15 0.06 0.25 Ermera Ermera Mirtutu 330 1,973 6.86 331.98 45.09 0.45 0.11 0.04 0.24 Ermera Ermera Poetete 1,356 8,828 7.06 439.40 45.61 0.43 0.10 0.04 0.24 Ermera Ermera Ponilala 533 3,372 8.63 319.63 43.42 0.47 0.11 0.04 0.23 Ermera Ermera Raimerhei 377 2,244 5.50 686.65 41.56 0.54 0.15 0.06 0.26 Ermera Ermera Riheu 305 2,032 9.60 409.99 49.05 0.36 0.08 0.02 0.23 Ermera Ermera Talimoro 902 6,340 8.17 494.93 53.71 0.29 0.06 0.02 0.24 Ermera Hatolia Asulau 343 2,060 8.38 326.21 42.99 0.48 0.12 0.04 0.22 Ermera Hatolia Ailelo 384 2,455 7.95 243.45 43.16 0.48 0.12 0.04 0.23 Ermera Hatolia Coliate-Leotelo 665 3,937 8.08 347.49 42.55 0.50 0.13 0.05 0.24 Ermera Hatolia Fatuessi 756 4,754 6.97 300.59 42.78 0.49 0.13 0.04 0.24 Ermera Hatolia Fatubolu 701 4,735 5.89 254.99 36.79 0.63 0.18 0.07 0.23 Ermera Hatolia Hatolia 510 3,037 8.69 489.57 46.45 0.40 0.09 0.03 0.22 Ermera Hatolia Leimeacraic 247 1,272 8.57 327.45 51.53 0.32 0.07 0.02 0.24 Ermera Hatolia Lemia 124 608 7.43 269.55 43.20 0.47 0.11 0.04 0.22 Sorimbalu Ermera Hatolia Lissapat 560 3,676 6.73 387.80 42.24 0.51 0.13 0.05 0.24 Ermera Hatolia Manusae 713 4,524 5.12 418.82 41.95 0.52 0.13 0.05 0.24 Ermera Hatolia Mau-Ubu 267 1,690 8.50 347.08 42.93 0.48 0.12 0.04 0.23 Ermera Hatolia Samara 101 545 6.73 566.89 63.87 0.19 0.04 0.01 0.24 Ermera Hatolia Urahou 565 3,327 6.00 264.12 44.03 0.47 0.12 0.04 0.24 Ermera Letefoho Catrai Leten 366 1,935 6.95 334.27 46.40 0.43 0.11 0.04 0.25 Ermera Letefoho Ducurai 791 4,622 6.86 380.25 38.46 0.60 0.17 0.06 0.23 Ermera Letefoho Eraulo 363 2,098 6.20 247.12 41.75 0.51 0.12 0.04 0.22 Ermera Letefoho Goulolo 230 1,414 6.89 323.56 42.01 0.53 0.15 0.06 0.26 Ermera Letefoho Hatugau 314 1,675 7.50 383.18 46.78 0.39 0.09 0.03 0.22 Ermera Letefoho Haupu 903 5,009 7.28 443.10 42.72 0.49 0.12 0.04 0.23 Ermera Letefoho Catrai-Craic 470 2,659 6.30 513.42 42.01 0.52 0.14 0.05 0.25 Ermera Letefoho Lauana 496 2,686 5.88 854.17 48.86 0.39 0.09 0.03 0.26 Ermera Railaco Deleco 85 503 6.93 208.21 36.24 0.65 0.20 0.08 0.22 Ermera Railaco Fatuquero 477 2,234 4.05 3,074.85 68.87 0.24 0.06 0.02 0.32 Ermera Railaco Lihu 290 1,918 7.92 274.16 45.51 0.44 0.11 0.04 0.24 Ermera Railaco Matata 229 1,354 5.68 360.16 44.86 0.47 0.12 0.05 0.25 69 Area Identification Number of Expenditure Poverty Poverty Poverty Gini Population Rate Gap Severity Ratio District Subdistrict Suco HH Indiv. Min Max Mean Mean Mean Mean Mean Ermera Railaco Railaco Craic 193 1,435 9.61 580.43 47.70 0.41 0.09 0.03 0.24 Ermera Railaco Railaco Leten 205 1,379 5.39 210.61 39.15 0.57 0.16 0.06 0.23 Ermera Railaco Samalete 174 1,108 10.86 167.91 41.48 0.49 0.12 0.04 0.18 Ermera Railaco Tara╟O 93 558 6.95 532.87 43.59 0.49 0.12 0.04 0.24 Ermera Railaco Tocoluli 204 1,315 8.87 476.06 47.85 0.37 0.08 0.03 0.22 Lautém Iliomar Ailebere 165 808 9.08 324.92 47.01 0.37 0.07 0.02 0.21 Lautém Iliomar Cainliu 235 1,185 4.00 1,143.87 52.91 0.33 0.08 0.03 0.26 Lautém Iliomar Fuat 96 575 8.60 528.95 57.99 0.29 0.07 0.02 0.27 Lautém Iliomar Iliomar I 416 1,902 8.76 436.66 50.24 0.34 0.07 0.02 0.24 Lautém Iliomar Iliomar Ii 239 1,253 6.51 447.61 50.37 0.33 0.07 0.02 0.23 Lautém Iliomar Tirilolo 337 1,726 4.71 595.25 50.84 0.39 0.10 0.04 0.28 Lautém Lautém Baduro 174 977 6.89 289.46 48.06 0.39 0.09 0.03 0.24 Lautém Lautém Com 482 2,348 3.81 1,061.20 53.70 0.34 0.08 0.03 0.28 Lautém Lautém Daudare 330 1,677 6.87 360.51 58.00 0.28 0.07 0.02 0.28 Lautém Lautém Euquisi 187 931 1.44 2,358.93 90.80 0.08 0.02 0.01 0.28 Lautém Lautém Ililai 127 998 0.16 139,204.73 329.10 0.20 0.10 0.06 0.61 Lautém Lautém Maina I 245 1,362 6.94 613.29 50.15 0.36 0.08 0.03 0.25 Lautém Lautém Maina Ii 401 1,951 5.99 427.40 54.27 0.30 0.07 0.02 0.26 Lautém Lautém Pairara 371 2,164 2.68 4,080.85 68.92 0.22 0.05 0.02 0.32 Lautém Lautém Parlamento 426 2,342 6.33 447.79 54.28 0.30 0.07 0.02 0.26 Lautém Lautém Serelau 214 1,234 5.47 721.92 52.42 0.37 0.10 0.04 0.29 Lautém Lospalos Bauro 471 2,432 5.29 868.36 59.00 0.27 0.06 0.02 0.28 Lautém Lospalos Cacavem 205 974 5.08 623.93 60.92 0.31 0.09 0.04 0.32 Lautém Lospalos Fuiloro 2,683 16,466 4.69 844.94 69.99 0.13 0.02 0.01 0.26 Lautém Lospalos Home 328 1,933 8.21 640.13 58.52 0.23 0.05 0.02 0.25 Lautém Lospalos Leuro 180 812 7.00 403.83 59.07 0.26 0.06 0.02 0.27 Lautém Lospalos Lore I 540 2,582 5.62 356.86 46.70 0.41 0.10 0.04 0.25 Lautém Lospalos Lore Ii 163 811 10.47 300.15 49.19 0.33 0.07 0.02 0.21 Lautém Lospalos Muapitine 350 1,763 6.73 632.90 60.17 0.24 0.06 0.02 0.27 Lautém Lospalos Raça 226 1,162 10.48 404.90 63.86 0.16 0.03 0.01 0.23 Lautém Lospalos Souro 445 1,987 4.98 895.40 49.39 0.40 0.11 0.04 0.27 Lautém Luro Afabubu 78 439 11.55 318.16 62.52 0.19 0.04 0.01 0.23 Lautém Luro Baricafa 190 1,013 6.28 393.87 50.81 0.39 0.10 0.04 0.28 Lautém Luro Cotamutu 346 1,983 6.22 643.77 49.57 0.38 0.09 0.03 0.26 Lautém Luro Lacawa 110 645 4.66 411.14 45.66 0.49 0.16 0.07 0.31 Lautém Luro Luro 415 2,233 6.28 645.92 53.83 0.36 0.09 0.04 0.29 Lautém Luro Wairoce 175 811 5.31 374.63 49.73 0.38 0.10 0.04 0.26 Lautém Tutuala Mehara 443 2,262 6.12 1,380.45 73.43 0.17 0.04 0.01 0.30 Lautém Tutuala Tutuala 256 1,244 8.98 415.91 54.74 0.27 0.06 0.02 0.24 Liquiça Bazartete Fahilebo 200 1,190 9.45 248.33 44.36 0.53 0.14 0.05 0.22 Liquiça Bazartete Fatumasi 273 1,544 10.85 438.35 61.99 0.24 0.05 0.01 0.23 Liquiça Bazartete Lauhata 533 3,620 7.39 359.68 44.70 0.52 0.13 0.05 0.23 Liquiça Bazartete Leorema 986 5,405 6.74 377.52 47.33 0.48 0.13 0.05 0.25 Liquiça Bazartete Maumeta 677 4,306 6.98 444.38 50.13 0.41 0.10 0.03 0.22 Liquiça Bazartete Metagou 280 1,677 12.08 275.64 42.05 0.58 0.14 0.05 0.19 70 Area Identification Number of Expenditure Poverty Poverty Poverty Gini Population Rate Gap Severity Ratio District Subdistrict Suco HH Indiv. Min Max Mean Mean Mean Mean Mean Liquiça Bazartete Motaulun 390 2,337 7.23 279.02 51.99 0.38 0.09 0.03 0.23 Liquiça Bazartete Tibar 702 4,171 9.19 469.85 60.76 0.26 0.05 0.02 0.23 Liquiça Bazartete Ulmera 544 3,523 5.84 340.03 44.40 0.53 0.15 0.05 0.23 Liquiça Liquiça Açumano 328 1,911 11.61 321.67 47.59 0.44 0.09 0.03 0.19 Liquiça Liquiça Darulete 283 1,868 8.20 240.79 42.70 0.58 0.16 0.06 0.23 Liquiça Liquiça Dato 1,407 9,246 8.93 339.98 58.14 0.28 0.06 0.02 0.23 Liquiça Liquiça Hatuquessi 506 2,899 6.17 274.03 39.19 0.65 0.19 0.07 0.22 Liquiça Liquiça Leoteala 424 2,543 8.41 227.39 41.17 0.60 0.16 0.06 0.21 Liquiça Liquiça Loidahar 477 2,795 8.28 349.77 43.35 0.56 0.15 0.06 0.23 Liquiça Liquiça Luculai 134 793 10.77 211.83 44.20 0.52 0.13 0.04 0.20 Liquiça Maubara Gugleur 691 3,693 8.58 257.20 43.69 0.53 