WDR ON POVERTY AND DEVELOPMENT 2000/01 Stiglitz Summer Research Workshop on Poverty, Washington DC POLICY REFORM AND INCOME Giovanni Andrea Cornia July 6 July 8, 1999
Policy reform and income inequality by Giovanni Andrea Cornia, WIDER The last two decades have seen the emergence and consolida of an economic paradigm (the 'Washington Consensus') which emphasizes macroeconomic stability, domestic liberalisa, privatiza, removal of barriers to internaal trade and financial flows, and the search for market solus also in the provision of public goods. This approach is claimed to reduce rentseeking, increase competi, offer opportunit for export and growth to poor developing countr and promote the convergence of their living standards with those of the advanced nas. It is also claimed that, within countr, the distributive impact of these polic is neutral, as evidence supposedly shows that income distribu is stable and that there is no associa between growth and inequality. For instance, Deininger and Squire note that 'Decadal averages of inequality indexes across regions...are relatively stable through time, but they differ substantially across regions...'. Likewise, Li-Squire-Zou (LSZ) conclude "... there is no evidence of a time trend in... 65 of our sample". Yet, during the last two decades only China and a few E.Asian countr grew fast enough to achieve convergence. Between-country inequality rose moderately between the late 1960s and 1980 and rapidly since 1980. Stud of the distribu of world income among individuals arrive at similar results, especially if China is remouved from the sample or is devided into its rural and urban sector. Growing polarisa among countr was accompanied by a surge in inequality in most OECD, all transial and many developing countr. This conclusion is based on a review of country stud on inequality and an econometric analysis of trends for 77 countr accounting for most of the world popula and GDP-. Inequality rosen in 45 of the 77 countr analyzed, stopped declining in 4, was untrended in 12 and declined in 16. Weighing these changes by popula size and GDP- strengthens the conclusions. These results differ substantially from those of LSZ. The annex table analyzes the sources of these differences by juxtaposing the LSZ results (colum 2) with those obtained by changing - one at the time - the funcal forms used for the interpola; enlarging the sample from 49 to 77 countr; and updating the time ser until 1998. The comparison shows that the difference mainly depends on the choice of the funcal form for the interpola of the trend. Differences in country coverage account for a small propor and for the updating of the ser until 1998 explains 20 percent of the difference. To sum up, while inequality declined in several countr during the Golden Age, inequality trend reversals have occurred with increasing frequency since the 1980s. Out the 43 countr with a rising trend, the reversal occurred in one over 1960-65, in 4 over 1966-75, in 6 over 1975-80, in 8 over 1980-85, in 10 over 1985-90 and in 14 after 1990. The rise - of less than 5 Gini points in six countr, by 5-10 points in 20, by 10-20 points in 23, over 20 points in six transial econom - was universal in EE-FSU, almost universal in L.A., very common in the OECD, and frequent in South, S.E. and E. Asia. Such generalized increase in 'vertical inequality' may prove to be incompatible with the poverty reduc objectives adopted by the World Bank. The achievement of a given poverty reduc target under condis of high inequality requires faster growth than under condis of greater equity. This is however difficult to achieve due to macroeconomic and environmental constraints. Poverty reduc is further affected by the fact that growth is influenced by the past and current distribu of income and assets. In addi, rises in 'horizontal inequality' - i.e. inequality among social classes, ethnic, religious groups, (especially if accompanied by low within-group vertical inequality) were found to be the most important cause of the rise in civil conflicts observed since the mid 1980s. Horizontal inequality can concern the distribu of income, public jobs, educaal opportunit, assets and state rents. The wisdespread recent increase in inequality may thus not only lead to stagnant growth but also to social tensions. 2
Tradial causes of income inequality, such as land concentra and the 'curse of natural resources' are unlikely to explain the recent rise in income concentra. This is most likely due to the shift in polic which began in the 1980s and which entailed an erosion of labour institus; trade and finacial liberaliza; privatiza; a decline in the redistributive role of the state and so on. Here I would like to focus on two of these changes: - the fall in the labour share and the rise in earnings inequality, - and the rise in remunera of capital and in the financial rent. (i) The rise in wage inequality has often been ascribed to technology. 'New technolog' - it is said - generate a demand for skills and earnings distribus more skewed than that emanating from 'old technolog'. Second, informa technolog diminish the cost of monitoring unskilled workers and reduce the wage premia needed to ensure their efficient performance. Third, new technolog replace labour and affect the funcal distribu of income. Yet, much of the rise in earnings inequality depends on educa and training polic, i.e. polic facilitating the adjustment of labour supply to its demand. Comparisons between SKorea and Brazil in the 1960-70s and Canada and the US in the 1980-90s show that inequality rose in Brazil and US but not in SKorea and Canada because the latter adopted vigorous polic to promote secondary and post-secondary educa. Public expenditure and other measures in educa - and the development of financial markets -do contain the rise in earnings inequality. - Greater earnings inequality and the fall in the wage share are influenced also by changes in labour institus towards greater wage flexibility, minimum wages, unionisa and collective bargaining. The liberalisa of the labour market was expected to generate fast employment growth and an increase in wage dispersion. The impact on overall inequality was to depend on whether the 'wage inequality effect' or the 'employment-crea effect' prevailed. In the US the fall in unionisa accounted for about 20 percent of the increase in earnings inequality. In EE, LA and the US, the decline of minimum wages relative to the average wages is closely associated with the rise in overall inequality. The