Inequality and Growth: What the Distribution Does

For several decades the standard treatment of distribution in macroeconomics was to set it aside. The argument was defensible on its own terms: growth was the first-order phenomenon, distribution was a second-order matter for public finance, and the two could be studied separately. The empirical work of the last twenty years has made that separation difficult to maintain.

Why the mean is not enough

Consumption responses to income differ sharply across the wealth distribution. Households with little liquid wealth spend nearly all of a transitory income increase; households with substantial buffers spend very little of it. This is not a subtle effect. It means the aggregate response to a stimulus depends on who receives it, and a model with one household by construction cannot produce that result.

The same logic applies to interest rates, to credit conditions and to price shocks. In each case the aggregate is a weighted sum over a distribution whose shape is doing the work. Once you accept that, you are propagating a distribution through time, which is a computational problem rather than an analytical one - see agent-based modelling.

Mechanisms that generate skew

Wealth distributions are consistently more unequal than income distributions, and the tail is consistently heavier than a normal distribution would produce. Three mechanisms account for most of it, and all three are multiplicative rather than additive:

• Returns compound. Wealth grows in proportion to itself, and small persistent differences in return rates produce large differences in level over a lifetime.
• Returns are correlated with wealth. Larger portfolios access asset classes, fee structures and diversification that smaller ones cannot, so the growth rate itself rises with the level.
• Inheritance resets nothing. Transfers across generations preserve position, so the process does not restart each generation.

Multiplicative processes with these features generate power-law tails almost automatically. That the observed tail is close to a power law is therefore weak evidence for any particular story; several mechanisms produce the same signature, which is a recurring identification problem discussed in validation and calibration.

Does inequality affect growth?

The honest summary is that the sign depends on the channel, and the empirical literature has moved several times. Channels that plausibly reduce growth: credit constraints preventing productive investment in human capital, and political capture producing rules that protect incumbents. Channels that plausibly increase it: returns to effort and risk-taking. Cross-country regressions have not settled this, partly because the underlying relationship is unlikely to be linear and partly because inequality measured as a single Gini number discards the information that matters. Whether the gap is between the middle and the bottom or between the top and everyone else changes the mechanism entirely.

Networks again

Opportunity is transmitted through contacts, and contact networks are strongly stratified by class. Job information, credit access and business relationships move along ties that mostly connect people who already resemble each other. This gives inequality a structural persistence that income-based models miss, and it is why the topic connects to network structure as much as to macroeconomics. It also makes distributional questions dependent on measurement infrastructure that is itself uneven - a point developed in big data and public policy.