Is Poverty Measurement More Like a Routine Checkup or Battlefield Triage?

Poverty fighters unable to attend yesterday’s Massachusetts Avenue Development Seminar here at CGD (video) missed a lively debate between James Foster and Martin Ravallion on the new Multidimensional Poverty Index (MPI) developed by Foster and co-authors (available online here).  This new approach to measuring poverty, which incorporates the capabilities approach into classic Foster-Greer-Thorbecke poverty measurement, has already been adopted  by Mexico and is under consideration in Columbia and at the UNDP.Foster, a professor at George Washington University  and one of the inventors of the FGT poverty measurement framework, opened with a clear and helpful exposition of the principles behind the MPI.  Essentially, he proposes that the measure of the number of “poor” people in a country – or in the world – be constructed by first choosing a set of relevant dimensions on which to rank people (e.g. access to health care, years of schooling completed, nutritional status, as well as consumption), deciding a cutoff poverty threshold for each dimension, and then counting the “total poor” by adding up the weighted number of people below the cutoff on at least k of the different dimensions.  (Omitted dimensions implicitly get a weight of zero.)Martin Ravallion, of the World Bank’s Development Research Group (his paper is also available for download here), then pointed out that the weights used to aggregate Foster’s multiple dimensions are essentially arbitrary, and that in lieu of arbitrary weights one might take one of two approaches:  (1) use prices (or shadow prices) for weights, or (2) simply present governments with several different measures of poverty, one for each of the relevant dimensions, leaving the weighting of those dimensions to the government.  To illustrate his point, Ravallion asked the audience to remember their last appointment with a physician for a general physical examination – a “routine checkup.”  He asked, “Would you really want the doctor to summarize all the information on all the tests with a single composite index of your health?”Then came James Foster’s rebuttal.  He chose to respond to Ravallion’s example from the medical setting by pointing out to the audience, “In the medical setting, doctors and nurses frequently must assign priority to patients in a process known as `triage.’  When assigning this priority, these medical professionals are essentially collapsing all the dimensions of a person’s health into a single index.”To me, this is the crucial distinction.  Implicit in Martin’s question to the audience is the presumption that neither the family doctor nor the patient has a well-specified valuation function with which to aggregate all the dimensions of health.  So the physician typically discusses each dimension in turn, focusing on those which seem below par.  Martin is suggesting that governments should do the same, separately measuring income poverty, nutritional status, educational attainment, and health service access, and should then decide how to appropriately intervene in each arena.  In the physical examination example, where the situation is not life threatening, both the doctor and the patient presume that the patient’s valuation of the information deserves priority, since presuming otherwise would be unnecessarily paternalistic.Implicit in James’ use of the triage example is the existence of a clear objective function which the medical professional on the battlefield is assumed to be able to assess:  the probability that seeing the physician will materially change the patient’s chances of survival over the next few minutes.  In the triage example, all presume that the medical professional must weight the various dimensions of health and make the decision paternalistically – in order to save the most lives with the available time and resources.Which of these implicit scenarios do you think is a more appropriate analogy for the situation faced by institutions that measure poverty?  Your answer to that question may determine whether you will prefer Foster’s proposed MPI to the traditional alternatives described by Martin Ravallion.To me, a way to move forward on poverty measurement was suggested by James Foster’s remark in passing that, “An MPI index tells us what people need to do to get out of poverty.”  But does it really?With household level panel data long enough to follow a child from birth to maturity, one could use a family’s early scores on all the dimensions of the MPI in order to explain the subsequent escape from poverty of that family’s grown children.  (One such dataset is analyzed in this paper.)  For the country context in which the study is performed (and subject to the usual conditions for a regression to estimate causal pathways), the multiple regression coefficients would provide weights for an MPI which has the specific meaning that Foster suggests would be useful.  Such a function would be comparable to a “triage” assessment on the battlefield – and when associated with the costs of changing the various dimensions, would provide policymakers with clear guidance on how to increase the probability that a child will do well as an adult.It is unlikely that the coefficients on the various dimensions of early family well-being (i.e. health care, education, nutritional status, etc.) would be proportional to the prices (or shadow prices) at which the family could potentially have purchased those dimensions.  Divergence of the coefficients from proportionality with the prices could be due to any number of market imperfections in the early family’s environment or simply to the early family’s  seeking objectives other than the child’s future escape from poverty.  Use of a price-weighted poverty index pays obeisance to the sovereignty of the family’s preferences and decisions, with the caveat that price-distortions be remedied.  Use of the hypothetical child-success-regression-coefficient-weighted MPI adopts the paternalistic view that government should maximize the chances that today’s children will escape poverty as adults, regardless of the preferences of the parents.  (For an overview of weighting options, see this paper by Decancq and Lugo from the Oxford Poverty & Human Development Initiative web site.  Foster’s MPI is an example of Decancq and Lugo’s approaches #4 or #5, while my regression-derived weights fit into category #8, hedonic weights.)Which would I prefer?  I like the idea of the child-success-regression-weighted MPI, because I can imagine parents supporting it.  They might endorse the government taking actions in the long-term best interest of their children, knowing that they would not themselves choose to take the same actions.  But I am much less enthusiastic about an expert-weighted MPI, even if the experts believe that the weights capture the same idea.  In the absence of evidence about the contribution to a child’s future success of each dimension of poverty, I would prefer to present the government with the full results of its routine check-up, and leave the weighting to them.