THE FINAL BLOW TO THE IDEA OF THE UTILITY OF POVERTY?

A strain of thought dating back to the mercantilists has essentially argued that, whatever moral position one takes about poverty, a more unequal initial distribution of income allows a higher long-run mean income for any given initial mean.

The precise form of this argument evolved over time, although incentives always played a role. Mercantilists worried about adverse effects of higher wages on work effort and export competitiveness. Later arguments switched to the idea that aggregate savings constrained growth. By this view, in a fully employed (closed) economy, capital accu­mulation was constrained by aggregate domestic savings, and saving is something rich people naturally do more of than poor people. Thus—the argument went—efforts to redistribute income in favor of the poor risked retarding growth and (hence) had ambig­uous implications for poverty reduction.

The neoclassical theory of economic growth, as represented by the Solow (1956) model, was interpreted by some observers as implying that there was an automatic self-correcting process whereby a high initial level of poverty would eventually be reduced by economic growth. By this argument, countries starting out with a low mean income (and hence high absolute poverty rate) would tend to have a higher marginal product of capital (given that they had so much less capital per worker and that there are diminishing returns), which would entail a higher rate of economic growth when compared to growing high income countries with a similar rate of investment. And so the initially poorer country would eventually catch up.

Because the Solow model is an aggregate model, with no heterogeneity, it was ques­tionable to use it to argue that poverty would be self-correcting. There was no inequality in this model.^[489] And, even in his aggregate model, Solow was well aware of the potential for a “poverty trap” (though he did not use that term). Indeed, the original (1956) paper outlined one possible trap, arising from assumed nonlinearities in how population growth rates depend on mean income, with population growth falling at low incomes but rising with higher incomes, then tapering off at higher incomes. A country in a stable equilib­rium but at low income would then need a large gain in capital per worker to escape the trap and move to a sustainably positive growth path.

The twentieth century saw another set of ideas, which challenged the “utility of poverty.” (Recall that there was an early hint of this challenge in Marshall (1890)). It appears to have been long understood that rich people saved a greater share of income than poor people, who were often assumed to save nothing (as in the models of Kalecki, 1942 and Kaldor, 1955). It would then have been only a small step to the conclusion that a higher poverty rate at a given mean income would yield lower aggregate savings and (hence) a lower growth rate in any economy for which aggregate savings constrained growth. But that conclusion was never drawn to my knowledge. It was, however, under­stood at least starting in the 1930s that the same property of the savings function implied a growth-equity tradeoff, whereby higher inequality would generate higher savings and (hence) higher growth. Keynes (1936, Chapter 24) questioned the existence of such a tradeoff.

In the 1990s, a new set of ideas emerged that seriously questioned the instrumental case for poverty and inequality even in a fully employed economy. By this view, poor and/or unequal societies stifled investment, invention, and reform.⁹⁷ These ideas opened up a new window to the potential role of antipoverty policies in economic development.

One argument about why poverty would self-perpetuate in the absence of effective policies related to the idea that poverty would foster a high rate of population growth which would (in turn) entail lower growth. The last step in this argument is an impli­cation of the Solow model discussed above. In that model, a higher rate of growth of the labor force dilutes the capital stock. A higher rate of population growth thus acts in a similar way to a higher rate of depreciation in lowering the steady-state level of capital per worker and (hence) mean income.⁹⁸ But what about the first step? The modern ver­sion of this argument emphasizes the role played by inequality. An undeniably important dimension of inequality in the world is that people living in poorer families tend to be less healthy and to die sooner. This and other factors—including a dependence on children for old-age support and inequalities in maternal education—play a key role in generating another socioeconomic gradient: fertility rates tend to be higher in poor families. On balance, the natural rate of population growth also tends to be higher for the poor. Thus, we can expect lower rates of progress against poverty in countries with higher population growth rates, and there is some supportive evidence for this view.⁹⁹

An influential strain of thought in the late twentieth-century literature also pointed to the implications of borrowing constraints associated with asymmetric information and the inability to write binding enforceable contracts.

⁷ See Loury (1981), Banerjee and Newman (1993), Perotti (1996), Hoff (1996), Aghion et al. (1999a,b), Bardhan et al. (2000), Ghatak and Jiang (2002), Banerjee and Duflo (2003), Azariadis (2006), and World Bank (2006, Chapter 5). Voitchovsky (2009) provides a survey of the arguments and evidence for how the initial level of inequality influences the subsequent growth rate.

