MULTIDIMENSIONAL POVERTY MEASUREMENT WORKSHOP POVERTY AND INEQUALITY MEASUREMENT AND ANALYSIS PRACTICE GROUP (PIMA PG)

Size: px
Start display at page:

Download "MULTIDIMENSIONAL POVERTY MEASUREMENT WORKSHOP POVERTY AND INEQUALITY MEASUREMENT AND ANALYSIS PRACTICE GROUP (PIMA PG)"

Transcription

1 MULTIDIMENSIONAL POVERTY MEASUREMENT WORKSHOP POVERTY AND INEQUALITY MEASUREMENT AND ANALYSIS PRACTICE GROUP (PIMA PG) Date: August 30 th, 2010 Time: 9h00 16h00 Chair: Peter Lanjouw/Jaime Saavedra 3. Panel Session Objective Gabriel Demombynes (Kenya Office AFTP2), Samuel Freije (LCSPP), Rinku Murgai (Delhi Office SASEP), Ruslan Yemtsov (HDNSP) and Nobuo Yoshida (PRMPR) shared their thoughts on conceptual issues or in country experiences they have had in the recent past. Gabriel Demombynes He has doubts about these multidimensional poverty indices, as he has expressed in the current blog debate. Similar Indices o Reading the literature on multidimensional poverty, it can be established that the natural antecessor of this index is the Unsatisfied Basic Needs approach (NBI) that was developed a year ago (or years ago?), especially in Latin America. The NBI, too, was based on a single data set, multiple deprivations at a household level, and a set of weights for each dimension studied. o One of the problems with the NBI was that the weighting scheme implied different tradeoffs between dimensions, which most of the time was difficult to defend. Assuming for instance, equal weights in an index with only two dimensions (access to sanitation and education), will say that having a toilet is considered equivalent to sending a child to school. o Given the difficulty in choosing weights and several exercises that showed an unsatisfactory behavior of the NBI (Ecuador), the World Bank advocated that governments in Latin America should not rely in this type of index. o It is not clear how the MPI is different than the NBI, as they seem to have the same problems. o Implicitly, decisions on how to allocate resources is based on the different weights that politicians give to dimensions. For example, how much is spent on infrastructure policies vs. educational programs. However, those discussions must remain in the political arena and not incorporated to the poverty measurement. Demand in countries like Sudan and Kenya o Sudan is actually more like two countries in one, North Sudan and South Sudan. o Constructing an MPI in a poor setting (Sudan and Kenya) is challenging because of the difference in institutional settings. o North Sudan There is a small crowd of academics who can understand the multidimensional poverty methodology and construct an index. However, given the nature of the regime, they would find weights in such a way that figures show an improvement in poverty trends especially in inequality. When James Foster was saying that there are no incentives to manipulate the index because no improvements could be shown in time, he wasn t thinking about Sudan. The regime would, in time, change weights and definitions if necessary to show an improvement in figures. o South Sudan There is an open government to discussion and ideas, but with a low capacity. At the moment, they are implementing a Referendum and the construction of an MPI would be way down in their list of priorities. 1

2 o o South Sudan is dominated by donors and if they ask for it, the Government will agree to construct an MPI. However, since there is no internal capacity to estimate the MPI, they would hire international consultants to do so. If this is the case, the point of making the index according to country specific needs would be difficult to achieve. Noone knows a country and its needs best than residents and members of that country. Kenya It is also an open government, but by Sub-Saharan Africa standards, they have a large pool of educated people that could implement an MPI. Given the latest political transformations, an MPI would be used to allocate resources. But what is gained from distributing funds basing decisions on an MPI instead of the traditional political discussion that would use a consumption-based poverty measurement? Additionally, who will be part of a committee that decides the weights? Interests to benefit certain sectors through transfers will play an important role in the construction process of the index. Even though it has been stressed that the MPI is easily understandable, in reality this may not be the case. Given the amount of freedom in choosing dimensions, weights and cutoffs, this will make the index accessible to only a few. Samuel Freije In Latin America, Mexico has been the front runner in the construction of an MPI. However, Colombia and Chile have started thinking about it and organized conferences to address the topic. The Case of Bolivia o Current president, Evo Morales, has a National Development Plan based on the lemma Vivir Bien (Live Good) which is a multidimensional concept. They want to emphasize the need of working in dimensions that previous governments had not addressed before. Thus, they need a multidimensional poverty measure. o There have been ideological and practical reasons why Bolivia wants to pursue a multidimensional poverty measure. Ideologically, they want to go beyond and against the western concept of poverty/well being and are even unwilling to work with economists. In practical terms, poverty headcounts have been stable over the past two decades while there have been improvements in other fronts. For instance, health indicators or school enrollment, which are closely related to poverty, have been improving but these changes are not being reflected in traditional poverty numbers. So, other things have changed but they don t show up. o The challenges of estimating an MPI number are huge They stated that they don t want experts to define the dimensions to be included in the index or the methods to aggregate. They are asking people around the country how to define happiness, vivir bien, etc. and what is well being. But differences and diversity within the country are huge, making it extremely difficult to compare answers and aggregate dimensions. People should first state what they expect and what is well being, and then experts will find a way to unify answers and construct the index. A problem arises when having such relevant measures for each group of the population. They will not likely be comparable within the country, because there will be differences in dimensions, in thresholds and in weights. Furthermore, comparisons in time can be problematic. On the other hand, there are multiple multidimensional measurements in Bolivia constructed by international institutions and organizations. When comparing results in terms of rankings, it is observed that they are highly correlated. So, what are the Bolivians gaining from constructing another measure which will show a similar ranking of the regions? 2

