WPS6367 Policy Research Working Paper 6367 The Income Lever and the Allocation of Aid Lidia Ceriani Paolo Verme The World Bank Middle East and North Africa Region Poverty Reduction and Economic Management Department February 2013 Policy Research Working Paper 6367 Abstract The paper develops a concept and a measure of the empirical application, this methodology can provide an monetary capacity of a country to reduce its own equitable tool to rank countries or regions when it comes poverty and shows how these tools can be used to guide to budget or aid allocations, whether it is the allocation budget allocations or the allocation of aid. The authors of social funds within the European Union (North- call this concept the income lever. Making use of tax North transfers) or the allocation of aid from rich to poor and distributive theory, the paper shows how different countries (North-South transfers). The findings indicate redistributive criteria correspond to the different that the allocation of social funds in the European Union normative criteria of the income lever. It then constructs follows closely the rank that results from the income various income lever indexes based on these criteria lever indexes proposed while the allocation of aid to Sub- and uses such indexes to rank countries according to Saharan African countries does not. their own capacity to reduce poverty. As shown in the This paper is a product of the Poverty Reduction and Economic Management Department, Middle East and North Africa Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http:// econ.worldbank.org. The authors may be contacted at pverme@worldbank.org or lidia.ceriani@unibocconi.it. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team The Income Lever and the Allocation of Aid Lidia Ceriani1 and Paolo Verme2 1 a Bocconi, Milano, Italy Universit` 2 The World Bank, Washington DC, USA Abstract The paper develops a concept and a measure of the monetary capacity of a country to reduce its own poverty and shows how these tools can be used to guide budget allocations or the allocation of aid. The authors call this concept the income lever. Making use of tax and distributive theory, the paper shows how different redistributive criteria correspond to the different normative criteria of the income lever. It then constructs various income lever indexes based on these criteria and uses such indexes to rank countries according to their own capacity to reduce poverty. As shown in the empirical application, this methodology can provide an equitable tool to rank countries or regions when it comes to budget or aid allocations, whether it is the allocation of social funds within the European Union (North-North transfers) or the allocation of aid from rich to poor countries (North-South transfers). The �ndings indicate that the allocation of social funds in the European Union follows closely the rank that results from the income lever indexes proposed while the allocation of aid to Sub-Saharan African countries does not. JEL: H2, H5, I3, O1, O2, F3 Keywords: Aid Distribution, Poverty Reduction, Redistribution Policies Sector Board: Poverty Reduction (POV) 1 1 Introduction The motivations behind aid flows across countries have varied over time and space. It could be argued that these motivations ranged (and still range) from pure altruistic motivations to pure sel�sh and opportunistic ones. This is not surprising. National governments manage tax payers’ resources and are accountable to the electorate. As elected governments have a mandate and a duty to pursue citizens’ interests - at home or abroad - these interests may span across multiple dimensions. Governments’ altruistic behavior normally derives from an internal demand for altruism, now prominent in many societies, while governments’ sel�sh behavior should be mostly understood as an internal demand for such behavior. This is, after all, the golden rule of democracy. The birth and development of multilateral organizations over the twentieth century con- tributed to change these simple dynamics. Multilateral organizations now play a super- national role not only in terms of coordinating national policies but also in terms of design- ing and guiding global policies in both normative and positive terms. The UN declaration of human rights and the UN international tribunal are two examples of how multilateral platforms have been able to shape and disseminate global values. One example of this phenomenon is the increasing attention devoted to global poverty. With the fall of the Berlin wall in 1989, the publication of the 1990 World Bank World De- velopment Report on poverty and the establishment of the Millennium Development Goals in 2000, multilateral and bilateral organizations have altered their policy priorities turning poverty reduction into the most prominent development objective. Poverty reeduction is not only the main overarching development objective for multilateral organizations such as the World Bank, the Asian Development Bank or the European Commission but also for bilateral aid organizations of rich countries such as France, Japan or the UK (OECD, 2005). This is a change vis-a-vis previous decades where investments, industrialization, agricultural development, �scal stability and debt reduction dominated at different times the global development agenda. This shift in emphasis towards poverty reduction has evidently affected the distribution of aid not only for poor countries but also for poor areas in rich countries. More than 2 ever before, individual countries and multilateral organizations distribute aid according to poverty or income criteria. For example, the World Bank classi�es countries in different income categories and accords to different categories different bene�ts and so does the European Union to allocate social funds across members’ countries and to allocate aid outside the Union. Beyond these global objectives, the reality of aid allocation is, of course, much more com- plex. It is still the result of a mixture of old and new criteria, altruistic and non, and a mixture of development indicators, policy capacity and geopolitical factors. However, poverty reduction can be safely described today as the ultimate development goal. And the motivation for aid largely resides in the idea that it can help low income countries to reduce poverty. What is mostly debated is not whether aid should exist or not but what is the optimal allocation of aid to minimize poverty given a budget constraint. In other words, whether the allocation of aid should be simply driven by criteria such as poverty or GDP per capita or should include other criteria such as the policy capacity of recipient countries or other criteria. This paper addresses this question introducing an additional criteria for the allocation of aid called the “Income Lever� (IL). Poverty measurement is characterized by the quanti�cation of the number of poor individuals or by the quanti�cation of the poverty gap, the total amount of resources necessary to end poverty. This is what is technically called the “focus axiom� or, in other words, the focus on only a part of the income distribution, the part below the poverty line. By focussing on the poor, no attention is given to the distribution of incomes above the poverty line. This paper argues that donors should also consider the distribution