Microeconomic determinants of inequality in Pakistan

Felix Naschold Cornell University & University of WyomingChristopher B. Barrett Cornell UniversityAAEA 27 July 2010A stochastic dominance approach to program evaluation And an application to child nutritional status in arid and semi-arid Kenya1A theoretical paper: program evaluation VS. stochastic dominanceEmpirical application: ALRMP evaluation projectAAEA paper comments, pleaseMotivationProgram Evaluation MethodsBy design they focus on mean. Ex: average treatment effectIn practice often interested in distributional impactLimited possibility for doing this by splitting sampleStochastic dominanceBy design look at entire distributionNow commonly used in snapshot welfare comparisonsBut not for program evaluation. Ex: differences-in-differencesThis paper merges the two Diff-in-Diff (DD) evaluation using stochastic dominance (SD)

2As much as this is true for many parts of the worlds in terms of MDGs, also for villages in rural India

What do we do about this as researchers?

One aspect: who moves in and out of poverty? Characterstics and targeting?2Main Contributions of this paperProposes DD-based SD method for program evaluationFirst application to evaluating welfare changes over timeSpecific application to new dataset on changes in child nutrition in arid and semi-arid lands (ASAL) of KenyaUnique, large dataset of 600,000+ observations collected by the Arid Lands Resource Management Project (ALRMP II)(one of) first to use Z-scores of Mid-upper arm circumference (MUAC)33Main Results4Methodology(relatively) straight-forward extension of SD to dynamic context: static SD results carry overInterpretation differs (as based on cdfs)Only up to second order SDEmpirical resultsChild malnutrition in Kenyan ASALs remains direNo average treatment effect of ALRMP expendituresDifferential impact with fewer negative changes in treatment sublocationsALRMP a nutritional safety net?

Program evaluation (PE) methods5Fundamental problem of PE: want to but cannot observe a persons outcomes in treatment and control state

Solution 1: make treatment and control look the same (randomization)Gives average treatment effect Solution 2: compare changes across treatment and control (Difference-in-Difference)Gives average treatment effect:

To next slide: So what is the key drawback of standard PE methods?5New PE method based on SD6Objective: to look beyond the average treatment effectApproach: SD compares entire distributions not just their summary statisticsTwo advantagesCircumvents (highly controversial) cut-off point. Examples: poverty line, MUAC Z-score cut-offUnifies analysis for broad classes of welfare indicatorsDefinition of Stochastic Dominance7First order: A FOD B up to iff

Sth order: A sth order dominates B iff

MUAC Z-scoreCumulative % of populationFA(x)FB(x)0xmax

SD and single differences8These SD dominance criteriaApply directly to single difference evaluation (across time OR across treatment and control groups)Do not directly apply to DDLiterature to date:Single paper: Verme (2010) on single differencesSD entirely absent from PE literature (e.g. Handbook of Development Economics)Expanding SD to DD estimation - Method9Practical importance: evaluate beyond-mean effect in non-experimental dataLet , G denote the set of probability density functions of . and The respective cdfs of changes are GA() and GB()Then A FOD B iff A Sth order dominates B iff

Expanding SD to DD estimation 2 differences in interpretation101. Cut-off point in terms of changes not levels.Cdf orders changes from most negative to most positive poverty blind or malnutrition blind.(Partial) remedy: run on subset of ever-poor/always-poor2. Interpretation of dominance ordersFOD: differences in distributions of changes between intervention and control sublocationsSOD: degree of concentration of these changes at lower end of distributionsTOD: additional weight to lower end of distribution. Sense in doing this for welfare changes irrespective of absolute welfare?

Setting and data11Arid and Semi-arid district in KenyaCharacterized by pastoralismHighest poverty incidences in Kenya, high infant mortality and malnutrition levels above emergency thresholdsDataFrom Arid Lands Resource Management Project Phase II28 districts, 128 sublocations, June 05- Aug 09, 600,000 obs.Welfare Indicator: MUAC Z-scoresSevere amount of malnutrition: 10 percent of children have Z-scores below -1.54 and -2.55 25 percent of children have Z-scores below -1.15 and -2.06

The pseudo panel used12Sublocation-specific pseudo panel 2005/06-2008/09Why pseudo-panel?Inconsistent child identifiersMUAC data not available for all children in all monthsGraduation out of and birth into the sampleHow?14 summary statistics mean & percentiles and poverty measuresFocus on malnourished childrenThus, present analysis median MUAC Z-score of children below 0Control and intervention according to project investmentResults: DD Regression13Pseudo panel regression model

No statistically significant average program impact

Results DD regression panel14(1)(2)(3)(4)(5)VARIABLESmedian of MUAC Z