0.13 0.04 0.20 Liquiça Maubara Guiço 320 1,983 9.49 203.25 35.94 0.73 0.22 0.08 0.19 Liquiça Maubara Lissadila 749 4,559 7.47 274.47 39.73 0.63 0.18 0.07 0.22 Liquiça Maubara Maubaralissa 354 1,969 8.77 250.63 41.21 0.60 0.15 0.05 0.20 Liquiça Maubara Vatuboro 493 2,791 8.28 222.63 45.39 0.50 0.13 0.04 0.22 Liquiça Maubara Vatuvou 693 4,175 8.75 305.58 43.38 0.55 0.14 0.05 0.22 Liquiça Maubara Vaviquinia 441 2,673 6.62 233.31 42.45 0.57 0.15 0.05 0.21 Manatuto Barique/Natarbora Abat Oan 207 1,357 10.94 222.72 49.47 0.55 0.14 0.05 0.20 Manatuto Barique/Natarbora Aubeon 240 1,258 7.28 244.28 47.17 0.61 0.17 0.07 0.23 Manatuto Barique/Natarbora Barique 86 438 9.02 234.52 42.02 0.71 0.22 0.09 0.21 Manatuto Barique/Natarbora Manehat 134 668 8.35 341.32 50.71 0.54 0.15 0.06 0.23 Manatuto Barique/Natarbora Uma Boco 280 1,553 9.11 344.95 52.88 0.50 0.13 0.05 0.23 Manatuto Laclo Hohorai 149 969 7.82 233.54 43.36 0.67 0.20 0.08 0.21 Manatuto Laclo Lacumesac 349 2,155 5.16 396.71 49.40 0.58 0.17 0.07 0.25 Manatuto Laclo Umacaduac 540 3,433 7.67 322.85 53.66 0.50 0.14 0.05 0.25 Manatuto Laclo Uma Naruc 199 1,183 9.67 420.91 62.83 0.36 0.09 0.03 0.25 Manatuto Laclubar Batara 345 2,367 9.97 299.18 52.10 0.51 0.13 0.05 0.22 Manatuto Laclubar Fatumaquerec 151 863 7.36 299.84 54.91 0.45 0.11 0.04 0.21 Manatuto Laclubar Funar 167 1,131 9.69 581.79 54.63 0.48 0.11 0.04 0.22 Manatuto Laclubar Manelima 313 2,191 7.74 354.20 46.72 0.61 0.17 0.07 0.22 Manatuto Laclubar Orlalan 686 4,839 8.43 431.14 49.94 0.55 0.14 0.05 0.22 Manatuto Laclubar Sanana'In 112 618 3.46 598.19 62.87 0.43 0.14 0.06 0.30 Manatuto Laleia Cairui 369 1,847 6.47 594.06 62.76 0.38 0.10 0.04 0.26 Manatuto Laleia Haturalan 190 995 13.88 332.34 70.46 0.21 0.04 0.01 0.20 Manatuto Laleia Lifau 160 847 14.45 309.72 76.10 0.15 0.03 0.01 0.19 Manatuto Manatuto Ailili 264 1,553 10.26 575.59 91.64 0.10 0.02 0.01 0.22 Manatuto Manatuto Aiteas 668 4,020 9.16 1,485.62 79.53 0.16 0.03 0.01 0.23 Manatuto Manatuto Cribas 358 2,435 6.88 396.00 49.40 0.57 0.17 0.07 0.25 Manatuto Manatuto Iliheu 295 1,744 6.49 496.49 58.54 0.43 0.12 0.04 0.25 Manatuto Manatuto Ma'Abat 117 750 15.82 675.23 80.87 0.14 0.02 0.01 0.20 Manatuto Manatuto Sau 598 3,890 10.11 501.60 76.84 0.16 0.03 0.01 0.22 Manatuto Soibada Fatumacerec 130 864 10.84 230.75 47.87 0.58 0.16 0.06 0.21 Manatuto Soibada Leo Hat 159 1,063 5.81 638.34 55.19 0.46 0.13 0.05 0.24 Manatuto Soibada Manlala 60 451 12.64 205.16 55.49 0.42 0.09 0.03 0.18 Manatuto Soibada Manufahi 83 537 9.11 246.26 60.79 0.32 0.07 0.02 0.18 71 Area Identification Number of Expenditure Poverty Poverty Poverty Gini Population Rate Gap Severity Ratio District Subdistrict Suco HH Indiv. Min Max Mean Mean Mean Mean Mean Manatuto Soibada Samoro 54 329 11.82 330.71 63.91 0.31 0.07 0.02 0.20 Manufahi Alas Aituha 131 734 12.62 200.63 48.76 0.46 0.11 0.04 0.19 Manufahi Alas Dotic 340 1,922 4.70 475.05 55.10 0.41 0.11 0.04 0.26 Manufahi Alas Mahaquidan 382 1,982 7.46 346.32 49.24 0.49 0.13 0.05 0.24 Manufahi Alas Taitudac 301 1,788 5.37 332.50 50.44 0.47 0.13 0.05 0.24 Manufahi Alas Uma Berloic 251 1,457 4.49 449.48 51.45 0.49 0.15 0.06 0.28 Manufahi Fatuberliu Bubususo 143 701 14.29 198.97 50.21 0.41 0.08 0.02 0.17 Manufahi Fatuberliu Caicasa 199 1,072 8.91 459.52 55.09 0.35 0.07 0.02 0.21 Manufahi Fatuberliu Clacuc 484 3,102 11.58 925.01 69.69 0.20 0.04 0.01 0.24 Manufahi Fatuberliu Fahinehan 223 1,328 12.34 304.70 57.55 0.28 0.05 0.01 0.19 Manufahi Fatuberliu Fatucahi 203 1,213 6.37 337.44 53.41 0.40 0.10 0.03 0.22 Manufahi Same Babulu 708 4,468 9.78 363.00 54.75 0.37 0.08 0.03 0.22 Manufahi Same Betano 1,054 5,753 7.07 577.87 59.97 0.33 0.08 0.03 0.26 Manufahi Same Daisua 499 2,719 8.28 318.50 49.63 0.47 0.11 0.04 0.22 Manufahi Same Grotu 163 810 8.37 266.82 47.10 0.50 0.12 0.04 0.19 Manufahi Same Holarua 1,167 6,871 9.78 376.62 54.41 0.39 0.09 0.03 0.23 Manufahi Same Letefoho 1,199 7,498 5.28 437.43 59.94 0.29 0.06 0.02 0.22 Manufahi Same Rotuto 173 848 10.32 251.96 44.10 0.58 0.15 0.05 0.20 Manufahi Same Tutuluro 282 1,631 9.24 263.02 51.66 0.40 0.08 0.03 0.19 Manufahi Turiscai Aitemua 107 817 8.94 223.93 44.68 0.56 0.16 0.06 0.22 Manufahi Turiscai Beremana 110 808 8.66 214.32 48.45 0.48 0.12 0.04 0.21 Manufahi Turiscai Caimauc 176 1,122 8.61 624.48 49.41 0.50 0.13 0.05 0.23 Manufahi Turiscai Fatucalo 59 393 8.15 188.48 43.42 0.60 0.18 0.08 0.24 Manufahi Turiscai Foholau 32 255 12.39 268.08 47.77 0.48 0.11 0.03 0.16 Manufahi Turiscai Lesuata 57 337 8.14 264.28 51.62 0.44 0.12 0.05 0.24 Manufahi Turiscai Liurai 92 599 8.83 309.23 47.62 0.50 0.13 0.05 0.22 Manufahi Turiscai Manumera 229 1,584 6.45 366.89 46.22 0.55 0.16 0.07 0.25 Manufahi Turiscai Matorec 64 457 11.32 181.35 48.72 0.47 0.12 0.04 0.21 Manufahi Turiscai Mindelo 85 593 11.98 217.80 47.93 0.48 0.11 0.04 0.19 Manufahi Turiscai Orana 110 753 10.19 204.96 48.23 0.48 0.11 0.04 0.20 Oecussi Nitibe Banafi 419 1,762 7.73 496.25 53.51 0.56 0.17 0.07 0.25 Oecussi Nitibe Bene-Ufe 580 2,735 6.52 446.77 49.88 0.62 0.20 0.08 0.26 Oecussi Nitibe Lela-Ufe 795 3,745 4.74 460.11 49.41 0.63 0.19 0.07 0.24 Oecussi Nitibe Suni-Ufe 447 1,815 6.60 410.38 55.02 0.54 0.16 0.07 0.26 Oecussi Nitibe Usi-Taco 462 2,139 4.36 419.11 50.90 0.60 0.20 0.09 0.27 Oecussi Oesilo Bobometo 1,585 7,287 7.69 516.90 52.46 0.56 0.16 0.06 0.23 Oecussi Oesilo Usi-Taqueno 216 833 6.38 317.21 52.34 0.58 0.18 0.07 0.26 Oecussi Oesilo Usi-Tacae 737 3,340 5.95 662.73 50.13 0.60 0.18 0.07 0.23 Oecussi Pante Macasar Bobocase 505 2,693 6.09 611.66 51.81 0.59 0.19 0.08 0.27 Oecussi Pante Macasar Costa 2,660 14,261 5.96 527.91 58.38 0.49 0.14 0.05 0.26 Oecussi Pante Macasar Cunha 918 4,493 5.73 367.39 50.00 0.61 0.19 0.08 0.25 Oecussi Pante Macasar Lalisuc 449 2,182 8.02 836.20 49.87 0.62 0.19 0.08 0.24 Oecussi Pante Macasar Lifau 469 2,505 6.45 377.70 53.34 0.56 0.17 0.07 0.25 Oecussi Pante Macasar Naimeco 923 4,809 6.06 440.65 46.31 0.68 0.22 0.10 0.25 Oecussi Pante Macasar Nipani 209 1,114 6.75 346.90 53.33 0.56 0.18 0.07 0.26 72 Area Identification Number of Expenditure Poverty Poverty Poverty Gini Population Rate Gap Severity Ratio District Subdistrict Suco HH Indiv. Min Max Mean Mean Mean Mean Mean Oecussi Pante Macasar Taiboco 1,151 5,124 5.48 304.85 45.27 0.69 0.23 0.10 0.25 Oecussi Passabe Abani 1,455 6,323 5.69 466.72 44.66 0.70 0.24 0.11 0.25 Oecussi Passabe Malelat 362 1,556 6.70 336.30 47.62 0.65 0.21 0.09 0.24 Viqueque Lacluta Ahic 223 1,247 8.89 538.28 57.36 0.38 0.10 0.04 0.25 Viqueque Lacluta Dilor 462 2,804 3.24 1,704.24 58.61 0.45 0.15 0.07 0.33 Viqueque Lacluta Laline 