Scribd Logo
Upload

Welcome to Scribd!

  • D.B (Econ) U 1,2 DR DV (C)

    Uploaded by

    Satakshi Nandy
    6 views26 pages

    Original Title

    d.b (Econ) u 1,2 Dr Dv(c)

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    6 views26 pages

    D.B (Econ) U 1,2 DR DV (C)

    Uploaded by

    Satakshi Nandy

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 26
    Chapter 6: . H Economie In 4 “Mali, ry unequal when we see it. If two Peo) t ciety is very a society i fit, that’s unequal. If the’ SYPpo e person has all of it, aw S unequal, f they Split § Seq and one R evaluate intermediate divisions (such as 39_5,~50, ang Ane cake o share aC e ao equal. We cee of precision. 40-60) with a fair am vy, however, once we have more than two indiy, Alll that goes inenneslints divisions of the cake. Is it Obvious} als and we ty fo eRe among three people with a 22-29-56 divine t compare a and in even tore complicated ones as Well, it mi ht be oe 7 In such cases, oe ure” inequality. This means that we develop’ oy vant to try and a that permit the ranking of income Or Wealth distribuio” inequalty i nt situations (countries, regions, Points of time, and so ms iy ean naturally arises: what are the Properties that a d ma index should satisfy? It is difficult to have complete y, ae ‘subject and there is none. If, to avoid controvers; only very weak criteria, then many inequality indices can be suggest, consistent with the criteria, but probably Siving very different Tesul used in actual inequality comparisons. If, hand, we i X79) On), Sirable” animity Y we lay do ed, each stricter criteria, but the criteria loses wide approval. As we will see, this Problem is endemi have a clear idea of you are identifying measure. If you are @ policy maker or | this form of identification can be useful or dangerous, de i erstand the underlying criteria Beneric individual, 4 We use the index i ++, 1. An income distribution received by each individual i i; t i= 'S @ description of hi hus, i 12. OW much j i Wy, My). ch income yj is le are interested jy, compari " TH Sistributions, To this end, we nee al t ii fees ’ oul Mequality in the form of applicable criteria, (1) Anonymity Pi who ig €arnin, Tinciple, From an e 8 the income, 8 the relative “inequality” of two income d to capture Some of our intuitive notion: z - tter thical point of view, it does not ma' Situation Wher ji arms € Debraj earns x and Rajiv “In this seq sure. You Mion, We refer to ‘neo called ot Ud repiace stn nome” as the crucial variable whose inequality we wish i be Sn individual, 12th, lifetime inco, e, an ise, the recipient interested b OU could », lace thi “hoot, ahd so on. Likewise, ; might is by household” or any other grouping that you 63. Measuring Economic Inequality os be viewed as identical (from the ¥* nich Debraj earns y and Rajiv earns a this sort of change (if x happens to eS difficult for him to persuade other oo ality in his society has deteriorated ee ane among people should not ma: ube ple of anonymity. Formally, rindi tw come distribution so that sould Point of view of inequality) to one x. Debraj may well be disgusted be larger than y), but it will be People that the overall degree of because of this. Thus permutations : tter for inequality judgments: this is this means that we can always arrange WSS SY, which is the equivalent of arranging individuals so that they are ranked from poorest to richest. (2) Population principle. Cloning the entire population (and their incomes) should not alter inequality. More formally, if we compare an income distri- bution over n people and another population of 2n people with the same income pattern repeated twice, there should be no difference in inequality among the two income distributions.° The population principle is a way of saying that population size does not matter: all that matters are the propor- tions of the population that earn different levels of income. Criteria 1 and 2 permit us to view income distributions in a slightly different way. Typically, no data set is rich enough to tell us the incomes of every single individual in the country. Thus the data are often presented in the following way. There is a set of income classes, where each class typically ‘S presented as a range of incomes; for example, “$100 per month or less,” “$300-400,” and so on. Figure 6.2 illustrates this procedure using a hypothetical example. A pop- ulation of people earn an income somewhere between zero and $1,000 in this example. The raw data are shown in the left panel of the figure. (You will almost never see data expressed like this for an actual population.) The “Nonymity principle tells us that we can number people in order of increas: Ng income and no useful information is lost. The population principle ee 'S that it does not matter how many people there are; we