Risk and the Evolution of Inequality in China in an Era of Globalization
31 Pages Posted: 3 Mar 2011
Date Written: March 1, 2007
Abstract
Changes in poverty rates within a country, whether due to globalization or some other source, can be usefully thought of as reflecting either changes in aggregate resources (growth) for the country as a whole, or changes in the withincountry distribution of these resources (inequality). Over the last twenty years China has experienced huge rates of economic growth, reducing poverty. Although at the same time China has experienced substantial increases in ruralurban and interregional inequality, the increase in the size of the Chinese economic pie has much more than offset any increase in inequality for the vast majority of China’s households. Ravallion and Chen (2004) report that although 17.6 percent of Chinese households were poor in 1985, the poverty rate had fallen by more than half by 2001.
Faced with evidence of high rates of aggregate growth and relatively modest increases in inequality, and with evidence that poor households have shared in the aggregate windfall, one might be tempted to conclude that China’s recent experience has had clear net benefits for almost all households. Yet this conclusion (while possibly correct) isn’t justified by the kinds of evidence given above. The kinds of changes described above are likely to involve a large increase in the risk faced by Chinese households. Would a typical household in China in the early eighties, given a choice between their “iron rice bowl” and the risky promises of economic reform, have willingly chosen the latter? We can’t know without some way of measuring the welfare costs of the increased risks actually borne by these households.
The welfare loss due to risk faced by households at a point in time is intimately related to changes in inequality in expenditures. In particular, riskaverse households with timeseparable preferences will tend to prefer to smooth shocks to income over time, so that even entirely transitory shocks to income will tend to have a permanent effect on future consumption expenditures. Thus, the same shocks to income that make next period’s consumption uncertain will also determine the household’s position in next period’s distribution of expenditures.
In this paper we exploit this link by using data on the evolution of expenditure inequality to estimate both household risk preferences and the welfare loss due to risk actually borne by urban Chinese households over the period 19852001, an era during which China’s economy has undergone dramatic reforms and experienced remarkable growth. Others have noted that increases in inequality imply that the rising tide of the aggregate Chinese economy has not lifted all boats equally (Kahn and Riskin 2001). Here we note that because households may change their position in the wealth (and expenditure) distribution, merely looking at changes in inequality will understate the displacement and (ex ante) welfare loss experienced by riskaverse households facing dramatic economic change.
Although the chief contribution of this paper is the application of a method to infer householdlevel idiosyncratic risk from aggregate data on the crosssectional distribution of consumption, it also has something to say about changes in inequality in the absence of this risk. In particular, we see that although there has been a notable increase in inequality among urban households, this increase is dwarfed by the increase in inequality between rural and urban households. We are also interested in documenting any relationship between globalization (as measured by changes in trade volume across sectors) and changes in urban inequality. In section 14.3 we show that after controlling for any effects that globalization may have on aggregate urban consumption the trade shocks we measure can’t account for any of the observed changes in inequality observed within the urban population.
Models having complete markets à la ArrowDebreu yield fully Pareto efficient outcomes; in such a model any changes in inequality must be preferred by all market participants, and so they yield little in terms of interesting policy implications. Complete market models that feature Gorman aggregable preferences (Wilson 1968) yield the very strong prediction that the distribution of consumption across households is invariant (see subsection 14.2.1 for an illustration and appendix A for a general treatment). More interesting are models in which some friction prevents allocations from being fully Pareto optimal, and that have enough dynamic structure to yield interesting predictions regarding the evolution of the distribution of consumption.
To estimate the importance of idiosyncratic risk we assume that all households have similar preferences, and that these preferences exhibit constant relative risk aversion (Arrow 1964). We further assume that all households have access to credit markets on equal terms, and that households exploit these credit markets to smooth their consumption over time, à la the permanent income hypothesis. Beyond this, we make no notably restrictive assumptions. We allow quite arbitrary forms of technology and shocks, and avoid the problem of measuring asset returns. Although this framework is quite general in several dimensions, we will show that conditional on the distribution of production shocks the model yields rather sharp predictions regarding the evolution of the distribution of resources across households. In particular, the model gives us the law of motion governing the inverse Lorenz curves that describe inequality in the economy; the idiosyncratic risk borne by households can be shown to depend entirely on the distribution of “relative surprises” experienced by the household.
The law of motion for inverse Lorenz curves allows us to make predictions about the sequence of Lorenz curves we would expect to observe, conditional on household risk preferences, on rates of aggregate economic growth, and on the distribution of unforecastable shocks facing households in different years, at different wealth levels, and in different occupations. By comparing realized and predicted Lorenz curves, we can estimate these preferences and distributions. This same procedure yields a Markov transition function mapping shares of consumption today into a probability distribution over possible shares tomorrow, and we use this object to calculate the risk borne by differently situated urban Chinese households in different years and to relate this risk to measures of globalization during this period.
The key to the empirical strategy of this paper involves exploiting the restrictions placed on data by Euler equations to make statements about the evolution of inequality. Related literature includes Deaton and Paxson (1994), who derive a martingale property from the consumption Euler equation and use several long panels of householdlevel expenditure data to argue that withincohort inequality in industrialized countries is increasing over time, and Storesletten, Telmer, and Yaron (2004), who use household panel data on expenditures from the United States and a more completely specified general equilibrium model to estimate a law of motion for the distribution of consumption. The central idea of those papers is to exploit intertemporal restrictions to estimate the law of motion for individual households’ consumption growth, and then in effect to integrate over households to infer what the law of motion is for the distribution of Risk and the Evolution of Inequality in China. For the reader who regards this assumption as unreasonable, we note that if some households are constrained so as to not have equal access to credit markets, then our estimates of risk for these households are likely to be underestimates. consumption across households. The present paper reverses these last two steps – we derive equations that impose intertemporal restrictions on individual households’ consumption growth, but then integrate over these equations to obtain restrictions on the law of motion for the distribution of consumption across households before taking these restrictions to the data. The cost of the procedure followed in this paper is that one can’t exploit all the information that would be available from the trajectories of consumption for many different individual households. The (closely related) benefit is that we can get by without panel data, using instead only a relatively limited set of data obtainable from repeated crosssectional surveys of household expenditures, of the sort that many countries conduct in order, for example, to compute consumer price indexes.
Keywords: Globalization, Inequality, Risk, China
Suggested Citation: Suggested Citation
Here is the Coronavirus
related research on SSRN
Recommended Papers

By David Dollar and Aart Kraay

Growth Still is Good for the Poor
By David Dollar, Tatjana Kleineberg, ...

What Can New Survey Data Tell Us About Recent Changes in Distribution and Poverty?
By Martin Ravallion and Shaohua Chen

By David Dollar and Aart Kraay

By David Dollar and Aart Kraay

How Did the World's Poorest Fare in the 1990s?
By Shaohua Chen and Martin Ravallion

Inequality and Growth: What Can the Data Say?
By Abhijit V. Banerjee and Esther Duflo

Inequality and Growth: What Can the Data Say?
By Abhijit V. Banerjee and Esther Duflo

True World Income Distribution, 1988 and 1993: First Calculation Based on Household Surveys Alone