In this part of the note, we explain why the model has revised up potential output. According to the model, the US economy has experienced a positive supply shock. Potential output is higher (both the level and the slope of Y*), driven by higher participation and a positive labor productivity shock. We discuss whether this can continue going forward.
Part I of this note is here. We proceed by Q&A.
If the reader has questions, please direct them to our FRB-US specialist: Tilda Horvath (tilda.horvath@underlyinginflation.com). Tilda has programmed FRB-US at the Board and managed the model for almost 20 years. We remind the reader that we run ad-hoc scenarios using FRB-US free of charge. Do not be shy.
Can you please show the relevant variables comparing the SEP to the current baseline?
The comparison is here. The document shows: the output gap, potential output growth (the slope of Y*), the level of potential output (Y*), trend of labor productivity growth, the level of aggregate hours, the level of civilian employment, the level of the labor force, the unemployment rate, the natural rate of unemployment (U*), the level of the labor force, and the participation rate. As usual, the SEP-consistent variables are in blue, while the current model baseline is in red. As we show below, both the SEP and the model revised up this round the level and the slope of Y*. The main difference is that the model is even more optimistic than the SEP-consistent database about Y*, both the level and the slope. For instance, Figure 1 shoes the level of Y* in the SEP- consistent database and in the model baseline. From this point of view, to us, the evolution of Y* seems more realistic in the model baseline than in the December SEP-consistent database.
Figure 1. Level of Y* in the SEP- consistent database and in the model baseline.
What happened between September and December?
Higher potential, both the level and the slope. At this link (here), the reader can see all relevant variables comparing the September SEP-consistent database to the December one (similar considerations apply to the “inconsistent” baseline). Figure 2 shows, as an example, the level of Y* (left panel) and the slope of Y*. In Figure 2, the blue line is the December SEP-consistent database, while the gray line is the September SEP-consistent (note: the level of Y* is multiplied by a factor in Figure 2 because the base year has changed in between September and December). The bottom line is that, as mentioned, there is a substantial upward revision, both in the level and the slope. Similar considerations apply to the model baseline (the “inconsistent” forecast).
Figure 2. Level and slope of potential output (Y*) in SEP-consistent databases.
Is this the first time it happens?
No, but this time is larger. Every time FRB-US is run, the model re-estimates some of the stars, such as Y* and U*. For the reader’s curiosity, the evolution of the output gap in the SEP-consistent databases can be seen here. It happens that this time the revision is large, and larger than usual. Please note that according to the model, the output gap was negative in 2023, despite the strong growth of real GDP.
Does this mean that the Okun’s law is broken?
The opposite. The reader can see the Okun law scatter plot here. Using a properly estimated model (with no priors or constraints to deliver the Okun law) results in a high correlation between the output gap and the unemployment gap. The Okun law is well and alive. The issue, if any, is the real-time estimate (and revision) of Y* and U*, given the conflicting signals of the data.
Ok but where does it come from the upward revision of Y* and its slope?
Higher participation, positive productivity shock. The model is specified and run in logs. Therefore, it is not possible to decompose the contributions to the level of Y*. In any case, it is pretty clear that the model has revised up Y* (mainly) because of an upward revision to the level of participation. On the other hand, it is possible to estimate the contributions to the revision of the slope of Y* (potential growth). Figure 3 shows the revision to the slope of Y* and its contributions. According to the model, the slope of Y* is about 20-25bps higher than 4 quarters ago. About half of the upward revision is attributed to higher participation, and about half to productivity. Please, note that the model results seem consistent with CBO’s new demographic estimates. As for productivity shocks, it is not possible to decompose the sources in the model because they are estimated as a residual (for the nerds, a document with the relevant FRB-US equations is here). As usual, we invite prudence in taking the model’s results at face value when it comes to productivity, given the possible large revisions. Still, the signal this time seems strong.
Figure 3. 4-quarter revision to Y* slope and contributions.
What is the implication of higher Y*?
Only good news, if you trust the model. A higher slope of Y* implies higher future growth, on average. Also, a higher level of Y* implies that the output gap is smaller (or negative) and there is lots of room for future growth. (For the record: FRB-US expects Q1 real GDP growth at 2.5% QoQ saar. The level of civilian employment is expected to expand at an average of 100.000 jobs per month in Q1). Finally, a higher productivity implies that the level of wage growth consistent with the 2% price target is higher (say from 3% nominal wage growth to about 3.5%).
Ultimately, can we trust the model?
The model is the model is the model. As usual, it is a model, although in case it is a very important one. In order to answer the question, we invite the reader to look at Figure 4 which plots actual and potential real GDP (in the FRB-US baseline). Does the level and the evolution of Y* look reasonable? To us, the answer seems “Yes”, especially because it is far from obvious how to treat the drop in published real GDP at the beginning of 2022. If the reader does not agree with the model, where (s)he would put Y* in 2022 and 2023? To our eyes, the only possible source of disagreement is that Y* could be a bit lower in 2022-2023; but even in this case, the main message would remain pretty similar. (Of course, we have abstracted from possible measurement issues and revisions to the level of real GDP in history).
Figure 4. Actual and Potential real GDP level in FRB-US baseline.
Conclusion
We have shown that according to FRB-US, the US economy might be experiencing a positive aggregate supply shock. In our experience, these signals are heavily scrutinized and discussed by the Fed staff (indeed, at the Board, there is a meeting -the so called “aggregate supply meeting”, where the staff discusses Y*, U*, and pi* each round). Given the above, the suspect is that the staff might be sending soon more dovish signals to the FOMC.