may normalize “verything to percentages. The right-hand panel gives us a common way to Put together this information. Income classes are on the horizontal axis and le Percentage of the population that falls into each income class is on t . "tical axis, Neither the names of people nor the actual numbers in eac! “ome class matter. “Warming: C\ tof the population wi may '8: Cloning only one segment Pt The ell affect our notions of inequality. Suppose that there are tered hhile keeping, the remainder unal nave two incomes, $100 and $1,000 ded the same PerceePlstion principle says that all income distributions are equally unequal prov ee ae 0 e low 3 inequayre® °f People earn $100. If the proportion of people eaming YY will, in general, be affected. Chapter 6: Economic Ineguay, i 176 9 ° - . | . 0 yan a “ j on i s” / | a | ® Lf | |e ee OF saa nto nrmnig EO Figure 62. Income distribution arranged by income classes. igure 6. . inci; t as population shares matter and (@) Relative eae Caton ioe do not it is possible to argue hs pene Eee matter and the absolute levels of these incomes aa not. If one income distribution is obtained from another by Scaling ev. erybody’s income up or down by the same percentage, then inequality should be no different across the two distributions. For instance, an income distri- bution over two people of ($1,000, $2,000) has the same inequality as ($2,000, $4,000), and this continues to be true if dollars are replaced by cruzeiros or yen. This is the relative income Principle: it is tantamount to the assertion that income levels, in and of themselves, have no meaning as far as inequality measurement is concerned. Certainly, absolute incomes are important in our overall assessment of development, although the distinction between “abso- lute" and “relative” in the context of inequality measurement may not be that easy to drawS With the a : nf data in a fo, it is now possible to prese wn. Both population and in- erent average income ley ery same hypothetical dat. “Isitas ree a ‘obuy the relative income pri Principle a: the i What we hs Sense, is the har o’ 28 the population principle? Not really. alters stand, our oo ei bappiness” oF utility, hoeve deep may be measured oF in the utilities of diferent ingineatY can be quantified at all forces us to make the RAMP: Sen (7g) Blt that ig requiseat® ©" be compared, (On the analytical framework tat asserts ca Roberts {1940} aa ee S¥Stematic egalitarian judgments, se, it neverthe Bes are proport inthe Felative income principle needs more ri use Chapter ProPOttional to ie atv Principle nee make ME shortfalls below sonhiet 8 will Partial fa This is a strong assumption. We lute OME Poverty line UY Make amends by studying the effects of abso ey els. Figure 6.3 shows how this ‘a that we used to generate Fig- ane ip Me peasuring Economic Inequality 7 ernst Farm aun wa ule ‘second Quinte proves ete ° 5 10 8 Fal b = se Porcentage of Total Income Earned Figure 6.3. Income distribution by population and income shares. ure 6.2. In Figure 6.3 we have divided the population into different equal- sized groups in order of poorest to richest. Because population size is unim- portant in the measurement of inequality (by the population principle), using percentages is enough. We use quintiles (one could use deciles as well de- pending on the richness of the original data). For each quintile, we record the income share earned by that quintile of the population. Because people have been arrayed from poorest to richest, these income shares increase as we move from the first to the fifth quintile. The relative income principle tells us that income shares are all we need in the measurement of inequality. (4) The Dalton principle. We are now in a position to state our final crite- tion for evaluating inequality. Formulated by Dalton [1920], ’ the criterion is ‘undamental to the construction of measures of inequality. Let (yi, Ya, ---, 7) € an income distribution and consider two incomes y; and y; with yi = ¥) ister of income from the “not richer” individual to the “not poorer” in- 0 ‘dual will be called a regressive transfer. The Dalton principle states that if ne income distribution can be achieved from another by constructing a Se ae of regressive transfers, then the former distribution must be deemed "é unequal than the latter. of f1OW far do these four criteria take us? Understanding this \ © next section. Before we do this, let us formally detine an inequality dyn. It is a rule that assigns a degree of inequality to each possible distri- °n of the national cake. In other words, it takes each income distribution fs also called the Pigou-Dalton principle: tanding, this is the task * also Pigou [1912], after whom the principle i Chapter 6; Econom iC Ie un thought of i tt 8 jue to it thet can Ee ieee equality sign’ # Ve aner