Trade and Development

Transcription

1 Trade and Development Tutorial Part I Summer Term 2017 Christoph Dörffel, M.Sc. Friedrich-Schiller-University Jena Department of Economics Chair of Economic Policy Prof. Dr. Freytag Summer 2017 Slides last updated: May 18, 2017 Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

2 Office: 5.36 (please make appointment by mail) Exam: TBD Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

3 General idea of the tutorial sessions We will......have a look into the measurement of Development This will give you clearer understanding of what development is, how it is measured and how to judge measures...will read some empirical papers about institutions, trade, foreign aid and development...and discuss them... reviewing the empirical literature will give you some idea of the theoretical models used to motivate the empirical works...a few papers will be required reading that we discuss in class and serve as introduction into the literature...talk about a special topic? Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

4 Outline 1 Introduction to Development Introduction to Development Measurement Inequality measurement Poverty measurement 2 A quick refresher on econometrics Endogeneity, IV, FE and alternative models How to read regression tables How to read an empirical paper 3 Aspects of Development Growth theory revisited Institutions and Development Foreign Aid and Development Trade and Development 4 Special Topic? Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

5 A small survey I made a small (anonymous) survey to get some feedback from you for the tutorial. It s 5 questions and shouldn t take more that 5 minutes: Go To: In case you want to talk about a special topic in the last sessions, those are the possible topics to chose from: Environmental impacts of trade Current account imbalances Economic impact of Trade Agreements (TTP, CETA) Econometrics issues (a practical guide to make your own empirical work) Foreign Aid allocation (as opposed to Aid effectiveness) Why is Africa not developing/industrializing? The informal sector in developing countries Microfinance/RCTs The survey will be open until Thursday May 4 (noon). Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

6 1 Introduction to Development Introduction to Development Measurement Inequality measurement Poverty measurement 2 A quick refresher on econometrics Endogeneity, IV, FE and alternative models How to read regression tables How to read an empirical paper 3 Aspects of Development Growth theory revisited Institutions and Development Foreign Aid and Development Trade and Development 4 Special Topic? Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

7 What is development? What do we talk about when we talk about development? How do we measure development? Definitions of classifications (p.c. income (ppp), HDI, SILICs, HIPCs) Poverty measures FGT P α, inequality measures Many many of institutional measures and indices, e.g. CPI, DPI, Polity IV, CoW, KOF Globalization Index, Freedom of the World, CPIA, ICRG Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

8 Development has many dimensions Developing countries mostly have structural features like: High population growth Younger populations than developed countries Large rural populations Large rural-to-urban migration flows Larger agricultural sector Informality Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

9 Most commonly used country classifications schemes World Bank 2017 income-classification low-income countries: equal to or less than 1025 USD lower middle-income countries: USD higher middle-income countries: USD high-income countries: USD or more By Human Development Index score low human development: HDI score less than 0.5 medium human development: HDI score high human development: HDI score 0.8 or above By debt: Heavily indebted poor countries (HIPC): debt-to-export ratio over 150%, or debt-to-government revenues over 250% Severely indebted low-/middle-income countries (SILIC/MIMIC): debt-to-gnp-ratio larger than 80% or debt-to-exports-ratio larger than 220% Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

10 And what is a developing country again? There is no generally agreed on definition The IMF/WEO makes a flexible distinction along GDP, export diversification and integration into the global financial system. But there are no strict criteria, i.e. no standing classification. Some economies such as Lithuania became developed overnight when joining the Euro Area (see ) When I say developing countries I use this synonym to ODA recipient, i.e. countries that get foreign aid. Those are all Low and Middle income countries according to the World-Bank classification, excluding G8 members, EU members and countries with firm entry date into the EU Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

11 And what is a developing country again? The UN has a definition for least developed countries (LDCs). There is a threshold where countries are eligible to be a LDC, and graduate from being a LDC: 3-year average GNI p.c. below 1035 USD plus scores on a Human Assets Index and a Economic Vulnerability Index which measures development achievements such as child mortality and economic indicators such as instability of agricultural production. If you re interested in the details: least-developed-country-category/ldc-criteria.htm The World Bank abandoned the classification altogether and prefers to use geographical distinctions: edition-world-development-indicators-out-three-featur Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

