Further Updating Poverty Mapping in Albania Gianni Betti*, Andrew Dabalen**, Celine Ferrè** and Laura Neri* * University of Siena, Italy, ** The World.

Further Updating Poverty Further Updating Poverty Mapping in AlbaniaMapping in Albania

Gianni Betti*, Andrew Dabalen**, Gianni Betti*, Andrew Dabalen**,

Celine Ferrè** and Laura Neri*Celine Ferrè** and Laura Neri** University of Siena, Italy, ** The World Bank, Washington, USA* University of Siena, Italy, ** The World Bank, Washington, USA

Poverty and Social Inclusion in the Western BalkansPoverty and Social Inclusion in the Western Balkans

WBalkans 2010, WBalkans 2010, Brussels, December 14-15, 2010Brussels, December 14-15, 2010

22

Scopes of the presentationScopes of the presentation

- Introduction on basic concepts of - Introduction on basic concepts of poverty mappingpoverty mapping

- Concepts of updating poverty mapping - Concepts of updating poverty mapping without new census datawithout new census data

- Application to Albania: 2002-2005-2008- Application to Albania: 2002-2005-2008 - Results on 2008 and comparisons with - Results on 2008 and comparisons with

2005 and 2002 are only reported in the 2005 and 2002 are only reported in the paper for sake of time restrictionpaper for sake of time restriction

33

THE METHODOLOGYCombines Census and Survey Data to produce disaggregated maps of poverty and inequality (Elbers, Lanjow and Lanjow, 2003, Econometrica).

THE APPLICATION HERE PROPOSEDCensus (2001) and LSMS (2002) in Albania, firstly updated to LSMS (2005), and then updated to LSMS (2008).

44

Citing the paper of Elbers, Lanjouw and Citing the paper of Elbers, Lanjouw and Lanjouw (ELLLanjouw (ELL, 2002 and 2003, 2002 and 2003))Poverty and inequality maps are spatial Poverty and inequality maps are spatial descriptions of the distribution of poverty descriptions of the distribution of poverty and inequality and are most useful to policy-and inequality and are most useful to policy-makers and researchers when they are makers and researchers when they are finely disaggregated, i.e. when they finely disaggregated, i.e. when they represent small geographic units, such as represent small geographic units, such as cities, municipalities, regions or other cities, municipalities, regions or other administrative partitions of a country.administrative partitions of a country.

What is “Poverty Mapping” ?What is “Poverty Mapping” ?

55

Figure 3Figure 3.. Head Count Ratio by Municipality, 2002 Head Count Ratio by Municipality, 2002..

66

BASIC IDEA OF “POVERTY MAPPING”BASIC IDEA OF “POVERTY MAPPING”

      To estimate a linear regression model with local       To estimate a linear regression model with local variance components on the LSMS data (the variance components on the LSMS data (the dependent variable is a monetary variable)dependent variable is a monetary variable) – – ESTIMATION (Stage 1) ESTIMATION (Stage 1)

      The distribution of the dependent variable is       The distribution of the dependent variable is used to generate the distribution for any used to generate the distribution for any subpopulation in the Census conditional to the subpopulation in the Census conditional to the observed dataobserved data – IMPUTATION or SIMULATION (Stage – IMPUTATION or SIMULATION (Stage 2)2)

      The variables used in the Census data and in       The variables used in the Census data and in the LSMS should comparable (i.e. same categories, the LSMS should comparable (i.e. same categories, etc…). A so-called Stage 0 is needed before etc…). A so-called Stage 0 is needed before estimation of linear regression model in Stage 1.estimation of linear regression model in Stage 1.

77

SStage 0: are the LSMS and the Census tage 0: are the LSMS and the Census comparable?comparable?  Fully analysis of the two data source to Fully analysis of the two data source to construct common variablesconstruct common variables in Albania we in Albania we have identified 38 common variableshave identified 38 common variables    Housing   Housing and and Dwelling conditions and Dwelling conditions and presence of durable goods (23)presence of durable goods (23)  Household head characteristics (8)  Household head characteristics (8)  Household socio-demographic characteristics   Household socio-demographic characteristics (7)(7)  Imputation for missing values in LSMS has been Imputation for missing values in LSMS has been done done ((IVE-wareIVE-ware, Raghunathan , Raghunathan et alet al. 2001). 2001)  Census and LSMS distribution should be Census and LSMS distribution should be comparedcompared

88

Stage 1: EstimationStage 1: Estimation

The modelThe model: it is a linear approximation to : it is a linear approximation to the conditional distribution of the the conditional distribution of the logarithm consumption expenditure (or logarithm consumption expenditure (or income) of income) of household household hh in in cluster cluster cc,,  

(1) (1) 

The error component is specified to allow The error component is specified to allow for a within cluster correlation in for a within cluster correlation in disturbances.disturbances.