166 943 5.97 265.89 46.39 0.57 0.18 0.08 0.26 Viqueque Lacluta Uma Tolu 386 1,795 4.00 517.52 52.40 0.49 0.15 0.06 0.28 Viqueque Ossu Builale 222 1,137 8.68 406.77 52.05 0.46 0.12 0.05 0.24 Viqueque Ossu Liaruca 241 1,008 6.11 2,125.32 61.16 0.40 0.12 0.05 0.32 Viqueque Ossu Loi-Huno 280 1,272 5.48 492.12 55.85 0.41 0.11 0.04 0.26 Viqueque Ossu Nahareca 469 2,030 6.66 581.17 54.98 0.39 0.09 0.03 0.23 Viqueque Ossu Ossorua 495 2,322 6.66 537.18 50.22 0.50 0.14 0.05 0.25 Viqueque Ossu Ossu De Cima 890 4,263 5.63 697.36 58.22 0.38 0.10 0.03 0.26 Viqueque Ossu Uabubo 577 2,711 6.23 386.77 52.27 0.45 0.12 0.04 0.24 Viqueque Ossu Uaigia 241 1,153 8.42 362.23 48.38 0.53 0.14 0.05 0.23 Viqueque Ossu Uaibobo 301 1,259 6.07 341.05 46.75 0.55 0.16 0.06 0.24 Viqueque Watulari Afaloicai 757 3,980 7.47 681.23 57.25 0.39 0.09 0.03 0.25 Viqueque Watulari Babulo 455 2,187 8.95 414.36 54.95 0.41 0.10 0.03 0.24 Viqueque Watulari Macadique 1,115 5,391 3.21 1,243.12 62.96 0.36 0.09 0.03 0.29 Viqueque Watulari Matahoi 923 4,492 8.11 571.45 57.33 0.37 0.09 0.03 0.25 Viqueque Watulari Uaitame 278 1,339 7.25 471.79 58.52 0.37 0.10 0.04 0.26 Viqueque Watulari Vessoru 330 1,519 4.80 730.31 50.99 0.50 0.15 0.06 0.27 Viqueque Uatucarbau Afaloicai 308 1,349 6.06 1,172.39 50.36 0.49 0.14 0.05 0.25 Viqueque Uatucarbau Bahatata 134 601 7.25 448.39 53.72 0.42 0.10 0.04 0.23 Viqueque Uatucarbau Irabin De Baixo 461 2,588 8.68 639.47 51.37 0.47 0.12 0.04 0.24 Viqueque Uatucarbau Irabin De Cima 139 660 8.96 643.03 71.34 0.25 0.06 0.02 0.26 Viqueque Uatucarbau Loi Ulu 126 652 8.89 524.63 60.49 0.36 0.09 0.03 0.26 Viqueque Uatucarbau Uani Uma 285 1,543 8.10 426.22 52.18 0.47 0.13 0.05 0.26 Viqueque Viqueque Bahalarauain 652 2,987 5.13 733.25 56.49 0.43 0.12 0.05 0.29 Viqueque Viqueque Bibileo 668 3,156 5.57 766.65 51.21 0.50 0.15 0.06 0.28 Viqueque Viqueque Caraubalo 1,131 6,572 6.99 504.05 66.44 0.26 0.06 0.02 0.25 Viqueque Viqueque Watu Dere 130 645 7.84 383.82 65.95 0.29 0.07 0.03 0.26 Viqueque Viqueque Luca 488 2,375 2.14 1,439.74 58.55 0.43 0.13 0.05 0.31 Viqueque Viqueque Maluro 181 816 7.46 374.12 59.57 0.35 0.09 0.03 0.25 Viqueque Viqueque Uai Mori 242 1,142 4.97 600.94 52.78 0.48 0.15 0.06 0.29 Viqueque Viqueque Uma Quic 402 1,981 5.05 763.72 53.99 0.47 0.14 0.06 0.29 Viqueque Viqueque Uma Uain Craic 782 4,454 6.79 683.94 62.73 0.35 0.09 0.03 0.28 Viqueque Viqueque Uma Uain Leten 356 1,580 4.74 442.39 53.88 0.46 0.14 0.06 0.28 73 Table 28: Suco-level Predicted Gender Indicators from 2014 TLSLS Labour Education Health DM DM DV DV DV Area Identification Force Gap Gap Index HCount Index HCount Count Gap District Subdistrict Suco Mean Mean Mean Mean Mean Mean Mean Mean Aileu Aileu Vila Aisirimou 0.53 0.36 0.53 -0.35 0.61 -0.47 0.36 1.55 Aileu Aileu Vila Bandudato 0.45 0.41 0.52 0.03 0.49 0.5 0.61 2.79 Aileu Aileu Vila Fahiria 0.48 0.42 0.53 0.05 0.49 0.22 0.54 2.51 Aileu Aileu Vila Fatubosa 0.48 0.4 0.51 0.17 0.45 0.31 0.56 2.35 Aileu Aileu Vila Hoholau 0.44 0.45 0.46 0.03 0.5 0.14 0.53 2.4 Aileu Aileu Vila Lahae 0.46 0.41 0.52 0.1 0.47 0.06 0.51 2.29 Aileu Aileu Vila Lausi 0.48 0.44 0.52 0.14 0.46 -0.11 0.46 2.22 Aileu Aileu Vila Saboria 0.48 0.42 0.57 0.18 0.45 0.1 0.51 2.18 Aileu Aileu Vila Seloi Craic 0.45 0.36 0.51 0.15 0.47 -0.07 0.48 2.24 Aileu Aileu Vila Seloi Malere 0.55 0.37 0.5 -0.06 0.51 -0.21 0.43 1.81 Aileu Aileu Vila Suco Liurai 0.45 0.41 0.47 0.09 0.47 0.2 0.54 2.45 Aileu Laulara Cotolau 0.52 0.41 0.53 0.12 0.46 0.03 0.5 2.17 Aileu Laulara Fatisi 0.45 0.44 0.51 0.13 0.47 -0.26 0.43 1.83 Aileu Laulara Madabeno 0.45 0.43 0.49 0.03 0.49 0.24 0.55 2.38 Aileu Laulara Talitu 0.46 0.46 0.53 0.01 0.5 -0.1 0.46 2.02 Aileu Laulara Tohumeta 0.45 0.43 0.57 -0.25 0.57 -0.41 0.4 1.7 Aileu Liquidoe Acubilitoho 0.45 0.42 0.47 -0.39 0.63 0.09 0.52 2.33 Aileu Liquidoe Bereleu 0.47 0.4 0.5 -0.33 0.61 -0.02 0.49 2.07 Aileu Liquidoe Betulau 0.4 0.43 0.49 -0.13 0.54 0.37 0.57 2.75 Aileu Liquidoe Fahisoi 0.47 0.45 0.51 -0.34 0.62 -0.11 0.47 1.92 Aileu Liquidoe Faturilau 0.41 0.43 0.54 -0.11 0.53 -0.07 0.48 2.08 Aileu Liquidoe Manucasa 0.4 0.42 0.52 0.1 0.46 -0.06 0.48 2.4 Aileu Liquidoe Namoleso 0.45 0.4 0.51 0.04 0.49 0.24 0.55 2.42 Aileu Remexio Acumau 0.49 0.38 0.51 0.08 0.48 -0.39 0.38 1.77 Aileu Remexio Fadabloco 0.43 0.43 0.47 0.14 0.47 0.33 0.57 2.76 Aileu Remexio Fahisoi 0.44 0.4 0.49 -0.08 0.53 0.14 0.53 2.49 Aileu Remexio Faturasa 0.43 0.4 0.46 0.45 0.38 0.15 0.54 2.65 Aileu Remexio Hautoho 0.47 0.42 0.48 -0.12 0.53 0.23 0.55 2.57 Aileu Remexio Maumeta 0.47 0.41 0.49 -0.07 0.52 -0.45 0.36 1.64 Aileu Remexio Suco Liurai 0.52 0.36 0.44 0.32 0.42 -0.74 0.3 1.01 Aileu Remexio Tulataqueo 0.41 0.42 0.44 0 0.5 0.37 0.59 2.59 Ainaro Ainaro Ainaro 0.56 0.39 0.48 -0.27 0.58 -0.52 0.34 1.42 Ainaro Ainaro Cassa 0.49 0.42 0.48 0.2 0.45 0.52 0.61 2.75 Ainaro Ainaro Manutasi 0.57 0.4 0.51 0.21 0.45 0.35 0.56 2.1 Ainaro Ainaro Mau-Nuno 0.54 0.46 0.51 0.11 0.47 0.14 0.53 2.2 Ainaro Ainaro Mau-Ulo 0.51 0.41 0.54 0.1 0.48 -0.52 0.36 1.55 Ainaro Ainaro Soro 0.56 0.41 0.49 -0.01 0.5 0.12 0.52 2.03 Ainaro Ainaro Suro-Craik 0.43 0.44 0.54 -0.22 0.59 -0.04 0.49 1.78 Ainaro Hatu-Udo Foho-Ai-Lico 0.49 0.46 0.5 0.21 0.44 0.32 0.56 2.47 Ainaro Hatu-Udo Leolima 0.5 0.46 0.54 0.2 0.45 0.12 0.52 2.28 Ainaro Hatu-Builico Mau-Chiga 0.44 0.45 0.5 -0.33 0.61 0.31 0.57 2.43 Ainaro Hatu-Builico Mulo 0.46 0.45 0.5 -0.36 0.63 0.25 0.55 2.27 Ainaro Hatu-Builico Nuno-Mogue 0.5 0.42 0.52 -0.35 0.62 0.25 0.55 2.36 74 Labour Education Health DM DM DV DV DV Area Identification Force Gap Gap Index HCount Index HCount Count Gap District Subdistrict Suco Mean Mean Mean Mean Mean Mean Mean Mean Ainaro Maubisse Aitutu 0.46 0.45 0.46 -0.17 0.55 0.41 0.58 2.55 Ainaro Maubisse Edi 0.44 0.44 0.43 -0.24 0.58 0.33 0.57 2.41 Ainaro Maubisse Fatu-Besi 0.46 0.42 0.46 0.06 0.49 0.3 0.56 2.5 Ainaro Maubisse Horai-Quic 0.43 0.42 0.52 -0.32 0.63 0.41 0.58 2.56 Ainaro Maubisse Suco Liurai 0.53 0.47 0.54 -0.26 0.59 0.59 0.62 2.7 Ainaro Maubisse Manelobas 0.41 0.44 0.42 -0.33 0.62 -0.12 0.47 1.81 Ainaro Maubisse Manetu 0.43 0.45 0.44 -0.31 0.62 0.35 0.58 2.46 Ainaro Maubisse Maubisse 0.49 0.42 0.5 -0.14 0.54 0.27 0.55 2.33 Ainaro Maubisse Maulau 0.44 0.43 0.46 -0.39 0.65 0.3 0.56 2.27 Baucau Baguia Afaloicai 0.48 0.37 0.48 0.45 0.38 -0.68 0.32 1.38 Baucau Baguia Alaua Craic 0.46 0.44 0.47 -0.48 0.67 -0.4 0.39 1.46 Baucau Baguia Alaua Leten 0.48 0.31 0.52 -0.53 0.69 -0.39 0.39 1.41 Baucau Baguia Defa Uassi 0.44 0.42 0.47 -0.48 0.7 -0.37 0.4 1.44 Baucau Baguia Haeconi 0.46 0.4 0.51 -0.39 0.64 -0.31 0.42 1.59 Baucau Baguia Lari Sula 0.46 0.46 0.44 -0.49 0.7 -0.32 0.41 1.52 Baucau Baguia Lavateri 0.48 0.41 0.52 -0.33 0.61 -0.02 0.49 1.8 Baucau Baguia Ossu-Huna 0.47 0.4 0.48 0.06 0.48 -0.08 0.48 2.02 Baucau Baguia Samalari 0.46 0.44 0.45 -0.38 0.64 0.17 0.54 2.01 Baucau Baguia Uacala 0.47 0.45 0.44 -0.36 0.64 0.12 0.52 1.78 Baucau Baucau Bahu 0.54 0.39 0.51 0.29 0.43 -0.16 0.44 2 Baucau Baucau Bucoli 0.5 0.41 0.48 0.12 0.47 -0.1 0.47 1.94 Baucau Baucau Buibau 0.51 0.4 0.49 0.68 0.32 -0.76 0.24 1.08 Baucau Baucau Buruma 0.53 0.41 0.49 0.09 0.48 -0.35 0.35 1.32 Baucau Baucau Caibada 0.52 0.37 0.47 0.3 0.43 -0.69 0.24 1 Baucau Baucau Gariuai 0.47 0.38 0.48 -0.05 0.52 -0.43 0.34 1.49 Baucau Baucau Samalari 0.43 0.4 0.5 -0.31 0.6 -1.05 0.12 0.24 Baucau Baucau Seical 0.5 0.41 0.49 0.1 0.47 -1.42 0.11 0.08 Baucau Baucau Trilolo 0.59 0.38 0.5 0.68 0.32 -0.31 0.39 1.59 Baucau Baucau Triloca 0.46 0.39 0.47 0.12 0.47 -0.22 0.43 1.91 Baucau Baucau Wailili 0.5 0.43 0.49 -0.19 0.56 -0.09 0.45 1.7 Baucau Laga Atelari 0.43 0.46 0.48 -0.31 0.6 -0.22 0.44 1.86 Baucau Laga Libagua 0.45 0.46 0.51 -0.25 0.59 -0.14 0.46 1.66 Baucau Laga Nunira 0.46 0.44 0.47 -0.08 0.52 -0.16 0.43 1.76 Baucau Laga Saelari 0.44 0.43 0.52 -0.2 0.57 -0.22 0.42 1.55 Baucau Laga Sagadati 0.44 0.45 0.49 -0.36 0.65 0.26 0.55 2.25 Baucau Laga Samalari 0.46 0.42 0.48 -0.32 0.62 -0.11 0.45 1.74 Baucau Laga Soba 0.51 0.43 0.53 0.01 0.5 -0.2 0.42 1.8 Baucau Laga Tequino Mata 0.52 0.42 0.48 -0.12 0.54 -0.5 0.33 1.22 Baucau Quelicai Abafala 0.45 0.43 0.51 -0.64 0.77 -0.36 0.39 1.36 Baucau Quelicai Abo 0.51 0.42 0.54 -0.49 0.69 0.1 0.51 1.7 Baucau Quelicai Afaça 0.48 0.45 0.49 -0.52 0.72 -0.69 0.26 0.83 Baucau Quelicai Baguia 0.48 0.43 0.5 -0.56 0.72 -0.65 0.28 0.93 Baucau Quelicai Bualale 0.43 0.45 0.55 -0.32 0.62 0.01 0.5 1.97 Baucau Quelicai Guruca 0.41 0.46 0.54 -0.27 0.6 -0.01 0.49 1.9 Baucau Quelicai Locoliu 0.43 0.47 0.54 -0.49 0.69 0.02 0.5 1.93 75 Labour Education Health DM DM DV DV DV Area Identification Force Gap Gap Index HCount Index HCount Count Gap District Subdistrict Suco Mean Mean Mean Mean Mean Mean Mean Mean Baucau Quelicai Laisorolai De Baixo 0.42 0.46 0.54 -0.29 0.62 0.07 0.51 2.09 Baucau Quelicai Laisorolai De Cima 0.43 0.48 0.57 -0.54 0.73 0.16 0.54 2.11 Baucau Quelicai Lelalai 0.43 0.44 0.56 -0.49 0.68 0.21 0.54 2.02 Baucau Quelicai Letemuno 0.45 0.44 0.54 -0.4 0.65 0.06 0.5 1.71 Baucau Quelicai Macalaco 0.44 0.43 0.54 -0.54 0.71 -0.06 0.47 1.69 Baucau Quelicai Maluro 0.42 0.43 0.55 -0.23 0.59 0.06 0.51 1.68 Baucau Quelicai Namanei 0.42 0.44 0.54 -0.59 0.75 -0.27 0.42 1.53 Baucau Quelicai Waitame 0.46 0.44 0.54 -0.54 0.73 0.13 0.52 2.03 Baucau Vemase Caicua 0.42 0.39 0.56 0.12 0.47 -0.93 0.15 0.82 Baucau Vemase Loilubo 0.47 0.41 0.48 -0.09 0.52 -0.17 0.45 1.99 Baucau Vemase Ossoala 0.43 0.43 0.53 0.58 0.35 -0.11 0.46 1.93 Baucau Vemase Ostico 0.42 0.36 0.45 0.18 0.46 -0.31 0.4 1.85 Baucau Vemase Uaigae 0.43 0.42 0.53 -0.18 0.55 -0.2 0.43 1.65 Baucau Vemase Uatu-Lari 0.46 0.4 0.48 -0.12 0.54 -0.06 0.47 1.73 Baucau Vemase Vemase 0.49 0.39 0.47 0.07 0.48 -0.44 0.32 1.41 Baucau Venilale Bado Ho'O 0.46 0.42 0.48 0.01 0.5 -0.17 0.44 1.76 Baucau Venilale Baha Mori 0.46 0.39 0.47 0.16 0.46 -0.3 0.39 1.65 Baucau Venilale Fatulia 0.48 0.45 0.52 0.14 0.46 -0.26 0.43 1.96 Baucau Venilale Uaiolo 0.4 0.43 0.45 -0.28 0.59 0.16 0.54 2.26 Baucau Venilale Uailaha 0.47 0.39 0.47 0.16 0.46 -0.34 0.4 1.68 Baucau Venilale Uataco 0.51 0.42 0.47 0.1 0.48 -0.13 0.46 1.92 Baucau Venilale Uma Ana Ico 0.45 0.47 0.48 0.2 0.45 -0.03 0.48 2 Baucau Venilale Uma Ana Ulu 0.48 0.43 0.48 -0.14 0.54 -0.15 0.45 1.89 Bobonaro Atabae Aidabaleten 0.48 0.41 0.46 0 0.5 -0.08 0.47 2.04 Bobonaro Atabae Atabae 0.44 0.42 0.48 0.26 0.44 -0.12 0.47 2.12 Bobonaro Atabae Hataz 0.45 0.44 0.47 0.1 0.47 -0.28 0.42 1.93 Bobonaro Atabae Rairobo 0.4 0.44 0.54 0.34 0.42 -0.27 0.43 1.9 Bobonaro Balibo Balibo Vila 0.46 0.45 0.51 0.2 0.45 0.12 0.52 2.44 Bobonaro Balibo Batugade 0.54 0.43 0.48 0.14 0.46 0.36 0.57 2.58 Bobonaro Balibo Cowa 0.43 0.47 0.55 0.04 0.49 -0.03 0.48 2.37 Bobonaro Balibo Leohito 0.44 0.43 0.51 0.23 0.44 0.35 0.58 2.63 Bobonaro Balibo Leolima 0.44 0.44 0.52 0.15 0.46 -0.27 0.43 1.96 Bobonaro Balibo Sanirin 0.48 0.44 0.52 0.07 0.48 -0.34 0.41 2.05 Bobonaro Bobonaro Ai-Assa 0.5 0.44 0.54 -0.11 0.54 0.35 0.57 2.57 Bobonaro Bobonaro Atu-Aben 0.5 0.44 0.53 -0.16 0.54 0.19 0.54 2.3 Bobonaro Bobonaro Bobonaro 0.5 0.41 0.47 0.08 0.48 0.18 0.54 2.51 Bobonaro Bobonaro Carabau 0.45 0.39 0.48 0.09 0.48 -0.09 0.48 2.37 Bobonaro Bobonaro Colimau 0.46 0.43 0.54 -0.27 0.6 -0.29 0.42 1.69 Bobonaro Bobonaro Cotabot 0.46 0.38 0.43 -0.01 0.51 -0.3 0.4 2.75 Bobonaro Bobonaro Ilat-Laun 0.44 0.41 0.53 0.1 0.47 0.15 0.53 2.56 Bobonaro Bobonaro Leber 0.46 0.46 0.55 0.12 0.47 0.29 0.56 2.61 Bobonaro Bobonaro Lour 0.49 0.46 0.55 -0.1 0.53 0.57 0.61 2.7 Bobonaro Bobonaro Lourba 0.49 0.42 0.47 -0.02 0.5 -0.12 0.47 2.12 Bobonaro Bobonaro Male-Ubu 0.45 0.36 0.5 -0.55 0.74 -0.19 0.45 1.94 Bobonaro Bobonaro Malilait 0.47 0.43 0.5 -0.29 0.59 -0.02 0.49 2.17 76 Labour Education Health DM DM DV DV DV Area Identification Force Gap Gap Index HCount Index HCount Count Gap District Subdistrict Suco Mean Mean Mean Mean Mean Mean Mean Mean Bobonaro Bobonaro Molop 0.45 0.41 0.46 0.19 0.46 0.12 0.52 2.38 Bobonaro Bobonaro Oe-Leu 0.54 0.39 0.48 0.11 0.47 0.26 0.56 2.42 Bobonaro Bobonaro Sibuni 0.46 0.44 0.47 -0.47 0.67 0.75 0.63 2.85 Bobonaro Bobonaro Soilesu 0.48 0.43 0.49 -0.07 0.52 0.19 0.54 2.37 Bobonaro Bobonaro Tapo 0.57 0.44 0.52 -0.17 0.55 0.56 0.61 2.72 Bobonaro Bobonaro Tebabui 0.41 0.38 0.5 0.1 0.47 -0.31 0.42 2.01 Bobonaro Cailaco Atudara 0.42 0.44 0.5 0.14 0.46 -0.16 0.45 2.16 Bobonaro Cailaco Dau Udo 0.4 0.44 0.45 0.17 0.46 -0.73 0.32 1.58 Bobonaro Cailaco Goulolo 0.41 0.47 0.46 0.13 0.46 -0.03 0.49 2.12 Bobonaro Cailaco Guenu Lai 0.43 0.47 0.5 0.03 0.49 -0.3 0.42 2.01 Bobonaro Cailaco Manapa 0.46 0.41 0.49 0.05 0.49 -0.24 0.43 1.87 Bobonaro Cailaco Meligo 0.46 0.43 0.49 -0.02 0.51 -0.19 0.44 1.84 Bobonaro Cailaco Purugoa 0.45 0.42 0.45 0.09 0.48 -0.24 0.43 2.07 Bobonaro Cailaco Raiheu 0.44 0.44 0.49 0.09 0.48 -0.08 0.48 2.23 Bobonaro Lolotoe Deudet 0.47 0.4 0.44 -0.02 0.5 -0.14 0.45 1.88 Bobonaro Lolotoe Gildapil 0.48 0.45 0.49 -0.14 0.54 0.41 0.58 2.52 Bobonaro Lolotoe Guda 0.44 0.43 0.55 0.06 0.48 0.13 0.52 2.1 Bobonaro Lolotoe Lebos 0.48 0.4 0.44 0 0.5 0.24 0.54 2.17 Bobonaro Lolotoe Lontas 0.42 0.4 0.5 -0.02 0.51 -0.15 0.47 1.89 Bobonaro Lolotoe Lupal 0.46 0.42 0.51 0.04 0.49 -0.06 0.48 1.99 Bobonaro Lolotoe Opa 0.45 0.39 0.48 0.01 0.5 -0.35 0.39 1.53 Bobonaro Maliana Holsa 0.6 0.36 