12 And what is a developing country again? Figure 1: LDCs according to UN-Definition picture source: Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

13 And what is a developing country again? Figure 2: Official Deleopment Assistance (ODA) recipient countries (2014, in % of GNI) picture source: Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

14 Measurements of development GDP and income Gross National Product (GDP) is a measure of how much a countries economy produces (imagine a physical pile of goods that is produced) Usually we take (real) per capita GDP as a simplified measure for the average income people get Gross National Income (GNI) real means inflation is netted out, to allow for meaningful comparisons over time To get meaningful cross-country comparisons, we need also correct for differences in purchasing power PPP (purchasing power parity) Data is very easily accessible online (World Bank, IMF, national statistics offices) Is (average) income a meaningful measure of development? Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

15 How does PPP work? To compare the GDPs of different countries we need to account for the different currencies in different countries One could simply use market exchange rate But they are often manipulated by governments, central banks, i.e. comparisons would be biased PPP exchange rates compare how much the same basket of goods and services costs in different countries and uses this as the exchange rate Using this exchange rate one unit of money would buy the same amount of goods and services in each country Hence, a higher GDP in ppp prices means a country can actually buy more goods and services (and therefore the exchange rate bias is mostly eradicated) The price data comes from surveys every 5 years asking for prices and consumption of goods and services of items Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

16 Other development measures Development and human well-being is multifaceted, and in general higher income highly correlates with other well being outcomes Those are e.g. literacy, life expectancy, infant mortality Especially life expectancy seems a good measure for well-being because living a longer life implies better living conditions (less hunger, decent nutrition, decent housing, better health) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

17 Figure 3: Correlation of income and life expectancy Source: Christoph World Development Dörffel, M.Sc. Indicators, 2016 Trade and Development Tutorial Summer / 68

18 Some cross-country correlations of income and... Life expectancy is increasing in p.c. income i.e. there s positive correlation The relationship is not linear Poorer countries have large increases in life expectancy when their income increases Adult literacy follows the same pattern: adult literacy increases in p.c. income Infant mortality decreases (slightly non-linear) with p.c. income Population growth decreases with p.c. income (linearly) The share of labor force in agriculture decreases with p.c. income (extremely non-linear) Always remember: CORRELATION IS NOT CAUSATION! Still: Those correlations tell us that p.c. income is in some sense a meaningful measure of development Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

19 Inequality, Poverty and more In the following we will deal in some detail with inequality and poverty measurement The technicalities will allow you to think about development measures and judge their usefulness in a more systematic way Inequality deals with distributions (of development) Poverty deals with deprivation (how many people live on less than 1.90$ a day?) We will also look at some more classic development measures and indices that can t be clearly put into the inequality or poverty category, but still are important for development Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

20 1 Introduction to Development Introduction to Development Measurement Inequality measurement Poverty measurement 2 A quick refresher on econometrics Endogeneity, IV, FE and alternative models How to read regression tables How to read an empirical paper 3 Aspects of Development Growth theory revisited Institutions and Development Foreign Aid and Development Trade and Development 4 Special Topic? Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

21 Inequality measures Comparing countries by income (or production) has shown to be a meaningful indicator However, it is an average over the whole population of each country Thus, it neglects a lot of data (information) that might have some interesting story to tell This is where inequality measures emerge in the picture They can give us an idea about the within picture of a country Usually we look at the income distribution of a country But other aspects can be important, too, like: wealth, assets, access to schools, hospitals, opportunities, capabilities Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

22 Reasons for caring about inequality Should people have access to same resources? Opportunities? Intrinsic value Economic incentives (functional role, labor efficiency) Societal aspects (perception of injustice, political stability) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

23 What is the problem with measuring inequality Consider 2 countries with incomes {50, 50} and {100, 0} Easy to say which one is more unequal But what about A = {20, 30, 50} and B = {22, 22, 56}? And what about population sizes? (e.g. Germany vs. China) There exist 4 mathematical principles that lets us make such comparisons To clarify the notation: e.g. A is a country and this country consists of a vector of individuals {a 1, a 2,..., a n } with incomes 20, 30, 50, etc.; those vectors of incomes is also called income distributions Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

