Social scientists have long debated the relationship between growth and poverty. One side in this discussion is represented by growth optimists, who believe in 'trickle down', that is, the notion that growth in average incomes automatically sinks down to benefit the poor. The opposing view puts the distribution of income and wealth at the centre stage, and argues that reductions in inequality are required to combat poverty. This includes adherents of the notion of 'immiserizing growth', that is, the idea that growth in average incomes may well occur at the same time as large groups of people are being increasingly impoverished. During the 1990s, the proliferation of quality data on income distribution from a number of countries has allowed rigorous empirical testing of standing debates such as this one.
Datt and Ravallion (1992) developed a method to decompose changes in poverty into a 'growth effect', stemming from change in average income, and a 'distribution effect', caused by shifts in the Lorenz curve holding average income constant. Using data from India and Brazil, they found the growth effect to explain the largest part of observed changes in poverty. Similar results have been found in a number of other developing countries by other researchers. White and Anderson (2001), looking, not at poverty, but at the income of the bottom 20 per cent, also found growth to be, on average, much more important than distributional change. Significant work has also been done based on cross-country comparisons of data 'spells', meaning instances where two or more comparable household surveys are available from the same country at different points of time. Such spells provide the data needed for detailed household-level analysis of growth, poverty, and inequality. Analyses based on spells have found that increases (decreases) in mean income tend to be strongly and significantly associated with falling (increasing) poverty rates (e.g. Ravallion 1995 , 2001).
Fields (2001 : 97-8) summarizes the literature this way: 'twenty years of research has shown convincingly that in a cross section of countries, those with higher per capita income or consumption have less poverty. The cross-sectional version of the absolute impoverishment hypothesis has been thoroughly discredited'. Moreover, there is substantial evidence to indicate that, usually, distributional change is too little and too slow to be relied upon for poverty reduction. Growth is, in practice, the main tool for fighting poverty (Bruno, Ravallion, and Squire 1998 ; Squire 1993). However, the