World Bank - Grant Funding Request (GFR)siteresources.worldbank.org/.../18489.pdf · · 2015-05-22World Bank - Grant Funding Request (GFR) Ref. : 18489 Status ... SRP - Survey to

World Bank - Grant Funding Request (GFR)

Ref. : 18489 Status : Cleared Printed on : 05/22/2015

GFR 18489 - SRP - Survey to Survey Computation Tools

Team Leader : 00000196700 - Mr Jose Antonio Cuesta Leiva

Summary InformationStatus ClearedTF Number/Status - Estimated Grant Start Date/Closing Date

07/01/2015 To 06/01/2017

Grant Amount 200,000.00 USD Beneficiary Country WorldImplemented by Bank Executed

Grant linked to P155744 - Survey to Survey ImputationProject Status : ACTVProduct Line : RF

Disbursing Fund Type Project/activity supportTF Usage - Managing Unit 9078 - GPVDRResponsible Cost Center 9078 - GPVDRContributing Managing UnitsWindow Manager Ms Elena Chi-Lin LeeFunding Window 0000007059 - SRP - Poverty and InequalitySub-Fund TF082580 - SRP - Poverty and InequalityTrustee TF072202 - World Bank Strategic Research ProgramDonor TF602001 - MULTIPLE DONORS

This GFR includes the following sections: Basic Data Info, Basic Data - TTL Comment, Description, ProjectInformation, Disbursement, Program Specific, Financing, Allowed Expenses, Sector/Theme, Attachments.

Internal Bank Report Page 1 of 15



Comments/Requests by TTLHai-Anh Dung is the co-TTL of this project but could not include his name in the space provided. Please add.

DESCRIPTION

1. What is the Development Objective (or main objective) of this Grant?

The task of monitoring poverty trends on a timely and consistent basis is difficult, if not impossible for a number ofdeveloping countries where household consumption data are neither frequently collected, nor constructed usingconsistent and transparent criteria. The Development Objective of this Grant is to explore and develop alternativemethods that can provide imputation-based estimates of poverty on a frequent and consistent basis in thesecontexts.

2. Summary description of Grant financed activities

Two sets of activities are proposed under this Grant. While the first set of activities focuses on extensions andrefinements of current imputation techniques, particularly in the presence of unexpected shocks and risks, thesecond set of activities is concerned with the application of these techniques to several different country-specificchallenges

3. (Optional question) What can/has been done to find an alternative source of financing, i.e. instead of a Bankadministered Grant?

4. What are the main risks related to the Grant financed activity ? Are there any potential conflicts of interest forthe Bank? How will these risks/conflicts be monitored and managed?

The risks associated with these activities are minimal. Some risks are related to data access which can potentiallydelay the proposed work, but we will minimize these risks in various ways, such as implementing the proposedactivities in a timely manner as soon as data are available or working closely with the client governments andencourage best practices with data sharing (e.g., through capacity building and training exercises).

SRP

1. How does (do) the objective(s) of this proposal align with the World Bank Group#s twin goals? What are the keythematic research questions (from the 2015 Call for Proposals) being addressed in this research?

The World Bank has recently proposed an ambitious goal of reducing the global extreme poverty rate to no more than3 percent by 2030. Such efforts are predicated on an ability to reliably assess and monitor progress since trackingpoverty trends can help us understand which policies work and which do not work, and how efficient they are.Producing reliable poverty estimates by conducting household expenditure (consumption) or income surveys,however, require significant financial and technical resources. As a result, consumption surveys are typicallyconducted every few years by statistical agencies, and poverty estimates are not available in the intervening yearsduring which surveys have not been implemented. Another challenge to tracking poverty trends is that survey designmay change over time, thus making consumption data and poverty estimates not comparable between differentrounds.

Our proposal has a very close alignment with both the Bank#s twin goals of poverty reduction and shared prosperityand the objectives of the SRP 2015 Call for Proposal. The SRP#s sub-theme of Conceptualization and Measurement ofWell-being (under the Poverty and Inequality Theme) specifically calls for a better understanding of #alternativetechniques for tracking economic well-being on a more frequent basis# to monitor poverty and prosperity without




full consumption data#, which is exactly our proposed topic of study. By offering to investigate a host of issues thatare both global and country-specific, our proposal is expected to produce valuable contributions on both themethodological and policy fronts that can significantly improve the quality of the Bank#s poverty measurement workand, by extension, on the diagnostic, analysis and policy recommendations in support of poverty reduction and sharedprosperity.

2. Provide a literature review & explain study#s intellectual merit. For Program proposal (PP), provide detailedrationale for the selection of issues within an overarching topic and how it helps to address critical problemsfacing Bank clients/operations.

