What do we really know about monetary policy?

Much of the talk in macroeconomic circles is whether central banks are likely to raise interest rates further, having tightened policy considerably over the past 18 months. Indeed, the ECB this week raised the depo rate to 4.0% - its highest since the ECB took over responsibility for monetary policy in 1999. Whether this is a wise move we will only be able to judge with the benefit of hindsight. But there are signs that the economy is struggling, with Germany not having grown at all in the first half of 2023 and the euro area as a whole posting meagre growth of 0.1% since 2022Q3.

The academic economics view

Central banks have, of course, prioritised curbing inflation over supporting growth in recent months with fiscal policy having done much of the heavy lifting to support households through the worst of the energy crisis. However, there remains a fundamental question about how much we really know about the linkages between interest rates and inflation – a topic tackled by John Cochrane in an excellent series of blog posts (here) and a subject I looked at in 2022. Rates are a blunt instrument to tackle inflation and Cochrane points out that the standard model in which higher rates slow the economy and bring inflation under control, albeit with a lag, is much less well founded than is often assumed.

Cochrane argues persuasively that the standard view is based primarily on the Fed’s experience in the early-1980s and that it has not been replicated since. Moreover, even this experience did not conform to what much of modern macro suggests because there was no lag between the tightening of monetary policy and the decline in inflation – it was an instantaneous process. This ought to raise a red flag for it suggests that something else was going on. He also calls into question the results from vector autoregression (VAR) models, which are the standard means of assessing the impacts of monetary policy on the economy. Much of the literature is based on the question of how the economy responds to unanticipated monetary shocks. But monetary shocks are not unanticipated – central banks are always reacting to something. If we could endogenise this process into the model, by knowing what factors trigger monetary actions, we would have a better approximation to the central bank reaction function. However, in doing so the process becomes far too cumbersome, and parsimonious reduced form VAR models may not be the appropriate method. Bottom line: I suspect much of the empirical literature operates with omitted variables which, as any econometrician will tell you, leads to biased results.

In another strange twist, the predominant macro paradigm currently in operation at central banks does not support the standard story. Standard new Keynesian models produce an instantaneous fall in inflation in response to monetary tightening, and then drifts higher over the medium-term (chart above). This is primarily due to the adoption of rational expectations in which policymakers adjust interest rates on the basis of expected inflation. If the model uses adaptive expectations, in which central banks react to past inflation, the new-Keynesian model can replicate the standard story.

Academics such as Cochrane argue that resorting to such expectations formation is sub-optimal because it results in unstable ad-hoc models from which economics has been trying to escape for the past 50 years. He then goes on to examine a number of tweaks which may add additional insight and allow the models to get closer to reality. However, I cannot help thinking that this is to miss the point. The economy is a complicated mechanism, perhaps more akin to a biosphere than a deterministic physical system, and the pursuit of simple solutions is an essentially fruitless exercise. In any case, as a practitioner rather than an academic, I am not averse to ad hocery. One of the lessons drummed into me at an early stage of my training was to consistently test whether the models we use match the data. If they do not, then it is likely that your model is wrong and you should find a better one. This is not to say I don’t admire the elegance of what academic economists produce, but if it produces outcomes at variance with how we think the economy works, we really need to go back to the drawing board.

And another thing …

I recently came across a neat paper published by the San Francisco Fed which questioned the long-run neutrality of monetary policy (summary here). The standard view is that monetary quantities do not impact real quantities in anything other than the short-run because they do not impact on the economy’s productive capacity, and as the economy adjusts to a stable equilibrium in the long-run so the monetary factors wash out. This is engrained in modern economics to the point that it is barely questioned. However, the authors of the paper took a look at more than a century of data across a wide range of economies, and found that “unanticipated” policy tightening (that word again) has impacts on output more than a decade later via its influence on total factor productivity and the capital stock. Interestingly, their results show asymmetric responses, with policy loosening having almost no long-term impact (chart below, taken from the San Francisco Fed blog).

That being the case, it suggests that the recent dramatic tightening by the Fed and ECB (which has tightened at the fastest pace in its relatively short history) may yet have a significant economic impact. Moreover, with governments having relatively little fiscal space to accommodate any slowdown, following the hit to public finances resulting from the pandemic and the energy shock, none of this screams for a positive economic outlook in the near-term.

Last word

Just as economics was forced to rethink in the wake of the 1930s depression and again following the inflation surge of the 1970s, it may be time to start thinking more deeply about how the economy actually works. If I have any advice to the academic macro profession, it would be to stop trying to find ever more elegant ways to make microfoundations work: Understand how real world businesses operate and incorporate this into the models. So long as the theory and evidence do not match up, as much of this post has demonstrated, it is difficult to conclude that we know enough about how the economy really works.