0.5 0.04 0.49 0.25 0.54 2.21 Bobonaro Maliana Lahomea 0.52 0.37 0.47 0.15 0.46 0.27 0.54 2.57 Bobonaro Maliana Odomau 0.55 0.33 0.48 -0.07 0.52 0.1 0.51 2.32 Bobonaro Maliana Raifun 0.43 0.4 0.51 0.05 0.48 -0.85 0.26 1.39 Bobonaro Maliana Ritabou 0.45 0.41 0.52 0.11 0.47 0 0.5 2.28 Bobonaro Maliana Saburai 0.45 0.44 0.44 -0.17 0.55 0.67 0.63 2.58 Bobonaro Maliana Tapo/Memo 0.51 0.41 0.5 0.06 0.49 0.55 0.6 2.71 Covalima Fatululic Fatululic 0.51 0.47 0.5 -0.05 0.52 -0.28 0.42 1.7 Covalima Fatululic Taroman 0.51 0.47 0.46 -0.27 0.61 0.14 0.52 1.93 Covalima Fatumean Belulik Leten 0.51 0.44 0.48 -0.1 0.53 -0.21 0.44 1.56 Covalima Fatumean Fatumea 0.49 0.48 0.57 -0.07 0.52 -0.1 0.47 1.81 Covalima Fatumean Nanu 0.47 0.43 0.5 0.37 0.4 -0.34 0.41 1.46 Covalima Forohem Dato Rua 0.45 0.5 0.46 0.23 0.44 0.35 0.57 2.65 Covalima Forohem Dato Tolu 0.45 0.46 0.46 -0.07 0.52 -0.44 0.37 1.36 Covalima Forohem Lactos 0.48 0.47 0.52 -0.02 0.51 -0.1 0.46 1.8 Covalima Forohem Fohoren 0.5 0.44 0.48 0.01 0.5 -0.11 0.46 1.62 Covalima Maukatar Belecasac 0.48 0.48 0.52 0.06 0.48 0.57 0.61 2.73 Covalima Maukatar Holpilat 0.5 0.45 0.48 0.09 0.48 0.46 0.59 2.58 Covalima Maukatar Matai 0.54 0.44 0.5 -0.13 0.54 0.33 0.56 2.75 Covalima Maukatar Ogues 0.51 0.44 0.49 0.15 0.46 0.35 0.56 2.47 Covalima Suai Beco 0.49 0.43 0.52 0.18 0.45 0.24 0.55 2.34 Covalima Suai Camenaça 0.57 0.38 0.51 -0.04 0.51 0.42 0.57 3.04 Covalima Suai Debos 0.58 0.39 0.5 -0.11 0.53 0.14 0.51 2.1 Covalima Suai Labarai 0.49 0.42 0.48 0.33 0.4 0.35 0.57 2.61 77 Labour Education Health DM DM DV DV DV Area Identification Force Gap Gap Index HCount Index HCount Count Gap District Subdistrict Suco Mean Mean Mean Mean Mean Mean Mean Mean Covalima Suai Suai Loro 0.51 0.44 0.49 -0.08 0.53 0.64 0.61 3.01 Covalima Tilomar Foholulic 0.52 0.43 0.48 0.12 0.47 0.55 0.6 2.65 Covalima Tilomar Casabauc 0.5 0.41 0.54 0.03 0.49 0.58 0.6 2.74 Covalima Tilomar Lalawa 0.49 0.42 0.52 0.14 0.46 0.66 0.6 2.88 Covalima Tilomar Maudemo 0.54 0.39 0.51 -0.14 0.54 -0.15 0.45 1.69 Covalima Zumalai Fatuleto 0.45 0.47 0.53 0.19 0.47 -0.15 0.46 2.2 Covalima Zumalai Lepo 0.42 0.45 0.5 0.32 0.43 -0.33 0.42 2.24 Covalima Zumalai Lour 0.45 0.46 0.48 0.14 0.47 0.34 0.57 2.58 Covalima Zumalai Mape 0.44 0.46 0.48 0.29 0.44 -0.46 0.37 1.63 Covalima Zumalai Raimea 0.44 0.46 0.5 0.18 0.45 0.14 0.52 2.53 Covalima Zumalai Tashilin 0.49 0.46 0.45 0.09 0.48 0.08 0.51 2.44 Covalima Zumalai Ucecai 0.51 0.47 0.45 0.34 0.4 -0.32 0.36 1.41 Covalima Zumalai Zulo 0.51 0.43 0.48 0.04 0.49 -0.01 0.49 2.28 Dili Atauro Beloi 0.49 0.43 0.53 -0.05 0.52 0.51 0.6 2.96 Dili Atauro Biceli 0.43 0.38 0.46 -0.05 0.51 0.47 0.6 2.99 Dili Atauro Macadade 0.45 0.44 0.45 -0.02 0.5 0.5 0.61 2.82 Dili Atauro Maquili 0.39 0.45 0.46 -0.11 0.53 0.49 0.59 2.6 Dili Atauro Atauro Vila/Maumeta 0.5 0.4 0.51 -0.24 0.57 -0.01 0.49 2.06 Dili Cristo Rei Balibar 0.5 0.4 0.51 0.16 0.45 0.08 0.51 2.32 Dili Cristo Rei Becora 0.58 0.35 0.49 -0.29 0.59 -0.74 0.26 1.24 Dili Cristo Rei Bidau Santana 0.58 0.35 0.49 0.02 0.5 -0.09 0.46 2.15 Dili Cristo Rei Camea 0.56 0.36 0.5 -0.04 0.51 -0.33 0.4 1.7 Dili Cristo Rei Culu Hun 0.59 0.35 0.48 -0.02 0.51 -0.37 0.38 1.91 Dili Cristo Rei Hera 0.52 0.4 0.5 -0.05 0.51 -0.03 0.49 2.09 Dili Cristo Rei Meti Aut 0.58 0.38 0.49 0.12 0.47 0.38 0.57 2.71 Dili Dom Aleixo Bairro Pite 0.6 0.3 0.48 -0.07 0.52 -0.29 0.4 1.87 Dili Dom Aleixo Comoro 0.6 0.29 0.47 -0.16 0.55 -0.55 0.31 1.34 Dili Dom Aleixo Fatuhada 0.6 0.29 0.46 -0.06 0.53 -0.68 0.26 1.2 Dili Dom Aleixo Kampung Alor 0.61 0.33 0.47 0.47 0.37 0.87 0.58 2.88 Dili Metinaro Duyung (Sereia) 0.52 0.44 0.53 -0.03 0.51 0.21 0.54 2.53 Dili Metinaro Sabuli 0.51 0.47 0.55 0.07 0.48 -0.32 0.41 1.79 Dili Nain Feto Acadiru Hun 0.61 0.39 0.5 0.61 0.37 0.31 0.54 2.61 Dili Nain Feto Bemori 0.59 0.37 0.46 0 0.5 0.03 0.49 2.42 Dili Nain Feto Bidau Lecidere 0.59 0.4 0.49 0.15 0.47 -0.87 0.21 1.53 Dili Nain Feto Gricenfor 0.64 0.29 0.47 -0.14 0.55 -0.26 0.4 1.73 Dili Nain Feto Lahane Oriental 0.59 0.38 0.53 -0.03 0.51 0.14 0.52 2.4 Dili Nain Feto Santa Cruz 0.59 0.35 0.47 0.13 0.47 -0.23 0.42 1.93 Dili Vera Cruz Caicoli 0.58 0.36 0.47 -0.11 0.54 -0.38 0.38 1.36 Dili Vera Cruz Colmera 0.61 0.33 0.46 0.68 0.36 0.28 0.54 2.39 Dili Vera Cruz Dare 0.5 0.41 0.51 0.08 0.47 0.09 0.51 2.3 Dili Vera Cruz Lahane Ocidental 0.59 0.42 0.53 0.1 0.47 0.31 0.55 2.49 Dili Vera Cruz Mascarenhas 0.59 0.36 0.48 -0.01 0.51 0.33 0.54 2.33 Dili Vera Cruz Motael 0.62 0.32 0.46 -0.14 0.55 -0.33 0.38 1.7 Dili Vera Cruz Vila Verde 0.59 0.35 0.49 0.32 0.43 -0.16 0.43 1.83 Ermera Atsabe Atara 0.47 0.4 0.51 -0.38 0.64 -0.12 0.47 2.19 78 Labour Education Health DM DM DV DV DV Area Identification Force Gap Gap Index HCount Index HCount Count Gap District Subdistrict Suco Mean Mean Mean Mean Mean Mean Mean Mean Ermera Atsabe Baboi Craic 0.47 0.43 0.49 -0.23 0.57 0.15 0.53 2.28 Ermera Atsabe Beboi Leten 0.47 0.44 0.53 -0.19 0.56 0.26 0.56 2.62 Ermera Atsabe Batumanu 0.41 0.39 0.47 -0.11 0.53 -0.07 0.48 2.25 Ermera Atsabe Lasaun 0.45 0.41 0.48 -0.43 0.65 0.41 0.59 2.56 Ermera Atsabe Laclo 0.47 0.41 0.54 -0.19 0.56 0.04 0.5 2.38 Ermera Atsabe Laubono 0.42 0.45 0.55 -0.38 0.65 0.03 0.5 2.38 Ermera Atsabe Leimea Leten 0.43 0.44 0.5 0.24 0.45 0.14 0.53 2.51 Ermera Atsabe Atadame/Malabe 0.44 0.43 0.54 -0.02 0.51 0.07 0.52 2.33 Ermera Atsabe Obulo 0.43 0.46 0.56 -0.25 0.58 0.26 0.56 2.48 Ermera Atsabe Paramin 0.48 0.43 0.55 -0.1 0.53 0.18 0.54 2.5 Ermera Atsabe Tiarlelo 0.39 0.47 0.53 -0.1 0.53 0.08 0.51 2.44 Ermera Ermera Estado 0.47 0.43 0.48 -0.09 0.53 0.13 0.53 2.34 Ermera Ermera Humboe 0.44 0.42 0.48 0.1 0.47 0.26 0.55 2.6 Ermera Ermera Lauala 0.48 0.38 0.46 0 0.5 -0.06 0.48 2.04 Ermera Ermera Leguimea 0.46 0.46 0.48 0.25 0.44 0.39 0.58 2.71 Ermera Ermera Mirtutu 0.49 0.46 0.47 0.16 0.46 0.17 0.53 2.36 Ermera Ermera Poetete 0.47 0.42 0.5 0.16 0.46 0.09 0.52 2.41 Ermera Ermera Ponilala 0.49 0.46 0.51 0.1 0.47 0.07 0.51 2.41 Ermera Ermera Raimerhei 0.47 0.39 0.48 0.07 0.48 0.15 0.53 2.4 Ermera Ermera Riheu 0.51 0.38 0.49 0.09 0.48 0.03 0.5 2.33 Ermera Ermera Talimoro 0.54 0.34 0.48 -0.1 0.53 -0.18 0.44 2.1 Ermera Hatolia Asulau 0.4 0.43 0.45 -0.11 0.53 -0.01 0.5 2.38 Ermera Hatolia Ailelo 0.48 0.44 0.46 0 0.5 -0.13 0.46 2.06 Ermera Hatolia Coliate-Leotelo 0.42 0.41 0.49 0.17 0.46 0.25 0.55 2.67 Ermera Hatolia Fatuessi 0.43 0.39 0.47 0.08 0.48 0.14 0.53 2.48 Ermera Hatolia Fatubolu 0.45 0.42 0.48 0.18 0.45 0.19 0.54 2.56 Ermera Hatolia Hatolia 0.38 0.44 0.46 0.11 0.47 -0.04 0.49 2.32 Ermera Hatolia Leimeacraic 0.4 0.47 0.45 -0.01 0.51 0.01 0.5 2.36 Ermera Hatolia Lemia Sorimbalu 