Estimation of poverty is a rather involved process where technical expertise is required, and often subtlemeasurement issues are confronted. In particular, if poverty estimates are to be compared over time, a crucialrequirement is that both the consumption aggregates and poverty lines be consistently constructed and be strictlycomparable across survey rounds. It turns out, however, that this seemingly undemanding condition is less oftensatisfied than one might think. (See also Deaton and Grosh (2000) and Crossley and Winter (forthcoming) for generalreviews on the influence of survey design on the quality of consumption data in developing and richer countriesrespectively). Another issue that commonly hinders the tracking of poverty over time is that consumption surveys aretypically conducted only occasionally (particularly in developing countries), and poverty estimates are not availablein the intervening years during which surveys have not been implemented. Furthermore, collecting, cleaning, andpreparing data for analysis can be a protracted process that, at times, can span multiple years from the start of fieldwork to the time when the data are ready for analysis. In all these cases, the challenge can be broadly regarded asone involving missing data: consumption data are available in one period but in the next period(s) are either notavailable, or are not comparable.The topic of imputing missing consumption data from one survey to another (i.e., survey-to-survey imputation) tobetter measure poverty has been studied by a small, but growing, literature. The estimation framework utilized bymost current economic studies that focus on poverty comparisons appears to be largely based on earlier workexploring the feasibility of survey-to-census imputation by Elbers, Lanjouw, and Lanjouw (2003). Applying thedistributions of the estimated model parameters of consumption from a household expenditure survey onto itsoverlapping variables with the census, and with appropriate adjustments for the error term, Elbers et al. (2003)predict consumption data into the census. These data can then be disaggregated to estimate poverty at loweradministrative levels than are possible using the household survey alone. Following this approach, Mathiassen (2009)proposes an exact expression for the standard errors instead of the Taylor approximation formula offered by Elbers etal. (2003); in addition to imputing consumption to predict poverty, she also imposes a stricter parametric functionalform on the error term and uses the probit model to directly estimate poverty headcounts. The latter difference issimilar to another approach employed by Tarozzi (2007), who proposes a two-step inverse probability weightingestimator where poverty can be estimated as a probit function of household consumption and other characteristics inthe target survey, with the relevant weights derived in the first step from the change in the distribution of householdcharacteristics across the two surveys.

On the empirical side, employing Elbers et al. (2003)#s framework, Stifel and Christiaensen (2007) combine thehousehold expenditure survey with more recent rounds of the Demographic and Health Survey (DHS) in Kenya toimpute household consumption in the latter. A more recent paper by Christiaensen et al. (2012) impute povertyestimates using data from several countries including China, Kenya, Russia, and Vietnam. Using seven rounds ofhousehold survey data from Uganda, Mathiassen (2013) also finds imputation-based poverty estimates to accuratelytrack the true poverty rates in most cases. In the same spirit, another approach is to combine a householdexpenditure survey and a more recent labor force survey to impute consumption into the latter and subsequently toestimate poverty. This approach has been implemented for Mozambique by Mathiassen (2009); Douidich et al. (2013)similarly take advantage of an almost identical design between the household expenditure survey and the LFSs inMorocco to impute poverty rates in the latter and find very encouraging results. (Also see Dang, Lanjouw, andSerajuddin (2014) for more literature review).

Most recently, Dang et al. (2014) develop a framework for survey-to-survey poverty imputation with several novelfeatures. Their framework imposes few restrictive assumptions, works with simple variance formulas, provides




guidance on the selection of control variables for model building, and can be generally applied to imputationinvolving surveys with either the same, or differing sampling designs.

Our proposed study will both employ and further refine this imputation framework. Our contributions are boththeoretical and highly policy relevant. On the theoretical front, we offer to provide new findings on the impacts ofunexpected shocks on imputation accuracy, which is a topic that receives little, if any, attention in the developmenteconomics literature. On the empirical front, our proposed study offers new insights on the interwoven connectionbetween subjective and objective well-being, the possibility of updating poverty estimates more frequently for alarge-size country, and the welfare status of refugees. The findings on these topical issues can be directly translatedinto policy advice and are highly relevant to the Bank operation in the cited countries. At the same, these empiricalresults can also help us sharpen our imputation tools and are readily generalizable to other similar contexts.