of incomes above the poverty line and the recipients’ monetary capacity to reduce their own poverty via internal redistribution of incomes. This monetary capacity depends, among other factors, on the relation between the welfare of the poor and the welfare of the non-poor in any given country. It is argued, therefore, that it is not enough to know whether a society is poor, it is also important to understand whether a society has the monetary capacity to reduce its own poverty. This is what we call the “Income Lever (IL)�. The purpose of this paper is to develop different indexes to measure the income lever, and to 3 provide different distributional interpretations of such indexes using tax and distributional theory. Two countries with identical poverty rates and identical populations below the poverty line may exhibit very different income distributions above the poverty line with one country being much richer than the other. Understandably, these two countries should not be treated equally from the donors’ community. We therefore switch attention from poverty measurement to the measurement of the IL, the relation between the welfare of the poor and the welfare of the non-poor. Neither the poverty rate nor GDP per capita, two of the most popular measures used to rank recipients’ countries, can capture this relation. The IL is therefore proposed as a complement to these indicators and in order to rank countries more accurately for the purpose of aid allocation. In the next section we review briefly the rationale for the allocation of aid discussing guiding principles and empirical practices. Section three outlines the theoreatical framework for the IL indexes proposed while section four describes a selection of indexes treating separately indexes that result in proportional, progressive and regressive redistribution of incomes. Section �ve provides two empirical illustrations, one applied to the European Union to illustrate the use of the IL indexes in the context of North-North transfers and one on Sub- Saharan Africa (SSA) to illustrate the use of the indexes in the context of North-South transfers. 2 Principles and practices behind the allocation of aid The literature on the allocation of aid has polarized around two streams of literature, one that focused on the discussion of what principles and criteria should be used to guide the allocation of aid (normative studies) and one that focused on the reality of aid allocation trying to determine the criteria that are used by donors de facto to allocate aid and the effectiveness of aid in achieving its objectives (positive studies). We briefly review these two strands of the literature in turn. Normative studies. Broadly speaking, aid allocation criteria include norms regarding ef- fectiveness, equity and transparency (see for example Guillaumont, 2008). Effectiveness regards the capacity of aid to improve on target indicators such as poverty. Equity regards 4 the capacity of aid to treat equal countries equally. And transparency regards the capacity to assess the equity and effectiveness of aid. In a recent note, Anderson and Waddington (2006) group contributions to the normative debate into three main allocation principles re- spectively based on: poverty-efficiency, equal opportunity or structural indicators, roughly reflecting in this way the broader norms described. The poverty-efficiency literature emphasizes the importance of considering the poverty level in conjunction with the policy capacity of a country so as to maximize the poverty reduction capacity of aid at the global level given a budget constraint. Collier and Dollar (2001) and Collier and Dollar (2002), for example, promote a poverty-efficiency aid allocation criteria. Poverty being equal, aid should increase with the policy capacity of a country, because aid can have larger growth effect in a better policy environment. Vice-versa, the policy capacity being equal, aid should increase with the level of poverty because the growth- promoting effect of aid has more poverty reduction capacity in a very poor country. The poverty-efficiency position is supported by influential institutions such as the World Bank (Wolfensohn, Stern, Goldin, Rogers, and Karlsson, 2002) or the OECD (OECD, 2003; OECD, 2006) although empirical evidence (among others, Nunnenkamp and Thiele, 2006) suggests that the broader donors’ community rarely follows this criterion. This position is also questioned by scholars arguing for transparency, who point out the necessity of predictable, or less volatile, flows of aid to individual countries (among others Eifert and Gelb, 2008, Fielding and Mavrotas, 2008, Bulir and Hamann, 2006, Bulir and Hamann, 2003). As the policy capacity of a country is hard to assess and also volatile, the poverty efficiency criteria would be difficult to implement and also subject to frequent changes. This would speak in favor of using standard statistics such as GDP per capita or poverty rates to rank countries and allocate aid so as to capture basic but well understood indicators. Guillaumont, 2008 argues that the poverty-efficiency position is not backed by evidence and that a structural vulnerability approach whereby countries are consid- ered based on their structural and exogenous constraints would be preferable, which is an argument for allocating more aid to the LDCs. A different stream of the literature (Llavador and Roemer, 2001; Cogneau and Naudet, 2007) focuses instead on equal opportunities. These authors believe that the distribution 5 of aid should equalize opportunities among recipient countries for achieving growth, con- trolling for the effort undertook to turn aid into economic growth. This is based on the Roemer’s inequality of opportunities literature when one should distinguish between cir- cumstances and efforts and try to give a premium to efforts. This approach emphasizes the role of equity although - as it was the case of the poverty-efficiency literature - it is difficult to implement in practice given that the distinction between efforts and circumstances is not simple to make with empirical data. For the sake of transparency and in order to comply to speci�c normative criteria, some organizations have tried to translate guiding principles into formulas used to rank countries for the purpose of the allocation of aid. The World Bank uses a Performance Based Allocation (PBA) formula including indicators such as GNI per capita, population size and a measure of the country performance. Unicef uses a formula that includes the size of the child population and the under-�ve mortality rate. The European Commission (EC) uses several different formulae combined with various normative criteria to allocate social funds within the union or aid funds outside the union with criteria being different according to different groups of countries such as the Africa Caribbean and Paci�c (ACP) countries, Mediterranean countries or new accession countries. These are examples of attempts to operationalize normative criteria but, in the overall panorama of aid, these criteria remain few and organization speci�c. Positive studies. The empirical literature on aid allocation has tried instead to answer two main questions: what are the actual determinants and motivations of aid flows and whether aid is effective in boosting growth and reducing poverty in developing