1.0 Introduction
"Theorizing aside, Keynesian policy conclusions, such as the wisdom of additional stimulus geared to money transfers, should come down to empirical evidence. And there is zero evidence that deficit-financed transfers raise GDP and employment—not to mention evidence for a multiplier of two." -- Robert Barro, Wall Street Journal
I find the above disingenuous, although cleverly worded to narrowly focus on transfer payments. Empirical evidence exists on Keynesian multipliers, thereby supporting the theory from which the policy recommendations are drawn.
In this post, I draw attention to four studies using Instrumental Variables (IVs) to estimate Keynesian multipliers. Currently, many economists find quite promising the use of IVs to emulate controlled experiments. For some more on the stimulus, that is, the 2009 American Recovery and Reinvestment Act, see Dylan Matthews' summaries of nine studies. I also have yet to make anything of Auerbach and Gorodnichenko (2011).
2.0 Methodology
The change in the ratio of public debt to GDP is increased by a rise in nominal interest rates, a fall in the rate of growth of real GDP, and a fall in the rate of inflation. The question addressed by this post is how to distinguish the effects of government policy on GDP from the effects of changes in GDP on government policy.
Some tax and government spending is endogenous to the economy. For example, a reduction in Gross Domestic Product (GDP) can result in:
- Lower taxes, since the income on which taxes are based has declined, and more government spending on predefined social programs
- Deliberate attempts at short-run government counter-cyclical increases in government spending and lower taxes.
One cannot easily use such endogenous changes to spending or taxes to identify subsequent causal effects from government spending and taxes on the overall level of economic activity. Accordingly, Christina D. Romer and David H. Romer (2010) consider exogenous changes to taxes, which generally fall into two categories:
- Taxes to supposedly lower or raise long run rate of growth of GDP
- Taxes to bring thedeficit closer to balance.
Romer and Romer consider quarterly USA data from 1945 to 2007. They look at narrative data, such as official reports, to classify changes in tax rates as endogenous or exogenous. The exogenous changes provide their IV which they can then use to identify subsequent causal changes in GDP.
David Cloyne (2011) applies this narrative approach for constructing a time series of exogenous tax changes to the United Kingdom. Jaime Guajardo, Daniel Leigh, and Andrea Pescatori (2011) extend this narrative approach across countries in the Organisation for Economic Co-Operation and Development (OECD).
Daniel Shoag adopts a different approach, applicable at the state level in the United States. He considers changes in state-level spending associated with windfall gains and losses in state pensions. A state pension is typically not invested in that individual state, and the dispersion of pension performance across states is exogenous, he argues, to the variation in the economic performance across states. So pension performance gives Shoag his IV. (His references include additional multiplier estimates, if I understand correctly.)
3.0 Results
- Romer and Romer find lowering taxes by 1% of GDP raises GDP by 3% in 2 1/2 years.
- Cloyne finds that lowering taxes by 1% of GDP raised GDP by 2.5% in three years.
- Guajardo et al find that fiscal consolidation (that is, raising taxes or lowering government spending) of one percent of GDP reduces real GDP by 0.62 percent.
- Shoag finds that a dollar of government spending generates $2.12 of personal income. Raising government spending by $35,000 generates another job in-state.
Update (12 October 2011): The September
issue of the
Journal of Economic Literature contains three articles debating the size of the multiplier.
Bibliography