value of oS be inter 1 Presence of of ty we a jon. A oe inequality index can preted as a fung ist ay, Thus 4 inequality form 1 = 1(Yar Yar Yn) . Bteap tion te of hy nceivable distributions of income (Yi, Yp,. My emer that the inequality measure Satisfy the anonymity waated formally a8 fOllows: the function 1 is completely 92" ciple can or mutations‘of the income distribution (y;, y,,..., Yn) amon tive to all Ewa “_.,m}. Similarly, the requirement of the population 2 individuals {1 i for every distributi i ple can be translated as saying that for every ibution (y;, y,, 21% defined over # e Ui TYy, Yar =r Yn) = Yay Yar oo Yi Yar Yar ee Ya)y so that cloning all members of the population and incomes has no eff: Thus by taking the lowest common multiple of the populations of any 3. lection of income distributions, we can always regard each distribution «5 effectively having the same population size. The relative income principle can be incorporated by requiring that for every positive number A, TY Yar +1 Yn) = TAY a AYar «+1 AYn)- Finally, I satisfies the Dalton transfer principle if, for every income distro tion (y;, Yo,..-, Yn) and every transfer 5 > 0, Te Yin Yr ee Yn) LY gy ee Yy 8p HS: whenever y, < Y}. 6.3.3. The Lorenz curve ; ous 5° ton ee s ou way to see what the four criteria of the prevere ie 's Pictures often speak more than words, and in : ict edie been ‘urement, there is a nice diagrammatic way © Phe Le imcome in any society. The resulting graph is calle ther 'S Very often used in economic research and dis¢ aa curve, which fore, it is : it worthwhile to invest a little time in order to ‘understand incom is order dep | Figure 64 shows Pe°Ple in a population in increasiny cumulat: S a typical Lorenz curve. On thé horizon ie Percentages of meraiat} «increasing O atiol come. Thy of the population arranged in inc the port US poi Points on that axis refer to the poorest 20% o Suppose we g * e tal axis, We of i” wring Economic Inequality 179 3 Meas 100% Cumulative Income EI o Lorenz Curve 10% oa Cumulative Population &°% 100% Figure 6.4. The Lorenz curve of an income distribution. the poorest half of the population, and so on. On the vertical axis, we mea- sure the percentage of national income accruing to any particular fraction of the population thus arranged. The point A, for example, corresponds to a value of 20% on the population axis and 10% on the income axis. The inter- pretation of this is that the poorest 20% of the population earns only 10% of overall income. Point B, on the other hand, corresponds to 80% on the pop- ulation axis and 70% on the income axis. This point, therefore contains the 'nformation that the “poorest” 80% enjoy 70% of the national ae on et 90% have 30% “quivalent way to describe this is from “above” the richest 20% have eo 5 Of gross income for themselves. The graph that connects all these points 1s called the Li orenz curve. . rt rf. Notice that the Lorenz curve begins and ends on the 45 Me ae wl earn 0% of national income by definition and the poorest aor vould i Fi o income. Ho ‘le population, and so must earn 100% of the income: F oe woul re} os same income! , eels evenuer) hed eae with the diagonal of the ‘ tly 10% of " coincide everywhere with the 45° line, : °% The pootest 10% (however selected) would then ave same 10%. In onenel income, whereas the richest 10% will also ae Cee exactly that (/O%4S, any cumulative fraction of the population vesses the relation” ship fraction of national wealth. Because the 45, Hine o~Pring inequality, the = X, it is our Lorenz curve in this case With increasing Chapter 6: Econom ic o fall below the diagonal in a loop that is a ie e Arg the diagram it cannot curve the other way 1 YS by right of mn is simply the contribution of the Person the curve at uy eae of national jmcon’. Teen we have Order, rest to richest, this “marginal contribution" cannot oie, s saying that the Lorenz curve can never Bet flatter fal ft to right. Swe . cioure 6.4, the “overall distance” between the 45° i Thus in Figui dlicative of the amount of inequality present in ne and the Lorenz curve Saar greater the extent of inequality, the further Fe Society that it sere the 45° line. Hence, even without writing down, Loren, a the measurement of inequality, we can obtain an intuitive n for. how much inequality there is by simply studying the Lorenz curve, Some of the conceptual problems encountered in the meas inequality can also be brought out with the aid of this diagram. In Fi 65, the Lorenz curves of two different income distributions, marked ti and L(2), are represented. Because the second curve L(2) lies entirely Sele the first one, it is natural to expect a good index to indicate Sreater inequali in the second case. Let’s try to understand why this is so. The fact that a4 lies above L(2) has the followi i i ul) ve