IMPORTANT: several models are estimated IMPORTANT: several models are estimated in terms of number of strata in the LSMS in terms of number of strata in the LSMS

survey.survey.

chTchch

Tchchch uxuxyEy |lnln

99

Stage 2: SimulationStage 2: Simulation

The estimates obtained are applied to the The estimates obtained are applied to the Census data to simulate the expenditure Census data to simulate the expenditure for each household in the Census.for each household in the Census.A certain number (i.e.100) of simulations A certain number (i.e.100) of simulations are conductedare conducted      The simulated values are      The simulated values are::

(4)(4)      The beta coefficients       The beta coefficients , are drawn , are drawn from a multivariate normal distribution from a multivariate normal distribution with mean with mean and variance covariance and variance covariance matrix equal to the one associated to matrix equal to the one associated to ..

ˆ exp Tch ch c chy x

~

1010

            For the residual, any specific For the residual, any specific distributional form assumption is avoided distributional form assumption is avoided so the residual are drawn directly from the so the residual are drawn directly from the estimated residuals. estimated residuals.

      For each of the simulated consumption       For each of the simulated consumption expenditure distributions a set of poverty expenditure distributions a set of poverty and inequality measures is calculated.and inequality measures is calculated.

      Mean over all the simulations      Mean over all the simulations point point estimatesestimatesStandard deviation over all the Standard deviation over all the

simulationssimulations bootstrapping standard error.bootstrapping standard error.

1111

Updating the 2002 Poverty MappingUpdating the 2002 Poverty Mapping

In 2005, Dabalèn and Ferrè have proposed to In 2005, Dabalèn and Ferrè have proposed to update the poverty mapping in two Phases:update the poverty mapping in two Phases:

First Phase: construct the so-called “counterfactual First Phase: construct the so-called “counterfactual population distribution”: this is population distribution”: this is the distribution that the distribution that would have prevailed in 2002 if the parameters of would have prevailed in 2002 if the parameters of consumption and the distribution of observable consumption and the distribution of observable and unobserved covariates were as they were in and unobserved covariates were as they were in 2005 (now 2008);2005 (now 2008);

Second Phase: apply the ELL methodology Second Phase: apply the ELL methodology described in the previous slides using the described in the previous slides using the “counterfactual population distribution” [CM].“counterfactual population distribution” [CM].

1212

CM – logic behind the model - 1CM – logic behind the model - 1

In this case, In this case, Dabalèn and Ferrè (2005) Dabalèn and Ferrè (2005) proposed to construct a counterfactual proposed to construct a counterfactual consumption distribution of the old consumption distribution of the old household survey, using information from household survey, using information from both the old and new household survey both the old and new household survey and match the corresponding estimates and match the corresponding estimates with the old census data, following the with the old census data, following the methodology proposed by Lemieux methodology proposed by Lemieux (2002). (2002).

1313

CM – logic behind the model - 2CM – logic behind the model - 2

To construct the counterfactual wealth To construct the counterfactual wealth distribution, firstly let’s consider a consumption distribution, firstly let’s consider a consumption model using the model using the newnew survey. survey.