3 o Finally, UNDP carried a survey in which they asked Bolivians what were the most important dimensions of well-being. Around 70% considered a good job and earnings. Thus, this is a dimension which must not be excluded. However, it is inevitable that governments will want to construct a multidimensional poverty measure. Thus, the challenge of international institutions is to find the best ways to construct these indices attempting to solve the tension between relevance and consistency (comparability across countries and in time). Rinku Murgai Reactions to the release of UNDP-OPHI MPI o The cross-country comparison exercise has been useful, in the sense that whenever there is an aggregate index and a ranking, the media picks it up. This, in turn, raises questions and issues at a governmental level, increasing awareness of specific deprivations. o The UNDP-OPHI numbers for India were a reality check, reflecting the perceptions of the middle class. The rapid rates of growth have not always translated into improvements in education, health, etc. The 55% MPI for India reflects that the median Indian person is still deprived in at least 30% of the dimensions. o However, there are still problems with the methodology and a broad agenda for future improvements. Nonetheless, the case of Mexico shows that these indices may be powerful. The case of India o Measurement of poverty and identification of the poor, in India are two disconnected exercises. o The usefulness of the tool relies on the fact that it may be powerful as a targeting device. o India has been estimating a targeting index based on information collected by a below poverty line census, but there are many issues regarding the survey design and the construction of the index methodology. o The new census will be available on 2011 and this is the opportunity to apply the Foster and Alkire methodology, because the index meets the requirements of simplicity, transparency and verifiablility. o Given that the survey is in a pilot stage, the question on how to start the exercise from scratch is relevant and the government is actually in the process of consultation. Ruslan Yemtsov Even though the index is replicable to any one, this may not be feasible because of data constraints. Many countries do not share the data. Additionally, presenting an index based on non-official data sets would be problematic when estimating official figures. The methodology presents a credible and appealing way of integrating different dimensions, but the cross-country exercise does not take into consideration the quality of data. For instance, some DHS have problems with anthropometric indicators and so that would affect the overall quality of the MPI. There is concern of moving towards multi-topic surveys that do not guarantee data quality. It is clear that poverty is multidimensional and people are thinking about it in this way. The challenge now is to find a mechanism to expand traditional poverty measurement. A possibility is to define poverty in different dimensions and cutoffs. For example, what is the appropriate cutoff for education? What is educational poverty? These can be defined in a normative way or through economic rationale. In this sense, what is the schooling level that will guarantee lifetime earnings? Or it can be thought of as the compliment of public and private spending. How much public spending and private spending are necessary so that a household can attain a defined something? On the other hand, different observable achievements besides nutrition can be used to calibrate the poverty line. In terms of targeting, there are risks. Traditionally, the idea is to identify income poor and target has been done through cash transfers, to those below a certain line, or to those who meet some kind of criteria. The concern is that many countries with weak institutional settings, unlike Mexico, will have an incentive to displace income 3