starts t his is the 4 move from le ‘Urement of Cumulative Income a: Figure 65 Cumulative Population ™* Sing th 8 the Lorenz curye to make judgments. uring Economic Inequality 3. Mews 181 mpiased” toward the poorest x% of the po ulati ava reason that L(1) should be judged inbre aguK a 3 nis criterion for inequality comparisons is known as the Lorenz criteric Ir says that if the Lorenz ae ofa distribution lies at every point oe ght of the Lorenz curve of some other distribution, the former should be ee ed to be more unequal than the latter. Just as We required an inequali ae to be consistent with the criteria of the Previous section, we ra ai ee be consistent with this Particular criterion. Thus an inequality pie Lorenz-consistent if, for every pair of income distributions (Ys Yar eee Yy) and (21) 22r-0+ 7 Zm)r . T(Vay Yar v1 Yn) & UZ, Zap. 26 Zp) whenever the Lorenz curve of (y;, y,..., y,) lies everywhere to the right of 24) Zay v0 Zn) " This is all very nice, but now confusion starts to set in. We just spent an entire section discussing four reasonable criteria for inequality comparisons and now we have introduced a fifth! Are these all independent restrictions that we have to observe? Fortunately, there is a neat connection between the four criteria of the previous section and the Lorenz criterion that we just introduced: an inequality measure is consistent with the Lorenz criterion if and only if it is simultaneously consistent with the anonymity, population, relative income, and Dalton principles. This observation is very useful. First, it shuts down our apparent expan- Sion of criteria by stating that the earlier four are together exactly equivalent to the Lorenz criterion. Second, and more important, it captures our four cri- teria in one clean picture that gives us exactly their joint content. In this way ‘ecan summarize our verbal ethical criteria in simple graphical form. The Preceding observation is so central to our understanding of inequal- tty that it is worth taking a minute to see why it is true. First, observe that the renz curve automatically incorporates the principles of anonymity, popu- tion, and telative income, because the curve drops all information on in- Come or Population magnitudes and retains only information about income be Population shares. What we need to understand is how the Dalton prin- “ble fits in. To see this carry out a thought experiment, Pick any ee ‘stribution and transfer some resources from people, say from the ae Population Percentile, to people around the eightieth population Feuer le. S is a regressive transfer, and the Dalton principle says that inequality 82S up as an ‘ : esult. ‘Bure 66 tells us what happens to the Loren: 1 ss i S the original Lorenz curve and the thinner c la z curve. The thicker curve urve shows us the Lorenz ‘ [1985]. For a useful discussion of the history of this result, see the survey by Foster [ Chapter 6: Economic In Ie 182 _——_—— —- Mality enzcune | old Lor | | Now Lorenz Curve | 80 | 300 40 Cumulative Shares of Income (%) 0 20 40 60 80 100 Cumulative Fractions of the Population (%) Figure 6.6. The Dalton principle and the Lorenz criterion, curve after the transfer of resources. What about the new curve? Well, noth- ing was disturbed until we get close to the fortieth percentile, and then, be- cause resources were transferred away, the share of this percentile falls. The new Lorenz curve therefore dips below and to the right of the old Lorenz curve at this point. What is more, it stays below for a while. Look at a point around the sixtieth population percentile. The income shares here are re- duced as well, even though the incomes of people around this point were not tampered with. The reason for the reduction is that Lorenz curves plot cumulative population shares on the horizontal axis and their cumulative in- come share on the vertical axis. Because people from the fortieth percentile citcth ponent the benefit of the eightieth percentile, the new share at ee forty a ene Population mark (and indeed, at all percentiles ae ao ra y) must also be lower than the older share. This state of @ a all effect of thy ne eightieth percentile comes along, at which point the on 10 exactly the ee vanishes, At this stage the cumulative shares ae curves again ae at which they were before. In other words, the bow - bowed to the ae after this point. In summary, the new Loren? cur at the Lorenz criterion att 2'4 (at least over an interval), which mean® converse a Nors the Dalton principle; that is, they aia cont Parison is true as well: if two Lorenz curves 4 Parable acc, in Figure 6.5, then i ae Lorenz criterion, as in the case of L(1) and on ‘ers leading from Lay ie Possible to construct a set of disequaliing ad i : e Of this chapter, /(2), We leave the details to an exercise at Measuritg Economic Inequality 