ReferencesCrossley, Thomas F. and Joachim K. Winter. (forthcoming). #Asking Households About Expenditures: What Have WeLearned?# in Carroll, C., T. F. Crossley and J. Sabelhaus. (Eds.). Improving the Measurement of ConsumerExpenditures. Studies in Income and Wealth, Volume 74. Chicago: University of Chicago Press.Dang, Hai-Anh and Minh Cong Nguyen. (2014). "POVIMP: Stata Module to Provide Poverty Estimates in the Absence ofActual Consumption Data." Statistical Software Components S457934. Boston College, Department of Economics.Dang, Hai-Anh, Peter Lanjouw, and Umar Serajuddin. (2014). #Updating Poverty Estimates at Frequent Intervals inthe Absence of Consumption Data: Methods and Illustration with Reference to a Middle-Income Country#, World BankPolicy Research Paper No. 7043. Washington DC: The World Bank.Deaton, Angus and Margaret Grosh. (2000). #Consumption#. In Margaret Grosh and Paul Glewwe. (Eds.), DesigningHousehold Survey Questionnaires for Developing Countries: Lessons from Ten Years of LSMS Experience. Washington,DC: The World Bank.Douidich, Mohamed, Abdeljaouad Ezzrari, Roy van der Weide, and Paolo Verme. (2013). #Estimating QuarterlyPoverty Rates Using Labor Force Surveys: A Primer.# Policy Research Working Paper No. 6466. Washington DC: TheWorld Bank.Elbers, Chris, Jean O. Lanjouw, and Peter Lanjouw. (2003). #Micro-Level Estimation of Poverty and Inequality.#Econometrica, 71(1): 355-364.Marotta, D., Yemtsov, R., El-Laithy, H., Abou-Ali, H., and Al-Shawarby, S. (2011) #Was Growth in Egypt between 2005and 2008 pro-poor. From Statistic to Dynamic Poverty Profile#. Working paper. Washington DC: The World Bank.Mathiassen, Astrid. (2009). #A Model Based Approach for Predicting Annual Poverty Rates without Expenditure Data#.Journal of Economic Inequality, 7:117#135.---. (2013). #Testing Prediction Performance of Poverty Models: Empirical Evidence from Uganda#. Review of Incomeand Wealth, 59(1): 91#112.Stifel, David and Luc Christiaensen. (2007). #Tracking Poverty Over Time in the Absence of Comparable ConsumptionData#. World Bank Economic Review, 21(2): 317-341.Tarozzi, Alessandro. (2007). #Calculating Comparable Statistics from Incomparable Surveys, With an Application toPoverty in India#. Journal of Business and Economic Statistics, 25(3): 314-336.World Bank (2011) #Arab Republic of Egypt. Poverty in Egypt 2008-09, Withstanding the Global Economic Crisis#,Social and Economic Development Group, Middle East and North Africa Region, Report No. 60249-EG. Washington DC:The World Bank.---. (2012). World Development Report 2013: Jobs. Washington DC: The World Bank.

3. Describe analytic design & methodology. Individual Proposal: elaborate on hypotheses, conceptual framework,data; PP: describe broad research design, components, and how they#re linked & coherently contribute to theoverall objective.

We propose to extend the current survey-to-survey poverty imputation framework on both the theoretical andempirical fronts. In particular, our proposed activities can be grouped under two following broad categories, one isrelated to the impacts of risks on poverty imputation, and the other related to country-specific or topical challenges.While this rough categorization is useful for presentation purposes, our proposed work under both categories are




complementary to each other, and can offer generalizability to other contexts as well as policy insights that arecountry-specific.

The role of risks and shocks in poverty imputationOne of the key assumptions traditionally made by cross-survey imputation studies is that the estimated parameters(i.e., estimates of beta#s and the variance of the error terms) in the household consumption model remain stableover time. This can be relaxed by assuming instead that the changes in the independent characteristics can wellcapture the changes in poverty over time (Dang, Lanjouw, and Serajuddin, 2014). Both versions of this assumption,however, may not hold when unexpected shocks occur that result in substantive changes to the economy and thedynamic relationship of these model parameters.

We propose to improve the analytical treatment of risks, shocks and resilience issues in existing cross-surveyimputation techniques, which remains an unexplored area in the literature. In particular, we seek to address thefollowing questions: (1) how sensitive existing cross-survey imputation techniques are to risks and shocks?; (2) howcan these cross-survey imputation techniques be improved to better address systemic and/or idiosyncratic risks andshocks?; and (3) can information on resilience and compensation strategies be integrated in cross-survey imputationtechniques to improve the accuracy of estimates?

To answer these questions, we will follow a two-pronged approach. Our first approach is to analyze alternativecross-survey imputation models that differ in the way they include variables related to risks, shocks and resiliencestrategies. For example, we will analyze whether and how the inclusion of employment vs health related shocks canaffect imputation-based estimates compared to those derived from the reference scenario where such variables arenot included. We will examine a diverse set of shocks such as natural disasters, political instability or commodityprice shocks, as well as the varying time periods in which these shocks take place.As an example, employment shocks are of particular interest as evidence abounds that improvements in the labormarket constitutes a major uplifting channel for household consumption growth and poverty reduction (see, e.g.,World Bank, 2012). We will seek to analyze the potential role of employment shocks on imputation-based povertyestimates using the household consumption survey (ENBC) and labor force survey (LFS) from Tunisia. Both surveyscollect data on individuals# current employment status and if these individuals do not work, their reasons for nothaving worked. Given high rates of job informality in Tunisia, the inability to work, say even for short periods of timedue to illness, may result in a non-negligible drop in welfare. One way to take advantage of the availability of suchquestions would be to test the extent to which their inclusion in the consumption model changes its predictivepower. In particular, comparing the variations in consumption explained (say, the R^2) in separate model runs withand without such regressors can provide useful information on the extent to which imputation techniques can beimproved when information on job-related shocks are collected in household surveys.

