Could Monetary Policy Mitigate the Real Effects of Oil Shocks?

Michael Levi (hat tip:Marginal Revolution) and Jeremy Kahn are among those who recently rediscovered some earlier research by Ben Bernanke and others that concluded that the economic downturns that followed historical oil price shocks could have been avoided if the Fed had followed a more expansionary monetary policy at the time. Here I call attention to some subsequent research that took another look at their evidence and reached a different conclusion.

Big oil price increases such as occurred in 1973-74, 1979, and 1990 were each followed by economic recessions in the United States. In a study published in Brookings Papers on Economic Activity in 1997, Ben Bernanke and Mark Watson (professors at Princeton University at the time) and Mark Gertler of New York University concluded that these economic downturns could have been avoided if the Fed had not allowed the fed funds rate to rise following the oil shocks. The figure below reproduces one of the important findings from their paper. In these figures, the price of oil is assumed to rise 10% above its previous 12-month high at month k = 0, with the horizontal axis registering the number of months k after the oil price increase. The solid line on the vertical axis plots how that increase in oil prices at date 0 would cause you to revise your forecast of real GDP k months into the future. The calculation is based on a historical summary of the dynamic relations among a set of 7 different macroeconomic variables assuming that you could get a reasonable forecast of each of those variables by looking at their values over the previous 7 months. These changes in forecasts were calculated using methods developed by recent Nobel laureate Christopher Sims that I briefly described last week. The figure suggests that we might see real GDP 0.25% lower about two years after the shock. The dashed line represents a counterfactual simulation that Bernanke and co-authors conducted assuming that the Fed deviated from the policy that it followed over this period and instead kept the fed funds rate from rising following the shock. This counterfactual simulation suggested that the Fed may have been able to avoid any contractionary effect of the oil price increase.

Horizontal axis: time in months since oil price increased 10% above its previous peak. Solid line: predicted effect on real GDP growth rate as determined from a vector autoregression with 7 monthly lags. Dashed line: counterfactual simulation assuming the Fed kept the overnight interest rate from rising. Source: Bernanke, Gertler, and Watson (1997), Figure 4.
ana1.gif

Wayne State University Professor Ana Herrera and I took a closer look at those results in a paper published in the Journal of Money, Credit and Banking in 2004. One thing we noticed was that the counterfactual policy being proposed would have been a pretty radical departure from what actually happened. For example, the Bernanke simulation would call for the Fed to have kept the fed funds rate at 4% all through 1973 and 1974, whereas in the actual event it rose above 13%. Ana and I applied some ideas proposed by Chris Sims in 1982, and elaborated on in 2003 by two of his students, Eric Leeper and Tao Zha (each of whom were to go on to become quite famous in their own right). Their framework allowed us to construct confidence intervals to summarize how far out of the ordinary a proposed counterfactual policy is relative to the data. The solid line in the figure below indicates the amount by which the proposed counterfactual policy intervention was supposed to have been able to change U.S. real GDP according to the Bernanke, Gertler, and Watson dynamic relations, indicating that real GDP could have been 6% higher by 1977. This effect turns out to be well outside the Leeper-Zha 95% confidence intervals, indicated by dashed lines in the figure below,

Implied effects of counterfactual policy (solid lines) and 95% confidence intervals (dashed lines). Source: Hamilton and Herrera (2004), Figure 3.
ana2.gif

A more important detail that Ana and I brought out was that the Bernanke, Gertler, and Watson simulations used a vector autoregression with 7 monthly lags, whereas most of the rest of the literature on oil shocks has recognized that some of the biggest effects take longer than this to be observed. It turns out that using the Bernanke, Gertler, and Watson data, one would reject the null hypothesis that 7 lags are enough, favoring an alternative of 12 lags with a p-value as small as one in a billion.

Ana and I then redid the Bernanke, Gertler, and Watson counterfactual simulations, doing everything exactly as in their original study, except this time using 12 lags instead of 7. The results are summarized in the figure below. The estimated effect of the oil shock itself (the solid line) is significantly bigger than one would conclude if one assumed that the effects can all be seen within 7 months. More importantly, even the very aggressively counterfactual monetary policy would appear to have made relatively little difference for the outcome.

Horizontal axis: time in months since oil price increased 10% above its previous peak. Solid line: predicted effect on real GDP growth rate as determined from a vector autoregression with 12 monthly lags. Dashed line: counterfactual simulation assuming the Fed kept the overnight interest rate from rising. Source: Hamilton and Herrera (2004), Figure 9.
ana3.gif

Here’s the inference that Ana and I drew from the exercise:

We conclude that the potential of monetary policy to avert the contractionary consequences of an oil price shock is not as great as suggested by the analysis of Bernanke, Gertler, and Watson. Oil shocks appear to have a greater effect on the economy than suggested by their VAR, and we are unpersuaded of the feasibility of implementing the monetary policy needed to offset even their small shocks.

This post originally appeared at Econbrowser and is reproduced with permission.