63: 183 Cumulative Income 25% 50% 75% 100% Cumulative Population Figure 6.7. Ambiguous comparisons: A Lorenz crossing. At this point, it looks like we are all. set. We have a set of criteria that has a clear diagrammatic reformulation. It appears we can compare Lorenz curves using these criteria, so there is apparently no more need to make a fuss about inequality measurement. Unfortunately, matters are a bit more complicated. Two Lorenz curves can cross. Figure 6.7 illustrates a Lorenz crossing. There are two income distribu- tions that are represented by the Lorenz curves L(1) and L(2) in the diagram. Observe that neither Lorenz curve is uniformly to the right of the other. For ‘wo income distributions that relate to each other in this fashion, the Lorenz {riterion does not apply. By the equivalence result discussed previously, it follows that our four principles cannot apply either, but what does it men for these criteria “not to apply”? It means that we cannot go omens ser “tion to the other by a sequence of Dalton regressive transfers, Put anole ay, there must be both “progressive” and “regressive” transfers in go!ns a e illustrates this a one distribution to the other. The following example illus int. Example Suppose that society consists of four individuals who earn ee 175, 125, 200, and 600. Now consider a second income distbution given ¥ (25, 175, 400, 400). Compare the two. We can “travel eae from the first ution to the second in the following manner. First trans sa 300 from the FetSon to the second: this is a regressive transfer: Next Wael ts ap Surth person to the third: this is a progressive transfer. Chapter 6: Econom;, ic neauaty these transfers are just a “co ° 7 Struct; “ m bution. OF conve occurred (eg, the two distributig. tov d ha’ 5 ONS tha the second dist nee on societies). Try another Construction, Trang pa not somethin four-perso the third: This is Fae Transfey Nov be fort pe first Pip third this i progressive: ‘Again transfer 54 5 5 to the sive as well. in, 150 from the four ond! this is progres ‘ h to the the fourtl .d distribution. TO} We arrive at te le al many imaginary ways to travel from the first to there are int is that they all necessarily involve at |e. jon, but ihe ptobtessive ae (Try it.) In other — See previous section are just Rot enough to Perm the four principles © case we have to somehow weigh iN our minds the a a esse transfer(s) against the “benefit” of the progressive eA These trade-offs iS almost impossible to quantify in a Way so Bete a in the example? Sure enough, they mirror What peut FF the comparison. The poorest 25% of the Population eam ie in the first distribution and only 2.5% of the income in the second, The opposite comparison holds when we et to the poorest 75%, of the population, who enjoy only 40% of the total income under the first distribution, but 60% of the income under the second distribution. Now go back and look at Figure 6.7 once again. You can see that L(1) and L(2) are Precisely the Lorenz curves for the two distributions in this example, Hence, there second distributi jive an one regressi nz curves provide a clean, visual image in a country. Figure 6.8 Provides several : cllfetent countries, By looking at these figures, with a little iment of income inequalities in different parts of the world, and Superimposit : ‘equalities actogs 0 court on of any two diagrams you can compare "enz curves for diffey 6.3.4 » Com lete m i _ pI! asures of imequality ity eS pre i ij Provide a Pictoria) Teprese, aken /. ere a fe two 'S and Tesearcherg v Policy rociet ntation of the degree of inequal- Problems with such a representation. First, are often int, : Saal ss and “rested in summarizing inequ’! tes get 987) ap a 7 for a fi Pate prog er e "Shorr This Princes "resi ‘ sensitivil former shes at ye (the pe: Which they call transfer seri TP OSie tran thicaly | ary at the lower end of the income distribution Me tay ¥ alloweg , 'BUing th, eancen the "0 rule ou “ahsfer lowed” to, tif both involve . sfer, then out anbigu Principle further teigh the latter. ineguat Hi haul om fae under the 7 ali Id come dow roadens the 5 ope of Somer is stil not enous? TeSsive h Principle, 63 Measuring, Economic Inequality Brazil 100 80 Cumulative Shares of income: 0 2) 40 Ot Cumulative Shares of Population (%4) Egypt i i 3 Cumulative Sh 0 2 40 6 8 100 Cumulative Shares of Population (‘4) Kenya 100 ) 6 3 8 40 Cumulative Shares of Income 3 ® 2% 4 6080100 Cumulative lative Shares of Population (%) Figuy re 6.8 Lorenz curves for different countries 185 Mexico 100 i 3 i i 5 9 2% © 6 ‘Cumulative Sha i of Population (4) India 100 —__ 80 40 20 z i 3 i 3 g ; a 0a 0 2 ©6406 100 Cumulative Shares of Population (%) Uganda 100 - B® S60 40 4 a 20 aan) a ol er ee ee “Cumulative Shares of Population (%) Source: Deininger-Squire data base; see