As another example, we can employ the household consumption data from Egypt (HIECS) to study how shocks such asfood price shocks can affect imputation accuracy. Notably, the year 2008 was characterized by large increases infood prices in Egypt that can `change the trend# of poverty in Egypt (World Bank, 2011). This can be well illustratedby the analysis by Marotta et al. (2011) that finds that poverty in 2008 decreased with respect to 2005. However, thisanalysis was based on the panel data collected before the food price crisis at the beginning of 2008. When data onhousehold consumption were collected after the crisis from other quarters of 2008, the poverty rate for this year wasfound to be actually higher than in 2005 (World Bank, 2011). We will make the most use of the detailed monthlyconsumption data to explore how the accuracy of imputation can possibly vary for the months with more spikes inprices versus the months without these sudden changes. This analysis will help provide some crude#but practicallyuseful and original#estimates of the potential bias of assuming away differences in regional and temporal pricevariations when imputing from one survey to another.

In a similar spirit, we will also analyze data from other countries such as Pakistan and Senegal#for which expenditurehousehold and/ or labor force surveys exist with different frequencies and both idiosyncratic and systemic shocks canbe traced in those surveys. This list of countries can grow when more data become available. We aim at coming upwith a range of estimates for the magnitude of the impacts of different types of shocks on imputation-based povertyestimates.




Our second approach is to further develop the theoretical framework offered by Dang et al. (2014) to integrateunexpected shocks in the estimation models. One promising direction is to probe more deeply into the decompositionof poverty change over time that can be attributed to the estimated returns to characteristics (i.e., beta#s) and theunexplained part (i.e., error terms) to understand better how risks may differentially affect each component, and towhat extent. We will implement both simulation exercises and analyze real survey data for this component. Whileeach of these two different approaches is promising in its own right, these are closely related and can supplementeach other.

Poverty imputation for topical challenges or country-specific contextsWe propose to investigate the application of poverty imputation to the three countries below, each of which offers auniquely defined set of technical challenges and policy lessons.

BhutanHappiness and different dimensions of non-monetary well-being such as health and education feature prominently inthe public policy arena in Bhutan. The National Planning Commission in Bhutan even screen new policies againstcriteria based on the different components of the Gross National Happiness (GNH) index, which is constructed from anationally representative household happiness survey. At the same time, this country also implements the regularLSMS-type household survey (i.e., Bhutan Living Standards Measurement Survey) to produce standardconsumption-based poverty measures. However, the happiness survey does not collect consumption data, andconversely, the consumption survey does not collect happiness data that allows for such studies. We thus propose touse imputation methods to impute poverty in the happiness survey, as well as happiness into the consumption survey,to produce better analysis on the linkage between poverty and happiness.

One specific technical challenge is that, happiness variables (e.g., with life or work) usually have a discrete(categorical) nature, which is quite different from the continuous household consumption variable. This change to thedependent variable would require corresponding changes or improvements to the existing imputation techniques toensure better estimates.

IndiaAccounting for 30 percent of the global poor, poverty reduction in India is not only important for this country but alsofor the whole world as well. Tracking poverty trends is priority work that receives much public attention in thiscountry. The main source of household survey data, the India#s National Sample Survey (NSS), only collects fullconsumption data#or the #thick# survey round#every few years. For the intervening years, the NSS collects areduced version of the consumption data#or the #thin# survey round#which is rarely used to provide official povertyestimates. We propose to further employ imputation methods to produce imputation-based poverty estimates for theyears where only the thin survey rounds are available. The lessons learnt will be relevant to both the public discourseon poverty reduction and will also help improve survey designs where a reduced list of consumption items may haveto be used, say, for costs-saving purposes.

SyriaRefugees are not captured by World Bank poverty statistics because regular national household surveys seldom, ifever, collect data on these population groups. Our proposed work on Syrian refugees in Iraq, Jordan, Lebanon andTurkey would help to fill this critical data gap by opening a window on what is currently a black box for the study ofpoverty. This work can better inform policy design that aims to improve the well-being of refugees and help, in turn,to mitigate the impact of refugee crises on hosting communities.

We aim to produce a full welfare and needs assessment for Syrian refugees based on cross-survey imputationtechniques. Our proposed study will contribute to enhance our understanding of the following questions: 1) Who arethe refugees? 2) How poor are the refugees? 3) How vulnerable are refugees and why? 4) How effective are refugeeassistance programs in addressing vulnerabilities and targeting, and how can they be improved? The study builds onan on-going World Bank-UNHCR welfare study on Syrian refugees and will expand this work by using cross-surveyimputation methods to pass from sample survey analysis to population wide analyses.