countries. With respect to the determinants of aid flows, Baulch (2003) analyzes the distribution of international development assistance using aid concentration curves. He �nds that among the six most generous bilateral donors (United States, Japan, Germany, United Kingdom, France and the Netherlands), only the United Kingdom and the Netherlands devolve the largest part of their aid budget to assistance to the poorest countries, while the other countries give large proportions of their aid budgets to relatively well-off middle income countries. The same bias is found in the analysis of the three main institutions providing multilateral development aid. The World Bank and other UN organizations show different 6 degree of progressivity in aid allocation while the European Union directs aid to middle- income countries where poor individuals account for the minor share of the world’s absolute poor. Alesina and Dollar (2000) also �nd that donor countries are inspired more by their political-strategic interests rather than by the extent of poverty, or the institutional and political environment of recipients. Chauvet (2002) investigates how aid allocation by donors is influenced by elite, violent, and social instabilities. The author �nds that the three different forms of socio-political instabilities have a different impact according to the type of aid (bilateral or multilateral) and the type of recipient countries (middle-income non-oil-exporting countries, low-income countries and oil-exporting countries). A recent contribution by Dreher, Nunnenkamp, and Thiele (2011) focuses on the difference in bilateral aid allocation patterns between new donors not members of the Development Assistance Committee (DAC) of the OECD and DAC members. Their �ndings indicate that both types of donors share the same shortfalls by not taking into account bad policy behaviors of recipient countries and by being influenced by commercial interest. According to Koch, Dreher, Nunnenkamp, and Thiele (2009) economic interests do not influence non-governmental organizations (NGOs) although NGOs seem affected by the geographical choices of the donor countries in which the NGOs are based, by the presence of other NGOs in the recipient countries and by sharing the same religion with the recipient country. With respect to the aid effectiveness literature, the results are mixed. Burnside and Dollar (2000) analyze a sample of 51 countries in the period 1970-1993, and �nd that aid has a positive impact on growth in countries with a good policy environment (accounting for �scal, monetary and trade policies), but not in countries with poor policies. These results have been later challenged, among others, by Easterly, Levine, and Roodman (2004) who �nd less clear-cut results by replicating the estimates. Also Dalgaard and Erickson (2009) �nd a modest effect of aid on growth by applying a theory-based calibration approach on a sample of Sub-Saharan African countries. Claessens, Cassimon, and Campenhout (2009) �nd that aid has improved over the years in terms of targeting, by increasingly focusing on poorer countries, and in terms of efficiency, by increasingly favoring countries with better policy environments. For an interesting literature review and meta-analysis on the topic, see Doucouliagos and Paldam (2008). 7 In essence, the normative debate and its advances are still fairly far from the actual criteria used by donors to allocate aid. Also, while poverty reduction has increased in importance among the overarching development priorities, the actual allocation of aid is not driven mainly by poverty. These outcomes are partly explained by the fact that the measurement of poverty requires instruments that have only recently become widely available. When the World Development Report on poverty was published in 1990 only a handful of countries carried out and made available household consumption surveys. Today, the World Bank has a database of over 1,700 household surveys many of which are for public use. We are now in a position to make much better use of micro data for allocating aid and targeting the poor whether we believe in efficiency, equity or structural criteria. The IL methodology proposed makes an optimal use of the micro data available and com- plies to efficiency, equity and transparency criteria. It improves on the simple measurement of poverty by extending the analysis to the non-poor. The concept can help to improve the policy capacities of countries by highlighting the potential for redistributive policies, it complies to the equal opportunity principle by allocating aid according to the monetary capacity of a country to address its own poverty and it is transparent in that responds to principles well established in the tax and distribution theory literature. 3 Framework The idea that individuals should contribute equally towards the expenses of the government dates back to Mill (1848) theory of equality of sacri�ce. For what reason ought equality to be the rule in matters of taxation? For the reason that it ought to be so in all affairs of government. As a government ought to make no distinction of persons or classes in the strength of their claims on it, whatever sacri�ces it requires from them should be made to bear as nearly as possible with the same pressure upon all, which, it must be observed, is the mode by which least sacri�ce is occasioned on the whole. If any one bears less than his fair share of the burthen, some other person must suffer more than his share, and the alleviation to the one is not, ceteris paribus, so great a good to him, as 8 the increased pressure upon the other is an evil. Equality of taxation, therefore, as a maxim of politics, means equality of sacri�ce. It means apportioning the contribution of each person towards the expenses of government so that he shall feel neither more nor less inconvenience from his share of the payment than every other person experiences from his. (Mill, 1848, Book V, Chapter II, On the General Principles of Taxation) Equality of sacri�ce can be interpreted as equality of absolute, proportional or marginal sacri�ce where sacri�ce is measured in terms of utility. It is absolute if everyone sacri�ces an equal amount of utility irrespective of the starting utility. It is proportional if everyone sacri�ce the same amount of utility relatively to the starting utility. It is marginal if everyone sacri�ces the same amount of marginal utility. The interpretation of equality of sacri�ce is a question for the social planner as well as the assumptions about the shape of the individual utility functions. In this paper paper we will not make any normative judgement on the type of sacri�ce or on the shape of the utility function. However, any normative choice that the social planner will make on these two grounds (type of sacri�ce and utility) will results in a proportional, progressive or regressive redistribution mechanism. This is what we focus on in this paper where we will construct different IL indexes that imply different redistribution mechanisms. How this equality of sacri�ce is then realized through transfers depends upon two as- sumptions. The �rst assumption concerns the form of the utility functions attributed to individuals and - as in orthodox economic theory - we can simply assume these functions to be linear in incomes and identical for all individuals