被占领巴勒斯坦领土的多层面贫困.docx

《被占领巴勒斯坦领土的多层面贫困.docx》由会员分享，可在线阅读，更多相关《被占领巴勒斯坦领土的多层面贫困.docx（32页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、Economic and Social Commission for Western AsiaNowcasting multidimensional poverty in the occupied Palestinian territoryUnited NationsBeirutDimension ofPovertyIndicator1Deprived if- 1Indicators weightEducational attainment - persons aged 19-50All household members aged 19-50 not completing secondary

2、 schoolddi = 1/30Quality of education - household with children age 6-17 years enrolled in schoolHousehold has any child aged 6-17 who had problems with education quality.Indicated a serious problem with the school in terms of poor teaching or lack of teachers or lack of books or lack of facilities.

3、ddi = 1/30HealthDisabilityAny household member having great difficulty in hearing, vision, movement, communication, OR understandingddi = 1/30Chronic diseaseAll household members aged 30+ suffering from a diagnosed chronic disease.ddi = 1/30Health insuranceHousehold lacking health insurance: (the he

4、ad OR any member has health insurance defined as NOT deprived)ddi = 1/30Health AccessHousehold lives more than 5 km away from the nearest doctor clinic or hospitalddi = 1/30EmploymentUnemploymentNone of adults aged 18+ currently employedddi = 1/30Employment benefitsWage earners aged 15-60 lacking pa

5、id sick leave, maternity leave or annual vacationddi = 1/30Quality of workHousehold has any working member 18+ who is currently an irregular wage employee, OR does not have a contract OR is a seasonal & causal worker OR has worked only 6 months during last 12 months.ddi = 1/30Youth NEETHousehold has

6、 any youth aged 18-24 who is not in school or training and unemployedddi = 1/30Housing conditions and access toAccess to piped waterDwelling is not connected to public networkddi = 1/30servicesfrequency of water and electricity supplyDisruption of water supply (daily) during the past yearddi = 1/30D

7、imension of PovertyIndicatorDeprived if-Indicators weightSafety and use of assetsVentilation problems in dwellingOvercrowdingTheft or damage to propertyDwelling suffers from noise, smoke or any other pollutantddi = 1/30More than 3 persons per sleeping room ddi = 1/30Stealing from household or damage

8、 of household property as a result of attacks last yeardh = 2/45Personal freedomOwnership and use of assetsInterpersonaland state violenceFreedom of movementHousehold lost land, house/building or ddi = 2/45 business establishment during the past year due to confiscation or demolitionHousehold was un

9、able to useagricultural land or private property due to restrictions of movementAny household member attacked or forcibly assaulted with or without a weapon last yearOR, any child or women hit or attacked by another family member during the past year.OR Injuries, deaths or torture in household from

10、state/settler violence during the past yearddi = 2/45A household member was not able to ddi = 1/15 visit family, relatives, or friends because of checkpoints, wall or travel restrictions during the past yearddi = 1/15dd1 = 1/15Monetary resourcesNational poverty lineHousehold is below the national po

11、verty ddi = 1/5 lineSource： Khawaja, Al-Saleh, Reece and Conconi, 2020.Control of womens income or womens participation in the labour marketAny women in household who does not have a separate bank account or does not control her use of income or earnings ORAny women in household does not work (or lo

12、ok for work) because of husband/father/brothers restrictionsThere are two steps in the computation of the multidimensional poverty index. In a first step, a deprivation cutoff,即,is chosen for each indicator kk E 1,2, .,22. Then, for each household ii G 1,2, ，皿, a score is assigned based on the indic

13、ators xx” =(琢，22) associated with this household. This score, ssn is given by: 5舟=11 (崎曲2级)X ddhh，where dd楠 represents the weight assigned to indicator kk, 优 1dd版=1 and 11(-) is an indicator function that takes a value of 1 if the argument is true and a value of 0 if it is false.A second step consis

14、ts of identifying multidimensional poverty by choosing a cutoff, cc 6 (0,1). This cutoff is fixed at 0.20 for the Palestinian multidimensional poverty index. A household is poor if its score is larger or equal to this cutoff. Note that in the empirical section, we will test the sensitivity of the re

15、sults on this poverty cutoffs choice. The multidimensional poverty headcount is then given by:1ram = _snn cc).；E 口The Palestinian multidimensional poverty index is:(3) MWJfJffl =11 (ss cc) x ss.nii ii=l iiiiIt is straightforward to perform the above exercise for each year for which a survey is avail

16、able. However, these surveys are not available every year. In this context, one needs to model these two indices value for the years without a survey. The rest of this section sets out the modelling approach proposed byMakdissi (2020) that we will adapt and use in the present paper.One can think of