The study will rely on two data sets. The first data set is the UNHCR profile Global registration system (proGres). Thisis the main global administrative database used by the UNHCR to register refugees. The data provided by the UNHCRinclude all records for the registered Syrian refugees, the first time that this agency shares data outside theorganization. The value of this database is its large size and the information on key socio-economic characteristics ofrefugees such as date of birth, place of birth, gender, date of flight, arrival date in Jordan, registration date,ethnicity, religion, specific needs, and vulnerabilities. But this database contains no variables on householdconsumption. The second dataset is the Home Visits (HV) database, which collects information on both income andexpenditure and richly includes almost 200 variables that can be used for welfare analysis. In addition, the HVdataset is also a large database that collects information on about a third of all registered refugees. However, amajor shortcoming is that this database focuses on specific targeting purposes and thus do not offer a random sampleof the refugee population.

Applying cross-survey imputation on the proGres and HV databases allows us to impute welfare data for allhouseholds not participating in the home visits program, which can usefully expand the study population bythree-folds and cover the entire population of Syrian refugees. Remarkably, Syrian refugees are also dispersed acrossseveral countries with very different economic conditions and policies towards refugees. We will thus add anadditional and new layer of analysis by implementing imputation from one country to another to investigatecounterfactual scenarios where we can simulate the impacts of the policies regarding refugees adopted in onecountry versus those in other countries. Similarly, we can also simulate alternative policies to those adopted by theUNHCR or the World Food Program (WFP).

We will also combine insights from the development economics literature with those of the humanitarian assistanceliterature. The average refugee status is currently estimated at 17 years and the Syrian refugees are expected to stayin refugee status for years to come. The developmental and humanitarian camps are traditionally separate spheres ofinterventions but this is not justified by the actual needs of the target population and creates a fracture in thecapacity of the international community to assist refugees. These #counterfactual# estimations should provide for thefirst time an assessment of welfare policies for refugees and provide some indications on the feasibility of alternativeshort-term humanitarian interventions and medium and long-term development policies. Bridging developmental andhumanitarian interventions is currently considered as the first priority by donors worldwide and this study canpotentially make original contributions and help shed light on this objective.

4. Describe Implementation arrangements. Identify timeline, key team members and their roles. If thepartnership is involved, describe the partnership arrangements, and the respective responsibility of Bank unitsand partners.

The co-TTLs will lead and supervise all activities and manage the resources. The project will be implemented in closecollaboration with colleagues at and outside the Bank. The proposed work will take place starting from the beginningof FY 2016, with research products and policy briefs to be produced every six months thereafter. The component onBhutan may start later after data from the happiness survey is obtained. Dissemination events (i.e., training courses,workshops) will be provided throughout the life of the project.

Our team members, in alphabetical order, and proposed responsibilities are as follows1) and 2): Jose Cuesta (Senior Economist, GPVDR) and Hai-Anh Dang (Economist, DECPI), co-TTLs: manage project,developing technical methods, writing papers, provide training and presentations3) # 7): Gabriel Lara Ibarra (Economist), Rinku Murgai (Lead Economist), Utz Pape, Carlos Sobrado (Senior Economist),Hiroki Uematsu (Economist), and Paolo Verme (Senior Economist) (GPVDR): developing technical methods, writingpapers, provide training and presentations8) Peter Lanjouw (Professor, University of Amsterdam): developing technical methods, writing papers, providetraining and presentations9) Martin Rama (Chief Economist, South Asia Chief Economist#s Office): provide overall guidance and collaborationwith Bhutan Center for Happiness Study, contribute to writing paper and policy advice10) Consultants: data preparation




5. Outline the expected outputs (working paper, publication, computational/analytical tools, datasets, etc.). Forboth individual and program proposal, please specify the expected date of delivery for each output.

The expected outputs include:# At least 4 policy research papers for publication in peer-reviewed journals (i.e., 1 to 2 method papers thatdiscuss poverty imputation that incorporates risks; 3 poverty imputation papers, one for each of the countriesBhutan, India, and Syria)# Several methodological briefs/ notes to guide Bank task teams in the use of the technique under differentrisk/shock scenarios; one methodological note will draw from the lessons learnt from all the countries under study# Tailor-made software modules (i.e., Stata algorithms) that automate estimation procedures# Several presentations/ training courses provided for

a) national statistical agencies (e.g., those in Vietnam have expressed interest) b) researchers/ university students (e.g., National Economics University, Vietnam, and University of

Amsterdam, the Netherlands, have expressed interest)# Conferences/ workshops (local and international) and dissemination events# In-house training for Bank staff working in poverty or related areas

We aim to improve significantly the existing poverty imputation techniques with our proposed activities. Beyond theresearch products that will contribute to academic knowledge, we will directly apply what we learn to improve thequality of Bank operation in our ongoing policy dialogue with governments.

