Which monetary policy goal is best?

Beware, for this is another one of my increasingly frequent philosophical posts. I hope I don’t bore my readers with these. I promise normal service will be resumed tomorrow.

One of the discussions I had after criticizing the Fiscal Theory of the Price Level (FTPL) the last time was about the degree of simplification in the model. As I mentioned in that post, every theory is a simplification of the world, but I fear that by assuming rational agents that discount future government deficits to a present value, the FTPL makes some heroic assumptions that are just not realistic. I argued in my conversation that we need a behavioural theory of inflation that explains sentiment swings in risk premia and expectations, not a rational theory of the net present value of fiscal deficits. To which the challenge to me was that I cannot be sure that a behavioural theory would be just as bad as a rational theory.

Well, unrelated to FTPL, Olena Kostyshyna and her colleagues from the Bank of Canada invited people from the street without experience in economics to a lab and let them participate in an artificial economy. This artificial economy consisted of 50 periods during each of which the participants had to form inflation expectations and the output gap for the coming period. Based on these expectations about how strong the economy was going to be and how high inflation was going to be, they then simulated the anticipated consumption and savings behaviour and the resulting impact on growth, inflation, and interest rates.

The beauty of these lab experiments was that they also simulated a central bank which follows different monetary policy targets:

• Inflation targeting (IT) like the Bank of England or the ECB follow

• A dual mandate (DM) of low inflation and full employment as in the case of the Federal Reserve

• Average inflation targeting (AIT) over 4 or 10 rolling periods, i.e. the new inflation targeting regime the Fed was starting to implement before the Covid pandemic.

• Price level targeting (PLT), an alternative monetary policy target sometimes proposed as a better solution than inflation targeting. In this monetary policy regime, the central bank does not target an annual inflation rate, but rather a specific price level. However, theoretical investigations show that the key to PLT being superior to IT lies in consumers forming forward-looking expectations that take into account long-term changes in price levels.

• Nominal GDP targeting is an alternative where the central bank would target a certain level of nominal GDP, thus allowing it to run inflation higher when real GDP is declining and vice versa.

For each of these monetary policy regimes the lab experiments provided a calm period for periods 1-19 when players could settle in. In period 20, the economy would then expect a significant growth shock that led to a deflationary impulse. In other words: a recession. For periods 21 to 50, the lab experiments then allowed to study how fast the economy would get out of recession, how fast inflation returns to normal and what the costs are overall for the economy.

The chart below shows the results for inflation in the IT and DM regimes for the individual simulated economies (blue lines) and a theoretical simulation using rational people. As you can see, for both inflation targeting and the dual mandate, inflation would recover relatively quickly after a recession and then be relatively stable. Under pure inflation targeting, inflation would be a bit lower and closer to the theoretical simulation in the inflation targeting regime, but the differences are relatively small. If a central bank follows average inflation targeting, inflation swings become larger and longer, thus reducing economic stability.

Inflation pathways under dual mandate (DM, left) and inflation targeting (IT, right)

Source: Kostyshyna et al. (2022)

And now look at price level targeting and nominal GDP targeting below. As is visible in the charts, in a normal calm economy, these approaches work reasonably well, but once the deflationary shock arrives, these monetary policies quickly force the central bank to cut interest rates to zero or into negative territory. And once the central bank is at this effective lower bound for interest rates, things go horribly wrong. Investors expect more and more deflation each year and quickly adjust their expectations for growth and inflation once they realise the central bank is powerless. The result is more and more deflationary trend-following behaviour and a catastrophic deflationary spiral. Note, how large the difference between a rational model simulation is and the true inflation path when the world meets real people. While the theoretical simulation in the case of both PLT and NGDP shows on average inflation fluctuating around the targeted rate of inflation (the red line fluctuates around the 0bps deviation from target), the real world shows enormous deviations.

And that is what I meant by theories that simplify the real world too much. Any theory based on rational and/or forward-looking agents is bound to fail because people simply don’t even remotely behave like these theories assume they do. So let me repeat that famous slogan again: Everything should be made as simple as possible, but not simpler. Any economic theory based on rational agents has automatically oversimplified the real world, in my view.

Inflation pathways under nominal GDP targeting (NGDP, top) and price level targeting (PLT, bottom)

Source: Kostyshyna et al. (2022)