17、the multidimensional poverty headcount and the Palestinian multidimensional poverty index at year tt as two functionals of the multidimensional cumulative distribution of indicators, NMu(xx), of that year. This multidimensional cumulative distribution function is formally defined as玲(XX)=帆(的小2,，m22)

18、=P巾1 - XX, XX2.的， zzlyear = tt-Proposing a model of A%(xx) would allow to nowcast or forecast these two indices “MMM% = 加刚船金卜(xx)电end MM=硼嬲伏xx)。In order to estimate counterfactuals of the multidimensional cumulative distributions 叫增(xx) for the years in which we do not have a survey, Makdissi (2020)

19、 propose to follow an insight of Khaled, Makdissi, Rao, and Yazbeck (2020) and build on Sklars Theorem 1 that stipulates that 1 岬(xx) can be expressed as:(4) WfC6(xx) = CCGC (,a ),.，FF (xx )。tt XX1,U 1H22,K221 .e. in terms of a 22-dimensional copula CC(-) and the univariate marginal distribution fun

20、ctions 取沁私)/kf L2, .,22. One can generally also recover M眄(xx) from the copula (T () and the ttttmarginal distribution functions 的林口。)，kk E (1,2, .,22. The main analytical advantage of looking at the problem with a copula approach is that it allows for the modelling of each one of the indicators kk

21、G 1,2, .,22 separately and the reconstruction of a counterfactual multidimensional cumulative distribution 哪(xx) from the counterfactuals of each indicator. The1 For more information on Sklars Theorem, see Hofert, Kojadinovicz Maehler and Yan (2018).only assumption that one needs in this context is

22、the stability of the copula.In our context, since we have some discrete indicators, the copula function is not unique. Some solutions in the statistical literature analyse situations in which some dimensions are discrete and others continuous. These solutions consist of choosing one of the potential

23、 copulas that fit the data. In order to account for survey weights, we follow Makdissi (2020) and adapt the checkerboard copula estimator from Genest, Neslehova, and Remillard (2017)Smith and Khaled (2012) propose another type of solution: a Bayesian latent approach to model the discrete dimensions.

24、 using a Hajek weighted estimator (annex 1 gives the details of the estimation procedure). The models construction implies estimating each indicators cumulative distribution function and the copula for the reference year. We can then use this model to build counterfactuals of the joint distribution

25、and indices when one or more indicators change following a simulated pattern or an assumed exogenous shock.This approach allows us to decompose the change between two observational years as induced by a change in marginal distributions and a change in the copula. For instance, let 力()(xx) represent

26、the multidimensional cumulative distribution function as the copula function of year tt0/ 缢)()and the marginal distribution of year tti,尸%匕叮(也效)，kk E 1,2, .,22. Using this notation, a change in the Palestinian for group between any year t and 2016 can be rewritten as:(5) 棚珊眄 (xx)6 岫邮叫xx)O=2016tl(xx)

27、 0-伸(xx)2016:2016tt:2016+J删棚极(XX):2016tt:ttwhere:(6) 硼M棚郴%16:2016(瓶)6 WW:2016(XX)=姗他删(assuming copula stability.andmffl觥2016(双)6牯(xx)=induced by change in the copula.Assuming stability of the copula allows us to estimate the indices associated with equation (8). If we have models for each univariate

28、 marginal distribution, we can estimate the change corresponding to equation (8).Let DD indicate the 22-dimensional domain of integration. The terms on the r.h.s. of equation (8) can be evaluated using the expressions:(7) 删嘏剧多眄(xx)O= J 11022tt:2016CC)d 股 2016(麻)删伽钏G航f创。22 dM 11(恤式ZZQ 出 2016/ kk=lCC)

29、X 卷%=1 ddkk 11- z编)翎雌2016(XX).These multiple integrals can be evaluated using Monte Carlo integration (technical details are in annex 2).2. Modelling the change of the 22 indicators in the State of PalestineFor our empirical analysis, we use thePalestinian Consumption and Expenditure survey (PECS) o

30、f 2011 and 2016. Twelve of the twenty-two indicators in table 1 are available in both surveys. The remaining ten indicators are only available in 2016. Inspection of equations (9) and (10) and the details of the Monte Carlo integration in annex 2 indicate that the only dynamic we need to estimate th

31、e counterfactual indices is a model of the change in the incidence of the deprivation in each dimension FF% (zz收). The estimated copula produces the remaining structure of the index.As shown in Makdissi (2020), for the context of Iraq, using the estimated copula of a year to model the multidimension