6. Describe the beneficiary of the research, the relevance for policy in developing (or transition) countries andfor WBG Operations. Outline dissemination plans, including plans to reach policy makers.

Tracking poverty and shared prosperity is one of key priorities of the Bank#s work that provides indispensable inputsin both operational and technical advisory and assistance (AAA) products. For example, the Systematic CountryDiagnostic (SCD) and Country Partnership Frameworks (CPF) in the Bank#s new engagement model with countriesrequire strong and updated evidence on poverty and shared prosperity. In the absence of available household surveys,cross-survey imputation is increasingly used as a substitute. To the extent that the precision of this technique can beimproved, the quality of the Bank#s work in policy dialogue and operations will accordingly improve.

Our proposed country studies also provide a diverse set of lessons learnt with both specific and generalizable insights.In particular, India is a country with a large population that affects poverty estimates for the world, Bhutan offers aunique experiment on the use of subjective well-being in measuring a country#s progress. The ongoing Syrian conflictdirectly affects strategic clients for the World Bank, and refugees constitute a major issue in the Bank#s policydialogue and operations with all these countries.

7. Describe the capacity building components, including the collaboration with local partners, researchers fromdeveloping countries.

As briefly reviewed in an earlier section, poverty imputation methods are increasingly used for the Bank#s work. Inparticular, the method developed by Dang et al. (2014) has been employed to predict poverty trends for the ongoingregional flagship on poverty in Africa and the national poverty assessment report for India (Dang and Lanjouw, 2015).This paper was also presented at the recent World Congress of International Economics (Jordan, 2014) and will bepresented at the 2nd Annual Meeting of the International Association for Applied Econometrics (Greece, 2015).

A tailor-made Stata routine (#povimp#) that automates the estimation procedures in this paper has been madepublicly available to the research community and be downloaded from within Stata (i.e., type #ssc install povimp#)or the SSC Archives website (https://ideas.repec.org/c/boc/bocode/s457934.html), which contributes to the existingglobal stock of knowledge. We plan to continue similar dissemination activities that provide research to both Bankstaff and researchers and others outside of the Bank on a wholesale basis. We have made a presentation onimputation methods for local researchers in Jordan and Vietnam, who expressed strong interest in further and longertraining courses. We plan to provide similar training activities to staff at statistical agencies and researchers in other




developing countries. (Please see Section 5 on #Expected Outputs# above).

Our country-specific work also has a strong built-in capacity building component. For example, the work on the Syrianrefugees stems from a full partnership established between the World Bank and the United Nations High Commissionfor Refugees (UNHCR) since 2013. These two organizations have worked together in the course of the past two yearsto find innovative solutions to the Syrian refugees# crisis. As the two leading organizations in the development andhumanitarian fields, each of these organizations have long established relations with all other key partners assistingrefugees, from the donors# community and the international financial system to the NGOs community and thehumanitarian organizations system. In terms of stakeholders involved, this is perhaps one of the most comprehensiveoperations ever undertaken towards refugees. The UNHCR in particular relies on a large network of local staff andNGOs from the registration of refugees to data collection via sample surveys. The World Bank, on the other hand,contributes to build analytical capacity within the UNHCR and other humanitarian organizations that can havepermanent effects on the capacity of these organizations to improve on their effectiveness. As one example, thetechnical assistance provided by the World Bank to the UNHCR has already resulted in improved targeting of Syrianrefugees in Jordan and Lebanon.

8. Document evidence of the consultation process with relevant research and operations units. E.g. consultationconducted, comments received, & how comments were addressed. TTLs should also describe plans to maintainoperational and research consultation.

We have very strong collaboration between DEC and operation. The TTL-ship of the project is joint between stafffrom DEC and the global Poverty Practice.

We provide below as some examples some quick snapshots of the ongoing email exchange on poverty imputationbetween our team and the country team. Similar email exchange with other teams or more details on our variousdiscussions in person or by phone can be provided upon request.

1) Emails from Rinku Murgai (Lead Economist, TTL for ongoing India Poverty Assessment Report) discussing variousissues with imputationEmail No. 1

From: Rinku Murgai [mailto:[email protected]]Sent: Thursday, July 17, 2014 8:46 AMTo: Hai-Anh H. DangCc: Peter F. Lanjouw; Ambar NarayanSubject: Re: Fw: BBL: Diagnostics of Poverty and Shared Prosperity in India, by Rinku Murgai and Ambar Narayan (May19, 1230 - 2 pm)Hi Hai-Anh,

My turn to apologize. Between us, we need to start responding earlier to each other if we're going to get this done!I'm so glad to hear that you can get back to the India work. Several of the pieces that Pete and you have beenworking on, or that we've discussed are very relevant to our work. These are the priorities from my vantage point:

1. Wrap up the analysis and write-up a short note that we can share with others, assessing whether the declinebetween 2009/10 and 2011/12 is "real"2. Move to the thin round analysis to address the question of whether the thin rounds are suitable for inclusion in theBank's global database on poverty3. Work on #2 in the list below.

