I am deeply skeptical about how much we can learn from econometrics, but if you do it right doing econometric studies can sometimes be an worthwhile effort.
Recently I have been doing a bit of econometrics myself in cooperation with my colleague Jens Pedersen. Or rather to be frank Jens has really been doing most of the work, while I have been providing ideas. Anyway we what we have tried to do has been to estimate a so-called structural VAR models to identify historical demand and supply shock to the US economy.
Our work is far from finished and my purpose with this post is not to report on the main results from our econometric experiments, but rather to discuss the importance of monetary policy rules in understanding the workings of the economy.
Impulse-response functions – what it tells about monetary policy rules
The standard ‘output’ when you are doing structural VAR models is graphs with so-called impulse-response functions. In the case of our work we for example got an impulse-response function for how a shock to aggregate demand (AD) impacts real GDP over time.
On Friday Jens sent me such a graph. My initial response was that the graph looked as it should. A positive shock to aggregate demand in period 0 caused real GDP (we used real industrial production as month proxy) to rise over a couple of months and then to fade after that.
However, something puzzled me about the results. Jens’ estimations showed that the impact of an AD shock was very short-lived. Hence, after less than half a year the impact of the AD on real GDP would have disappeared. The impact on the price level was similarly short-lived. Or said in another way the Phillips curve is vertical after only 6 months or so.
I was puzzled by the results because our results seemed to indicate that the ‘long-run’ really is not very long and certainly much shorter than other similar studies have indicated.
So why did we get these results? We soon found the reason. We had estimated our model on the Great Moderation period from 1985 to 2007. So I suggested to Jens that we tried to estimate the model going back to the period prior to the Great Moderation. That changed our results dramatically.
Jens re-estimated the model going back to 1948. Now suddenly the ‘long run’ was no longer 3-6 months, but rather 3-6 years. Suddenly the world looked very ‘Keynesian’ (in the macroeconomic textbook sense).
The results also indicated that AD shocks was lot more common in the period prior to the Great Moderation and if anything the Great Moderation period could best be described with a Real Business Cycle (RBC) model where most of the volatility in real GDP can be explained by supply shocks (AS).
I should stress that our results are very primary, but that said the results are not completely surprising. After all the Great Moderation was termed the Great Moderation exactly because that we during that period much less macroeconomic volatility – fewer and small AD shocks – than used to be the case.
What we are of course are missing when we do our estimations is that we do not explicitly model changes in the monetary policy regime. Hence, until 1971 the US operated a quasi-fixed exchange rate regime within the Bretton Woods system until President Nixon in 1971 effectively floated the dollar and left the Bretton Woods system. After a prolonged period of monetary and exchange policy limbo Paul Volcker in 1979 started moving the US towards a rule based monetary policy. That of course was the beginning of the Great Moderation.
This is of course extremely important for the results we get. If we look at the Great Moderation period it looks like the Federal Reserve effectively had an NGDP level target. This mean that the Fed effectively would offset any shock to money-velocity to keep NGDP on track. Keeping NGDP ‘on track’ basically means that the fed would counteract all positive and negative AD shocks. That would also mean that we would basically not observe any AD shocks – as fed policy would eliminate them and the AD shocks we do observe will be quite short-lived.
On the other hand if the fed operated a fixed exchange rate regime it would do nothing to offset shocks to money-velocity shocks and we would therefore observe a lot more AD shocks and the impact of these shocks would be much longer lasting.
Any economist should of course know this – it is the Lucas Critique – but most economists tend to forget this and unfortunately most university professors forget it when they are teaching macroeconomics.
Implications for how we teach macroeconomics
In most macroeconomic textbooks the students are presented with different models of the world. The students are told that it is basically an empirical question which models are more or less correct. The key empirical question – in the textbook – is whether prices and wages are sticky or flexible.
In the first model the students learn – the paleo-Keynesian model – that prices are fixed and there is no monetary policy and the supply side of the economy is very rudimentary (supply is completely determined by demand). Not surprisingly demand is everything and fiscal policy is extremely potent.
Then the IS/LM model is introduced. And now we suddenly get monetary policy in the model, but again the discussion boils down to an empirical discussion – not about price stickyness, but about the interest rates elasticity of investments and money demand.
But we are missing something and that of course is the implicit assumptions made in these models about the monetary policy regime.
When Keynes formulated what became what I here call the paleo-Keynesian model in General Theory (1936) he assumed that we where in a fixed exchange rate world and basically also that interest rates where stuck are zero. It is therefore not surprising that Keynes came to the conclusion that fiscal policy was extremely potent. However, he only got these results exactly because of his assumptions about monetary policy. Had he instead assumed that there was flexible exchange rate regime then he would have had to come to the conclusion that fiscal policy will not have any impact on aggregate demand. This of course is exactly the result we get in the traditional Mundell-Fleming model with floating exchange rates.
Similarly the result in the IS/LM dependents strongly on implicit assumption about the monetary policy regime. When we in the IS/LM model can show that the fiscal multiplier is greater than zero it is exactly because we assume that the money supply is fixed. On the other hand if had assumed that the central bank operates for example an NGDP level target then we would not have got that result. This is what I have shown in what I have called the IS/LM+ model. Here the central bank targets aggregate demand (or NGDP) and as a result the LM curve becomes vertical and the fiscal multiplier will be zero.
These two example demonstrate how important it is to be completely clear about what assumptions we have about the monetary policy regime. And it is of course shows why our structural VAR models gave so different conclusions about the importance of AD shocks depending on what estimation period we chose.
I think this discussion is extremely important when we talk about how to teach macroeconomics. Obviously prices are sticky, but they are not fixed forever. Everybody agrees on that. So lets to assume that in our models. However, that is not really the important question. The important discussion is about monetary policy regimes and students should be told that when we show that the fiscal multiplier is positive in both the standard paleo-Keynesian model and the in the IS/LM model then it is a result of the assumptions we make about monetary policy in these models.
Therefore, we should also teach economics students that the Real Business Cycle model could work very well to explain the world if monetary policy ensures than nominal GDP (and hence AD) is kept on track. Obviously if monetary policy “removes” most AD shocks as in the US during the Great Moderation then all we have is AS shocks and that of course can be described in a RBC style model.
However, the RBC model is doing a terrible job explaining what have been going on over the past 4-5 years exactly because central banks have failed to keep NGDP on track. As a consequence it is no surprise that even the most rudimentary Keynesian models seem to be making a comeback. The fact is, however, that it still all about monetary policy or said in another way central bankers have turned to world ‘keynesian’ again.
Concluding, the monetary policy rule is the last equation in the model. It is the equation in the model that determines whether we are in a paleo-Keynesian world or in a RBC world. That is the case in our models and that is the case in real life. I hope central bankers realises this so we can get out of the keynesian hell hole and back to world where the only macroeconomic concern is the supply side.