0.41 0.46 0.53 -0.02 0.51 0.15 0.53 2.62 Ermera Hatolia Lissapat 0.44 0.4 0.47 0.23 0.44 0.3 0.57 2.69 Ermera Hatolia Manusae 0.43 0.43 0.47 0.19 0.45 0.12 0.52 2.5 Ermera Hatolia Mau-Ubu 0.46 0.44 0.48 0.22 0.44 0.38 0.58 2.67 Ermera Hatolia Samara 0.39 0.46 0.44 -0.44 0.68 -0.55 0.36 1.62 Ermera Hatolia Urahou 0.42 0.42 0.46 0.05 0.49 0.41 0.59 2.82 Ermera Letefoho Catrai Leten 0.44 0.42 0.52 -0.07 0.53 0.31 0.56 2.46 Ermera Letefoho Ducurai 0.47 0.41 0.49 0.05 0.49 0.07 0.51 2.3 Ermera Letefoho Eraulo 0.44 0.41 0.47 0.18 0.45 0.1 0.52 2.4 Ermera Letefoho Goulolo 0.57 0.41 0.51 0.19 0.45 0.28 0.55 2.45 Ermera Letefoho Hatugau 0.44 0.43 0.49 -0.03 0.51 0.22 0.55 2.52 Ermera Letefoho Haupu 0.47 0.43 0.48 0.06 0.49 0.16 0.53 2.46 Ermera Letefoho Catrai-Craic 0.45 0.43 0.54 -0.02 0.51 0.26 0.56 2.56 Ermera Letefoho Lauana 0.43 0.41 0.51 -0.14 0.55 0.41 0.59 2.7 Ermera Railaco Deleco 0.45 0.39 0.45 0.27 0.44 -0.01 0.5 2.43 Ermera Railaco Fatuquero 0.57 0.34 0.56 0.07 0.48 -0.07 0.47 1.78 Ermera Railaco Lihu 0.57 0.4 0.5 -0.05 0.51 -0.04 0.48 2.12 79 Labour Education Health DM DM DV DV DV Area Identification Force Gap Gap Index HCount Index HCount Count Gap District Subdistrict Suco Mean Mean Mean Mean Mean Mean Mean Mean Ermera Railaco Matata 0.49 0.39 0.49 0.45 0.37 0.5 0.6 2.81 Ermera Railaco Railaco Craic 0.42 0.39 0.44 0.09 0.48 0.05 0.51 2.51 Ermera Railaco Railaco Leten 0.45 0.37 0.5 0.18 0.45 0.34 0.58 2.64 Ermera Railaco Samalete 0.43 0.38 0.49 -0.19 0.57 0.33 0.57 2.81 Ermera Railaco Tara╟O 0.52 0.44 0.48 -0.12 0.54 -0.07 0.47 2 Ermera Railaco Tocoluli 0.49 0.38 0.51 0.03 0.49 0.16 0.53 2.17 Lautém Iliomar Ailebere 0.46 0.44 0.53 0.04 0.49 0.13 0.52 2.02 Lautém Iliomar Cainliu 0.46 0.41 0.48 -0.05 0.52 -0.04 0.48 1.74 Lautém Iliomar Fuat 0.5 0.38 0.44 -0.34 0.61 0.3 0.56 2.54 Lautém Iliomar Iliomar I 0.5 0.43 0.52 -0.06 0.52 0.18 0.53 2.13 Lautém Iliomar Iliomar Ii 0.46 0.43 0.53 0.1 0.47 0.19 0.54 1.96 Lautém Iliomar Tirilolo 0.46 0.41 0.48 0.05 0.49 -0.31 0.4 1.5 Lautém Lautém Baduro 0.49 0.46 0.5 -0.12 0.54 -0.14 0.46 1.87 Lautém Lautém Com 0.52 0.48 0.53 0.23 0.43 0.36 0.57 2.61 Lautém Lautém Daudare 0.46 0.44 0.5 -0.01 0.51 -0.34 0.37 1.7 Lautém Lautém Euquisi 0.47 0.39 0.47 0.28 0.43 -1.38 0.1 0.1 Lautém Lautém Ililai 0.47 0.4 0.47 0.35 0.37 -1.85 0.21 -1.06 Lautém Lautém Maina I 0.51 0.48 0.54 -0.17 0.56 -0.05 0.48 2.11 Lautém Lautém Maina Ii 0.47 0.45 0.53 -0.13 0.54 -0.03 0.49 2.04 Lautém Lautém Pairara 0.57 0.46 0.56 -0.23 0.57 -0.31 0.41 1.64 Lautém Lautém Parlamento 0.54 0.5 0.53 -0.05 0.52 0.22 0.54 2.29 Lautém Lautém Serelau 0.49 0.43 0.48 0.12 0.47 -0.22 0.41 1.6 Lautém Lospalos Bauro 0.51 0.45 0.48 0.02 0.49 0.26 0.55 2.6 Lautém Lospalos Cacavem 0.48 0.44 0.46 0.22 0.44 0.1 0.52 2.32 Lautém Lospalos Fuiloro 0.55 0.37 0.5 -0.17 0.55 -0.1 0.47 2.13 Lautém Lospalos Home 0.49 0.43 0.49 0.31 0.41 0.11 0.51 2.54 Lautém Lospalos Leuro 0.48 0.46 0.48 0.22 0.44 0.18 0.54 2.64 Lautém Lospalos Lore I 0.48 0.42 0.49 -0.1 0.53 0.01 0.5 2.03 Lautém Lospalos Lore Ii 0.48 0.41 0.5 0 0.5 0.04 0.51 2.14 Lautém Lospalos Muapitine 0.54 0.45 0.5 0.12 0.47 -0.07 0.47 1.91 Lautém Lospalos Raça 0.49 0.44 0.47 -0.08 0.52 -0.18 0.45 2.16 Lautém Lospalos Souro 0.5 0.44 0.52 0.3 0.41 0.45 0.59 2.69 Lautém Luro Afabubu 0.51 0.43 0.54 0.12 0.47 -0.4 0.38 1.9 Lautém Luro Baricafa 0.46 0.39 0.49 -0.51 0.69 0.06 0.51 1.94 Lautém Luro Cotamutu 0.44 0.44 0.54 -0.53 0.71 0.2 0.54 2.14 Lautém Luro Lacawa 0.43 0.46 0.56 -0.49 0.67 -0.24 0.44 1.63 Lautém Luro Luro 0.47 0.43 0.47 -0.19 0.56 0.11 0.52 2.28 Lautém Luro Wairoce 0.46 0.46 0.55 -0.12 0.54 -0.1 0.47 1.99 Lautém Tutuala Mehara 0.52 0.41 0.47 -0.23 0.57 -0.11 0.46 2 Lautém Tutuala Tutuala 0.52 0.45 0.48 0.34 0.41 0.13 0.53 2.18 Liquiça Bazartete Fahilebo 0.48 0.44 0.51 0.16 0.46 0.36 0.57 2.4 Liquiça Bazartete Fatumasi 0.46 0.39 0.51 -0.02 0.5 -0.12 0.46 1.92 Liquiça Bazartete Lauhata 0.52 0.39 0.48 0.13 0.47 0.11 0.52 2.21 Liquiça Bazartete Leorema 0.47 0.4 0.48 0.45 0.39 0.08 0.51 2.43 Liquiça Bazartete Maumeta 0.56 0.4 0.49 0.06 0.49 -0.33 0.4 1.71 80 Labour Education Health DM DM DV DV DV Area Identification Force Gap Gap Index HCount Index HCount Count Gap District Subdistrict Suco Mean Mean Mean Mean Mean Mean Mean Mean Liquiça Bazartete Metagou 0.41 0.45 0.47 0.14 0.47 -0.1 0.48 2.18 Liquiça Bazartete Motaulun 0.46 0.42 0.49 -0.14 0.54 -0.06 0.48 2.14 Liquiça Bazartete Tibar 0.55 0.37 0.48 0 0.5 0.19 0.54 2.38 Liquiça Bazartete Ulmera 0.51 0.42 0.46 0.09 0.48 0.01 0.5 2.28 Liquiça Liquiça Açumano 0.43 0.36 0.56 0.21 0.45 0.02 0.5 2.54 Liquiça Liquiça Darulete 0.45 0.39 0.48 0.18 0.46 0 0.5 2.35 Liquiça Liquiça Dato 0.53 0.38 0.49 0.04 0.49 0.01 0.49 2.12 Liquiça Liquiça Hatuquessi 0.46 0.42 0.45 -0.08 0.52 0.17 0.54 2.49 Liquiça Liquiça Leoteala 0.43 0.45 0.52 0.15 0.47 0.18 0.54 2.62 Liquiça Liquiça Loidahar 0.46 0.41 0.48 -0.04 0.51 0.11 0.52 2.61 Liquiça Liquiça Luculai 0.45 0.43 0.47 0.19 0.46 -0.23 0.45 2.02 Liquiça Maubara Gugleur 0.43 0.43 0.45 -0.11 0.54 0.36 0.58 2.75 Liquiça Maubara Guiço 0.46 0.45 0.48 0.04 0.49 -0.05 0.49 2.3 Liquiça Maubara Lissadila 0.43 0.45 0.55 -0.05 0.51 0.54 0.61 3.2 Liquiça Maubara Maubaralissa 0.44 0.45 0.45 -0.13 0.54 0.14 0.53 2.42 Liquiça Maubara Vatuboro 0.48 0.41 0.47 0.01 0.5 0.04 0.51 2.45 Liquiça Maubara Vatuvou 0.46 0.43 0.51 -0.04 0.51 0.08 0.52 2.35 Liquiça Maubara Vaviquinia 0.52 0.47 0.49 -0.1 0.53 0.04 0.5 2.17 Manatuto Barique/Natarbora Abat Oan 0.54 0.38 0.5 0.24 0.42 0.28 0.55 2.82 Manatuto Barique/Natarbora Aubeon 0.51 0.42 0.47 0.05 0.49 0.47 0.58 3.04 Manatuto Barique/Natarbora Barique 0.48 0.49 0.55 -0.05 0.51 0.45 0.6 2.71 Manatuto Barique/Natarbora Manehat 0.5 0.43 0.48 0.21 0.44 0.32 0.57 2.43 Manatuto Barique/Natarbora Uma Boco 0.5 0.4 0.47 0 0.5 0.41 0.57 3.07 Manatuto Laclo Hohorai 0.43 0.43 0.56 -0.27 0.59 0.25 0.56 2.46 Manatuto Laclo Lacumesac 0.45 0.45 0.51 0.12 0.47 0.35 0.57 2.72 Manatuto Laclo Umacaduac 0.51 0.43 0.5 -0.16 0.55 0.15 0.53 2.55 Manatuto Laclo Uma Naruc 0.47 0.4 0.47 0 0.5 -0.16 0.45 1.9 Manatuto Laclubar Batara 0.4 0.45 0.49 0.09 0.48 0.42 0.59 2.83 Manatuto Laclubar Fatumaquerec 0.44 0.45 0.5 -0.37 0.63 0.65 0.63 2.75 Manatuto Laclubar Funar 0.44 0.42 0.55 -0.21 0.57 0.48 0.6 2.74 Manatuto Laclubar Manelima 0.45 0.43 0.45 -0.12 0.54 0.77 0.65 3.03 Manatuto Laclubar Orlalan 0.46 0.45 0.48 0.05 0.49 0.2 0.54 2.39 Manatuto Laclubar Sanana'In 0.37 0.42 0.48 -0.18 0.55 0.11 0.53 2.44 Manatuto Laleia Cairui 0.46 0.45 0.52 0.04 0.49 -0.11 0.46 2.17 Manatuto Laleia Haturalan 0.48 0.37 0.45 0.16 0.46 -0.67 0.27 1.43 