⁹⁸ Evidence of an adverse effect of population growth on GDP per capita growth can be found in Kelley and Schmidt (1995, 2001) and Williamson (2001).

⁹⁹ Evidence can be found in Eastwood and Lipton (1999, 2001), who regressed changes over time in pov­erty measures for a cross section of countries on the fertility rate (with various controls) and found an adverse demographic effect on poverty. Using time series data for India, Datt and Ravallion (1998) found evidence that higher rates of population growth increased poverty.

¹⁰⁰ Models with such features include Loury (1981), Galor and Zeira (1993), Benabou (1996), Aghion and Bolton (1997), and Banerjee and Duflo (2003).

The model outlined at the beginning of this chapter illustrates this point well in the special case in which the distribution of wealth (given production technologies) is such that the threshold is not binding (w_t > k^min for everyone). Mean future wealth in a grow­ing economy is then a weakly quasi-concave function of the distribution of current wealth. By standard properties of such functions, a mean-preserving increase in wealth inequality will entail lower mean wealth in the future, that is, a lower growth rate (Banerjee and Duflo, 2003).

Borrowing constraints is not the only way that inequality can matter to growth. Other models have also been proposed, implying that high inequality leads democratic govern­ments to implement distortionary redistributive policies, as in the model of Alesina and Rodrik (1994). Another class of models is based on the idea that high inequality restricts efficiency-enhancing cooperation, such that key public goods are underprovided, or desirable economic and political reforms are blocked.^[490] Rajan (2009a,b) provides an interesting analysis of how the two main types of economic reforms that are widely seen as key to poverty reduction, namely, making markets more competitive and expanding access to education, can be blocked in a democracy in which three classes—the rich oli­gopolists who benefit from market distortions, an educated middle class, and the unedu­cated poor supplying unskilled labor—strive to preserve their rents in the status quo. The model helps us understand the observations of Weiner (1991) and others about India’s relative lack of progress in attaining mass literacy.

A new interpretation of the long-term impacts of colonialism has identified adverse effects of initial inequality on policies and institutions; Engerman and Sokoloff (2006) provide an overview. The essence of this argument is that the geographic patterns of colonialism (notably between North and South America) implanted greater initial inequality and population heterogeneity in some colonies than others.

But is it inequality that matters to growth and poverty reduction or something else, such as poverty, the size of the middle class, or the extent of polarization? Inequality is obviously not the same thing as poverty; inequality can be reduced without a lower pov­erty measure by redistributing income among the nonpoor, and poverty can be reduced without lower inequality. (Similarly, efforts to help the middle class may do little to relieve current poverty.) In fact, there is another implication of credit market failures that has received less attention until recently. Although the literature has emphasized that higher inequality in such an economy implies lower growth, so, too, does higher current wealth poverty for a given mean wealth.^[491] Again, the point can be illustrated using the model outlined in Section 22.2. Plainly, a larger density of people near the zero wealth equilibrium will entail lower subsequent growth. What if the threshold is not binding? It is assumed that the poverty line does not exceed fe*Z(λ +1), and we can let H* denote the poverty rate (headcount index) at this maximum poverty line. Now consider the growth effect of a mean-preserving increase in the poverty rate. I assume that H* increases and that no individual with wealth less than fe*Z(λ +1) becomes better off.If this holds true, then we can say that poverty is unambiguously higher. Then the credit constraint implies that unambiguously higher poverty incidence—defined by any poverty line up to the minimum level of initial wealth needed to not be liquidity constrained—yields lower growth at a given level of mean current wealth. As this point does not appear to have been made in the literature, the Appendix demonstrates the point more formally.

101

This theory implies an aggregate efficiency cost of a high incidence of poverty. But note that the theoretical prediction concerns the level of poverty at a given initial value of mean wealth. Without controlling for the initial mean, the sign of the effect of higher poverty on growth is ambiguous (see the Appendix). Two opposing effects can be iden­tified. The first is the conditional convergence property described above, whereby coun­tries with a lower initial mean (and hence higher initial poverty) tend to have higher subsequent growth in a neoclassical growth model. Against this, there is an adverse dis­tributional effect of higher poverty. Which effect dominates is an empirical question, which we will return to later in the chapter.