in society. The second assumption regards instead our preferences for different redistributive criteria such as proportional, pro- gressive or regressive. As we will show in the next sections, different preferences correspond to different IL indexes. Let N be a population made of i = 1, 2, ..., n individuals, n ∈ N. Each i-th individual is denoted by an income level yi ∈ R+ , and let us assume that the poverty line is set at level zN . Without loss of generality, let us assume that y1 ≤ y2 ≤ · · · ≤ yn . The set of poor individuals in society N , QP N , is de�ned as the set of individuals whose income falls below the poverty line: QP N = {i ∈ N : yi < zN }. Symmetrically, the set of non-poor individuals 9 in society is N and Q∼ P N , is de�ned as the set of individuals whose income is greater or equal than the poverty line: Q∼ P N = {i ∈ N : yi ≥ zN }. To eradicate poverty, we should �ll the gap between each poor individual’s income and the poverty line. The total amount of income to eradicate poverty is therefore given by the sum of poverty gaps: GP N = (zN − yi ) (1) i∈QP N On the other hand, incomes in excess of the poverty line can be quanti�able by the sum of all the wealth gaps, de�ned as follows: G∼ P N = (yi − zN ) (2) i∈Q∼ N P This quantity can be considered as the upper-bound budget for any redistributive policy aiming at keeping above (or at) the poverty line all non-poor individuals. Figure 1 shows a simple illustration of the poverty and wealth gaps. On the y − axis we plotted income and on the x − axis we plotted the population share (this is what it is usually de�ned as Pen’s parade from Pen, 1971). The poverty line z divides the poverty gap region from the wealth gap region so that the �gure provides a �rst visual indication of the relative proportions of these two areas. The point qz denotes the share of poor individuals in the population (poverty headcount index). 4 Alternative income lever indexes 4.1 Proportional redistribution IL indexes As a �rst and simple index, we de�ne the income lever as a measure of the relative pro- portions between the poverty gap and the wealth gap as follows: GP ILprop N = N (3) G∼ N P 10 Figure 1: Poverty and excess-poverty gaps y 6 f (y ) G∼P z GP i -n qz 1 Note: On the y-axis we measure income levels, on the x-axis shares of the population. Therefore, qz stands for the share of individuals having an income level smaller than the poverty line z (in other words, qz is the headcount index). The area above f (y ) and below z is the sum of poverty gaps GP , while the area below f (y ) and above z is the sum of wealth gaps G∼P (note that the indicator on the x-axis is normalized by the population). The index described in equation (3) is increasing in the sum of poverty gaps, decreasing in the sum of wealth gaps, and it is scale invariant, which are all reasonable properties.1 Moreover, it has a straightforward distributional implication. It can be considered as the proportional tax-rate α that has to be levied on the incomes of each non-poor individual to defeat poverty: α(yi − zN ) = (zN − yi ) (4) i∈Q∼ N P i∈QP N Solving for α: i∈QP (zN − yi ) GP α= N = N = ILprop N (5) i∈Q∼ P (yi − zN ) G∼ N P N If GP ∼P P ∼P N ≤ GN , then α ∈ [0, 1], whereas if GN > GN , then α > 1. Note that this interpretation of the index reveals the normative statement embedded in the index ILprop N : we claim that each non-poor individual should contribute to poverty reduction equally 1 See Quiggin and Mahadevan (2010) for a characterization of this index. 11 and in a proportional way to her wealth gap. This is evidently a choice and a normative statement. As an example, imagine two populations, A and B , where n = 5, with the following income distributions sorted in ascending order of income: A = (2, 5, 6, 15, 17), B = (2, 5, 6, 9, 13), and let zA = zB = 6. It is evident that the two populations are equal in terms of poverty, given that standard poverty measures are subject to the focus axiom. However, from the perspective of the income lever as de�ned in equation 3, the two populations have a 5 5 different capacity to reduce poverty. Indeed, ILA = 20 is smaller than ILB = 10 , which places society A in a better position than society B when it comes to poverty reduction, simply because society A can rely on a larger wealth gap than society B. We can therefore make an ordinal ranking of the two societies based on the income lever. 4.2 Regressive redistribution IL indexes For each non-poor individual i ∈ Q∼ P ∼P be the wealth gap, with g ∼P = y − z . N , let gi i i N Let ||.|| be the cardinality of a generic set so that ||Q∼ P N || is the number of individuals belonging to the population set Q∼ P N (set of non-poor individuals). We can now imagine that all people who can afford a transfer without falling into poverty contribute equally, in absolute terms, to poverty eradication. In other words, we want to �nd the flat transfer fN that each non-poor individual in society N should provide to contribute to the eradication of poverty. This is equivalent to the average poverty gap to be levied on non-poor individuals who can bear the cost without crossing the poverty line: GP N fN = (6) ||Q∼ N P,f || where Q∼ N P,f is the set of non-poor individuals who can bear the cost of the transfer and it is de�ned as Q∼ N P,f = {i ∈ Q∼ P N : (yi − f ) ≥ z }. Note that the set of non-poor individuals Q∼ N P,f is endogenously de�ned as it depends upon the magnitude of the sum of poverty 12 gaps, the number of non-poor individuals, and the distribution of incomes above the poverty line. The level of f itself is a good measure of income lever because the smaller the flat transfer, the lighter the income lever. In other words, the weight needed to lift all poor individuals out of poverty is small. Therefore, we de�ne f as our second income lever index: GP ILf N latA = fN = N (7) ||Q∼ N P,f || Consider, for instance two societies A and B , where the distributions of incomes are A = (2, 5, 6, 9, 13) and B = (2, 5, 6, 7, 15). Let the poverty line z = 6. Using ILprop N , the two societies have the same income lever: ILprop A = ILprop B = 0.5. However, by using ILf lat N , we �nd that ILf A lat = 2.5 while ILf B lat = 5. We then �nd that society B ’s effort to reduce poverty is higher than society A’s, and therefore the income lever of society B is higher than the income lever of society A. Note that this second scheme forces the de�nition of a no tax area, a group of non-poor individuals who are not considered responsible for the eradication of poverty. The dimen- sion of this no-tax area is positively correlated with the dimension of the sum of poverty gaps and the inequality of the distribution of income among non-poor individuals, and negatively correlated with the number of non-poor individuals. Above the no-tax area, instead, each individual should transfer the same �xed amount. Figure 2 illustrates this new concept using the same coordinates as in Figure 1 All non-poor individuals are - in principle - called to contribute to the poverty eradication effort in equal amounts. However, to make sure that some of the non-poor will not fall into poverty as a consequence of the flat tax, a no tax area is derived based on the underlying distribution of incomes so that all individuals above the no tax areas will be able to afford the flat tax without falling below the poverty line. Note