Deininger and Squire (1996a] i Pe umber, something that is more co “quality ee when Lorenz curves cross, every coneei ings. Thus an inequality mea ing of ing “esiyable income distribution can me distributions. As we will see, te and quantifiable than a provide useful in- t a number for mplete rank- 5 not come nicrel they cannot sure that spits ou be thought of as a cor this completeness doe: Chapter 6: Economic Ineguy ! ality a . Taiwan Korea 100 /j : _ A ' ea | F Fe : oe i i | iF zo i Eo se i Q / 6 i = * 2 i 3 ee I ol =<. : 00 a i Cumulative Shares of Population (1) 100 ‘Cumulative Shares ‘of Population (%) United Kingdom Gnet eee ————— z zw : = 0 bo a ed . Za go : om 100 Cumulative Shares of Population (%) ° 2 40 60 8 100 Cumulative Shares of Population (%) Figure 6.8. (Continued). free of charge: it means that in some situations, inequality measures tend to disagree with one another. We now survey some commonly used inequality measures.'? We use the following notation. There are m distinct incomes, and in each income class j, the number of individuals earning that income is denoted by 1). Thus the total so os, am eae of people nis simply equal to })”", n,, where the symbol List cara enotes the Sum over the income classes 1 through m. The mea”! ome distribution is simply average income, or total income divided by the total number of People. Thus 1 nme i= The follow lowing (complete "See Sen [1 Plete) measures of equality are often used. . ee Sen [1997] for Iratmert ot teed ot 8 discussion of these ye ove °F economic inequ . ‘or nomic inequa and other measures, and for a compreher Ny. ; Measuring Economic Inequality fo 4 (1) The range: This value is given by the differen ° richest and the poorest individuals, divided by tl ly of the units in which income is measu 187 me in the incomes of ¢ mean to express it tl 4 ntl independel red. Thus the range-R is given by R=1(y,— 9) (6.1) pln We Quite obviously, this is a rather crude measure. It pays no attention, wh: goever, t0 people between the richest and the poorest on the income 5 a In particular, it fails to satisfy the Dalton principle, because, for eanle ‘ small transfer from the second poorest to the second most rich individval will keep the measure unchanged. However, the range is often used as a crude, though useful, measure when detailed information on income distri- bution is missing. : (2) The Kuznets ratios. Simon Kuznets introduced these ratios in his pi- oneering study of income distributions in developed and developing coun- tries. These ratios refer to the share of income owned by the poorest 20 or 40% of the population, or by the richest 10%, or more commonly to the ratio of the shares of income of the richest x% to the poorest y%, where x and y stand for numbers such as 10, 20, or 40. The ratios are essentially “pieces” of the Lorenz curve and, like the range, serve as a useful shorthand in situations where detailed income distribution data are missing. (3) The mean absolute deviation. This is our first measure that takes ad- vantage of the entire income distribution. The idea is simple: inequality is proportional to distance from the mean income. Therefore, simply take all income distances from the average income, add them up, and divide by to- tal income to express the average deviation as a fraction of total income. This means that the mean absolute deviation M is defined as 1 62 1 _ ) an 118 HI ue (neglecting negative of inequality that takes ne serious drawback: it where the notation | - | stands for the absolute val a Although M looks promising 2 a ae into account the overall income distribution, " seople is often insensitive to the Dalton principle. Suppost that there ae ont with incomes y, and y,, such that Yj is below the meat then a regressive lation and y, is above the mean income of the pop’ mulation measured by M. income transfer from y; tO Ye certainly raises inequality as i y and yx goes This is obvious from the formula, becatve a en ott However the Up and no other distance is altered, so M unan ig) t just those fers, no! alton principle is meant to apply to all regressive "ans Chapter 6: 7 re pter 6: Econ, Mic J, requal; 188 an to incomes above the it mean. y incomes y, and Yr that are both above the mean, coe example e two, say y,, to the other (higher) Make a ‘eke one. Cla. ahs he lower of thi s small enough so that both incomes stay abo: A e no difference in the sum of the fis ts mean 2, if solute dif ter bsolute deviati i ation will register no as ference ange j In from incomes below the me any two i fer from # the transfer i the transfer, there will b from mean income. The mean ai such a case, and so fails the Dalton principle. We must conc ‘onclude that us, t does the entire income distribution, the mean absolut fe deviation has 8 no asl 'y features as a quick estimate and is therefore