Credit market imperfections are not the only argument suggesting that poverty is a relevant parameter of the initial distribution. Lopez and Serven (2009) introduce a sub­sistence consumption requirement into the utility function in the model by Aghion et al. (1999a) and show that higher poverty incidence (failure to meet the subsistence require­ment) implies lower growth. Another example can be found in the theories that have postulated impatience for consumption (high time preference rates possibly associated with low life expectancy) and hence low savings and investment rates by poor people (see, for example, Azariadis, 2006). Here, too, although the theoretical literature has focused on initial inequality, it can also be argued that a higher initial incidence of poverty means a higher proportion of impatient consumers and hence lower growth.

The potential inefficiency of poverty is starkly obvious when one considers how work productivity is likely to be affected by past nutritional intakes, as these determine the stock of human capital.^[492] As noted in Section 22.2, only when nutritional intake is high enough will it be possible to do any work, but diminishing returns to work will set in later; see the model in Dasgupta and Ray (1986). Poverty’s effects on the nutrition of young children in poor families are also of special concern. A sizable body of research suggests that poor nutrition (both food energy intakes and micronutrients) in the early years of life retards children’s growth, cognitive and learning abilities, schooling attainment, work productivity, and likely earnings in adulthood.^[493] The health environment also matters. Chronic undernutrition in children can stem from either low nutritional intake or low nutritional absorption due to constant fecal-oral contamination,^{[494] [495] such as due to the lack of clean drinking water. This can mean that direct nutritional supplementation does little or nothing to improve children’s nutritional status (such as measured by stunting) until the health environment improves.1 This type of argument can be broadened to include other aspects of child development that have lasting impacts on learning ability and earn­ings as an adult (Cunha and Heckman, 2007). And the handicap ofpoverty can emerge in the prenatal period. Maternal and prenatal conditions are now also thought to matter to child development and (hence) economic outcomes later in life (Currie, 2011; Dasgupta, 2011). By implication, having a larger share ofthe population who were born in and grow up in poverty (including living in poor health environments) will have a lasting negative impact on an economy’s aggregate output. Poverty will perpetuate.}

102

In another strain of thinking about how poverty can perpetuate, Mani et al. (2013) present evidence from both experimental and observational studies, suggesting that pov­erty reduces cognitive ability. The evidence is consistent with the view that, given that human cognitive capacity is physically limited, the concerns generated by poverty crowd out thinking about other things relevant to personal economic advancement.

There are also theoretical arguments involving market and institutional development, although this is not a topic that has received as much attention in the literature. Although past theories have often believed credit market failures to be exogenous, poverty may well be a deeper causative factor in financial development (as well as an outcome of the lack of financial development). For example, given a fixed cost of lending (both for each loan and for setting up the lending institution), liquidity constraints can emerge as the norm in very poor societies.

Some of the theoretical literature has also pointed to the possibilities for multiple equilibria associated with a nonconvexity in the production possibility set, as in Figure 22.1. As noted already, in poor countries, the nutritional requirements for work can readily generate such nonlinearity in the dynamics, as argued by Dasgupta (1997). Such a model predicts that a large exogenous income gain may be needed to attain a per­manently higher income and that seemingly similar aggregate shocks can have dissimilar outcomes; growth models with such features are also discussed in Day (1992) and Azariadis (1996, 2006) among others. Sachs (2005a,b) has invoked such models to argue that a large expansion of development aid would be needed to ensure a permanently higher average income in currently poor countries.

Some of the empirical literature on economic growth has found that higher initial inequality impedes growth.^[496] And the effect is quantitatively large, as well as statistically significant. Consider the two most recent published studies at the time of this writing. Herzer and Vollmer (2012) find that a 1% point increase in the Gini index results in a decrease in long-term mean income of 0.013%; when normalized by standard deviations, this is about half the growth impact ofthe investment share. Berg et al. (2012) also find that more unequal countries tend to have less sustained spells of growth, and this effect is also quite large; a 1% point higher Gini index is associated with a decline in the length of the growth spell of 11—15%.

Not all the evidence has been supportive.^[497] The main reason why some studies have been less supportive appears to be that they have allowed for additive country-level fixed effects in growth rates. This specification addresses the problem of time-invariant latent heterogeneity in growth rates. However, it may well have little power to detect the true relationships given that the changes over time in growth rates will almost certainly have a low signal-to-noise ratio. Simulation studies have found that the coefficients on growth determinants are heavily biased toward zero in fixed-effects growth regressions (Hauk and Wacziarg, 2009).