erosion of the minwage in the USA, is estimated to account for about 30 percent of the rise in earnings concentra. Earnings concentra did not increase in countr with collective bargaining institus, adequate minimum wages and social protec systems. Evidence from LA shows that minimum wages do not affect growth while improve poverty and inequality. Earnings inequality is explained also by a rapid rise in the highest wages (a rise possibly related to the FIRE sector) and in interindustrial wage dispersion not-justified by differential rises in labour productivity. Different approaches in this area are needed. - In many developing countr the wage share declined and earnings inequality risen during recession and adjustment. Macro stabilisa has been attained mainly through demand management measures. Measures to stimulate supply have also been introduced but require time and resources to produce results, and have little chance to succeed in undiversified econom with weak institus and infrastructure. While yielding rapid results in terms of macroeconomic balance, demand management generates recessions of varying dura. Unlike in the OECD, inequality in developing countr rises during recessions and falls during recover. Economic slumps in industrialized countr have a greater impact on profits than wages because of the stickiness of the latter and social safety nets cushion the loss of wage income. In low and middle-income countr, wages are downward flexible, social safety nets less developed and labour stockpiling less frequent. Thus, wages (esp. unskilled wages) fall faster than GDP/capita and profits, the wage share declines and income inequality grows. Interest rate and fiscal policy should be conscious of these effects. - Recent evidence points also to the correla between financial crises and earnings inequality, particularly in countr with weak labour institus. In L.America and Asia, for instance, 3
financial crises raised inequality in 73 and 62 percent of the time, while Finland, Norway and Spain experienced a sequence of crises without experiencing increased inequality thereafter. Strengthening financial regula and the establishment of social insurance should reduce the frequency and impact of financial crises. - Probably the most important prescrip of the stabilisa packages promoted by the Washington Consensus is controlling infla. affected by high infla experience a worsening of distribu, as the poor are least able to index their incomes and maintain the real value of their assets, and unskilled labour is especially vulnerable to lay-offs in recessions. Yet, the rate of infla targeted by orthodox programmes is single digit, even though the literature shows that - below the threshold of 40 percent a year - infla is not costly. Second, such ambitious stabilisa targets are sought by means of large rises in interest rates and budget cuts (which have considerable distributive effects). Third, the monetary approach to the control of infla can generate side-effects which require that austerity measures be kept in place for many years in order to push infla below 20 per cent a year. (ii) The rise in real interest rate - and the development of the financial sector - have increased the share of output ascribed to the financial rent and reduced that received by wages and profits. In 1982 US interest rates rose sharply, and pushed upward those of most other countr. With the explosion of the debt crisis, the heavily indebted developing countr had to borrow heavily also from the domestic bond markets. This forced interest rates up. This movement was reinforced by the IMF policy of demanding large increases in interest rates in crisis countr and by the wispread liberalisa of the domestic financial sector of the 1980s. In most countr, real interest rates rose from a long term value of 2-3 percent to much higher - often two digit - levels during the 1980-90s. In Canada, US and UK, real rates rose from -1 per cent in 1976 to 5-6 per cent in 1982-4 period. In developing econom often rose by 10-15 points. As a result, interest payments on public and private domestic debt rose rapidly. All this had the effect of pushing a number of governments into a 'vicious circle' in which the rise in interest rates augmented the cost of debt servicing, which further pushed upward deficits and indebtedness. As a result, with the liberalisa of domestic financial markets, interest payments on public debt rose rapidly, reaching in the early-mid 1990s, levels close to 15 percent of GDP in a number of middle income and industrialized countr. Financial deregula has thus led to a substantial increase in the rate of return to financial capital, a rapid accumula of public debt, an increase in the share of GDP accruing to nonwage incomes, the emergence of a new class of rentiers and the redistribu of taxed labor income to bondholders. Since 1980, non-labor incomes increased in most G7 countr (by between 5 and 10 percent of GDP) in parallel with the increase in interest rates and with an increase in average rates of return on assets. ****************** In conclusion, in view of the widespread increase in inequality observed during the last twenty years - and of its negative impact on growth, poverty, inequality and social stability - adjustment and development polic must be more attentive to their distributive impact. Here we have discussed in particular polic which affect - one way or another - the relative remunera of capital and labour. In this and other areas it is necessary to explore alternative adjustment and development trajector involving different impacts on income distribu. Disincentive effects and free riding may occurr at very low level of earnings inequality, as in some former socialist countr, but high inequality of income or assets is likely to generate slower growth and much slower poverty reduc. Too 'high' or too 'low' inequality would thus reduce growth, which would be invariant within an 'efficient inequality range'. 4
RISING of which: Analysis of differences between the LSZ and WIDER estimates LSZ SAMPLE 49 countr Linear trends only LSZ Period of LSZ SAMPLE Best Fit LSZ Period (1947-94) WIDER SAMPLE 77 LSZ Period WIDER SAMPLE 77 WIID Period (1939-98) of GDP- GDP- GDP- 20 47 53 71 66 60 68 77 73 85 Continuously Rising 20 15 4 10 33 40 29 18 9 12 U-shaped pattern.. 24 43 56 15 13 34 40 58 70 Stable-Rising.. 6 2 2 17 7 5 16 6 3 Rising-Stable.. 2 4 3 1 0 0 3 0 0 SLOWDOWN IN INEQ. DECLINE FALLING NO TREND/ STABLE.. 4 23 5 3 21 5 7 22 6 14 35 15 16 29 19 27 15 6 9 66 14 9 8 2 0 0 1 0 0 5