(5)(5) Where denotes consumption in year 2005, Where denotes consumption in year 2005, ii

indexes the household, is a parameter (that indexes the household, is a parameter (that captures the “returns” to or “price” of covariates captures the “returns” to or “price” of covariates in 2005), is a vector of covariates and, is in 2005), is a vector of covariates and, is unobserved component of consumption. unobserved component of consumption.

iii Xy ,05,0505,05 )ln(

05y

05

05X 05

1414

CM – logic behind the model - 3CM – logic behind the model - 3 Note that using this Note that using this newnew survey, without survey, without

additional adjustment, and applying the ELL additional adjustment, and applying the ELL estimator would be problematic because the estimator would be problematic because the returns to covariates, the parameter may have returns to covariates, the parameter may have changed between 2002 and 2005. In addition, changed between 2002 and 2005. In addition, the profile of the population – that is covariates the profile of the population – that is covariates such as education levels, age composition, and such as education levels, age composition, and so on – may also have changed. Finally, the so on – may also have changed. Finally, the returns to unobserved covariates may also have returns to unobserved covariates may also have changed. To recreate a consumption distribution changed. To recreate a consumption distribution that resembles consumption of 2002, CM would that resembles consumption of 2002, CM would have to account for these changes. Therefore, have to account for these changes. Therefore, the counterfactual consumption distribution has the counterfactual consumption distribution has three steps.three steps.

1515

CM – First step - 1CM – First step - 1

The first step is to create a consumption The first step is to create a consumption distribution that would have prevailed in 2002 if distribution that would have prevailed in 2002 if the parameters were as in 2005. That is,the parameters were as in 2005. That is,

(6)(6)

Equation (6) accounts for changes in the Equation (6) accounts for changes in the parameters of covariates, by using the parameters of covariates, by using the estimated parameters from the estimated parameters from the newnew survey to survey to estimate consumption distribution in the estimate consumption distribution in the oldold survey. survey.

iipi Xy ,02,05,02

ˆ)ln(

1616

CM – First step - 2CM – First step - 2 However, in addition to these parameters, However, in addition to these parameters,

levels of covariates may have changed levels of covariates may have changed because, for instance, the population is now because, for instance, the population is now more educated, etc… more educated, etc…

1717

CM – Second step - 1CM – Second step - 1 Instead, the CM methodology creates a score that Instead, the CM methodology creates a score that

reduces the dimension of the data, by stacking the reduces the dimension of the data, by stacking the newnew and and oldold surveys, and then by running a probit model: surveys, and then by running a probit model:

(7)(7)

In principle, a large set of observable household level In principle, a large set of observable household level characteristics can be included, , and also the characteristics can be included, , and also the migration status of the household, , or any suitable migration status of the household, , or any suitable variables that capture the scale of migration, which is variables that capture the scale of migration, which is of crucial concern when trying to update poverty maps. of crucial concern when trying to update poverty maps.

itZ

itM

itmitzititit MZMZsurveyobP ),|2005(Pr

1818

CM – Second step - 2CM – Second step - 2 Equation (7) allows us to obtain a propensity Equation (7) allows us to obtain a propensity

score – the predicted probability of being in score – the predicted probability of being in period - conditional on the period - conditional on the observable characteristics. observable characteristics.

(8)(8)

Where is the unconditional probability that Where is the unconditional probability that an observation belongs to period an observation belongs to period tt or the share or the share of year 2005 observations in total observations of year 2005 observations in total observations (that is, both years). (that is, both years).

t

t

it

itit P

P

P

P

1

1

}2005,2002{t

tP

1919

CM – Second step - 3CM – Second step - 3

In this framework, accounting for In this framework, accounting for changes in the distribution of observable changes in the distribution of observable characteristics is equivalent to characteristics is equivalent to reweighing the consumption distribution reweighing the consumption distribution estimated in equation (6), so that the CM estimated in equation (6), so that the CM model becomes to be as:model becomes to be as:

(9)(9)ipi

ri yy ,02,02,02 )ln()ln(

2020

CM – Third step - 1CM – Third step - 1 The only step remaining is to add a measure of The only step remaining is to add a measure of

the unobserved component of consumption. If the unobserved component of consumption. If the dispersion in unobserved consumption is the dispersion in unobserved consumption is due to random events that are unrelated to due to random events that are unrelated to systematic differences across households, systematic differences across households, then there would be nothing more to say about then there would be nothing more to say about the error term. However, one reason to add a the error term. However, one reason to add a measure of the unobserved consumption is measure of the unobserved consumption is that the residual is unlikely to be just a random that the residual is unlikely to be just a random component of consumption. component of consumption.