a b, a bad m eas 'SUre of compensator’ inequality. (4) The coefficient of variation: One way to avoid the in Sensitivity 'Y Of the mean absolute deviation is by giving more wei oe family statistical measure that a wt deviations from ee ow Appendix 2}, which squares all deviations Hae standard itself, this is effectively th ee ree aero the mumbo viations from the nea Th conc attaching greater weight t ings seviatiir divided by me The coefficient of variation (C) is just eee ean, so that only relative nese ae T. Thus (6.3) The measure C, it turns o1 principle an ut, has satisfactor ; Dalton banetee oa a Lorena: consistent In peneadag 7 satishes all four He implies a trendies eras a transfer from ‘ Re oe satisfies the i.e., (y; —)], which i a smaller number [i » where yj S yi- decreases fe i ne the square of the anges Y% - i to a larger one registers an increase wh 'e smaller number. The net effect i er ey, Be than eS check by trying out va en such a regressive transf ct is ithat C invariably rious examples that this is Eee mad Ca ie case. (The Gini coeffi in empyri icient) We that Aa fang the oD, coctticient Th to a measure that(is widely used i ct eon on ee EE a equality isthe saan simply. ible ne an litenenoe betwee that can conceivaty ae ily totals the (absolute) differences. It i #8 in- ing by population 2 ee made. The aeons of “two-persan inequalities” Pairs) as well as ee (because all paire are ath is normalized BY ik : ‘ome. In symbols, the Gini pea G is ven y (6.4) Ce ante & ninely — Yel- = curing Economic Inequality _3, Measurilg e 63 Iso “ey _ OW oo i ” ne Mn Me Yor y, Figure 6.9. The Lorenz consistency of the Gini coefficient. : The double summation sign signifies that we firs each j constant, and then sum over all the js. Thi of income differences (weighted by the number that because each ly — yy! is counted twice (a; expression is finally divided by 2 as well as by normalizes. . The Gini coefficient has pleasing properties. It satisfies all four princi- ples and is therefore Lorenz-consistent, just like the coefficient of variation. Figure 6.9 shows us why the Gini coefficient is consistent with the Lorenz criterion. In this figure, we arrange everybody's incomes from lowest to high- est. Now take two incomes, say y, and y,, with Y; SY, and transfer some small amount" 6 from y; to y;,. Figure 6.9 shows us how these two incomes change. Now let us see how the Gini coefficient has altered as a result of this regressive transfer. All we have to do is see the change in those pairs in which j or k figure. Consider incomes to the left of y;. Because y; has come down, the difference between these incomes and y, ha3 narrowed by 5. This narrowing is exactly counterbalanced by the fact that y, has gone up by the same amount, so the distance between y, and incomes to the left of y; has gone up by an equal amount. The same argument holds for incomes to the night of y,: the distance between them and y, narrows, but the distance to y, g0es up by the same amount, so all these effects cancel. This leaves us with incomes between y, and y,. However, the pairwise distance between these comes and both y, and y, has gone up. So has the distance between y; ane 4. Thus the overall effect is an increase in the Gini coefficient. This shows why the Gini coefficient is Lorenz-consistent. ; There is another interesting property of the Gini coefficient that pean very closely indeed to the Lorenz curve. Recall that the more owes oe the Lorenz curve, the higher is our intuitive perception Seaet aaa eeer the Tonerat the Gini coefficient is precisely the Tie ot Naren of the triangle wis curve and the 45. line of perfect equality, to ©w the 45° line se, the first two are very cri’ have thus surveyed five indexes Pier ie detailed data ude but nevertheless useful indicators of inequ ily, both the coefficient *€ unavailable. The third should not be used. Finally, "We Argum, th t sum over all ks, holding. S is like summing all Pairs of such pairs, njn,). Notice gain as "Ye yj), the whole the population and income ing order of income: The nividuals in ascene pieces and applying g of i take 6 smalll so as to preserve the ranking ransfer into smal u Bment for larger values of 5 follows by breaking uP the “gic in the text. Chapter 6: Economic ic Inet 190 Gini coefficient (G) appear to be Perfect, an (C) and the + four principles (or what is equ 0 by ou a puzzle. If both C and G gives 1 es? Why not just one? i Y sa ivalent, Satis, ati Org f varia a icf, NZ ° ATE Satistacy, going factory indexes, SONS as t this 2 s aur consistent) ee use Ea peta Lorenz crossings. We have just a Ths DMRS us back fa sistent This means ih when Loreng ry, both C and G oF both C and G give us exactly the same ranking jot" can be ae ‘with . " . Ca the Lorenz criterion. The problem arises wine Wt they both a In that