Numerous studies have found that government spending multipliers have a very low value. Indeed one recent study found that in a country with characteristics like the U.S., it was significantly negative. Yet Keynesian economists are more desperate than ever to show that government spending can be effective in promoting economic recovery. Last week the Federal Reserve Bank of San Francisco released a summary of an NBERworking paper written by two of its economists. One article in the financial media called attention to the release under the screaming headline “STUDY: Every $1 Of Infrastructure Spending Boosts the Economy by $2.”
The paper in question purports to demonstrate that Federal government spending on infrastructure projects has a much higher “multiplier” effect on real income than previously thought. According to the authors of the paper, Fed economists Sylvain Leduc and Daniel Wilson:
We find that unanticipated increases in highway spending have positive but temporary effects on GSP, both in the short and medium run. . . . We also assess how much bang each additional buck of highway spending creates by calculating the multiplier, that is, the magnitude of the effect of each dollar of infrastructure spending on economic activity. We find that the multiplier is at least two. In other words, for each dollar of federal highway grants received by a state, that state’s GSP [Gross State Product] rises by at least two dollars. . . . Over a 10-year horizon, our results imply an average highway grants multiplier of about two.
Needless to say the study has numerous flaws. It narrowly focuses on highway spending funded by Federal grants. Actually, its focus is even more narrow because the spending variable is “unanticipated” increases in highway spending, not highway spending itself. The authors seek to capture spending “shocks,” defined as “unanticipated events that affect economic activity,” because only spending that is not anticipated and adjusted to in advance can reveal the full impact of spending on the economy. And indeed LeDuc and Wilson assure us that they carefully construct a “variable” according to the latest methodology for capturing errors in forecasting each state’s future highway grants. This statistical construction is supposed to be a proxy for changes in volatile and subjective expectations about the political, economic, and financial variables that affect actual highway outlays. Forgive my deep skepticism, but I don’t think so.
Furthermore, the study uses Gross State Product (GSP), or the total value added by production in each state. But GSP is a notoriously unreliable statistic, even more so than GDP. Each state is a small open economy with completely open borders with respect to the movements of goods and factors (capital and labor) to and from other states. It is therefore virtually impossible to accurately calculate the value of a state’s exports and imports or to ascertain the income earned by its resident workers and investors in other states.
Most important, the study is disingenuous, because it uses a “New Keynesian model” that combines the traditional pure aggregate-demand effect with a supply-side productivity effect supposed to occur after the new highways have been completed and are put to use as “public capital” by the private sector to increase the output of goods and services. In fact if we inspect the graphs displaying the results of the study carefully, we find that the short-run or impact effect of increased spending yields an average multiplier of between 1.5 and 3 in the first two years and then becomes negative in years 3 to 5 yielding an overall pure aggregate demand multiplier of around zero or less after 5 years. It is then that, according to LeDuc and Wilson’s findings, the productivity effect of the newly completed roads kicks in causing the multiplier to become high in years 6 to 8. This productivity effect of the enlarged capital stock is large enough to more than offset the zero or negative spending multiplier and yield a New Keynesian hybrid spending/productivity multiplier of about 2 over a ten-year horizon.
Even if we accept that the productivity effects of spending can be captured 6 or 8 years down the road, this study hardly demonstrates the Keynesian doctrine that government spending per se raises income and promotes economic recovery. Rather it illustrates yet again the venerable Austrian insights that capital goods take “time to build” and increase productivity. The question then becomes whether “public capital” constructed willy-nilly by government officials for political purposes and without recourse to economic calculation is more productive than the private capital investment it crowded out and that would have been undertaken by entrepreneurs risking their own wealth and guided by profits and losses. I do not believe the answer to this question is in doubt.