Thread by @jasonfurman

How to forecast inflation? Four approaches, different answers from each:

1. Micro: component-by-component
2. Macro: slack
3. Macro: quantity theory
4. Macro: nominal GDP - real GDP

A 🧵 on thinking about these.

MICRO: COMPONENT-BY-COMPONENT. Some people are skilled/experienced at this, I'm not one of them. For example, might think used car prices temporarily spiked and will fall when the temporary bottleneck ends up airfares still need to rise.

This can be very useful for shorter-run forecasting but ultimately the microeconomic stories can only tell you about relative prices.

Two countries, one might have 1% average inflation and the other 10% average inflation and no micro story can distinguish them.

Right now the micro stories are telling us that inflation rates will come down from what we've experienced in the first five months of this year. The main reason being that a lot of the May inflation was in freakish items with huge increases that will reverse.

MACRO: SLACK. This is the main mental model I use and most of the people I engage with us. It says if some slack measure is low then some nominal measure will be high or higher or higher than expected.

Endless debates over the measure of slack, the nominal measure, magnitudes.

The slack framework is the main reason people have expected inflation to stay low. They have emphasized the flatness of the Phillips curve, that is unemployment or employment or whatever can vary a lot without much change in inflation.

Some others look at the slack framework and get worried:

--The curve may unflatten if expectations unanchor
--NAIRU temporarily high so no slack
--Openings a better RHS variable so no slack
--Ray Fair's use of the inverse unemployment rate

cowles.yale.edu/sites/default/…

But on balance I would say the slack framework mostly says not much additional inflation but it still leaves from for debate.

MACRO: QUANTITY THEORY. This is not a framework I use or most anyone I engage with uses. It says inflation will be related to the growth of some monetary aggregate. For example, M2 is up 15-20% so inflation should be up about that amount too.

Most economists place little weight on this model because it has worked terribly in the past. I expect it to be just as useless this year. But if inflation is above 10% this year I promise to travel to each of the quantity theory adherents in person to apologize to them.

MACRO: NOMINAL GDP - REAL GDP. This isn't a particularly standard approach, in a normal year I wouldn't recommend it, but at the current moment with so many other relationships out of whack I find it a useful perspective.

The "theory" is simply that if you can forecast nominal GDP growth and real GDP growth you can just subtract to get price growth (GDP prices, but close enough).

That is true by definition but is it useful?

You can make it useful by trying to forecast. Nominal demand is related to variables like incomes and interest rates. With nominal disposable personal income for 2021 likely ~15% higher than in 2019 and interest rates low likely demand in 2021-Q4 above pre-pandemic trends.

Real GDP = hours worked * output/hour.

The former looks like it will be below trend at the end of this year. The later is a huge wildcard (more on that in another thread in the future). But overall put this at pre-pandemic trend by 2021-Q4.

You can debate the inputs the nominal/real GDP forecast framework but generally this mentality leads you to inflation forecasts like 3% or 4% and possibly even higher in 2021 and 2022.

CONCLUSION. Ex post all of these frameworks can be made correct. The question is how useful are they ex ante. They can all be implemented in different ways but to overgeneralize would say that I'm finding the nominal GDP - real GDP the most useful mental crutch right now.