There are a number of remaining issues in this literature. The bulk of the literature has used consumption or income inequality measures. Theoretical arguments based on bor­rowing constraints point to the importance of asset inequality, not income inequality per se. There is evidence of adverse effects of asset inequality on growth.^{[498] [499]}

The aspect of initial distribution that has received almost all the attention in the empirical literature is inequality, as typically measured by the Gini index of (relative) inequality. The popularity of the Gini index appears to owe more to its availability in secondary data compilations on income and consumption inequality measures than to any intrinsic relevance to the economic arguments.¹¹ However, as Lopez and Serven (2009) observe, the significance of the Gini index in past studies may reflect an omitted variable bias, given that one expects that inequality will be highly correlated with poverty at a given mean.

There are also issues about the relevant control variables when studying the effect of initial distribution on growth. The specification choices in past work testing for effects of initial distribution have lacked clear justification in terms of the theories predicting such effects. Consider three popular predictors of growth, namely, human development, the investment share, and financial development. Of the first predictor, basic schooling and health attainments (often significant in growth regressions) are arguably one of the chan­nels linking initial distribution to growth. Indeed, that is the link in the original papers of Loury (1981) and Galor and Zeira (1993).^[500] The second predictor, one of the most robust predictors of growth rates, is the share of investment in GDP (Levine and Renelt, 1992); yet, arguably one of the main channels through which distribution affects growth is via aggregate investment, and this investment is one of the channels identified in the theoretical literature. Finally, consider private credit (as a share of GDP), which has been used as a measure of “financial sector development” in explaining growth and poverty reduction (Beck et al., 2000, 2007). The theories discussed above based on borrowing constraints suggest that the aggregate flow of credit in the economy depends on the initial distribution.

Although the theories and evidence reviewed above point to inequality and/or pov­erty as the relevant parameters of the initial distribution, yet another strain of the literature has pointed to various reasons why the size of a country’s middle class can matter to the fortunes of those not (yet) so lucky to be middle class. It has been argued that a larger middle class promotes economic growth by fostering entrepreneurship, shifting the com­position of consumer demand, and making it more politically feasible to attain policy reforms and institutional changes conducive to growth.^[501] This has been an issue in India, where, since the 1970s, it has been argued that “inequality” constrained the growth ofthe manufacturing sector by limiting the size ofthe domestic market for consumer goods; see, for example, the discussion in Bardhan (1984b, Chapter 4). Here, too, it can be argued that it was not inequality per se that was the culprit but the relatively small middle class, or (more or less equivalently) the extent of absolute poverty that generated the domestic demand constraint in a relatively closed economy. The argument has been heard less in the more open economies. However, the Indian middle class has also been seen to promote reform (Sridharan, 2004). Using cross-country regressions, Easterly (2001) finds that a larger income share controlled by the middle three quintiles is a significant predic­tor of rates of economic growth.

So we have three main contenders for the distributional parameter most relevant to growth: inequality, poverty, and the size of the middle class. The fact that very few encompassing tests are found in the literature and that these different measures of dis­tribution are not independent, leaves one in doubt about what aspect of distribution really matters. As already noted, when the initial value of mean income is included in a growth regression alongside initial inequality, but initial poverty is an excluded but still relevant variable, the inequality measure may pick up the effect of poverty rather than inequality per se. Similarly, the main way the middle class expands in a devel­oping country is almost certainly through poverty reduction, so it is unclear whether it is a high incidence of poverty or a small middle class that impedes growth. Similarly, a relative concept of the “middle class,” such as the income share of middle quintiles, will probably be highly correlated with a relative inequality measure, clouding the interpretation.

Possibly, the strongest evidence to date to support the view that it is poverty not inequality per se that impedes growth in developing countries comes from an observa­tion made by Ravallion (2012b), namely, that we see convergence in average living standards among developing countries and greater progress against poverty in faster growing economies, yet we do not see poverty convergence; the poorest countries are not enjoying higher proportionate rates of poverty reduction. Ravallion resolves this paradox by arguing that a high initial incidence of poverty, at a given initial mean, impedes subsequent growth (this theory is compatible with a number of the theories outlined above). This is shown to be consistent with data for almost 100 developing countries, which reveal an adverse effect on consumption growth of high initial poverty incidence at a given initial mean. Ravallion finds that high poverty at a given initial mean matters more than inequality or measures of the middle class or polarization. Also, starting with a high incidence of poverty limits progress against poverty at any given growth rate. For many poor countries, the growth advantage of starting out with a low mean is lost due to their high poverty rates. That does not, however, imply that any antipoverty policy will promote growth. That will depend on many factors, as dis­cussed in the next section.

