The Fiscal Multiplier: Another Stab at Fiscal Neutrality

Apparently, using Japan as an example for the study of fiscal multipliers carries baggage. There seems to be a thousand reasons why Japan in the 1990s was not a good laboratory for fiscal experiments. I am sympathetic: Fiscal stimulus tends to occur at economic turning points and causality is difficult to determine.

I tried to think examples of large changes in fiscal expenditures that were not intended as stimulus (and where quotas were not imposed). The best example I could think of was the rapid increase in government spending associated with the ramp up in cold-war military spending in the early 1980s.

Of course, 1980 is also convenient because it is home to the largest post-war decline in output in the United States. I think once again the timing of changes in fiscal spending versus the timing in changes of GDP is informative.

Take a look at the picture immediately below. The picture shows real GDP for the United States and real federal expenditures on government consumption and government investment. I have placed a black vertical bar at what I believe is the break point in Federal spending. Both series are indexed to 100 at this point, the fourth quarter of 1979.

Importantly, this spending is not down (at least directly) for economic reasons. All of the 1979 and almost all of the 1980 increase in government spending is defense spending. [As an interesting aside, this ramp up is associated with Reagan. But, Reagan was not elected until November 1980 and did not take office until January 1981: the beginning of the buildup predates him.]
Just as we saw in the Japanese case, GDP did not fall until after the increase in spending. The fall in GDP could not have caused the increase in G. And, G did not cause GDP to stagnate forever, eventually, around the first quarter of 1983, the economy began to grow. Perhaps the economy adjust to the faster growth rate of G or perhaps the relative growth rate of G fell.

In the next picture, I show the same picture on the same scale except that I now scale government expenditures by aggregate GDP. The red line then depicts the ratio G/GDP. In this space, the story is easier to tell. Government spending, particularly defense spending, was decreasing in relative terms from the wind-down following the end of our involvement in Vietnam. In 1979 as our embassy was overrun in Tehran and as Soviet tanks rolled through Afghanistan, defense spending increased sharply and continued to grow relative to the rest of the economy through the end of 1982 where it leveled out. In 1984, defense spending fell and G began once again to grow at the same pace as overall GDP.

I have placed a dashed black vertical line at the leveling off point for government spending. I find it remarkable that the two series, GDP and G, turn in the exact same quarter. Again, G is not responding to the upturn in growth, the turning point is too quick, the timing too tight for fiscal policy.

A multiplier calculated over this interval would either be zero or negative depending on the exact start and end points. That is, government spending increased and private demand decreased by an exact equal and opposite amount. Remember, there were no quotas; this is simply a result of private agents responding to changes in real relative prices. (We will deal with Volcker below.)

I find this example completely convincing. I have a feeling it will not move pro-fiscal policy advocates. What we need to compute the multiplier is the counterfactual: what would private demand have looked like in the absence of fiscal spending. Again, if you simply know the multiplier is 1, then you can compute the path of private demand in the absence of the expansion of government spending.

Fortunately in this case, we do have a counterfactual: our neighbor to the North. Canada is a very similar country in terms of manufacturing structure and customs to the United States. It is not a perfect counterfactual; Canada is rich in commodities and during this period there were large swings in commodity prices. Nonetheless, we press on.

This picture below adds real Canadian GDP also indexed to the fourth quarter of 1979. Before we look at the time between the vertical black bars, look how similarly the growth rates of GDP in the two countries behaves outside of this interval. On this scale, I have trouble seeing any systematic differences.

Now, look at the relative behavior during the ramp up in defense spending in the United States. Canadian GDP continues to grow and grow robustly until the third quarter of 1981. The picture below shows the ration of Canadian to U.S. GDP. The gap between U.S. and Canadian GDP reaches 5.3 percent in the second quarter of 1981. I cannot think of a better counterfactual.

There is one final issue to deal with during this episode. Most economists know that the 1980 recessions were caused by monetary policy: the Volcker Stabilization. In order to bring inflation under control in the early 1980s Volcker cranked real interest rates up to nearly 10 percent. The increase in real interest rates pushed inflation down at the expense of economic growth.

Monetary policy may have caused the increase in interest rates. I believe it was the time path of government spending itself. If we pretend the United States is a closed economy, we can use the intertemporal Euler equation to generate a path of interest rates consistent with GDP and government spending. In particular, 1 = (1+r) β E[U’(Ct+1) / U’(Ct)] and in the closed economy (ignoring investment for simplicity only) Ct = Yt - Gt. In the tradition of macro finance, we can plug in observed values of Y and G and produce a series for r.

The following picture gives the results. I assume utility is CRRA with modest risk aversion. (In a standard utility function σ=1.5, a very low value for macro.) I use the three year forward growth rate of GDP.

Notice how well the model-based interest rate matches the real interest rate, over this time period. In particular, the timing of the large increase in interest rates in late 1979 and early 1980 is nearly identical. The model accounts for more than 60 percent of the Volker tightening. And it does so using only the simplest form of the intertemporal Euler equation.

This result is not consistent with the monetary policy story. In the Volker story, interest rates rise today and GDP falls in the future. What the picture is telling us is that during the Volker Stabilization, GDP was low today and rose in the future.