that a flat tax is a regressive tax. Let us de�ne fi the average transfer levied on the i − th non-poor individual: 13 Figure 2: Flat Transfer y 6 f (y ) G∼P,f z+f z GP no tax area i -n qz qz+f 1 ¯ fn fi = (8) yi ¯ Then, it is straightforward to show that fi is decreasing with incomes: ¯ dfi −fn = 2 ≤0 (9) dyi yi Therefore, any ordering between countries based on (7) relies implicitly on a regressive redistribution hypothesis: the marginal tax rate to be levied on non-poor individuals is decreasing with the wealth gaps. This is again a normative choice that the social planner is called to make. Note that the proposed ILf N lat is just one of the possible indexes we can think of based on the de�nition of the flat transfer f . For instance, another reasonable index can be: GP ILf N latB =f (10) G∼P,f ∗ where G∼P,f = ∼P i:yi ≥z +f (gi − f ). Index ILf N lat is the weighted version of ILf N latA . The weight stands for the additional contributive capacity of the new set of non-poor individuals 14 after taking away the flat transfer. In other words, for a given flat transfer, the society where the sum of excess poverty gap minus the transfer for the new set of non-poor is larger, has the lower income lever. Consider for instance two societies C and D, where the distributions of incomes are C = (2, 5, 6, 7, 8, 13) and D = (2, 5, 6, 7.5, 8.5, 12). Let the poverty line z = 6. The sum poverty gaps is the same as previous examples: ILprop C = ILprop D = 0.5, ILf C latA = ILf D latA = 2.5, but ILf C latB = 2.5 ∗ 5 4 < ILf D latB = 2.5 ∗ 5 3.5 . 4.3 Progressive redistribution IL indexes The measure de�ned in (3) may be interpreted as the average tax rate to be levied on the sum of wealth gaps, therefore underlying a proportional redistribution scheme. The measure de�ned in (7), instead, may be interpreted as the flat-tax to be levied on non- poor individuals, therefore underlying a regressive redistribution scheme. A different set of income lever measures can be de�ned starting from a progressive idea of redistributive scheme. We can imagine that not all the individuals above the poverty line are to be considered responsible for transferring income to poor individuals, but just the rich individuals. We can de�ne as rich the subset of non-poor individuals whose income is higher than a thresh- old r, the richness line. Let Q∼ N P,r be the set of all individuals whose income lies above r: Q∼ N P,r = {i : yi ≥ r} , (11) and let G∼P,r be the sum of richness gaps gi r: G∼P,r = r gi = (yi − r). (12) i∈Q∼ N P,r i∈Q∼ N P,r The richness line can be determined in different ways, and the discussion about the topic is broad and largely beyond the scope of this paper. Medeiros (2006) and Eisenhauer (2011), 15 for example, de�ne the richness line as a function of the poverty line. Other scholars de�ne the richness line as a function of the median income (see Peichl, Schaefer, and Scheicher, 2010 or Brzezinski, 2010) while Peichl and Pestel (2011) de�ne the richness threshold as the 80th percentile. Once the richness line is de�ned, then the income lever becomes: GP ILprog N A = N (13) G∼P,r Note that ILprog N A has a twofold interpretation. First, it may be seen as the proportional tax rate δ to be applied on all wealth gaps above a no-tax area de�ned by the richness line. Second, it may be interpreted as the marginal tax rate on the bracket of wealth gaps larger than r, the �rst bracket (0 − r) being taxed with a marginal tax rate equal to zero. Differently from ILreg N , this new measure implies a progressive transfer scheme. For ¯ is increasing in each non-poor individual exceeding the no-tax area, the average transfer t incomes of non-poor individuals: δ (yi −r) ¯ dt d yi δr = = 2 > 0. (14) dyi dyi yi Consider the two societies A and B with incomes A = (3, 5, 6, 8, 9, 12, 15) and B = (3, 5, 6, 7, 11, 13, 13). Let the poverty line be z = 6. The two societies have the same income lever with ILprop A = ILprop B = 0.2. They also have the same income lever as mea- sured by ILreg reg A = ILB = 0.05. However, the two societies would appear different with a progressive income lever scheme since ILprog A A = 4 12 > ILprog B A = 4 13 . The progressive scheme described can be visualized in Figure 3 using the same axes as in Figures 1 and 2. As in Figure 2, we have a no-tax area but unlike Figure 2 the no-tax area is de�ned by the choice of the richness line and not by the implicit distribution of incomes. In this case, it is the social planner who de�nes the richness line and the group of people thought to be liable for poverty reduction. ILprogA is not the only income lever index that can be obtained by setting the richness line r. We may want to take into account, for instance, also the proportion of individuals who are taxed over all non-poor individuals: the same value of ILprogA , in fact, can be 16 Figure 3: Progressive Transfers, setting a richness line y 6 f (y ) G∼P r r z GP no tax area i -n qz qr 1 obtained by taxing all possible non-poor individuals (and in this case ILprop = ILprogA ), or just a small subset of individuals. In this second case, the income lever is smaller as the same tax effort (in the form of a proportional tax) is levied on a smaller percentage of rich individuals. If we believe that the richest individuals in society are responsible for eradicating poverty, we may want to collect revenues by leveling the income distribution from the top of the income scale until the point where enough revenues are collected to �ll the poverty gap. In this case, the set of rich individuals to be taxed is found endogenously. We can proceed in a recursive way. In the �rst step we tax away the difference between the richest income and the second richest income (yn − yn−1 ). If yn − yn−1 ≥ GP N the recursive process ends at the �rst step, where only one individual is responsible for eradicating poverty. Otherwise, we go on by taxing the second richest individual in the original distribution by (yn−1 − yn−2 ). If (yn + yn−1 ) − 2(yn−2 ) ≥ GP N the recursive process ends, and we are taxing two individuals. Otherwise, we include the third richest individual and so on, until we reach the last step m de�ned as:    n  m = min j : j (yn−j +1 − yn−j ) ≥ GP n (15)   j =1 17 Note that m corresponds to the minimum number of richest individuals that the govern- ment should target for collecting enough revenues to �ll the poverty gap. A straightforward index of income lever can therefore be de�ned as the proportion of taxed individuals on the total number of non-poor individuals: the highest this proportion, the heavier the income lever: m ILprogB = (16) ||Q∼P || Consider the following example. Let us assume that society A has a sum of poverty gaps GP A = 7 and that the income distribution among rich individuals is (6,7,8,9,10,11). The following matrix shows the stepwise procedure. steps individual incomes GP -Revenues y1 y2 y3 y4 y5 y6 0 6 7 8 9 10 11 7 1 6 7 8 9 10 10 7-(11-10) =5 2 6 7 8 9 9 9 7- [(11-10) + 2(10-9)] =4 3 6 7 8 8 8 8 7- [(11-10) + 2(10-9) +3(9-8)] 4 6 7 7 7 7 7 7- [(11-10) + 2(10-9) +3(9-8) + 4(8-7)]=-3 The recursive process in the example ends at step 4: the �rst four richest individuals are taxed out of 5 non-poor individuals: ILprogB n 1 =4 5. Note that this procedure mimics what it is done in Medeiros (2006) in order to de�ne endogenously an affluence line, which can be considered by itself