Should the federal government borrow money to jump-start the economy in periods of high unemployment? Economists and policy makers who favor deficit-financed stimulus argue that a dollar of government spending is apt to have a bigger impact when lots of people are out of work, stimulating more than a dollar of economic activity. 

Michael Owyang, Valerie Ramey, and Sarah Zubairy—economists at the Federal Reserve Bank of St. Louis, the University of California at San Diego, and the Bank of Canada, respectively—tested that hypothesis by examining historical evidence. They found that multipliers, which indicate the amount of stimulus generated by one dollar of spending, do appear to be higher during times of slack in Canada, but not in the United States.

The researchers, who reported their results at the January meeting of the American Economic Association, looked at gross domestic product, government spending, population, and the unemployment rate from 1890 through 2010 in the U.S. and from 1921 through 2011 in Canada. They compared multipliers for periods of high unemployment—above 6.5 percent in the U.S. and above 7 percent in Canada—to multipliers for periods when unemployment was below those thresholds.

In Canada, multipliers seemed to be substantially higher during periods of high unemployment: about 1.6 on average, compared to 0.44 for periods of relatively low unemployment. But in the United States, the effect looked quite different. Multipliers were always below 1.0, where a dollar of government spending results in a dollar of economic activity, and they were slightly lower during high unemployment, ranging from about 0.54 to about 0.64, compared to a range of 0.63 to 0.78 below the 6.5 percent unemployment threshold. That suggests stimulus does not generally provide benefits commensurate with cost and is even less effective when the economy is struggling. 

The researchers tested other unemployment thresholds and got similar results. “[We] find no evidence that multipliers are higher during periods of slack in quarterly U.S. data from 1890 to 2010,” they conclude.