Of course, I'd be more than happy if you could work on #2 and #3 simultaneously. #3 is of greater personal interestand use for our work in the India program; #3 is a corporate priority and we made a commitment to deliver that lastFY, so can't delay that too much longer.




On 1, thanks for sending the decompositions. I'm not sure how to interpret those which reveals that I don'tunderstand the imputation method sufficiently well. I thought, for instance, for the 2004-05 to 2009-10 imputations,you are estimating a regression for 2004-05, and then applying the coefficients from that regression to the newvector of characteristics in 2009/10 and using that to predict poverty rates. If that is the case, why will there be acomponent due to changes in coefficients? I realize this question is probably dumb -- grateful if you could clarify.

Also, did you have a chance to do the backward imputations to check if we get similar results if we run the modelsfrom 2009 to 2004, and 2011 to 2009?

Please let me know if you'd like to discuss on the phone. I promise to be a more regular correspondent on this work!

thanks,

Rinku

Email No. 2From: Rinku MurgaiSent: Friday, March 20, 2015 9:40 AMTo: Hai-Anh H. Dang; Ambar NarayanCc: Peter F. Lanjouw; Peter Lanjouw ([email protected]); Urmila ChatterjeeSubject: RE: URGENT PLEASEHI Hai-Anh,

Many thanks for sending these results. Interesting to see the profiles with respect to education and caste coming outso clearly. The main source of household income ones are a little harder to understand, but I have to re-adjust mymind now thinking about this as a panel, rather than repeated cross section.

On the graphs, could you clarify what the lines represent? I expected them to be the sum of off-diagonal elements inTable 5 from the paper. But that doesn#t seem to be the case.

Could I also bother you with one more request? The figures show the percent of households (or population?) in eachgroup that moves up or down. For the same categories # i.e., education, caste and main source of household income# could you also produce graphs that show the distribution of movers across different values of each of these. Forexample, of all those who move up one or two brackets, what % have primary, secondary or tertiary education (somaybe a bar graph, with one column for moving up, one for moving down, one for staying poor, one for stayingmiddle class and one for overall population).. And similarly for caste and household source of income.

On your cohorts question, both versions are interesting. Keeping age fixed, or keeping sticking with one cohortacross periods. I think if the question is # was there more or less mobility between the two periods # the way you#vedone the analysis at present makes sense (ie.., keeping age group fixed).

Thanks also for clarifications on my other two questions, and for all your work on this paper.

Cheers,

Rinku

PS I am out of town next week, but will be meeting Pete next Friday in Delhi, and can catch up with him then if thereare any last pending issues.

2) Exchange with Shinya Takamatsu (Consultant, Team member for ongoing Africa Regional Poverty Flagship Report)on capacity buildingEmail No. 1




On Nov 11, 2014, at 4:58 PM, "Shinya Takamatsu" <[email protected]> wrote:Hi Hai-Anh,

Would you mind having a meeting tomorrow to get your feedback on the modeling for S2S imputation? Please let meknow your availability.

I am implementing the S2S imputation for African countries, but I want to learn from you on the variables I shouldconsider to include, etc.

Shinya

Email No. 2From: Shinya TakamatsuSent: Wednesday, November 12, 2014 6:46 PMTo: Hai-Anh H. DangCc: Andrew L. DabalenSubject: RE: Meeting on S2S imputationHai-Anh,Thank you very much for your valuable time. The meeting was very valuable. After reading your paper carefully andimplanting it to one country, I will get back to you.Shinya

3) Endorsement lettersEmail 1:From: Andrew L. DabalenSent: Wednesday, April 29, 2015 9:01 AMTo: Hai-Anh H. DangSubject: RE: SRP proposal with poverty imputation

Hi Hai-Anh,

Many thanks for sharing this concept note on imputation methods to obtain more frequent poverty data. We havebenefited in important ways from the research you and your co-authors have done in the past in many of our workand I am therefore a strong advocate for this work. As an example, we have used your work (with others) on Senegalin our poverty assessment to understand the profiles of the chronic poor, their potential size and how the plannedsafety nets could use this information for designing beneficiaries. This is very useful work indeed. We are also usingthese techniques to understand the size and profile of the chronic and transient poor in the Africa Poverty flagshipreport. There are many more contexts where we have used these techniques.

