Robust empirical evidence proves contractionary monetary policy works

This paper revisits one of the fundamental questions of macroeconomics: Does monetary policy matter? It is a question that lies at the heart of any model of short-run macroeconomic fluctuations. If monetary policy matters, then it is vital to include a channel through which changes in aggregate demand have real effects. It is also a question that is once again at the forefront of policy discussions. The Federal Reserve is in the middle of an aggressive tightening, in hopes of lowering inflation. There appear to be widely disparate views about the likely effects of its actions on real activity, and about how quickly inflation might fall. “Does monetary policy matter?” is also a question that is inherently hard to answer. Like so many empirical questions in economics, omitted variable bias is a central issue. Both monetary policy actions and real economic activity are likely to be influenced by other variables. Anything that affects output—expectations, fiscal policy, financial stress—is also likely to drive decisions by the Federal Reserve. As Kareken and Solow (1963) pointed out long ago, in the extreme case where monetary policymakers use policy to successfully counteract other forces affecting output, one would find monetary policy variables moving all around and output not changing. A simple regression of output on an indicator of monetary policy would naively and incorrectly conclude that monetary policy didn’t matter. That is, if countercyclical monetary policy actions are common, the estimated impact of monetary policy will be biased toward zero. In 1989, we wrote a paper on the effects of monetary policy using a technique we termed the “narrative approach” (Romer and Romer 1989). This is an empirical technique where one gathers systematic evidence from contemporaneous qualitative sources (such as newspapers, government reports, and policy meeting transcripts), and incorporates it into statistical analysis. In the case of our 1989 study, the idea was to deal with omitted variable bias by using the plentiful narrative record on the motivation for monetary policy actions. We used the records of the Federal Reserve to identify a subset of monetary actions that were not motivated by other factors affecting output. We argued that the behavior of output following these shocks would provide relatively unbiased estimates of the impact of monetary policy. The subtitle of that paper, “A New Test in the Spirit of Friedman and Schwartz,” captured our intellectual debt to Milton Friedman and Anna Jacobson Schwartz. Friedman and Schwartz pioneered the narrative approach in their seminal work, A Monetary History of the United States (1963). We like to think that we have built on their insights and added some modern rigor to the approach. Over our careers, we have used the narrative approach in a number of studies. For example, we used it to separate tax changes into those taken for countercyclical reasons or because spending was increasing, and those taken for more exogenous reasons (Romer and Romer 2010). We found that the estimated response of output to tax changes was faster and larger when we limited the analysis to exogenous changes. Others have also used the narrative approach to achieve macroeconomic identification. James Hamilton used it to argue that many large oil price increases resulted from factors outside the U.S. macroeconomy (Hamilton 1985). This suggested that the correlation he found between oil price shocks and recessions reflected a largely causal relationship (Hamilton 1983). Valerie Ramey used narrative sources to identify when news of changes in defense spending became known (Ramey 2011). This information allowed her to better estimate the size and timing of the effects of government spending. Andrew Jalil used the narrative approach to identify prewar financial crises in the United States, and to isolate a subset that were due to factors uncorrelated with macroeconomic conditions—such as managerial malfeasance or other idiosyncratic shocks (Jalil 2015). Incorporating this additional information allowed him to identify the true causal impact of financial crises. This address is an opportunity to reflect on the narrative approach, and to revisit our first paper using it. The heart of the address involves redoing our 1989 study. We feel we have learned a great deal about the pitfalls of narrative research and how to do it better over the past three-plus decades. In Section I, we discuss those lessons, and apply them to the identification of monetary policy shocks in the United States since 1946. Having redone the narrative work, we then incorporate the new evidence into a statistical framework. In Section II, we discuss both our methodology and our findings about the impact of monetary policy on real economic indicators, such as the unemployment rate and real GDP. We find that a contractionary monetary policy shock increases the unemployment rate by 1.6 percentage points and reduces real GDP by 4.4 percent, relative to the no-shock baseline. The effects of monetary policy are highly statistically significant. We find that our improved series on monetary policy shocks results in slightly smaller estimates of the real effects than we found in our 1989 study, but greater statistical precision. We also examine the impact of monetary shocks on inflation. We find that contractionary monetary policy reduces inflation by about 1.5 percentage points, but the effects develop slowly and are less precisely estimated than those for real activity.