What FRB-US Says After Q1 GDP Figures

The model-based forecast

No cuts in 2023. Figure 1 shows the forecast of the model (the orange line) together with the December SEP (the blue line) for four variables: the FF rate, real GDP growth (YoY), the unemployment rate, and core PCE price inflation (YoY).

Note: the publicly available FRB-US dataset refers to the December SEP and the variables are updated only through Q3:2022. Therefore, we have updated the relevant variables up to Q1:2023 to run the model. When the Fed Board will publish the new dataset consistent with the March SEP (should be next week), we will update the model forecast again. Most likely, the final results will be slightly different, but today’s update gives us a clear guidance. We will let you know only in case the final results will be different compared to the ones presented here. As for the SEP lines in Figure 1, the March SEP forecast was very similar to the December SEP, therefore it does not alter the conclusions of this note.

Once the model is allowed to see the Q1 GDP/prices, the “unconstained” FRB-US expects YoY real GDP growth (Figure 1) to remain positive in 2023, roughly in line with both the December and the March SEP. Technically, the model does not expect a recession (properly defined) in 2023 but it expects anemic growth this year and slightly negative in 2024 and 2025. This happens because: (i) the model is taking signal from the weak investment figures, (ii) it takes time for a shock to show up in real GDP growth space, and (iii) the FF rate remains elevated in 2023. Also, as just mentioned, according to the model the Fed raises the FF rate above 5% (peak at 5.5% in Q3:2023) and starts cutting only at the beginning of 2024 after the unemployment rate starts raising. The reason is simple: the starting point of core PCE price inflation is so high (and the process is so persistent) that the Fed has to be sure that disinflation is on its way before cutting (ultimately, in Q1 core PCE grew at about 5% ar).

Figure 1. FRB-US model-based forecast (orange) and December SEP (blue)

Note: the blue line is the model fit of the December SEP. The orange line is the model forecast taking out the “add factors” that the Fed staff had to impose to several variables in order to fit the median dots of the SEP. The dataset for the simulations is updated through Q1. The reader can find an explanation of the publicly available FRB-US and the “unconstrained” version (the orange line) here.

Some Q&As

The real difference between the blue and the orange line

The model-based forecast presented in Figure 1 (“simulation”, the orange line) is internally consistent by definition: we run the model and plot the path of the FF rate implied by the Taylor rule. On the other hand, the median dot of the SEP (the blue line in Figure 1) is neither necessarily internally consistent nor intertemporally so. In other words, what the model is suggesting is something like this: if the SEP was internally consistent, conditional on a different forecast (lower U, higher pi), the median dot would have been higher. In other words, it is very hard to fit/explain the SEP using FRB-US and that is why the Fed staff needed to put a large number of add factors, as we have explained here and here. If there is anything that we have (painfully) learned in the last 2 years is that the FOMC has underestimated the persistency of inflation at each meeting and they ended up revising up the terminal rate over and over again; therefore, it is unsurprising that a model continues to signal a higher terminal rate (and we suspect it will continue to do so until the model will be downwardly surprised by the inflation data).

Why is the FF rate in the simulation not even higher?

The reason why the peak (terminal rate) of the orange line is at 5.5% and not higher is because it is based on a Taylor rule with a loss function. Put it simply: in the model the Fed is already facing a mild recession and the unemployment rate is expected to rise. Getting a faster disinflation would require an additional cost on the labor market, which the model evaluates as sub-optimal.

Can the inflation rate turn out to be less persistent than the orange line?

Yes, as usual it can; it is a model. Not only, but in the publicly available version of FRB-US the inflation coefficients are calibrated based on the latest estimated exercise (which dates back to 2018). For this reason, the orange line that you see in Figure 1 is based on the persistency that we re-estimate constantly based on our “main” Phillips curve model. Therefore, as any other model, if we will get weaker than expected readings, the model will start revising down its forecast. Having said so, as previously explained, at this point we would need a string of downward surprises for the models to revise down the estimated persistency of the process, given what happened in the last 2 years.

How high is uncertainty and what happens if the Fed stops hiking?

Stochastic simulations show uncertainty is elevated and calling a recession is like flipping a coin. We have run stochastic simulations using FRB-US. Results are shown in figure 2. Following standard procedures of the Fed staff (in case, #AskTilda tilda.horvath@underlyinginflation.com), we draw randomly from historical errors (1970q1 to 2017q4) for 54 variables and 5,000 replications. Simulations start in Q1:2023 and we impose no uncertainly on the FF rate in Q1 and in Q2. The main takeaway in Figure 2 is that once the model is asked to evaluate the uncertainty around real GDP growth, the answer is pretty clear: there is no statistical way to distinguish the forecasted recession in 2024/2025 from a low growth environment. The reader should notice that, instead, the model suggests a pretty convincing case for the unemployment to gradually increase, and for core PCE price inflation to remain above target at the end of the medium-term. Put it differently: while so many are “sure” a recession is coming, we think that the FOMC has very little guidance from the models in this moment and ultimately, we would not be surprised either way, mild recession or not. (The reader can find the implied path of the 5y yield as well as the 10y and the 30y here).

Figure 2. Stochastic simulations using FRB-US

Conclusion

What can or should a central banker do when uncertainty is very elevated and there is little indication core inflation is on its way back to target? According to FRB-US, the answer is straightforward: raise the FF rate above 5% and remain there until it becomes clear that the Fed can successfully disinflate the US economy. Are we going to have a recession? Despite the sell-side and the Fed staff calls, we remain agnostic because the simulations suggest there is no statistical ground for such calls. Therefore, until we will have some real good news on the inflation front, it will remain very difficult to see the Fed cutting rates, (mild) recession or not.

Contact

Want something more tailored?

We provide tailored consulting on ad-hoc projects.