Manatuto Laleia Lifau 0.51 0.39 0.48 0.2 0.44 -0.89 0.17 0.83 Manatuto Manatuto Ailili 0.55 0.4 0.52 0 0.5 0.05 0.49 2.69 Manatuto Manatuto Aiteas 0.55 0.31 0.47 0.12 0.47 -0.35 0.36 1.87 Manatuto Manatuto Cribas 0.44 0.43 0.48 -0.01 0.51 -0.35 0.41 1.85 Manatuto Manatuto Iliheu 0.42 0.45 0.48 0.3 0.4 0.13 0.51 2.49 Manatuto Manatuto Ma'Abat 0.51 0.29 0.46 0.07 0.48 -1.65 0.07 0.39 Manatuto Manatuto Sau 0.58 0.37 0.49 0.02 0.49 0.34 0.54 2.77 Manatuto Soibada Fatumacerec 0.51 0.42 0.5 0.25 0.43 0.36 0.57 2.67 Manatuto Soibada Leo Hat 0.5 0.4 0.47 -0.09 0.52 0 0.49 1.97 Manatuto Soibada Manlala 0.48 0.44 0.49 -0.27 0.57 -0.11 0.46 1.44 81 Labour Education Health DM DM DV DV DV Area Identification Force Gap Gap Index HCount Index HCount Count Gap District Subdistrict Suco Mean Mean Mean Mean Mean Mean Mean Mean Manatuto Soibada Manufahi 0.41 0.4 0.52 -0.06 0.52 -0.53 0.36 1.72 Manatuto Soibada Samoro 0.53 0.44 0.5 -0.06 0.52 -0.12 0.46 2.26 Manufahi Alas Aituha 0.48 0.44 0.48 0.01 0.5 0.45 0.59 2.68 Manufahi Alas Dotic 0.49 0.4 0.52 0.1 0.47 0.42 0.58 2.63 Manufahi Alas Mahaquidan 0.48 0.42 0.46 -0.01 0.51 0.33 0.57 2.72 Manufahi Alas Taitudac 0.48 0.39 0.45 -0.1 0.53 -0.36 0.4 1.62 Manufahi Alas Uma Berloic 0.47 0.4 0.48 0.01 0.5 0.8 0.65 3.92 Manufahi Fatuberliu Bubususo 0.47 0.49 0.44 -0.1 0.54 0.22 0.55 2.32 Manufahi Fatuberliu Caicasa 0.47 0.46 0.49 -0.09 0.53 0.52 0.6 2.69 Manufahi Fatuberliu Clacuc 0.49 0.42 0.51 -0.12 0.53 -0.22 0.43 2.12 Manufahi Fatuberliu Fahinehan 0.49 0.47 0.43 0.08 0.48 0.55 0.59 2.42 Manufahi Fatuberliu Fatucahi 0.51 0.41 0.51 -0.39 0.63 -0.33 0.4 1.51 Manufahi Same Babulu 0.53 0.4 0.48 0.1 0.47 0.14 0.52 2.17 Manufahi Same Betano 0.51 0.43 0.5 0.08 0.48 0.47 0.58 2.96 Manufahi Same Daisua 0.49 0.45 0.48 0.16 0.45 0.46 0.6 2.76 Manufahi Same Grotu 0.46 0.47 0.45 0.04 0.49 0.43 0.59 2.63 Manufahi Same Holarua 0.47 0.42 0.45 0.02 0.5 0.37 0.57 2.56 Manufahi Same Letefoho 0.57 0.35 0.49 -0.09 0.52 0.1 0.51 2.22 Manufahi Same Rotuto 0.48 0.41 0.57 -0.38 0.63 0.28 0.56 2.32 Manufahi Same Tutuluro 0.45 0.44 0.46 0.01 0.5 0.26 0.55 2.48 Manufahi Turiscai Aitemua 0.47 0.46 0.44 -0.19 0.57 -0.06 0.48 1.86 Manufahi Turiscai Beremana 0.44 0.39 0.55 -0.5 0.68 0.27 0.57 2.16 Manufahi Turiscai Caimauc 0.46 0.42 0.48 -0.23 0.57 -0.19 0.44 1.84 Manufahi Turiscai Fatucalo 0.46 0.48 0.42 -0.5 0.69 0.3 0.56 2.2 Manufahi Turiscai Foholau 0.39 0.38 0.4 -0.36 0.64 0.29 0.57 2.33 Manufahi Turiscai Lesuata 0.44 0.4 0.48 -0.4 0.64 -0.28 0.43 1.91 Manufahi Turiscai Liurai 0.41 0.38 0.53 -0.41 0.65 0.27 0.56 2.44 Manufahi Turiscai Manumera 0.55 0.4 0.47 -0.22 0.56 -0.71 0.31 1.01 Manufahi Turiscai Matorec 0.45 0.45 0.46 -0.42 0.66 0.15 0.53 2.26 Manufahi Turiscai Mindelo 0.46 0.48 0.48 -0.16 0.55 0.28 0.56 2.55 Manufahi Turiscai Orana 0.46 0.39 0.43 -0.35 0.63 0.46 0.6 2.41 Oecussi Nitibe Banafi 0.42 0.44 0.5 -0.19 0.57 0.86 0.69 3.34 Oecussi Nitibe Bene-Ufe 0.45 0.44 0.52 -0.1 0.53 0.48 0.6 2.77 Oecussi Nitibe Lela-Ufe 0.42 0.45 0.52 0.07 0.48 1.01 0.71 3.42 Oecussi Nitibe Suni-Ufe 0.39 0.43 0.52 -0.09 0.53 0.94 0.69 3.27 Oecussi Nitibe Usi-Taco 0.43 0.44 0.54 -0.1 0.53 1.02 0.72 3.46 Oecussi Oesilo Bobometo 0.45 0.44 0.52 -0.13 0.54 0.65 0.63 2.98 Oecussi Oesilo Usi-Taqueno 0.41 0.43 0.56 -0.28 0.6 1.06 0.7 3.64 Oecussi Oesilo Usi-Tacae 0.44 0.45 0.51 -0.11 0.54 0.88 0.68 3.23 Oecussi Pante Macasar Bobocase 0.46 0.42 0.51 -0.18 0.55 0.4 0.59 2.73 Oecussi Pante Macasar Costa 0.55 0.37 0.51 -0.18 0.55 0.04 0.5 2.08 Oecussi Pante Macasar Cunha 0.47 0.39 0.52 0.17 0.45 0.35 0.57 2.64 Oecussi Pante Macasar Lalisuc 0.48 0.38 0.46 0.08 0.48 -0.06 0.48 2.05 Oecussi Pante Macasar Lifau 0.45 0.41 0.56 -0.02 0.5 0.61 0.63 3.37 Oecussi Pante Macasar Naimeco 0.46 0.41 0.52 -0.22 0.58 0.51 0.61 2.8 82 Labour Education Health DM DM DV DV DV Area Identification Force Gap Gap Index HCount Index HCount Count Gap District Subdistrict Suco Mean Mean Mean Mean Mean Mean Mean Mean Oecussi Pante Macasar Nipani 0.46 0.4 0.49 -0.17 0.56 0.57 0.63 3.34 Oecussi Pante Macasar Taiboco 0.43 0.44 0.55 0.18 0.44 1.07 0.72 3.55 Oecussi Passabe Abani 0.44 0.43 0.52 -0.16 0.56 0.77 0.66 3.05 Oecussi Passabe Malelat 0.41 0.44 0.5 -0.03 0.52 0.99 0.69 3.35 Viqueque Lacluta Ahic 0.45 0.44 0.54 0.19 0.45 0.03 0.51 2.23 Viqueque Lacluta Dilor 0.52 0.41 0.5 -0.24 0.58 -0.33 0.41 1.49 Viqueque Lacluta Laline 0.48 0.44 0.55 0.25 0.43 0.66 0.64 2.96 Viqueque Lacluta Uma Tolu 0.46 0.45 0.48 -0.16 0.55 0.54 0.62 2.69 Viqueque Ossu Builale 0.45 0.44 0.53 0.25 0.44 -0.07 0.48 2.2 Viqueque Ossu Liaruca 0.48 0.45 0.56 0.22 0.45 -0.42 0.39 1.71 Viqueque Ossu Loi-Huno 0.48 0.42 0.5 0.01 0.5 0.07 0.51 2.03 Viqueque Ossu Nahareca 0.41 0.43 0.53 0.06 0.49 0.07 0.51 2.32 Viqueque Ossu Ossorua 0.5 0.46 0.55 0.13 0.47 -0.34 0.41 1.93 Viqueque Ossu Ossu De Cima 0.5 0.41 0.47 -0.02 0.51 -0.36 0.39 1.8 Viqueque Ossu Uabubo 0.49 0.44 0.5 0.02 0.49 -0.21 0.44 1.8 Viqueque Ossu Uaigia 0.48 0.42 0.52 0.04 0.49 -0.14 0.46 2.1 Viqueque Ossu Uaibobo 0.43 0.44 0.55 -0.32 0.61 -0.01 0.49 2.15 Viqueque Watulari Afaloicai 0.5 0.43 0.5 -0.1 0.53 -0.24 0.43 1.78 Viqueque Watulari Babulo 0.5 0.42 0.52 -0.01 0.5 0.05 0.5 2.05 Viqueque Watulari Macadique 0.47 0.44 0.48 -0.24 0.58 -0.48 0.37 1.69 Viqueque Watulari Matahoi 0.51 0.39 0.51 -0.25 0.58 -0.58 0.33 1.23 Viqueque Watulari Uaitame 0.49 0.41 0.5 0.39 0.4 -0.18 0.45 2.1 Viqueque Watulari Vessoru 0.51 0.41 0.52 0.14 0.47 0.03 0.5 2.44 Viqueque Uatucarbau Afaloicai 0.47 0.46 0.54 -0.12 0.54 -0.13 0.46 1.81 Viqueque Uatucarbau Bahatata 0.45 0.46 0.56 0.01 0.51 -0.02 0.49 1.92 Viqueque Uatucarbau Irabin De Baixo 0.51 0.49 0.53 -0.01 0.51 -0.62 0.32 1.31 Viqueque Uatucarbau Irabin De Cima 0.48 0.56 0.54 0.14 0.47 -0.57 0.35 1.41 Viqueque Uatucarbau Loi Ulu 0.43 0.53 0.58 -0.04 0.53 -0.21 0.44 2.06 Viqueque Uatucarbau Uani Uma 0.5 0.5 0.53 -0.23 0.57 -0.32 0.4 1.65 Viqueque Viqueque Bahalarauain 0.51 0.4 0.5 0.08 0.48 0.38 0.57 2.56 Viqueque Viqueque Bibileo 0.48 0.47 0.52 0.03 0.49 0.19 0.54 2.41 Viqueque Viqueque Caraubalo 0.58 0.4 0.49 -0.27 0.58 -0.23 0.43 1.8 Viqueque Viqueque Watu Dere 0.5 0.45 0.49 0.03 0.49 0.06 0.51 2.44 Viqueque Viqueque Luca 0.49 0.48 0.56 0.14 0.47 0.23 0.55 2.4 Viqueque Viqueque Maluro 0.5 0.45 0.51 0.33 0.41 0.72 0.62 2.97 Viqueque Viqueque Uai Mori 0.45 0.43 0.55 0.35 0.41 -0.06 0.48 1.95 Viqueque Viqueque Uma Quic 0.56 0.35 0.46 -0.04 0.51 0.01 0.5 1.9 Viqueque Viqueque Uma Uain Craic 0.52 0.43 0.46 0.01 0.5 -0.25 0.43 1.83 Viqueque Viqueque Uma Uain Leten 0.48 0.44 0.47 0.25 0.43 0.32 0.56 2.55 83 Figure 19: Suco-Level Estimates of Poverty Rate and 95% CI Figure 20: Suco-Level Estimates of Labor Force Index and 95% CI 84 Figure 21: Suco-Level Estimates of Health Index and 95% CI Figure 22: Suco-Level Estimates of Education Index and 95% CI 85 Figure 23: Suco-Level Estimats of Decision-Making Autonomy (DM) Index and 95% CI Figure 24: Suco-Level Estimates of Domestic Violence (DV) Index and 95% CI 86 Annex Suco Level Maps (Full Page Version) List of Figures Figure 1: Illuminated Areas in Timor Leste, 2013 DMSP Satellite F18 (using 5% Luminosity Threshold to restrict to non-ephemeral lit areas) ............................................................................................. 