4 poverty measures and move strictly to multidimensional indices. But all the cash transfers programs are based on income/consumption poor and so they will probably be ineffective when used to target multidimensional poor, identified with indices that do not include an income/consumption measure. If the movement is to multidimensional indices, other type of social programs must be developed. Nobuo Yoshida What can the WB and other international/multilateral organizations do as advisors to national governments? o There are still problems and ambiguities with income/consumption poverty, but plenty of studies have been made and advice can be provided. Useful recommendations and guidance can be given. o The MPI is, however, still in its initial stages and advice is difficult to give. Are countries the only players that can create an MPI? Governments can create bad quality indicators. How can academia and international organizations find a better MPI? How can we determine if an MPI is a good measure or not? What kind of weighting mechanism can be advised? o What does an MPI represent? Correlations between GDP per capita and USD 1.25 day poverty and GDP per capita and MPI are high, especially for less poor countries. In this case, either measure will work. However, much less correlation is found for poorest 20% countries and so, the MPI can play an important role by providing governments a clearer picture. The MPI could be a useful measure to verify progress in achieving goals, if there are high correlations with perceptions. How to set weights? o Setting weights by asking perceptions of a highly illiterate population, may not useful. This stresses the importance of finding alternative mechanisms to define the relative importance of each dimension. A possibility is to use subjective poverty questions and with data and statistical methods understand relationships between variables. An objective method to determine weights by measuring people s perceptions can be found, using data and avoiding expert judgment. Questions and Answers Questions/comments by the public o Answers by the presenters Does this approach lead to a better development policy? Final Remarks by Gonzalo Hernandez It is evident that there is a demand for multidimensional poverty indicators. Probably not in countries like Sudan, that have bigger problems, but around the globe, it is a fact. What can the WB do to answer these demands? The answer cannot be that the WB is not engaging in those practices. How can cutoff be determined? o Both economic theory and laws may be useful to set them. These should be established based on all of the available information. Investigation of all available sources is needed. Mexico is using the multidimensional poverty measurement to complement traditional figures, so as to get a clear picture of poverty and have a better diagnosis. They are not dismissing income poverty, because it also contains important information. 4

5 Final remarks by James Foster Lessons can be learned from different exercises or previous experiences, but certainly answers are not going to be homogenous for different clients. It is necessary to step out of the comfort zone of consumption/income poverty, where there are specific sets of recommendations for every step along the way. Hopefully, this level of comfort increases as further discussions and different applications are carried out. Final remarks by Sabina Alkire Multidimensional poverty measurement is far behind income poverty in building the conventions and capacity required in specific contexts. Now is a time to try to learn about the experience of people who once pursued the measurement of income poverty. It is essential to have better indicators and quality, especially in the health dimension. How can we develop health modules to construct health indices at the household level? The deficit in health and education are indices that reflect quality. Work (informality, work environment) and empowerment (freedom, social structures and discrimination) are two essential dimensions that have to be included in future work. There is an interface about politics and the construction of measures. Who is to decide about the relative importance of each dimension? At the end, it is the politicians, but as economists, the challenge lies in finding suitable statistical procedures to determine an appropriate way for setting weights. Final remarks by Maria Emma Santos There are two main questions from the discussion: How do we advise countries and what type of questions to include in surveys? Unfortunately, the authors don t have a final answer to these and it is a work to be done in the future. Final remarks by Peter Lanjouw The multidimensional poverty measurement discussion is much on the radar of circles involved in poverty analysis. It is important that at the Bank, we spend some time collectively thinking and debating the measurement involved in a multidimensional poverty analysis. Today s discussion was successful and enriching, as we learned about the motivations and technicalities of an approach. This will hopefully be an ongoing discussion. Multidimensional poverty analysis is very different from a multidimensional poverty index. Everyone agrees that poverty is multidimensional. The question that was discussed at the workshop, however, was if this multidimensionality should be measured in a single index or if it is better to analyze it separately. An analogy to the health of a person can be made. There are different tests that determine the health of a person: cholesterol, electrocardiogram, echocardiogram and many other indicators. Should all of these be combined in a single index to determine if a person is healthy or not? Or should they be analyzed separately to come up with an overall assessment? How do we think about an index? There are certain purposes for which a unique index may be useful and powerful. For example, raising awareness and stimulating debate. There may be less consensus that when designing a particular policy or intervention, a unique index is the way to go. Although there may be ongoing room for more debate and discussion on this issue. Atkinson and Bourguignon recognized the need of weights and arbitrariness. It was difficult at that time and this discussion has reminded us of how difficult it still is today. Should we set definite weights or a range? This can further be discussed. How much can we expect from a multidimensional poverty index? Take inexistence data and try to get additional insights to current exercises. Is this index going to give more detail and prove to be more useful than studying each dimension separately? 5

6 Empirical demands of the application. What do the indicators represent? What is it? What are the indicators that best capture it? Is it feasible to collect all of the indicators simultaneously? Linkage of surveys? Is it the way we want to go? Questions that remain unanswered. Final remarks by Jaime Saavedra There is a demand to improve ways in which different indicators are monitored. How to display the information? A dashboard? An only indicator? There are huge challenges ahead. 6