The New York Fed has published an update of its DSGE model forecast. The most relevant news compared to last update (March FOMC round) is that the model has a higher forecast for core PCE price inflation in 2022, and a marginally higher forecast in 2023 and 2024 but a much lower forecast for real GDP growth. Indeed, the model is now forecasting negative real output growth this year (-0.6% Q4/Q4) and the next (-0.5% Q4/Q4). In other words, the NY Fed DSGE model is now expecting a protracted recession.
How to reconcile the NY Fed DSGE model forecast with other inflation forecasts, including the forecast of our “main Phillips curve” model forecast, the SEP forecast, and the Fed staff forecast inferred from the FOMC minutes?
Keep in mind
The evolution of the inflation forecast across models reveals that: (i) our “main model” is a reliable indicator of future Fed staff and SEP forecasts, (ii) monetary policy consistent with 2 percent inflation by the end of the medium-term is now more consistent with a recession than positive real output growth as signalled by the NY Fed DSGE model.
We discuss what signal to take from each forecast and what can “go wrong” going forward.
The facts
The figures below show the evolution of core PCE price inflation forecast of our “main Phillips curve” model (the black line), the SEP (the red line), the Fed staff (the green line), and the NY Fed DSGE model (the yellow line). (reminder: our “main” model mimics the Fed staff main Phillips curve model, as described in Detmeister et al. (2014). The model is an augmented (anchored) Phillips curve model with flat long-term expectations). The top figure refers to the evolution of the (Q4/Q4) 2022 forecast, the figure in the middle to the 2023 forecast, and the figure at the bottom to the 2024 forecast. All figures show the evolution of the forecast from the November 2021 FOMC round to the June 2022 FOMC round.
Evolution of core PCE price inflation forecast (2022 Q4/Q4, %)
Evolution of core PCE price inflation forecast (2023 Q4/Q4, %)
Evolution of core PCE price inflation forecast (2024 Q4/Q4, %)
Three takeaways:
- At the end of 2021 the “main” model was forecasting core PCE price inflation well above 3% in 2022. Since then, the model has revised up its 2022 forecast and it is now projecting core inflation to be around 4% at the end of this year.
- The forecast of the “main” model has constantly been ahead of the curve. The SEP forecast has converged to the “main” model at the March 2022 FOMC and it has followed the model closely since then. The Fed staff forecast has also converged towards the “main” model forecast, although more gradually (in our view, consistent with the procedures of the Fed staff forecast which, for the current year, tends to mix incoming data with model-based predictions). Finally, the NY Fed DSGE model forecast is also converging toward the “main” model forecast but remains a bit below it.
- The 2023 (2024) forecast of the “main” model has remained stable around 3% (2½%). The alternative 2023 forecasts have increased somewhat but remain below the “main” model. In other words, the “main” model continues to signal upside risks around the other forecasts going forward.
What explains the differences between the forecasts?
First of all, the “main” model and the NY Fed DSGE forecasts are model-based forecasts, while the SEP and the Fed staff are judgmental forecasts (that is, they are informed by the models but the Fed staff and the FOMC participants can deviate from them judgmentally).
Second, we need to distinguish between the current year (2022 in this moment) and the years beyond the current (2023 and 2024). The reason is that the forecasts are constructed using different procedures. Specifically, the current year forecast is a mix of incoming data and model-based predictions, while the forecast of the years beyond the current tends to be based on the fundamentals (in the case of core PCE price inflation the fundamentals are the amount of slack in the economy, underlying inflation, and supply-side shocks including the passthrough from the Dollar). Therefore, for the current year the “main” model tends to anticipate the SEP and the Fed staff forecast because it is more sensitive to the incoming data and less inertial in updating the forecast. As for the years beyond the current, by construction, all forecasts tend to revert to target at the end of the medium-term, unless one assumes that expectations will de-anchor (which is still not the case of any forecast).
On top of this, the models have different assumptions for the path of the FF rate and consequently the path of other macro variables (see next).
Why the “main” model forecast is higher than that the other forecasts at the end of the medium-term and what is the assumed path of the FF rate across models?
The “main” model is run assuming a path of the FF rate based on the latest published SEP (and updated between SEPs using the Fed staff (2021) Taylor rule). Until recently, the path of the FF rate was very gentle. Given the level and estimated persistence of core PCE price inflation (and the path of the FF rate), the “main” model delivers a forecast that converges gradually towards 2 percent but remains materially above it in 2024. In other words, the difference between the “main” model forecast in 2023 and 2024 and the SEP (and the Fed staff) forecast can be taken as a signal of the risks around the SEP forecast (or a signal of the SEP inconsistency), conditional on the assumed path of the FF rate. For this reason, we have stressed that if the goal of the committee was to bring inflation down to target by 2024, the path of monetary policy was too gentle (i.e. the ex-post path of the FF rate would have been steeper than implied by the dots plot, as it did so far).
On the other hand, the forecast of the NY Fed DSGE model is constructed using the opposite approach: “what is the path of the FF rate and other macro variables that is consistent with inflation at target by the end of the medium-term?”. For this reason, the model has delivered a forecast close to target in 2024 in each round since last November. Many Fed watchers have been surprised by the latest results of the model. However, given what we have explained in this note and given previous simulations using the model (see for instance “Disinflation Policies with a Flat Phillips Curve”), the latest run should not come as a surprise: the Fed has little room to dis-inflate the US economy without a recession (see “Impulse Responses to a Cost-Push Shock under Alternative Monetary Policy Strategies”) because the model estimates a flat Phillips curve.
Will the Fed send the US economy into a recession?
In our view, nobody at the Fed would like to send the US economy into a recession. In fact, we are sure that the Fed staff and FOMC participants would like to deliver a soft landing. Having said so, the models forecasts show the trade-off the Fed is facing. The Fed has to act because the upside risks (including to long-term expectations) are too high. But the monetary policy that brings inflation down to 2 percent at the end of the medium-term is now more consistent with a recession than a path of positive real output growth.
What will happen going forward?
In our view, the FOMC knows that the risks of de-anchoring expectations are real. To the extent that long-term inflation expectations will continue to move upward, we think the Fed will be very determined at the cost of facing a recession. However, should expectations moderate, the Fed can be a bit more patient and tolerate an inflation rate above target (around 2½ percent) next year or the following. For this reason, it is unsurprising that different models are estimating very different recession probabilities within the next 24 months (see Fed note by Michael Kiley).
What can go wrong? Can the models be wrong?
Yes, as usual the models can be wrong. Specifically, the models can be wrong in estimating the persistency of the inflation process (and therefore the amount of monetary effort to control it). The NY Fed DSGE model estimates that the rise in inflation is mostly accounted for by large cost-push shocks and not by demand shocks (see “Drivers of Inflation: The New York Fed DSGE Model’s Perspective”). For this reason, the disinflation is very costly in the NY Fed DSGE model. A recent SF Fed note by Adam Shapiro (“How Much Do Supply and Demand Drive Inflation?”) paints a slightly more favorable picture and argues that demand factors are responsible for about one-third of the runup in prices (with supply driving half of the increase and the rest being ambiguous).
Unfortunately, nobody seems to have a good model for the global supply chain and most forecasters underestimated the issue in the last two years. Therefore, nobody seems to have a reliable estimate of if, when, and to what extent supply will normalize. What we know is that without “luck” (that is, large deflationary pressures in the goods sector driven by oversupply that offset the increase in services inflation), all models suggest that the Fed needs to remain aggressive for the time being, even risking a recession.