24 Inequality Principles 1 Anonymity (or Symmetry) Principle: We don t care about the identity of the person who has an income, i.e. {100, 200, 300} and {300, 200, 100} are identically unequal 2 Population Principle: Cloning a population leaves inequality unchanged, i.e. {100, 200, 300} and {100, 100, 200, 200, 300, 300} have identical inequality 3 Relative Income Principle (Scale Invariance): Multiplying each income of a given distribution with the same positive constant leaves inequality unchanged. (We only care about relative incomes). I.e. distributions {100, 200, 300} and {200, 400, 600} have identical inequality. 4 Dalton Principle: Assume you have 2 income distributions A and B that are not equal to start with. If you can make an regressive transfer within A, such that A and B are equal afterwards, then B has greater inequality. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

25 Digression: Quick example for the Dalton Principle Assume the starting distributions are A = {a 1, a 2 } = {90, 110} and B = {b 1, b 2 } = {80, 120} Principles 1-3 won t bring us any further here But we can make a regressive transfer, i.e. take 10 from the poorer person in A (a 1 = 90) and give it to richer person in A (a 2 = 110). Then the new A = {80, 120} = B. We artificially worsened the situation in A to get to A = B The conclusion is: A has less inequality than B Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

26 Definition of Lorenz curve Figure 4: Lorenz curve Cumulative income: Y i = i i=1 y i x-axis: Cumulative fraction of the population: p i = i n y-axis: Share in total income s i = Y i Y n Example: point V The lower 50% of the population earns 20% of total income (s i = 0.2 and p i = 0.5 ) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

27 The Lorenz Criterion, and Principle We can directly connect the Lorenz curve to the 4 principles: Lorenz Criterion An income distribution Y has strictly higher inequality according to the Lorenz Criterion than Y if L Y L Y for all p [0, 1]. In words: If the Lorenz curve of Y is always below the one of Y. Lorenz Principle Greater inequality according to the Lorenz Criterion is equivalent to greater inequality according to the Anonymity, Population, Relative Income and Dalton principles. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

28 This helps us in a lot of cases But what if the Lorenz Criterion fails, i.e. Lorenz Curves cross? Then we can t make an comparison based on the Lorenz Criterion Formally, an inequality ranking according to Lorenz Criterion is not complete It is transitive, though. Meaning: An ordering that is not complete but transitive is called a quasi-ordering, so the 4 principles induce a Lorenz quasi-ordering (Sen, 1997, p. 143). Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

29 Not complete but transitive ordering Figure 5: Lorenz curves We can say: L y L z and L x L z but not L y L x so: you cant say whether the whole income vector of x is more unequal than y or vice versa You can induce rankings for part of distribution and every individual (transitivity) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

30 Not complete but transitive ordering That s why we use complete measures to reduce an entire Lorenz curve (income distribution) to one single number Then you get a complete ordering, but different measures (might) give you different orderings Complete measures of inequality are... E.g. the Gini-coefficient, normalized Range, Kuznets ratio, mean absolute deviation, coefficient of variation. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

31 Inequality measures & Lorenz-consistency Strict Lorenz consistency A complete measure of inequality K strictly satisfies the four principles of inequality comparisons if (i) greater inequality according to the four principles implies a strictly greater value of K and (ii) identical inequality according to the four principles implies an identical value of K. In this case, K is said to be strictly Lorenz-consistent Weak Lorenz consistency A complete measure of inequality K weakly satisfies the four principles of inequality comparisons if (i) greater inequality according to the four principles implies a greater or equal value of K and (ii) identical inequality according to the four principles implies an identical value of K. In this case, K is said to be weakly Lorenz-consistent Violation of Lorenz Criterion A complete measure of inequality K violates the Lorenz Criterion if cases exist where any of the following is true: (i) Y has greater inequality according to the Lorenz Criterion than Y while K(Y ) is strictly smaller than K(Y ). (ii) Y and Y have identical inequality according to the Lorenz Criterion while K(Y ) does not equal K(Y ). A measure of inequality (weakly) satisfying our four principles is also called a measure of relative inequality (Sen, 1997, p. 139f.). Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