case, it is possible for the Gini coe: Lorenz curves cFOss: ” tion to give contradictory rankings. Thi the coefficient of ae t that our intuitive sense of inequality puta reflection of the ae we should probably not rely entirely on incomplete. In such ine vali, but rely on a whole set of measures, fy one particular eer ety the two Lorenz curves as well, may bea ee ota example, consider two societies, each Consisting of only ae Let the distribution of income in the two Societies be (3,12, 12) and (4,9, 14), respectively. You can easily check that for the first of our hy- pothetical societies, the coefficient of variation is 0.27, whereas it is 0.26 for the second. Using C as an index, therefore, we reach the conclusion that the first society is more unequal than the second. However, if we calculate the Gini coefficient, the values come out to be 0.22 and 0.25, Tespectively. On the basis of the latter measure, therefore, inequality seems to be higher in the second society compared to the first.! To be sure, this isn’t just a hypothetical Possibility. Such contradictory movements of inequality indexes occur in real life as well. Consider, for instance, the study by Weisskoff [1970] on inequality variations in Puerto woe and Mexico during the 1950s. Table 6.1, put together from Thea 'udy by Fields [1980], illustrates the ambiguities that arise. ¢ table is remarkable in its varied movements of inequality measures. intries, there is some ambiguity. In Puerto Rico, for in- 40% and the richest 5% of the Population lost income nN tw ficient and 'S is nothin is &ssentially is the use of log variance as an inequality garithm of incomes. Although it is easy ‘agrees with the Dalton principle in some on by Henri Theil and known a the Th Ur principles a epee X looks bizarre at first, it turns out to be the ol er eel 2 convenient decomposability principle thet peri ts makes t Y into between ‘Omposability principle that pernut 4 he Theil in, fae BOUP andl within-group components (Foster (1983). Thi Situations where we want to decompose inequality inte arming: Ther ‘oithin and actoss ethnic, religious, caste, occupations: IS connec : ection be fon’ 8 number achieves nee? 8 value of, say, 0.25 achieved by the Gini cveticet a : , SAL this "ome change OLing differ ge That's lke comparing apples and oranges. All ths “es in the movements ofthese tnleres as the stb een a y; A Measuring Econontic Inequality 6. Table 6.1. Changes in inequality in Puerto Rico, Argentina, and Mexic, Income ; Income share 0 shar Coeff. of richest o a Country/date Gini variation (%) (%) Puerto Rico 1953 0.415 1.152 24 155 1963 0.449 1.0354 22.0) 1371 Argentina 1953 0.412 1.612 272 18.1 1959 0.4634 1.8874 31.84 16.44 1961 0434)t 1.605) 24Nt I74yt Mexico 1950 0.526 2.500 40.0 143 1957 0.5514 1.652) 37.04 M3t 1963 0543)t 1.380) 2B8LL 10.144 TT Source: Fields [1980]. Note: First arrow indicates a change in inequality from the previous observation; the second arrow indicates the chai ange in inequality from two observations ago. share, a clear sign that the Lorenz curve has crossed. This doesn’t necessarily mean that the Gini coefficient and the coefficient of variation disagree, but they do in fact. In the case of Argentina, there is no evidence from the income shares of the richest and poorest that the Lorenz curves have crossed, but they must have, at least over the period 1953-61. Do you see why? Finally, look at Mexico for the period 1957-63. In this case, both the Gini coefficient and the coefficient of variation agree, but it’s also clear from the movement of income shares that the Lorenz curves have crossed (check this). It should be abundantly clear, therefore, that unless we have a clear case where a Lorenz romparison can be made‘we should consult a variety of inequality measures before making a judgment.) Clearly we have a dilemma here: the result of our comparison is sensitive '° the choice of the index, but we have no clear intuitive reason to prefer one Over the other. There are two ways out of this dilemma. The first, as we said “lore, is to examine our notion of inequality more closely and to come up With stricter criteria after such introspection. The result, as was pointed out, jullinevitably be subjective and controversial. The second escape isto realize at human thought and ideas abound with incomplete orderings)everyone Bt€es that Shakespeare is a greater writer than the Saturday columnist in the ‘Ocal Newspaper; however, you and | might disagree whether he is greater than Tagore or Tolstoy, and even | may not be very sure myself Relative a “quality, like relative literary strength, may be perfectly discernible some 0 1p ime and difficult to judge in other cases. We can learn to live with that * Society manages to significantly increase economic fairness and humane “~

    You might also like