32、al poverty index of another year under the stability of the copula assumption produces numbers that are close to the estimated numbers if one uses the correct marginal distributions of each deprivation indicator. In other words, the model produces a counterfactual multidimensional poverty index that

33、 is as good as the counterfactual projection in each dimension.Since 喘/%) E 0,1, we propose to calibrate the dynamic of F端口(如)for the twelve indicators for which we have two years of observations on a two-parameter logistic function:(8) F 型 Q ) =1.的 kkl+eeaakk+即时配where is the value of an index at ye

34、ar tt. Depending on the indicator, this index may be time, per capita state-wide (regional) production, or one of the political indices: the demolition index, or the occupation severity index. The state-wide and regional real per capita GDP from 2011 to 2019 are obtained from the PCBS.where is the v

35、alue of an index at year tt. Depending on the indicator, this index may be time, per capita state-wide (regional) production, or one of the political indices: the demolition index, or the occupation severity index. The state-wide and regional real per capita GDP from 2011 to 2019 are obtained from t

36、he PCBS. It is important to note that these estimates exclude the parts of Jerusalem governorate which were annexed by Israel in 1967. The values for 2020 and 2021 are obtained by applying the growth projections for 2020 and 2021 produced in the World Bank economic update for West Bank and Gaza in O

37、ctober 2020. We construct the political indices using information from the Country Office in the occupied Palestinian territory (oPt) of the United Nations Office for the Coordination of Humanitarian Affairs (OCHA). The demolition index is just the number of demolitions in the year. divided by the a

38、nnual number of demolitions geometric average during the period of analysis. We use this index for the state-wide projections and the West Banks regional projection since demolitions due to colonization is a West Bankspecific phenomenon. At the state-wide level, the occupation severity index is a we

39、ightedgeometric average of the demolition index (0.70) and an index of the number of Palestinian injuries (0.30) due to conflict.geometric average of the demolition index (0.70) and an index of the number of Palestinian injuries (0.30) due to conflict. This index is the ratio of yearly injuries to t

40、he geometric average of injuries during the period of analysis. For the West Bank regional occupation severity index, we use a similar index but using only the West Banks injuries to construct the index of annual injuries.For Gaza, we need to substitute the demolition index by another index capturin

41、g the specificity of Gazas occupation compared to the West Bank. Israel has imposed a blockade on Gaza since 2007. To capture variation in this blockades intensity, we pick the quantity of an essential non-food item entering the Strip: hygiene and cleaning supply. We use the information on the numbe

42、r of truckloads of hygiene and cleaning supplyThese estimates can be found at and Gazas estimated population Rainbow/Documents/%D8%A7%D9%84%D9%85%D8%AD%D8%A7%D9%81%D8%B8%D8%A7%D8 %AA%20%D8%A7%D9%86%D8%AC%D9%84%D9%8A%D8%B2%D9%8A%2097-2017.html. to get an estimate of the number of truckloads per 100,0

43、00 habitants. We then construct the index as the ratio number, the geometric average of the number of truckloads per 100,000 habitants per year for analysis to the annual number of truckloads for the year of interest.For the ten indicators for which we have only one year of observations, we can only

44、 calibrate one of the above logistic parameters. To fully identify the dynamics of these indicators, we will assume that the semi-elasticity (0依)of the odds-ratio FF刑(zzkk)/(l FFMV (zzQ) is equal to the ，侏附XXkbttsame semi-elasticity for a close enough indicator for which we have two years ofobservat

45、ions and for which dynamics is linked with the same index.We have a vector of 33 parameters to calibrate to capture our simple model of the 22 univariate marginal distributions dynamics. Let M represents the vector of these 34 calibrated parameters. We follow an insight from Abdelkhalek and Dufour (

46、1998, 2006) and Dufour, Kichian and Khalaf (2013) and assume that the vector of parameters to be calibrated is a function of many stochastic variables. In other words, 0。=硼(朕2011，朕2016), a function of the fourteen stochastic indicators in 2011 and 2016 and the ten additional ones appearing only in t

47、he 2016 wave of the survey. To account for sampling variability, we recalibrate this vector at each bootstrap iteration.A. Baseline modelling frameworkThe baseline modelling framework will produce simulations for all years between 2011 and 2021. Since the only years for which we have a survey of all

48、 the indicators is 2016, the simulated model indices will be compared with the estimated indices for that year.The following indicators for which we have two years of observation will be linked with the index of per capita state-wide (regional) production (below, we give our wellbeing definition of the ill-being indicators):(1) Hi： All children aged 6-17 enrolled in school.H2: All children aged 6-17 enrolled in school andnever repeated a school year.(2) H7: H