32 Complete Measures of inequality (I) The Gini-coefficient G = 1 2N 2 µ N i=1 j=1 N y i y j, where µ = 1 N N i=1 y i is mean income. Graphically, that is G = A A+B where A is there area above the Lorenz curve, B the area below, in a Lorenz curve graph (see Figure 4). Is strictly Lorenz-consistent. The Normalized Range R = Sample Range Mean = 1 µ (y max y min ) Is weakly Lorenz-consistent. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

33 Complete Measures of inequality (II) Kuznets-ratio Ratio of the shares of income of the richest x% to the poorest y% where x and y stand for numbers such as 20, 40, or 90. Example: the ratio. This corresponds to taking selected points on the Lorenz curve. Is weakly Lorenz-consistent. Normalized Mean Absolute Deviation Is weakly Lorenz consistent. M = MAD Mean = 1 µn N y i µ i=1 Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

34 Complete Measures of inequality (III) Coefficient of Variation CV = STD Mean = 1 1 µ N N (y i µ) 2 i=1 Mean Logarithmic Deviation MLD = 1 N N ln i=1 ( ) yi µ Theil Index T = 1 N N i=1 ( ) yi ln µ ( ) yi µ All strictly Lorenz-consistent. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

35 Specifics MLD and T are so called generalized entropy measures. Class of Generalized Entropy measures [ 1 1 n ( ) α yi GE(α) = α 2 1] α n µ with α = 1 we get the Theil Index, with α = 0 we get the Mean Log Deviation GE is sensitive to changes at the lower end of the distribution for α s close to 0, equally sensitive to changes across the whole distribution when α = 1 (Theil) sensitive to changes at the upper end of the income distribution for α s larger than 1 i=1 Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

36 Subgroup decomposability GE-class inequality measures are is subgroup decomposable i.e. you could split up your sample in any way and get 3 measures:... 2 for each group (within-group inequality)... and one that measures inequality across the two groups (accross-group inequality) Adding those 3 together gives the GE index for the whole sample again. GE measures are the only measures (we talked about) which has this property Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

37 Summing up inequality measurement (I) In the literature you usually encounter the Gini, sometimes the Theil index and MLD. For all of this measures: They rely on extensive and high quality data of incomes. Those are usually compiled through surveys (such as Censuses or the Socio-Economic Panel in Germany). Imaginable, data quality is a problem in poor countries Also, very rich people are reluctant to tell someone their real income. This biases those inequality measures on the upper part of the distribution. Some studies therefore use data from tax-authorities which might give a better picture of inequality. (Anand, Segal, et al. 2014) Then again, richer people have higher incentives to avoid taxation...e.g. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

38 Summing up inequality measurement (II) In some of the literature (or in the press) you will encounter measures like the 90/50, or 80/20 ratio. This means comparing 2 persons in the distribution and judging inequality. While this is easy to grasp, calculate and can give a first hint about inequality, it neglects most of the data The principles above can be guideline to judge such measures, and demonstrate that measures like the Gini-coefficient or the Theil-index are preferable Empirical inequality-growth literature deals with two mean topics: (i) Does growth increase inequality over time? (ii) Do countries with higher inequality grow slower? Note here, that growth is a proxy for changes in the development of a country...lastly some results from the empirical literature in next few slides... Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

39 Historical global inequality Figure 6: Historically, inequality within has decreased and between countries has increased, from Bourguignon and Morrisson (2002) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

40 This trend might have reversed since 1990 from Anand, Segal, et al. (2014) Figure 7: Between and within-country inequality It seems the long-term trend in Bourguignon and Morrisson (2002) has reversed since I.e. Inequality between country decreases (slightly) and inequality within countries increases Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

41 The elephant graph Figure 8: The elephant graph, from Lakner and Milanovic (2013) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

42 Figure 9: Measuring top incomes is a real difficulty, from Anand, Segal, et al. (2014) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

43 Readings Most of the discussion about inequality measurement is based on Ray (1998) chapters 6 & 7. The axioms, Lorenz curve, complete measures of inequality in ch. 6 Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

44 1 Introduction to Development Introduction to Development Measurement Inequality measurement Poverty measurement 2 A quick refresher on econometrics Endogeneity, IV, FE and alternative models How to read regression tables How to read an empirical paper 3 Aspects of Development Growth theory revisited Institutions and Development Foreign Aid and Development Trade and Development 4 Special Topic? Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