as an indicator of income lever: the highest the affluence line, the smaller the income lever. In particular, given m as de�ned in equation (15), the affluence line aN can be de�ned as: m k=1 k (yn−k+1 − yn−k ) − GP N aN = yn−m + max 0, (17) m 3 In the example given above, for instance, the affluence line aA would be 7 + 4 = 7.75. Instead of using the affluence line, we may think of using the ratio between the poverty line and the affluence line, so that the index becomes scale invariant: 18 zN ILprogC = (18) aN As it was the case for Figures 2 and 3, in Figure 4 we also have a threshold that separates non-poor tax-payers from non-poor non tax-payers and we also have a no-tax area. The difference from previous examples is that the process starts from the top, a sort of reversed Rawlsian approach, where the richest individual is taxed �rst for the income distance with the second richest individual and this is followed by the second richest, the third and the qa − th richest individual until the full poverty gap is recovered. This mechanism boils down to a progressive transfer, where progressivity is obtained by �xing a tax allowance in the measure of a, and a proportional tax rate of 1. Figure 4: Progressive Transfers, �nding an affluence line y 6 f (y ) GP a z GP no tax area i -n qz qa 1 5 Empirical illustrations As an illustration of the indexes developed in the previous section, we present here the results of the analysis on a sample of European member states (Cyprus, Czech Repub- lic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovenia and Slovakia) and a sample of African countries (Benin, Cote d’Ivoire, Cameroon, Ethiopia, Mali, Mauritania, Uganda, South Africa). We present �gures for poverty (headcount and sum of poverty 19 gaps) and income lever indexes developed in the previous sections. The following Table 1 summarizes all indexes. Table 1: indexes used the empirical illustration Poverty indexes ||Qp || P0 = n n z −y ˜ P1 = i=1 z , i ˜i = z for each i ∈ Q∼P where y Income Lever indexes GP ILprop = G∼P GP ILf latA = ||Q∼P,f || GP ILprogA = G∼P,r m n ILprogB = ||Q∼P || , where m = min j : j =1 j (yn−j +1 − yn−j ) ≥ GP n m z k(yn−k+1 −yn−k )−GP ILprogC = aN , where aN = yn−m + max 0, k=1 m N 5.1 Ranking EU nations for North-North transfers The European Union �nances the economic imbalances between different regions using the European Regional Development Fund - ERDF (EC, 2006). During the period 2007-2013, a total of 347.410 billions Euros (2007 prices) have been allocated to �nancing regional policies designed to meet the objectives of Convergence, Regional Competitiveness and Employment and Territorial Cooperation. We focus here on the Convergence objective which aims at helping the least-developed member states and regions by improving condi- tions for growth and employment. The resources allocated to this objective are equivalent to 81.5% of the total ERDF funds. Eligible regions (at NUTS2 level) are those with a per-capita GDP level below 75% of the European average2 . Nearly all regions of the new member states fall under this de�nition and most of Southern Italy, East Germany, Greece 2 These regions receive 70.5% of the resources. The remaining funds are allocated to regions where per capita GDP has risen above 75% of the European average and to member States whose per capita Gross National Income is below 90% of the European average and which are running economic convergence programmes. 20 and Portugal, much of Spain, and some parts of the United Kingdom. Among our sam- ple of countries, all regions are included, with the exception of Prague (Czech Republic), ´ Bratislava (Slovakia) and the following Hungarian regions: Eszak-Alf¨ el-Dun´ old, D´ ul, ant´ ´ ag, K¨ Eszak-Magyarorsz´ ant´ ozp-Dun´ ant´ ul, Nyugat-Dun´ ul. We use data from the 2008 round of the European Union’s Survey on Income and Living Conditions (EU-SILC). As income variable, we use the household equivalent disposable income using the modi�ed OECD scale. We then deflate incomes using the harmonized consumer price indexes provided by Eurostat (2011), in order to have real and comparable incomes. As poverty line z , we use 11 USD per day which is the threshold set by the World Bank for these set of countries. Since income data are expressed in euros, we use the exchange rate euro-dollar (1.43 euro per 1 dollar in 2008). Following Medeiros (2006) and Eisenhauer (2011), we de�ne the richness line as a function of the poverty line. In particular, considering that the absolute poverty line z may be interpreted as 60% of the median income ym of an hypothetical income distribution, we set the richness line at 200% of ym (see Peichl, Schaefer, and Scheicher, 2010). Table 2 summarizes the values of the different indexes (P 0 , P 1 , ILprop , ILf latA , ILprogA , ILprogB , ILprogC ) for the sample of European countries. Recall that P 0 stands for the percentage of poor individuals in the population (Headcount index) and P 1 is the poverty gap index. ILprop can be interpreted as the average tax rate to be levied on non-poor individuals to collect enough revenue to rise all poor individuals to the poverty line. For Romania we see that the index is greater than 1 (3.62). It means that, to defeat poverty, the income of non-poor individuals should almost quadruplicate. Estonia and Hungaria have the same proportion of poor individuals and almost the same poverty gap index. Note however that in Estonia it is sufficient a proportional tax rate of 5.5% levied on the non-poor to solve poverty, while in Hungaria the tax rate should be �xed at 8.8%. ILf latA is the flat tax f to be levied on all non-poor individuals who can bear this cost without crossing the poverty line. For Romania no flat-tax can be collected without having non-poor individuals crossing the poverty line. 21 ILprogA can be interpreted as the tax rate to be applied on all wealth gaps above a no-tax area de�ned by a deduction r, which is the exogenous richness line. In Romania nobody is above the richness line, therefore the sum of richness gaps is zero and the index cannot be computed. ILprogB is the smallest proportion of rich individuals that alone can defeat poverty by transferring all their incomes above the endogenous affluence line. A value of the index greater than one means that the number of rich individuals should be a multiple of the actual number of individuals above the poverty line, and therefore it is not possible for the country to defeat poverty alone (and this is the case for Romania). Finally, ILprogC is the ratio between the poverty line and the endogenous affluence line: the smaller the ratio, the larger the gaps between the two lines. If the affluence line in a country is much higher than the poverty line, this means that the distribution of income above the poverty line is very unequal. Table 2: Poverty and Income Lever indexes and ranks (1-higher to 8-lower) for a selection of European countries, 2008 Country P0 P1 ILprop ILf lat ILprogA ILprogB ILprogC index r index r index r index r index r index r index r Czech Republic 0.05 8 0.009 8 0.007 8 24.97 8 0.060 8 0.053 8 0.177 8 Estonia 0.20 5 0.053 5 0.057 6 190.88 5 0.535 7 0.063 6 0.251 7 Hungary 0.20 6 0.050 6 0.088 4 180.40 6 1.878 2 0.063 4 0.349 3 Latvia 0.35 3 0.130 2 0.176 3 643.58 2 1.159 4 0.154 3 0.318 4 Lithuania 0.35 2 0.111 3 0.181 2 550.44 3 1.385 3 0.154 2 