The arguments summarized above about why poverty can bring lasting efficiency costs do not require the existence of a poverty trap. However, when a poverty trap is present, the cost of poverty can rise greatly. So it is important to ask whether such traps have economic significance. On a priori grounds, it is highly plausible that threshold effects exist. Biology alone makes this plausible; unless one can support the nutritional needs of the body at rest, it will be impossible to do any work. Whether this is of eco­nomic significance in practice (even in poor economies) is another matter. As Deaton (2006) points out (in reviewing Fogel, 2004), human caloric requirements can be covered with seemingly modest spending on food staples.^[502] However, this is not conclusive. Environmental enteropathy can generate quite low nutrition absorption rates given the persistent fecal-oral contamination of the environments in which many people live. In effect, the implicit price of an absorbed calorie capable of fueling work effort is higher, possibly far higher, than the nominal price. Furthermore, we have also learned that work productivity depends on the personal history of nutrition and health, as argued by Dasgupta (2011). Someone whose growth is stunted due to a long history of undernutrition—low intakes and/or low absorption—can be in current nutritional bal­ance (able to afford current food energy requirements) but have such low productivity that a poverty trap emerges. It may not be a strict threshold, as in Figure 22.1, but a smoother, S-shaped function.

Other sources of threshold effects are also plausible on a priori grounds, such as the fact that a minimum level of schooling is essential before schooling can be a viable route out of poverty (recalling the story of Sunil in Boo, 2012). One can also interpret the aforemen­tioned arguments about how poverty reduces cognitive functions as stemming from bio­logical threshold effects—that a minimum level of time not worrying about the financial and other stresses created by poverty is needed to escape poverty (Mani et al., 2013).

In testing for threshold effects, some of the literature has looked for lumpiness in non­human capital requirements. The results have been mixed. Mesnard and Ravallion (2006) find evidence of nonlinear wealth effects on new business start-ups in Tunisia, but they do not find signs of thresholds effects. Nor do McKenzie and Woodruff (2006) find any sign of nonconvexities in production at low levels among Mexican microenterprises. In one of the few studies using wealth data, Barrett et al. (2006) do find evidence of the nonconvexity in asset data for rural Kenya and Madagascar.^[503]

It can also be difficult to detect theoretically plausible threshold effects on dynamics in standard microdata sets (Day, 1992). For one thing, depending on the frequency of the observations over time in the data, the existence of the unstable “middle” equilibrium (point B in Figure 22.1) can generate attrition—the destitute simply drop out of the data (including by becoming homeless) (Lokshin and Ravallion, 2004). Also, there will be high social returns and risk-sharing arrangements to prevent most people falling into the trap. The trap is still there, but it may only be evident in extreme situations when those social relationships break down, as Ravallion (1997b) argues is the case during famines.

A testable implication of the models based on credit market failures is that individual wealth should be an increasing concave function of its own past value. In principle, this can be tested on suitable micropanel data, though most data sets only show consumption or income, not wealth. Lokshin and Ravallion (2004) provide supporting evidence of concavity in panel data on incomes for Hungary and Russia while Jalan and Ravallion (2004) do so using panel data for China. These studies do not find the threshold properties in the empirical income dynamics that would be needed for a poverty trap. Using similar methods, but arguably a better identification strategy, Dercon and Outes (2013) find evi­dence of a low, unstable equilibrium in the income dynamics for a long panel of house­holds in rural India.

Microempirical support for the claim that there are efficiency costs of poor nutrition and health care for children in poor families has come from a number of studies. In a recent example, an impact evaluation by Macours et al. (2008) of a conditional cash trans­fer (CCT) scheme in Nicaragua found that randomly assigned transfers to poor families improved the cognitive outcomes of children through higher intakes of nutrition-rich foods and better health care. This echoes a number of findings about the benefits to dis­advantaged children of efforts to compensate for family poverty.

The upshot of all this data is that present-day thinking is both more optimistic about the prospects of eliminating poverty through an expanding economy and more cognizant of the conditionalities in the gains to poor people from economic growth. Under the right conditions, growth can be a powerful force against poverty. Those conditions per­tain in large part to aspects of both the initial distribution and how it evolves. As we will see in the following section, the focus of much antipoverty policy has shifted over time toward efforts to ensure that the conditions in place will allow poor people to contribute to an expanding overall economy, and so escape poverty permanently.

22.9.