13 Figure 2: National-level Poverty Mapping Model Estimation ....................................................................... 14 Figure 3: National Model Estimation ............................................................................................................. 16 Figure 4: Rural Sector Poverty Mapping Model estimation ........................................................................... 18 Figure 5: Urban Sector Poverty Mapping Model estimation .......................................................................... 19 Figure 6: Suco-Level Mean Differences of People in Male-Headed vs Female-Headed Households ........... 22 Figure 7: Suco-Level Relative Differences of People in Male-Headed vs Female-Headed Households ....... 23 Figure 8: Proportion of the Population in Households Where the Index of Male-Female Education Gaps Indicates Female Disadvantage ...................................................................................................... 27 Figure 9: Relationship Between Gender Education Disadvantage and Suco-Level Mean Consumption ...... 27 Figure 10: Comparison Between Direct Estimates from the Census and Survey-to-Census Imputed Values, in terms of the Proportion of the Population in Households, Where the Index of Male- Female Education Gaps Indicates Female Disadvantage ............................................................... 29 Figure 11: Proportion of the Population in Households, Where the Index of Male-Female Health Gaps Indicates Female Disadvantage ...................................................................................................... 31 Figure 12: Relationship Between Gender Health and Suco-Level Mean Consumption ................................ 31 Figure 13: Proportion of the Population in Households Where the Index of Male-Female Labour Force Gaps Indicates Female Disadvantage ............................................................................................. 35 Figure 14: Relationship Between Labour Force Disadvantage and Suco-Level Mean Consumption ........... 35 Figure 15: Proportion of the Population in Households, Where the Index of Female Decision Making (DM) Autonomy Indicates Female Disadvantage .......................................................................... 38 Figure 16: Proportion of the Population in Households Where the Index of Female Experience of Types of Domestic Violence (DV) Indicates Female Disadvantage ........................................................ 39 Figure 17: Relationship Between Share of Population Living in Households With High Domestic Violence Index and Average Predicted Headcount Poverty Rate by Suco .................................... 39 Figure 18: Suco Average of the Predicted Number of Types of Domestic Violence (DV) Reported by Females ........................................................................................................................................... 40 Figure 19. Suco-Level Estimates of Poverty Rate and 95% CI ...................................................................... 84 87 Figure 20. Suco-Level Estimates of Labor Force Index and 95% CI ............................................................. 84 Figure 21. Suco-Level Estimates of Health Index and 95% CI ...................................................................... 85 Figure 22. Suco-Level Estimates of Education Index and 95% CI ................................................................ 85 Figure 23. Suco-Level Estimats of Decision-Making Autonomy (DM) Index and 95% CI .......................... 86 Figure 24. Suco-Level Estimates of Domestic Violence (DV) Index and 95% CI ........................................ 86 88 Figure 25: Illuminated Areas in Timor Leste, 2013 DMSP Satellite F18 (using 5% Luminosity Threshold to restrict to non-ephemeral lit areas) 89 Figure 26: National-level Poverty Mapping Model Estimation, (a) Predicted Headcount Poverty Rate 90 Figure 27: National-level Poverty Mapping Model Estimation, (b) Number in Poor Households 91 Figure 28: National Model Estimation, (a) Predicted Poverty Gap Index 92 Figure 29: National Model Estimation, (b) Poverty Severity Index 93 Figure 30: Rural Sector Poverty Mapping Model estimation, (a) Predicted Headcount Poverty Rate 94 Figure 31: Rural Sector Poverty Mapping Model estimation, (b) Number in Poor Households 95 Figure 32: Urban Sector Poverty Mapping Model estimation, (a) Predicted Headcount Poverty Rate/ 96 Figure 33: Urban Sector Poverty Mapping Model estimation, (b) Number in Poor Households 97 Figure 34: Suco-Level Mean Differences of People in Male-Headed vs Female-Headed Households, (a) Predicted Poverty Headcount Rate 98 Figure 35: Suco-Level Mean Differences of People in Male-Headed vs Female-Headed Households, (b) Poverty Gap Index 99 Figure 36: Suco-Level Relative Differences of People in Male-Headed vs Female-Headed Households, (a) Predicted Poverty Headcount Rate 100 Figure 37: Suco-Level Relative Differences of People in Male-Headed vs Female-Headed Households, (b) Poverty Gap Index 101 Figure 38: Proportion of the Population in Households Where the Index of Male-Female Education Gaps Indicates Female Disadvantage 102 Figure 39: Proportion of the Population in Households, Where the Index of Male-Female Health Gaps Indicates Female Disadvantage 103 Figure 40: Proportion of the Population in Households Where the Index of Male-Female Labour Force Gaps Indicates Female Disadvantage 104 Figure 41: Proportion of the Population in Households, Where the Index of Female Decision Making (DM) Autonomy Indicates Female Disadvantage 105 Figure 42: Proportion of the Population in Households Where the Index of Female Experience of Types of Domestic Violence (DV) Indicates Female Disadvantage 106 Figure 43: Suco Average of the Predicted Number of Types of Domestic Violence (DV) Reported by Females 107