2121

CM – Third step - 2CM – Third step - 2 CM first estimates a consumption model for the CM first estimates a consumption model for the

2002 data, and ranks all the households on the 2002 data, and ranks all the households on the basis of the residual distribution for that year. basis of the residual distribution for that year. Then CM assigns to each household in year Then CM assigns to each household in year 2002, the value of ranked residual from the 2002, the value of ranked residual from the empirical distribution of residuals in year 2005 empirical distribution of residuals in year 2005 (equation (5)) that corresponds to the year (equation (5)) that corresponds to the year 2002 rank. We now have the counterfactual 2002 rank. We now have the counterfactual consumption, the consumption that would have consumption, the consumption that would have been observed in 2002, if the parameters, the been observed in 2002, if the parameters, the distribution of covariates and the unmeasured distribution of covariates and the unmeasured determinants of consumption are as in 2005. determinants of consumption are as in 2005.

2222

CM – Third step - 3CM – Third step - 3 From equations (5) and (9), this From equations (5) and (9), this

counterfactual wealth distribution can be counterfactual wealth distribution can be rewritten as:rewritten as:

(10)(10)

Where, denote the value of the ranked Where, denote the value of the ranked residual in 2005 assigned to a household residual in 2005 assigned to a household with the same residual rank in year 2002. with the same residual rank in year 2002.

)ˆ()(ln()ln( 05,0205,02,05,02,02,02r

iiri

pii

ci Xyy

ri,05

2323

Here we have further updated the Here we have further updated the poverty mapping using the new poverty mapping using the new LSMS conducted in 2008.LSMS conducted in 2008.

Clearly the counterfactual Clearly the counterfactual distribution corresponding to the distribution corresponding to the 2008 is less accurate comparing 2008 is less accurate comparing to the one of 2005.to the one of 2005.

However, the results are still However, the results are still good, and the errors still under good, and the errors still under control.control.

2424

IMPLEMENTATION OF THE POVERTY IMPLEMENTATION OF THE POVERTY MAPPING IN ALBANIAMAPPING IN ALBANIA

THE DATATHE DATA::  

Population and Housing Census (2001)Population and Housing Census (2001)        Reference Time: 31 March 2001      Reference Time: 31 March 2001      Number of Households: 726      Number of Households: 726,,895895      Number of Persons: 3      Number of Persons: 3,,069069,,275275      Collected Information: Building,       Collected Information: Building,

Dwellings, Household, Individuals.Dwellings, Household, Individuals.

2525

Living Standard Measurement Study (LSMS, Living Standard Measurement Study (LSMS, 2002, 2005 & 2008)2002, 2005 & 2008)        Reference Time: Spring 2002, 2005       Reference Time: Spring 2002, 2005 and 2008and 2008      Sampling Frame: 4 Strata, 450 PSU      Sampling Frame: 4 Strata, 450 PSUss

(corresponding to the EA in the (corresponding to the EA in the Census), 8 Household per PSUCensus), 8 Household per PSU

      Number of Households: 3599      Number of Households: 3599

      Collected Information: Household,       Collected Information: Household, Food Consumption, Diary, Food Consumption, Diary, Community, Community, PricePrice

2626

POVERTY MEASURES:POVERTY MEASURES:  The procedure for estimating the poverty The procedure for estimating the poverty measures has been applied for the whole of measures has been applied for the whole of Albania and disaggregated at seven levels:Albania and disaggregated at seven levels:a)  Rural – urban level;a)  Rural – urban level;b) The four strata used in sampling the b) The four strata used in sampling the LSMS;LSMS;c)  The six strata for which the linear c)  The six strata for which the linear regression models have been estimated;regression models have been estimated;d) The 12 Prefectures (or Counties);d) The 12 Prefectures (or Counties);e)  The 36 Districts;e)  The 36 Districts;f)  The 374 Communes/Municipalities;f)  The 374 Communes/Municipalities;g)  The 11 Mini-municipalities in which the g)  The 11 Mini-municipalities in which the city of Tirana is divided.city of Tirana is divided.

2727

Future work:Future work:

New Poverty Mapping using the fresh 2011 New Poverty Mapping using the fresh 2011 Census and the fresh 2011 LSMSCensus and the fresh 2011 LSMS

THANK YOU FOR YOUR ATTENTION!!THANK YOU FOR YOUR ATTENTION!!