KGP/World income distribution: past, present and future.

Transcription

1 KGP/World income distribution: past, present and future. Lecture notes based on C.I. Jones, Evolution of the World Income Distribution, JEP11,3,1997, pp and R.E. Lucas, Some Macroeconomics for the 21st Century, JEP 14,1,2000,pp and Xavier Sala -i - Martin, The Myth of Exploding Income Inequality in Europe and the World in H. Kierzkowski (ed.), Europe and Globalization, PalgraveMacmillan, JEP stands for Journal of Economic Perspectives Pre-industrial per capita income differentials are generally believed to be quite small. An analysis of real urban wage differentials in Europe around 1500 suggests that the ratio of real wages in high wage cities, such as London, to low wage cities, such as Istanbul, was 2 to 1 and remained in that order until around 1800, when divergence increased. Rural income differentials are not well researched for the pre-industrial period, but labour productivity differentials between advanced regions in Europe and less advanced is of the same order of magnitude, say 2 or 3 to 1. Rising income and productivity differentials are therefore clearly a phenomenon of the last 200 years. L. Pritchett suggests that the ratio of per capita income in the richest to the poorest economy was almost 9 to 1 by 1870 and 45 to 1 in The simple explanation is that not all economies have benefited from the advantages of modern economic growth, but the difficult part of that explanation is of course to explain why some economies continue to be unaffected by growth. However, a series of recent papers to be reviewed in what follows suggest that the trend towards inequality on a world scale has, or will be, broken. There is some confusion as to recent trends in world income distribution because different measures are used which give widely different results. Figure 1.2 below which is reproduced from the paper by Sala-i-Martin cited above is an appropriate starting point. 1

2 Three measures of global income inequality are presented suggesting increasing, decreasing and income inequality reverting to a constant level. All three measures look at variance of log per capita income but the one that displays increasing inequality is based on unweighted variance, which measn that the per capita income of a small nation counts as much as that of, say, China. The implication is that the a unit increase of per capita income for every one of 1200 millions Chinese is given the same weight as a unit increase of per capita income of 0.5 million Luxembourgians. The fact that China s 1.2 billions are catching up will, in other words, not be appropriately represented in the unweighted variance. If variance instead is measured so that each nation s population size is taken into account, then a different picture emerges. We arrive at the falling across country inequality curve. What s driving this result is the fact that some populous countries, such as China and India, has experienced faster per capita income growth in recent decades than the world average. However, across country inequality ignores within country inequality. It is possible that early industrializers with fast income growth also experience increasing income inequality. The solid curve, global inequality in the figure above, indicates that declining across-country inequality is counterveiled by increasing within country inequality. My conjecture is, however, that as the development process continues in the fast growing new industrializing nations, there will be falling within country inequality which implies that global inequality will decline as well. 2

3 The Jones paper looks at the evolution of income distribution in the world since 1960 and makes predictions based on different assumptions regarding convergence speed. First look at his Figure 3 above, that depicts the density of population weighted GDP per worker relative to the US GDP per worker. Density roughly means the frequency of people at a particular relative income, the horizontal axis. The area under the curve to the left of a particular point, say 0.04, on the relative income axis shows the share of people of world total at and below that particular relative income, that is 4 per cent of the US GDP per head. The historical evolution between 1960 and 1988 can be seen as one of modestly decreasing inequality, the density concentration and the median person is moving to the right. Please, note that within country distribution is ignored in this analysis. The evolution of world relative income suggests a narrowing of the gap to US and a twin peak distribution. For example the 75th percentile of population had 22 per cent of the US income in 1960 but 40 in What can you say about the future of income distribution? Will the trend of decreasing inequality, that is convergence, continue? Jones suggests an approach to speculate about future income distribution taking into account that economies have different propensities or probabilities to have (a) slow growth, (b) intermediate growth or (c) fast growth. By classifying a sample of 121 nations into initial relative income level and growth experience in a matrix of transition probabilities 3

4 was constructed. It turned out that growth differed much according to initial income. Nations with relative income of between 10 and 40 per cent of US level had extremely low probability to grow slowly, for example, while rich countries tended to grow at intermediate rates. Perhaps this is the long run growth rate for economies at or close to the technological frontier. For really poor economies growth experience was mixed and the risk of remaining in slow growth was quite high, but the chance of experiencing high growth was significant. These observed historical transition probabilities formed the basis for an estimate of a the stable (hypothetical) long term distribution of relative income using Markov processes in which the transition from an initial income state to all possible income states including the initial one was reiterated over and over again using the historical transition probabilities until you reach a stable state). The result indicates a narrowing of income gaps in the world, as documented in Table 2 reproduced here from Jones. While in 1960 only 20 per cent of nations had an income at 40 per cent or more of US income 50 percent enjoyed that in the predicted case. Also, the very poor, with income equal to or less than 10 per cent of the US income, became less numerous declining from 34 per cent of the sample to 16 per cent. It must be pointed out that the historical transition probabilities used in the simulation were generated by economic policies which need not be constant over time. Lucas paper is a simulation based on a simple model in which technology diffusion is the active ingredient and on the belief that there is unconditional convergence. Economies differ only in when they start growing and in that the later you start the faster you grow. This is a mechanical application of Gerschenkron s famous expression the advantages of backwardness. 4

5 There is an early industrialiser (UK) starting to grow around 1800 at a constant rate of 2 per cent per year. Nations starting to grow later will have an underlying growth rate of 2 per cent as well, plus a growth rate proportional to the income gap to the leading economy. As the income gap declines the latecomer s growth rate converges to 2 per cent as well. As you can see this is the idea we have discussed before as the negative relationship between initial income and subsequent growth. The other element in the argument is that the proportion of economies that experiences the transition from no-growth to growth is increasing over time, first slowly up to around 1945 but then very rapidly suggesting that only 15 per cent of the economies of the world is in a no growth phase in The risk of remaining a zero growth economy becomes smaller over time since the stock of available technical knowledge increases. The demonstration effect of specific country success stories, as indicated by high income per head, is present. (You are recommended to consult the article to get the details). Combining these two arguments you can simulate world growth rates and the standard deviation of income per person across nations as illustrated in Lucas Figure 3. Both curves start at zero year 1800 because at that date all economies were assumed to have the same income and no growth. Income differences then first increased slowly because early on only a small number of nations started to grow. The transition probability to growth was still quite low. But as that transition probability increased both average growth rates and inequality increased faster for a period, the 20 th century. Economies entering a growth phase pushed up the average growth rate above the steady state 2 per cent per year since they were catching up with the leader economy, UK. The later you started the faster your growth. However when almost all nations were growing by the end of the 20 th century, the average growth began to converge to the steady state growth of 2 per cent per year. Growth rates converge because income differentials between leaders and followers have fallen which imply that income variability measured as standard deviation of income also falls. 5

6 Lucas simulation is extremely simplified - but also quite attractive making use of the plausible notion that an economy that is starting out as very poor can imitate a lot of known technologies and thereby increase its growth. It has essentially the same message as the Jones paper: income inequality on a world scale has peaked and from now on convergence will dominate. In other words the world might become a better place to live in, which seems to be a nice conclusion at the time I am writing this just before Christmas. 6