I think the country cases will be especially useful. I am especially excited about the proposed work on India. Many ofour clients are seriously interested in more frequent #light# surveys that would allow them to track poverty (and aselect list of indicators) more frequently (annually, at least). The unresolved issue is how we should design suchsurveys. I think drawing on solid research results would be helpful and what you are proposing would go a long wayto helping on that. The results on India in particular, but others as well, will be particularly informative on how weshould advise countries that are thinking about tracking poverty more frequently in a credible way.

Therefore, I am very supportive of this concept note and hope that it will be funded.

Let me know if I can be of further assistance,

Andrew




Email 2:

From: Gero CarlettoSent: Tuesday, April 28, 2015 9:16 PMTo: Hai-Anh H. DangSubject: Your proposal on survey-to-survey imputation

Dear Hai-Anh,I read with great interest your draft proposal on survey-to-survey imputation for improved poverty and sharedprosperity diagnostics. I find the proposed work extremely important and of great relevance to the work of the Bankand our development partners. The demand for high-frequency data puts increasing pressure on countries and poorlyfunded data producers which can be partly relieved through the application of rigorous imputation methods. As youand other colleagues have already demonstrated in some of your previous work, the judicious use of these techniquesin data scarce environment brings great value added and enhance enormously our ability to monitor progress ineradicating extreme poverty. Partners like UNICEF and IFAD have been quite supporting of the Bank#s effort formethodological improvements in this area and would greatly welcome your research and use the findings.Best of luck with the proposal submission.Best,gero

Email 3:

From: Christina E. Malmberg CalvoSent: Sunday, February 15, 2015 11:29 AMTo: Jose Antonio Cuesta LeivaCc: Maria Arribas Banos; Ana L. Revenga; Hai-Anh H. Dang; Paolo Verme; Gabriel Lara Ibarra; Rinku Murgai; HirokiUematsu; Emmanuel Skoufias; Pamela Gaye C. Gunio; Clara SerrainoSubject: Re: For Ana's clearance RE: Launch of the Call for Proposals for the Strategic Research Program (SRP)

Jose,

As discussed, this proposal is a good example of how we would combine SRP funds w/GE and CE sources to seekalternative ways of estimating poverty more frequently and without full C data. The team is great. Pls remember todiscuss w/Ghazala and David on anything PK and keep Tara in the loop on Syria.This is a winner!C

FINANCING

Is retroactive financing required? No

Transfer Schedule Amount in Grant Curr Condition 07/01/2015 200,000.00

Comment on Transfer Condition:

DISBURSEMENTS

Disbursement Summary in USD




Date From Date To Amount in USD Amount in USD 07/01/2015 12/31/2015 50,000.00 50,000.0001/01/2016 06/30/2016 50,000.00 50,000.0007/01/2016 12/31/2016 40,000.00 40,000.0001/01/2017 06/30/2017 40,000.00 40,000.0007/01/2017 10/01/2017 20,000.00 20,000.00

ALLOWED EXPENSES

Commitment Item Group Indicative in USD Indicative USD Equivalent Associated overheads 0.00 0.00Consultant fees 0.00 0.00Contractual services 0.00 0.00Equipment cost lease 0.00 0.00Equipment cost purch 0.00 0.00Extended term consul 0.00 0.00Media & wrkshp costs 0.00 0.00Staff Costs 0.00 0.00Temporary Suppt Staf 0.00 0.00Travel expenses 0.00 0.00

SECTORS

Description Percentage

THEMES

Description Percentage

RELATED PROJECT INFORMATION

Basic Project Information




Company code Team Leader 00196700 - Mr Jose Antonio Cuesta Leiva

IBRD Status Active Region/Cty 1W-World Project Type RF-Research Services Project Description Survey to Survey Imputation Project Definition P155744

Project DescriptionP155744-DESCO

Project Milestones

Usage Description Basic Forecast Actual02310 AIS Sign-off (B)02320 Activity Implementation Start (B)02330 Draft to Director (N)02340 Paper to Director (Y)02350 Paper to VP (Y)02360 VP Meeting (Y)02370 Paper to MD (Y)02380 Completion/Director Approval (N)02390 Board Approval (Y)02400 ACS (B)02410 Publication (N)

Project Financing Total Cost 0.00

Finance 0.00

Financing Gap 0.00

Source Amount FinancedBORR 0.00

DOCUMENTS ATTACHEDSRP_individual grant_survey imputation_1stround.docxcvApril 2015_short.docSRP Budget Plan_SurveytoSurvey .xlsxShort Bio and Publications_Rinku.pdfSRP Budget Plan_SurveytoSurvey .xlsx




cv-dang.pdfJoseCuesta_cv2015.pdf


World Bank - Grant Funding Request (GFR)siteresources.worldbank.org/.../18489.pdf · · 2015-05-22World Bank - Grant Funding Request (GFR) Ref. : 18489 Status ... SRP - Survey to

Documents