45 Turning to poverty Why do we care about poverty? Obviously, there s an intrinsic value for understanding why people suffer and how we might help We will deal with the functional implications of poverty for development, i.e. how does poverty affect individual s incentives, and their human capital Poverty is an absolute concept: We define a poverty line that divides people into being poor or not Note the contrast to inequality Relative Poverty measures are really inequality measures ( at-risk-of-poverty-rate, below 60% of median income) How to measure poverty? Income? Consumption? Assets? Nutrition? Capabilities? What s the important unit of observation? The individual or households? Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

46 Measuring poverty (I) Again, Sen (1997) gives us some help He argues there are 2 distinct issues: 1. Identification: We need to determine who counts as poor, i.e. set a poverty line 2. Aggregation: We need to evaluate the situation of the poor vs. non-poor The World Bank defined the poverty line of 1.90$ a day (700$/year) in Before it was 1.25$ from 2005 on, and 1$ since 1990 The line is supposed to reflect minimum nutrition, clothing and shelter needs This is of course in PPP dollars, so the goods you could buy with 1.90$ in the U.S. to meet your basic nutrition, clothing and shelter needs are the same in other countries with other currencies This solves the problem of identification. Aggregation is more complicated... Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

47 Measuring poverty (II) Fortunately, we can borrow the first two principles from inequality measurement (Anonymity and Population) The third is easy: 3. Focus principle: We don t care about the incomes of those who are above the poverty line With those 3 we can judge whether 2 distributions have identical poverty, or not 4. Monotonicity Principle: An increase (decrease) in the income of poor individual decreases (increases) poverty (including individuals at the poverty line) 5. Transfer principle: A transfer from a poor individual to any weakly richer individual increases poverty Note: weakly means greater or equal ; this is a regressive transfer among the poor, or from a poor to a non-poor See Sen (1976) for Monotonicity and Transfer principles. The other 3 principles are implied. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

48 Quick example for the poverty principles Assume p = 10, and income distributions X = {6, 12, 12} Y = {16, 6} With Anonymity and Population we get X = {6, 6, 12, 12, 12, 12} Y = {6, 6, 6, 16, 16, 16} and with Focus and Monotonicity we can say that Y has greater poverty than X. But what about: W = {3, 3, 12} Z = {0, 6, 15} we need the Transfer principle here transfer 3 from w 1 to w 2 to make: W = {0, 6, 15} = Z and therefore Z has greater poverty than W. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

49 Measuring poverty (III) Similar to what we did with the Lorenz curve and our inequality principles, we can have a graphic representation of poverty For this, we need the cumulative distribution function (CDF) of our income distribution and check whether the CDFs of two countries lie above/below each other This is called First-Order Stochastic Dominance and Second-Order Stochastic Dominance. FSD is the weaker form and does not include the Transfer principle (But let s spare the technicalities and turn to some actual measures next...) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

50 Measuring poverty (IV) The FGT-Class of poverty measures Foster, Green, Thorbecke (1994) derived a measure that is very useful. From Foster et al. (1984, p. 763): P α (y; z) = 1 n q i=1 ( gi ) α, z where z is the poverty line, g i = z y i is the individuals distance from the poverty line (or income shortfall), q is the number of poor household (hh s below z), n is the total number of households. For α = 0 this is: P 0 (y; z) = 1 n q i=1 ( gi ) 0 1 = z n q 1 i=1 which is simply the share of people below the poverty line, i.e. Headcount Ratio (HCR) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

51 Measuring poverty (V) The FGT-Class of poverty measures For α = 1 this is: P 1 (y; z) = 1 n q i=1 ( gi ) 1 z which is called the poverty gap (index). It describes the average distance of all individuals below the poverty line to the poverty line. For α = 2 this is: P 2 (y; z) = 1 q ( gi ) 2 n z which could be called the squared poverty gap (index). It describes the average squared distance of all individuals below the poverty line to the poverty line. P 0 violates the Transfer principle, only weakly satisfies Monotonicity P 1 only weakly satisfies the Transfer principle P 2 strictly satisfies all principles Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68 i=1