0.359 2 Poland 0.24 4 0.064 4 0.085 5 244.61 4 0.698 6 0.063 5 0.255 6 Romania 0.80 1 0.376 1 3.620 1 - 1 - 1 3.750 1 2.606 1 Slovakia 0.10 7 0.028 7 0.036 7 89.62 7 0.740 5 0.056 7 0.283 5 Source: our elaboration on EU-SILC 2008 cross-section data. ‘r’ indicates rank. Table 2 also shows the ranking r of the the set of countries considered, which can be used for the purpose of aid allocation. In general, a donor may want to select the index that best corresponds to widely accepted normative criteria. While proportional and progressive criteria are usually the norm in tax systems worldwide and therefore justi�able in the context of aid allocation, regressive criteria are much less popular and more difficult to defend. One can also look at the consistency across indexes. The Czech Republic, for example, comes consistently in the last place while Romania comes in �rst place across all 22 indexes. In these two cases, the ranking is consistent across indexes and easier to use for the purpose of aid allocation. Is this ranking related to the actual distribution of EU social funds? Table 3 shows the indicative allocation of cohesion funds across the eight countries considered and the rank, which can be compared to the rank in Table 2. In �rst and last position we �nd Romania and the Czech Republic and this matches perfectly with all indicators in Table 2. The IL index which better approximates the rank of the actual distribution of funds is ILprogA , where all ranks are the same, except for Slovakia and Estionia. Table 3: Indicative allocation of Cohesion Fund, 200713 (2007 prices, in % GDP) Country Cohesion Fund (%GDP) rank Czech Republic 6.69 8 Estonia 7.17 5 Hungary 8.69 2 Latvia 7.32 4 Lithuania 8.02 3 Poland 7.13 6 Romania 18.94 1 Slovakia 7.11 7 Source: our elaboration on European Union (2007) and Eurostat for GDP 6 Ranking poor countries for North-South transfers Table 4 summarizes poverty and income lever indexes together with the country ranks obtained by computing each index for eight African countries based on the Global Income Distribution Database (GIDD). This is a database constructed by the World Bank that homogenized a number of standard variables including income and consumption across household consumption surveys worldwide.3 The welfare variable used is household equiv- alent consumption at Purchasing Power Parity (PPP) in USD where the equivalence scale is the square root of household size. To compute the indexes, we de�ne as poverty line z the World Bank official poverty line of 1.25 USD/day PPP. Similarly to what we have 3 For a full description of the data see Ackah, Bussolo, Hoynos, and Medvedev (2008) 23 done with the selection of European countries, we set the richness line r = 2z/0.6 = 4.16 USD/day PPP. Note that Mali and Uganda have the same headcount index (P 0 =0.5). Nevertheless, in Uganda a proportional income tax of 30% should be levied on the non-poor to solve poverty, while in Mali the proportional tax rate would be higher: 38% (see ILprop ). We can therefore conclude that poverty is more burdensome in Mali, where the contribution effort of non- poor individuals would be higher. The income lever index ILf lat stands for the average transfer payed by all non-poor in- dividuals who would remain non-poor after the tax. The higher this flat tax, the higher the income lever. Note, for instance, that Benin has a poverty incidence of 0.45, which is slightly higher than Ethiopia (0.40). On the other hand, the flat tax that would eliminate poverty in Benin is almost three times higher than the one in Ethiopia, as the income lever index ILprop indicates: 23.23 dollars in Benin against 8.31 dollars in Ethiopia. The income lever index ILprogA may be interpreted as the proportional tax rate that has to be levied on individuals above the richness line r. A value greater than one, as it is the case for Benin, Ethiopia, Mali and Uganda, means that there is no possibility of a poverty-solving redistribution by targeting individuals above r. On the other hand, even if Cote d’Ivoire and South Africa have the same poverty headcount (P 0 = 0.25), South Africa has a higher poverty reduction power than Cote d’Ivoire since it is sufficient to levy a proportional tax rate of less than 5% to eradicate poverty versus a proportional tax rate of 12% needed in Cote d’Ivoire (see ILprogA in Table 4). Recall that ILprogB shows the proportion of individuals above the endogenous affluence line over the total number of non-poor individuals (the individuals above the poverty line). The affluence line is de�ned in such a way that revenues collected by flattering the highest incomes at the affluence line equal the sum of poverty gaps. The smaller the percentage of individuals who bear the transfer, the lighter the burden of poverty reduction (and the smaller ILprogB ). For instance, Mali and Uganda have the same poverty headcount (P 0 = 0.5), but in Mali the percentage of non-poor individuals to be taxed is double that in Uganda (ILprogB is 0.4 in Mali and 0.2 in Uganda). 24 Finally, ILprogC is the ratio between the poverty line and the affluence line: in Mali and Ethiopia the endogenous affluence line is roughly two times the poverty line. In South Africa, the affluence line is twenty times the poverty line. This result suggest that in South Africa the distribution above the poverty line is very unequal. A small minority (the 6.7% of individuals above the poverty line, as indicated by ILprogB ) could transfer all income above the affluence line to eradicate poverty and they would still have an income level twenty times higher than the poverty line. Note that the same small minority is needed in Cote d’Ivoire (where IlprogB is still 0.67), but they would end up with an income level just ten times higher than the poverty line. Table 4: Poverty and Income Lever indexes and ranks (1-higher to 8-lower) for a selection of African countries Country (year) P0 P1 ILprop ILf lat ILprogA ILprogB ILprogC index r index r index r index r index r index r index r Benin (2003) 0.45 3 0.152 3 0.315 2 23.13 3 2.319 3 0.273 2 0.451 3 Ivory Coast (2002) 0.25 6 0.067 7 0.045 6 3.39 7 0.119 7 0.067 6 0.099 7 Cameroon (2001) 0.35 5 0.095 5 0.089 5 6.00 5 0.267 5 0.077 5 0.131 6 Ethiopia (2000) 0.40 4 0.098 4 0.235 4 8.31 4 4.599 1 0.167 4 0.484 2 Mali (2001) 0.50 1 0.187 1 0.376 1 28.42 2 3.070 2 0.400 1 0.496 1 Mauritania (2000) 0.20 8 0.059 8 0.043 7 2.97 8 0.150 6 0.063 8 0.137 5 Uganda (2002) 0.50 1 0.179 2 0.295 3 34.01 1 1.094 4 0.200 3 0.318 4 South-Africa (2000) 0.25 6 0.090 6 0.030 8 4.88 6 0.046 8 0.067 7 0.047 8 Source: our elaboration on Global Income Distribution Database (GIDD). ‘r’ indicates rank. As a �nal exercise, one may want to ask the question of whether the IL indexes proposed proxy any of the criteria currently used to allocate aid. In Table 5, we show the average net official development assistance (ODA) in percentage of the gross national income re- ceived by the same African countries considered before. The net ODA comprises grants or loans to developing countries and territories on the OECD/DAC list of aid recipients that are undertaken by the official sector with promotion of economic development and welfare as the main objective and at concessional �nancial terms. Among the selected countries, Mauritania received the most generous amount of aid, followed by Ethiopia. Note that Mauritania is the least poor country (see estimates in Table 4) and Ethiopia is the fourth poorest country out of the eight considered). More generally none of the criteria summarized in Table 4 could justify the distribution of aid listed in Table 5. 