Why Greater Risk Sensitivity Does Not Always Mean More Saving

Greater sensitivity to adverse outcomes might seem certain to encourage precautionary saving. In the research preprint “Risk Sensitivity and Stationary Equilibrium in Incomplete Markets,” Qingyin Ma, John Stachurski and Bin Wu identify conditions under which that intuition holds—and construct examples in which it fails. Their results concern a theoretical economy, not measured changes in household behavior.

The central distinction is whether income today provides information about income tomorrow. With independent income, the authors establish a consistent direction: greater risk sensitivity raises saving at fixed prices and lowers the smallest and largest equilibrium interest rates. With income that predicts future prospects, comparisons between intermediate levels of risk sensitivity can run the other way.

What risk sensitivity changes

The model contains households facing individual labor-income uncertainty. They cannot borrow against future earnings, but they can carry savings forward in a risk-free asset. A competitive firm uses the capital supplied by households to produce output.

Households evaluate uncertain future utility using a risk-sensitive recursion. At zero risk sensitivity, this becomes the familiar expected-utility benchmark. As sensitivity increases, adverse continuation outcomes receive greater decision weight. The same preferences have a robustness interpretation: households evaluate unfavorable departures from a benchmark income model, with a penalty for those departures that decreases as risk sensitivity rises.

This parameter does not change preferences over deterministic consumption streams. Nor is it simply another name for patience. The paper supplies a nonlinear example in which no single adjustment to the discount factor reproduces the risk-sensitive household’s saving choices across different wealth levels.

When the precautionary intuition holds

Under independent income with positive probabilities for every income state, current productivity offers no information about future income. Successor states then differ in wealth: poorer states have lower continuation utility and a greater marginal value of additional resources.

Greater risk sensitivity strengthens the weight on those states. The authors prove that, holding prices fixed, individual saving cannot fall. It rises strictly wherever saving under the lower sensitivity parameter is already positive. The borrowing constraint allows some households to save nothing under both parameters.

The comparison carries through to the long-run wealth distribution and aggregate saving. Aggregate saving and wealth rise strictly unless both economies have zero aggregate saving. At equilibrium, capital demand is positive, excluding that exception. Both the lowest and highest equilibrium interest rates therefore fall strictly. This statement does not require equilibrium to be unique.

Why predictive income complicates the story

The paper also studies stochastically monotone Markov income: higher current productivity shifts the distribution of future productivity upward. Here, continuation utility depends on both wealth and productivity.

Increasing risk sensitivity changes not only the weights assigned to future states but also the continuation values being weighted. Those values can change unevenly across states.

A three-period example isolates the mechanism. One income state offers a much larger distant upside than another. Increasing risk sensitivity reduces the value of that upside disproportionately, narrowing the continuation-value gap between the states. Over a particular parameter comparison, this narrowing reduces the relative decision weight on the lower-income state enough to eliminate initial saving.

The authors also construct an infinite-horizon example with strictly positive, monotone income transitions. At fixed prices, aggregate saving is positive at one risk-sensitivity level and zero at a higher one. Adding a specified production technology produces a reversal in both extremal equilibrium interest rates.

These counterexamples challenge an earlier published proposition that saving rises with risk sensitivity whenever income is stochastically monotone.

Reversals have boundaries

The authors do not conclude that predictive income makes every comparison ambiguous. They prove that the usual saving and interest-rate orderings hold whenever the lower parameter is sufficiently small or the upper parameter sufficiently large. In particular, every positive sensitivity level yields lower extremal equilibrium rates than the expected-utility benchmark under the maintained assumptions. Reversals can nevertheless occur between intermediate positive values.

An exact decomposition separates three forces: direct changes in decision weights, additional reweighting caused by changing continuation values, and changes in future marginal values of wealth. At a positive baseline saving choice, their combined sign determines whether saving falls.

The paper also establishes stationary equilibrium existence for every nonnegative sensitivity level under its conditions, including impatience, vanishing absolute risk aversion and positive, monotone income transitions. Equilibrium need not be unique. Its continuity result ensures that limits of equilibrium rates remain equilibria as parameters converge.

The contribution is therefore conditional rather than a forecast: the structure of income prospects matters, and stronger concern about adverse outcomes alone does not determine how saving or equilibrium rates change.