52 Measuring poverty (VI) The FGT-Class of poverty measures What is with P > 2? heavier penalty on individuals further away from the poverty line FGT-measures are also subgroup decomposable Pα total = J j=1 n j n Pj α Which is just the sum of P-measure of each subgroup, weighted by their population share Note: there is no across-group-term in contrast to the MLD measure for inequality Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

53 Measuring poverty (VII) So far we looked only at poverty measures that measure poverty along one dimension, e.g. income However, if your interested in peoples capabilities other measures might be better Let s look at two of them, the Multidimensional Poverty Index (MPI) from Alkire and Santos (2010) (UNDP and OPHI), and The Human Development Index (HDI), which is not a classic poverty measure by itself but a development index It was development at the UN Development Program (UNDP) for the 1990 UN Development Report by Mahbub ul Haq and his team, working (allegedly) alongside (but certainly inspired by) Amartya Sen and his capability approach This illustrates that poverty is obviously a very important aspect of under-/development but not the only one Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

54 Measuring poverty (VIII) MPI 1 Dimension: Health 1 Child mortality (child died in the last 5 years in hh?) 2 Nutrition (any adult or child in hh stunted?) 2 Dimension: Education 1 Years of Schooling (at least 1 hh member 6 years schooling or more?) 2 Schooling Attendance (school-aged kids going to school til class 8?) 3 Dimension: Living Standard 1 Cooking fuel (not cooking with dung, wood or charcoal?) 2 Sanitation (got improved sanitation facility?) 3 Water (access to safe drinking water supply not further than 15min away?) 4 Electricity (have it?) 5 Floor (dirt, sand or dung floor?) 6 Assets (have at least one of: TV, radio, fridge, motorbike, car, truck, phone?) Dimensions are then equally weighted (1/3), variables are also equally weighted within each dimension (1/2 and 1/6). Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

55 Measuring poverty (IX) MPI (II) Each of the 10 indicators are binary (deprived=1 / not deprived=0) A household counts as multidimensional poor if it is 33.33% or more deprived The first part of the MPI is then headcount of all (multidimensional) poor people, i.e. H = q n, where q are number of people that are multidimensional poor, and n the total number of people H = q n The second part is called intensity of poverty. So for all hh s that are deprived (33% threshold) it measures how much they are deprived on average q 1 c i q A = The PMI is then the product of the headcount and intensity MPI = H A Source: United Nations Development Programme UNDP (2016). Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

56 Measuring poverty (X) HDI The HDI also has 3 dimensions, but less variables: The dimensions are HDI = (I health I education I income ) Health (var: life expectancy in years) 2 Education (vars: expexted years of schooling, mean years of schooling) 3 Standard of Living (var: p.c. GNI (PPP)) Each dimension index is calculated by: I dim = actual value minimum value maximum value minimum value Min/max values are defined by UN: life expectancy: 20/85, expected schooling: 0/18, schooling: 0/15, p.c. GNI: 100/75000 GNI is in logs because of decreasing marginal benefits of more income Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

57 Measuring poverty (XI) HDI (II) A quick example (taken from Human Development Report 2016 Technical Notes) Observed data in Georgia: Life expectancy at birth (years) 75.0; expected years of schooling = 13.9; Mean years of schooling = 12.2; GNI p.c. = It s just plugging in some numbers... I health = = I exp school = = I school = = I educ = = I gni = ln(8856) ln(100) ln(75000) ln(100) = HDI Georgia = ( ) 1 3 = Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

58 Measuring poverty (XII) HDI & MPI maps HDI map: year=1980 MPI map: StatPlanet.html Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

59 Measuring poverty (XIII) HDI discussion What are advantages and disadvantages of the HDI? The HDI is a relatively simple summery measure in key dimensions of development (health, access to knowledge and standard of living) It can reveal that a country can do better in terms of health and education achievements than you might expect from looking at only income Disparities in income is greater in that disparities in other development indicators (health and education are inputs to production) Income gains might not lead to human development/increases in human well-being (growth without development) It has mixed data sources (surveys, national accounts) Statements about true quality of life are limited (e.g. security, human rights, civil liberties) It makes no statement about in/equality Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

60 Measuring poverty (XIV) HDI & MPI discussion The MPI is more complicated but maybe tells the more important story: what is the actual situation of the people It uses only survey data (and one single survey to calculate one country-year data points) more consistent in itself B/c of this, overall, there are (a lot) less data points (years and countries covered) than for the HDI In addition, survey data has its own specific problems. Questions about food consumption, and clothing purchases are complicated, people have short memories (recall periods) and difficult to compare over countries; and surveys are expensive. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

