Modern fiscal policy doesn’t work the way it did in the days of John Maynard Keynes and President Franklin Delano Roosevelt. When Roosevelt increased federal spending from 1941 to 1945, the money went almost exclusively to direct purchases of goods and services. Today, fiscal stimulus happens mainly through taxes and transfers.

Tax revenues fall automatically in recessions, and governments back that up with lower tax rates and/ or new credits and deductions. On the spending side, extra outlays on unemployment benefits and other transfers greatly exceed extra outlays on infrastructure and other purchases. This modern kind of fiscal stimulus is supposed to work by stabilizing disposable income. Stabilize that, the thinking goes, and you stabilize output and employment.

But is that right? In a new working paper, Ricardo Reis of Columbia University and Alisdair McKay of Boston University say no. They find that stabilizing aggregate disposable income plays a “negligible role” in stabilizing the economy as a whole. Transfer payments can indeed stabilize output, they find, but mainly through a different channel -- not by changing disposable income in the aggregate, but by changing its distribution. Fiscal policy, in other words, is all about inequality.

“It’s the redistribution that has a lot of kick,” Reis said in an interview. “The usual argument for transfers is basically Keynesian. We find that has very low impact in our model.”

Reis and McKay reach this conclusion by building a complex macroeconomic model calibrated to U.S. data, but the intuition isn't all that complicated. Transfer payments yield the highest amount of stabilization per dollar when focused on people who can't effectively insure themselves against macroeconomic volatility -- namely, people with little savings to draw on and limited opportunities to borrow.

“When you look at the different programs, we find that food stamps and similar programs are really the ones that work, because they are being targeted to individuals who are up against their borrowing constraints and aren't going to work less because they are already unemployed," Reis told me. "They have very high marginal propensities to consume and are very underinsured, so it can very stimulative."

They also find -- this is a surprise -- that fiscal policy as currently designed does little to stabilize the economy. The most effective transfer programs, Reis says, constitute a small share of all transfers. “When we look at the whole set of stabilizers in the U.S., it turns out that even though food stamps are a plus, all of the other ones have near-zero impact. That means we’re not stabilizing very much,” Reis said.

The upshot: If the U.S. wants milder recessions in the future, its most effective fiscal policy options are food stamps, Temporary Aid for Needy Families and unemployment insurance. If the study is right, a better safety net for the poor turns out to be the best safety net for the whole economy.

(Evan Soltas is a contributor to the Ticker. Follow him on Twitter.)