Policy Endogeneity and Other Research Challenges

World Bank Summer Research Workshop on Poverty and Development, June 30, 1999

I have three issues ? for WDR and/or for a future research agenda on poverty and inequality.

Why poverty persists: What if policy is endogenous?

Consider the proposition that the design and implementation of policies to broaden economic opportunities is likely to be more difficult in societies with high proportions of poor and/or high income inequality. In other words, for example, it may be that high-inequality societies are more likely to maintain and tolerate what might be called laissez-faire or neutral economic and social policies, which do not alter over time the existing distribution of assets or income (some might thus compare the U.S. and Western Europe, or Brazil and Malaysia), all other things the same. Or, even where the design of policy is pro-poor in broadening opportunities, its implementation may be institutionally more difficult if the poor are more costly to reach (in rural areas at least or because they bring fewer complementary inputs to public investments in say the health of their children), if their effective demand (say for education) is low due to liquidity constraints and poor capital markets or to poor information etc. That makes broadening the poor?s opportunities look more costly and less efficient, reducing its political attractiveness.

If broadening opportunities is necessary to reduce poverty and inequality, then this proposition says that current inequality can generate a policy and institutional bias toward future inequality, a kind of policy-endogenous "inequality trap". Then there needs to be an external shock, or external pressure (international financial institution conditionality?, pressures of more globalized markets changing the rules of the game), or a change in the internal political rules (more democracy, more political empowerment of those lacking economic opportunities) to shift out of the bad equilibrium.

Moreover, this suggests that one of the costs of tolerating high levels of inequality and resulting social and economic distance across income groups is that the role of the state in creating opportunities for the poor to be productive is itself undermined (with secondary negative effects on growth itself if limited opportunities for the poor and inequality are bad for growth). Maybe even societies? tolerance for inequality expands over time, changing social norms and reducing social demand for pro-poor policy.

Seeing policy as endogenous is the business of political scientists, and is not economists? strong suit. So though there are political economy models of distribution and redistribution (median voter, interest groups, etc.) there is limited research (that I know of) on the nature and dimension of the relationships of this posited bad equilibrium, and on the "independent" variables (shocks, political changes, recent market reforms with other objectives) that change the relationships and how they change it. In very simple terms, and focussing on social policy, a simultaneous equations model might look like:

  1. Social policy effort = f(predicted inequality, current political rules, external pressures, market reforms, other possible factors such as ethnic differentiation)
  2. Current income inequality = f(inequality of assets, current shocks, predicted current policy, other)

(To identify this sort of model: Current shocks don?t affect social policy effort, e.g. latter is defined over last 5 years; current political rules don?t affect income inequality.)

There are all kinds of problems with this simple model: definition of variables, availability of data, and identification/endogeneity (as above suggests). Some thoughts are:

  • The World Bank research community could usefully develop a country-level index of "social policy effort". (Only the World Bank has the global data base and the necessary analytic expertise.) A more general measure of pro-poor policy effort would be ideal but much more complicated. An index of social policy effort would be useful as a descriptive measure (would change over time within countries) and as an independent variable affecting growth, poverty reduction, and inequality (analogous to "openness" and other indices reflecting degrees of distortion like black market premium as policy variables affecting growth) even if never sufficiently well measured to be a reliable dependent variable. Note could be disaggregated (education, health, etc.) for testing alternative program effects on different outcomes. For measures of social policy effort, current indices like the Human Development Index do not serve this purpose since they measure outcomes, not policy effort or inputs. (Nor are available measures like public expenditures on health and education the answer.)
  • The World Bank could commission academic work on the political rules and mechanisms that seem to be associated with so-called "empowerment": parliamentary vs. presidential; existence of various checks and balances; independent judiciary; fiscal decentralization; multi-party systems; first past the post vs. proportional voting and party lists; regulatory and fiscal regimes that affect civil society groups, labor unions, and ethnic-based politics, etc. Otherwise the concept of empowerment seems to boil down to please to encourage meetings and talk to NGOs, and might even seem to bypass rather than strengthen representative government.
  • The inequality variable needs to be unbundled to reflect and test, better than does median voter as representative agent, the relative political clout of different income groups under different circumstances (middle class).

Measuring "opportunity": proxies and indicators

The use of the word opportunity in relation to reducing poverty is itself important. Broadening opportunity implies reconciling improved distribution with growth, equity with efficiency. To put meat on the bones of a nice idea requires some effort at measuring opportunity. Thoughts on proxies and indicators are:

  • Inequality of assets. Is there a close correlation of societies where assets are not equally distributed with societies where opportunities are not broadly available? Does greater equality in the distribution of education imply greater opportunity especially for the currently poor, independent of changes in the returns to education at different levels? A usual characteristic of the poor is that they lack sufficient assets (land, other physical capital, education, other dimensions of human capital); but lack of assets is clearly not an all-encompassing cause of poverty, especially in the short run, when shocks can suddenly deplete the value of some assets.
  • Society-wide measures of discrimination. Could such measures be developed as partial indicators of the degree of "opportunity" across countries? (There is much less research in developing countries on caste, race, ethnic discrimination than on sex discrimination?)
  • Social mobility: lifetime and intergenerational. Inequality measures change slowly (especially the distribution of assets, e.g. education); market reforms and other policy changes may on balance increase opportunities, generating greater mobility upward and downward but little change in snapshot measures of inequality.
  • Returns to unskilled labor (the principal asset of the poor). Reduction of poverty in East Asia miracle economies can be linked to combination of good distribution of assets, and non-declining returns to labor, so that as opportunities for new and higher returns to certain investments (in farms, in children?s education) arose, the poor could exploit them by working more hours often at higher pay.

Mainstreaming poverty and inequality reduction into economywide policies

Policies that protect ("security") and enhance opportunities for everyone are likely to reduce poverty in a sustainable manner ? success over long periods and sustainable politically. Work needed on:

  • Welfare effects for different income groups of alternative macro regimes, with macro objective (stability) as given. E.g. what mix of acceptable (given macro objective) monetary, exchange rate, fiscal policy following a financial shock is least harmful to the poor, in initial period (12 months or even shorter, since temporary shocks can have long-term costs for the poor via school drop-outs, lifetime employment effects for new entrants etc.) (Note reliance on tight monetary policy to restore confidence is neither necessary nor politically feasible in OECD economies.)
  • Effects on different groups of economywide market reforms, especially trade liberalization, privatization, financial market deregulation. Depends heavily on initial conditions, structure of economy, and details of policy design and implementation. Issues are clear but there is little empirical work, surprisingly few good anecdotes.
  • Tax vs. expenditure policy to target the poor?s needs. In shock-prone economies (external, natural disaster, policy volatility itself), does reliance on expenditures rather than taxes to achieve distributive goals work, given that in the short run expenditures are more easily reduced than taxes raised (Mexico raised the VAT rate in 1995, but isn?t that the exception?)