In this note, we update FRB-US through Q3 and we run a scenario shocking the term premium to assess the implications of the recent runup in long-term yields. We show that the September SEP has come closer to the model. FRB-US predicts that the 10y yield will remain elevated at least until the end of 2024. We also show that the 10y yield (a proxy of “financial conditions”) is doing “a lot of the heavy lift” for the Fed because the recent runup implies about 6 tenths of lower growth in 2024. It remains very hard to see lower long-term yields from here.
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.
Why we take FRB-US very seriously
FRB-US is the model of reference for simulations at the Board and proved to be reliable. If the reader is unfamiliar with the “inconsistent” FRB-US model, (s)he can refer to our notes: here, and here. We take the forecast of this model very seriously because the model has proven to be reliable. For instance, post-SVB the model suggested that it was virtually impossible for the Fed to cut rates in 2023 (see our note here). Then, it forecasted correctly the June SEP dots (see here), the super strong Q3 real GDP figures, and the bear steepening (see here).
The update and the new assumptions
This update has been challenging. The publicly available FRB-US dataset has been created at the beginning of September and posted on-line at the beginning of October. Therefore, the public version does not include neither the BEA annual revision, nor the recent runup in yields, making that version almost useless. Therefore, we updated each relevant variable of the GDP report in history, we came up with an estimate of Q3 GDP, and put add factors to the yield curve. Specifically, we are working under the assumption that real GDP has grown 4.5% (QoQ saar) in Q3, and that the 10y yield will average 4.7% this quarter. (We reminder the reader that we run scenarios on demand using your own alternative assumptions, and that in the second part of this note we have simulated a shock to the term premium).
The new baseline
The SEP FF rate now overlaps with the model. Figure 1 shows the “FRB-US inconsistent” model forecast (red line) vs the September SEP (blue line). An Excel file with all data shown in Figure 1 as well as the previous runs of the model can be downloaded here. According to the model, real GDP growth (YoY) is expected at 2.3% at the end of this year, with Q4 (QoQ saar) expected at 1.1% (an Excel file containing GDP contributions from the model is available here). Going forward, the model expects a sharp deceleration in 2024 with below trend growth at around 1% -on average- throughout the forecasting horizon. Compared to the previous run (here), the current run is broadly similar; the only difference is that growth is a bit stronger in 2023 as Q3 is a bit stronger than the model own forecast stopping the sample in Q2 (4.5% vs 3.9% QoQ saar). The model is in line with the new SEP for the evolution of the unemployment rate until 2024:H1; after that, because growth is projected to be anemic, the model forecasts an increase of the unemployment rate that reaches almost 5% in 2026. As for core inflation, the story remains the same: all our models, including the “inconsistent” FRB-US, project a deceleration of core PCE price inflation going forward but it remains above the Fed target (to 2.5%) even in 2026. Putting everything together, FRB-US is more pessimistic than the SEP in terms of growth and inflation dynamics but the two effects offset each other and the path of the FF rate is now identical. Compared to the previous run, the red line of the FF rate in Figure 1 did not move much, while the blue line (the SEP) closed the gap. In other words, the model correctly predicted the September dots. Finally, the model forecast for real return on equity, the 5y yield, the 10y yield, and the 30y yield can be seen here. For brevity, we do not comment this feature of the forecast; we only say that: (i) the SEP consistent forecast appears already unrealistic, and (ii) the model expects the 10y yield to remain very high at least until the end of 2024.
Figure 1. SEP forecast (blue line) and “inconsistent” FRB-US simulation (red line)
Note: Real GDP growth and core inflation are expressed as YoY. Core inflation is core PCE price inflation. The blue line shows the latest SEP. The orange line shows the “inconsistent FRB-US” forecast (the current baseline), that is the model-based forecast removing the “add-factors” put by the Fed staff to match the latest SEP.