25 Table 5: Net ODA received (% of GNI), average 2000-2003 Country ODA (% GNI) rank Benin 9.77 5 Ivory Coast 4.33 7 Cameroon 5.54 6 Ethiopia 14.67 2 Mali 13.33 4 Mauritania 20.17 1 Uganda 14.09 3 South-Africa 0.40 8 Source: our elaboration on WB opendata 7 Conclusion In a global scenario where rich countries are facing increasing budget constraints, the debate on the optimal allocation of aid has become more important than ever. Exploiting at best new opportunities offered by the widespread availability of micro data, the paper has developed a concept and a measure of the monetary capacity of a country to reduce its own poverty de�ned as the Income Lever (IL). We constructed various income lever indexes based on different distributive criteria and used such indexes to rank countries according to their own monetary poverty reduction capacity. As shown in the empirical application, this methodology can provide an efficient, equitable and transparent tool to rank countries or regions when it comes to budget or aid allocations, whether it is the allocation of social funds within the European Union (North-North transfers) or the allocation of aid from rich to poor countries (North-South transfers). 26 References Ackah, C., M. Bussolo, R. D. Hoynos, and D. Medvedev (2008): “A new dataset on global income distribution,� Mimeo, Development Prospects Group, The Wold Bank. Alesina, A., and D. Dollar (2000): “Who Gives Foreign Aid to Whom and Why?,� Journal of Economic Growth, 5(1), 33–63. Anderson, E., and H. Waddington (2006): “Aid and the MDG poverty target: how much is required and how should it be allocated?,� Working Paper 275, Overseas Devel- opment Institute. Baulch, B. (2003): “Aid for the poorest? The distribution and maldistribution of inter- national development assistance,� CPRC Working Paper 35, Institute of Development Studies. Brzezinski, M. (2010): “Income Affluence in Poland,� Social Indicators Research, 99(2), 285–299. Bulir, A., and J. Hamann (2003): “Aid Volatility: an empirical assessment,� Staff papers, IMF. (2006): “Volatility of Development Aid: From the Frying Pan into the Fire?,� Working Paper 06/65, IMF. Burnside, C., and D. Dollar (2000): “Aid, Policies, and Growth,� The American Economic Review, 90(4), 847–868. Chauvet, L. (2002): “Socio-political instability and the allocation of international aid by donors,� European Journal of Political Economy, 19, 33–59. Claessens, S., D. Cassimon, and B. V. Campenhout (2009): “Evidence on changes in aid allocation criteria,� The Wold Bank Economic Review, 23(2), 185–208. Cogneau, D., and J.-D. Naudet (2007): “Who Deserves Aid? Equality of Opportunity, International Aid, and Poverty Reduction,� World Development, 35(1), 104–120. 27 Collier, P., and D. Dollar (2001): “Can the world cut poverty in half? How policy reform and effective aid can meet international development goals,� World Development, 29(11). (2002): “Aid allocation and poverty reduction,� European Economic Review, 46(8), 1475–1500. Dalgaard, C.-J., and L. Erickson (2009): “Reasonable Expectations and the First Millennium Development Goal: How Much Can Aid Achieve?,� World Development, 37(7), 1170 – 1181. Doucouliagos, H., and M. Paldam (2008): “Aid effectiveness on growth: A meta study,� European Journal of Political Economy, 24(1), 1 – 24. Dreher, A., P. Nunnenkamp, and R. Thiele (2011): “Are ‘New’ Donors Different? Comparing the Allocation of Bilateral Aid Between nonDAC and DAC Donor Countries,� World Development, 39(11), 1950–1968. Easterly, W., R. Levine, and D. Roodman (2004): “Aid, Policies, and Growth: Comment,� The American Economic Review, 94(3), 774–780. EC (2006): “European Regional Development Fund,� Regulation (EC) 1080/2006, Euro- pean Parliament and European Council. Eifert, B., and A. Gelb (2008): “Reforming Aid: Toward More Predictable, Performance-Based Financing for Development,� World Development, 36(10), 2067 – 2081. Eisenhauer, J. G. (2011): “The rich, the poor, and the middle class: Thresholds and intensity indices,� Research in Economics, 65(4), 294 – 304. European Union (2007): “Cohesion policy 2007-13, Commentaries and official texts,� Guide, European Union, Regional Policy. Fielding, D., and G. Mavrotas (2008): “Aid Volatility and Donor-Recipient Charac- teristics in Difficult Partnership Countries,� Economica, 75(299), 481–494. 28 Guillaumont, P. (2008): “Adapting aid allocation criteria to development goals,� An essay for the 2008 development cooperation forum united nations and social council, Mimeo. Koch, D.-J., A. Dreher, P. Nunnenkamp, and R. Thiele (2009): “Keeping a Low Pro�le: What Determines the Allocation of Aid by Non-Governmental Organizations?,� World Developmenta, 37(5), 902–918. Llavador, H. G., and J. E. Roemer (2001): “An equal-opportunity approach to the allocation of international aid,� Journal of Development Economics, 64(1), 147–171. Medeiros, M. (2006): “The rich and the poor: the construction of an affluence line from the poverty line,� Social Indicators Research, 78(1), 1–18. Mill, J. S. (1848): Principles of Political Economy. Longmans, Green and Co., London. Mitra, T., and E. A. Ok (1996): “Personal Income Taxation and the Principle of Equal Sacri�ce Revisited,� International Economic Review, 37(4), pp. 925–948. Musgrave, R. (1985): “A brief history of �scal doctrine,� in Handbook of Public Eco- nomics, ed. by A. J. Auerbach, and M. Feldstein, vol. 1 of Handbook of Public Economics, chap. 1, pp. 1–59. Elsevier. Nunnenkamp, P., and R. Thiele (2006): “Targeting Aid to the Needy and Deserving: Nothing But Promises?,� World Economy, 29(9), 1177 – 1201. OECD (2003): Harmonizing donor practicies for effective aid delivery. OECD Publishing. (2005): “Aid allocation criteria: managing for development results and difficult partnerships,� Unclassi�ed DCD(2005)4, OECD. (2006): Harmonizing donor practicies for effective aid delivery, vol. II: budget sup- port, sector wide approaches and capacity development in public �nancial management. OECD Publishing. (2011): Geographical Distribution of Financial Flows to Developing Coun- tries: Disbursements, Commitments, Country indicators 2005-2009. OECD Publishing, http://dx.doi.org/10.1787/�n flows dev2011enfr. 29 Peichl, A., and N. Pestel (2011): “Multidimensional Affluence: Theory and Applica- tions to Germany and the US,� Discussion Papers 5926, Institute for the Study of Labor (IZA). Peichl, A., T. Schaefer, and C. Scheicher (2010): “Measuring richness and poverty: A micro data application to Europe and Germany,� Review of Income and Wealth, 56(3), 597–619. Pen, J. (1971): Income Distribution. Allen Lane The Penguin Press, London. Quiggin, J., and R. Mahadevan (2010): “The Poverty Burden: A Measure of the Dif- �culty of Ending Extreme Poverty,� Australian Public Policy Program Working Papers P10-2, Risk and Sustainable Management Group, University of Queensland. Wolfensohn, J. D., N. Stern, I. Goldin, H. Rogers, and M. Karlsson (2002): A case for aid: Building a consensus for development assistance. World Bank Publisher. Young, H. P. (1990): “Progressive Taxation and Equal Sacri�ce,� The American Eco- nomic Review, 80(1), 253–266. 30