61 Measuring poverty (XV) HDI & MPI discussion Both the HDI and MPI have another problem: because they are aggregates from multiple sub-components, changes can offset each other. The HDI could remain unchanged if some country gets better in the health dimension, but worse in the education dimension (and so on). The MPI could remain unchanged if the poverty headcount decreases but the poverty intensity increases. Lastly, we can t really handle the HDI with our principles (because there is no identification). The headcount-part (H) of the MPI violates the transfer principle (see our discussion of the P 0 ) and therefore the overall MPI violates it, too. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

62 Measuring poverty (XVI) Summary It is important to look at different measures because they might produce different rankings. E.g. if you rank countries by GDP and the HDI, Sri Lanka has a differential of +5, Guatemala, -20 Pakistan -28 ranks (Ray, 1998, p. 29). Different measures try to catch different aspects of development, e.g. inequality, poverty, health, education, corruption and and and So each of them might be useful in different situations...lastly some results from the empirical literature in next few slides... Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

63 Empirics of global poverty Some evidence from Chen and Ravallion (2013) (and Chen and Ravallion (2010)). Figure 10: Absolute poverty is declining Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

64 Empirics of global poverty Figure 11: First-order stochastic dominance Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

65 Empirics of global poverty Figure 12: Poverty gaps from Chen and Ravallion (2013) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

66 Empirics of global poverty Figure 13: Africa is lagging behind, East Asia a success story b/c of China (South Asia is mostly India) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

67 Empirics of global poverty Figure 14: Absolute poverty is declining more in higher growth countries, but relative poverty isn t (anywhere) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

68 What other (development) data is there? Corruption: Corruption Perceptions Index Economic Freedom: Economic Freedom of the World (Fraser Institute), Index of Economic Freedom (Heritage Foundation) Globalization: KOF Index of Globalization Polity: Polity IV, Freedom in the World (Freedomhouse) Composite Institutional measures: CPIA (Country Policy and Institutional Assessment), ICRG (International Country Risk Guide) (both contain sub-indices for corruption and more), State Fragility Index (Center for Systemic Peace), Fragile States Index (Fund for Peace) Correlates of War project, IMF, World Bank have many more useful data The Human Development Report reports indices for different aspects: Inequality-adjusted Human Development Index (IHDI), Gender Development Index (GDI), Gender Inequality Index (GII) Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

69 Readings If not cited otherwise of the discussion about poverty measurement is based on Ray (1998) chapter 8 and the appendix in ch. 8. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

70 References I Alkire, S. and M. E. Santos (2010). Acute multidimensional poverty: A new index for developing countries. In: OPHI Working Paper No. 38. Anand, S., P. Segal, et al. (2014). The global distribution of income. In: International Development Institute (ed.), Working Paper 2014, p. 01. Bourguignon, F. and C. Morrisson (2002). Inequality among world citizens: In: The American Economic Review 92.4, pp Chen, S. and M. Ravallion (2010). The developing world is poorer than we thought, but no less successful in the fight against poverty. In: The Quarterly Journal of Economics 125.4, pp (2013). More relatively-poor people in a less absolutely-poor world. In: Review of Income and Wealth 59.1, pp Foster, J., J. Greer, and E. Thorbecke (1984). A class of decomposable poverty measures. In: Econometrica: Journal of the Econometric Society, pp Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68

71 References II Lakner, C. and B. Milanovic (2013). Global Income Distribution: From the Fall of the Berlin Wall to the Great Recession. In: World Bank Policy Research Working Paper No Ray, D. (1998). Development economics. Princeton University Press. Sen, A. (1976). Poverty: an ordinal approach to measurement. In: Econometrica: Journal of the Econometric Society, pp (1997). On economic inequality. (expanded 1997 edition). Oxford University Press. United Nations Development Programme UNDP (2016). Human Development Report 2016 Technical Notes. In: url: undp.org/sites/default/files/hdr2016_technical_notes.pdf. Christoph Dörffel, M.Sc. Trade and Development Tutorial Summer / 68