Stochastic simulations
Low probability of a recession, FF rate above 6%, and core inflation at target. Figure 2 shows the stochastic simulations around the “inconsistent” baseline (the red line) shown in Figure 1. Following standard procedures of the Fed staff, we draw randomly from historical errors (1970q1 to 2017q4) for 54 variables and 5,000 replications. The simulations continue to show that: (i) calling a recession in this environment is like a flip of a coin (in fact, if anything the probability has dropped a bit since last update), and (ii) conditional on the estimated persistency, the model sees a relatively low probability of getting back to the inflation target by the end of the medium-term. A PDF containing all numbers of Figure 2 is here; the PDF contains also the estimated probabilities of the 5y yield (variable “RG5”), the 10y yield (“RG10”) and the 30y yield (“RG30”) to be above or below a certain threshold at the end of each calendar year in the forecast.
Figure 2. Stochastic simulations around the baseline.
Note: Real GDP growth and core inflation are expressed as YoY. Core inflation is core PCE price inflation.
The term premium
A shock to the term premium does “a lot of the heavy lift for the Fed” and results in about 6 tenths of lower growth. Given the recent sharp runup in yields, we have used FRB-US to assess the implications for real GDP growth and the FF rate. Before describing the exercise and the results, we remind the reader that, in our experience, the Fed staff and the FOMC members (including Powell) proxy “financial conditions” with the 10y yield. Therefore, the increase in yields in the last 2-3 months is equivalent of a severe tightening in financial conditions.
We have assessed the implications of what happened by simulating in the model a term premium shock equivalent to a 100bps on the 10y yield (75bps on the 5y yield, and 35bps on the 30y yield). The scenario is in deviation from the SEP (blue line in Figure 1), not from the “inconsistent” baseline (red line in Figure 1). Figure 3 shows the VAR-based (PING program) impulse response functions of the output gap (left panel), and core PCE price inflation (mid panel) to the term premium shock (right panel). According to the model, a 100bps shock to the 10y yield in this moment results in a drop of the output gap of about 6 tenths, reached in 2024. As for core PCE price inflation, the model expects a deceleration of about 5bps as a consequence of the runup in yields. A PDF containing IRFs to other shocks (including a shock to the FF rate, government spending, the equity premium, oil prices, productivity, and the exchange rate) is here.
Figure 3. Impulse response functions to a term premium shock
Figure 4 shows the latest SEP (the blue line) and the resulting path of the FF rate and real GDP growth (green lines) under the term premium shock. As mentioned above, a 100bps shock to the 10y yield is a significant tightening in financial conditions and results in a deceleration of real GDP growth of about 6 tenths in 2024 (the difference between the green line and the blue line in the right panel, Figure 4). Consequently, the evolution of the unemployment rate (not shown) in the green scenario is less favorable (higher unemployment), while core PCE price inflation disinflates only a touch more. For this reason, the terminal rate in Figure 4 is similar, but the FF rate is lower than in the SEP at the end of the forecast because the output gap is lower in the simulated scenario. Please note that in Figure 1 the FF rate (red line, the baseline) is on top of the SEP (the blue line) because we have already folded in the tightening in the baseline. In other words, without the runup in yields, the FF rate (red line in Figure 1) would still signal upside risks around the SEP; but precisely because yields have gone up, the Fed needs to keep the FF rate elevated for a bit less time going forward. According to the model, the run of the 10y yield is equivalent to a gap in the path of the FF rate of about 50bps at the end of the forecast (the gap shown in the left panel of Figure 4). As they say, “the 10y has done a lot of the heavy lift for the Fed”.
Figure 4. FF rate and real GDP growth in SEP (blue) vs term premium shock scenario (green)
Note: the figure shows the path of the FF rate and real GDP growth consistent with the September 2023 SEP (the blue line) and under a term premium shock (the green line).
Conclusion
The Fed can hold and wait. FRB-US suggests that the FOMC has come closer to the model with the September SEP. The runup in long-term yields has then closed the gap. The model is more pessimistic than the FOMC in terms of growth, the path of the unemployment rate, and core inflation. However, when putting everything together, the path of the FF rate in the SEP is now identical to the model. For this reason, net of additional upward surprises, we do not see the Fed hiking again soon. The strategy implies keeping “financial conditions” (that is, the 10y) elevated for long. As we discussed in private and in-person meetings, we think this strategy is risky because the fiscal authority is not cooperating. In this environment, it remains very difficult for us to see long-term yields drifting lower any time soon, unless “something will break” (which, to us, is too vague to mean anything).