We have introduced a set of Phillips curve models to forecast wage inflation.
This set of models will help us monitoring the risks around the (Fed staff) medium-term judgmental forecast and they will be part of our “Pre FOMC Meeting Package” going forward.
Technical aspects of the “main” model
The “main” model is an augmented wage Phillips curve in the spirit of Detmeister et al. (2014).
Technically, a measure of wage growth is regressed on:
- Its own lags.
- A measure of underlying/trend inflation (pi*).
- A measure of structural/trend productivity.
- A measure of slack.
Assumptions behind the forecast
We assume a level of unemployment gap of -1.4 percent by the end of 2024, an average labor productivity of 1.5 percent per year, and that core PCE price inflation returns close to target by the end of 2024. Also, we impose the “anchored” restriction to the model (the autoregressive coefficients and the coefficient of pi* sum up to 1), and that in the long-run real wages growth at the same rate of structural productivity.
Results in ECI space
The results of the model for the Employment Cost Index (ECI) are reported below. Our ECI model expects the Q4/Q4 of ECI at 4.8% in 2022, followed by a deceleration in 2023 (4.1%), and in 2024 (3.7%).
What explains the “high” forecast throughout the entire medium term? There are several factors that contribute. First, under the assumption of no recession, the labor market remains tight and contributes by about 3 tenths by the end of the horizon. Second, because actual inflation is running high in 2022 and converges back to target only in 2024, pi* gradually trends higher putting upper pressure on wage growth going forward. Finally, (and most importantly) the model estimates a high persistence of wage inflation. In our ECI model, the sum of the autoregressive coefficients is 0.82. This implies that any large innovation to wage growth is reabsorbed only with time (or alternatively with a recession).
A. C.I. calculated from parameters distributions
Note: the confidence intervals (C.I.) are estimated using 100,000 draws from the estimated parameters distributions to simulate the path of ECI wage inflation going forward.
First quarter of forecast: 2022:Q2.
B. C.I. calculated using quasi-out-of-sample methods
Note: the confidence intervals (C.I.) are estimated using quasi-out-of-sample methods (estimate the model over a sub-sample, forecast, and calculate the root mean squared forecast errors).
First quarter of forecast: 2022:Q2.
Thick modelling approach
We have assessed model uncertainty with a “thick modelling” exercise. In particular, we have assessed the uncertainty of the number of lags of the dependent variable, the choice of a moving average for pi* and structural productivity, and the unemployment gap as a measure of slack. The total number of alternative models included in our “thick modelling” wage Phillips curve exercise is around 500 (importantly, as alternative measures of slack we have considered the following variables in deviation from their Z-score: (i) the employment to population, (ii) capacity utilization, (iii) job openings ratio, (iv) labor force participation, (v) and quits rates). Results are shown below.
According to our thick modelling approach, ECI wage growth is expected a bit above 4% this year, and around 3.5% in both 2023 and 2024.
What explains the different forecast between the ECI main model and the thick modelling? The thick modelling exercise is in line with the main ECI Phillips curve model at the end of the medium term but it is a bit lower in 2022 and 2023. The reason is that in the main ECI model we use the unemployment gap, while in the thick modelling exercise we consider alternative measures of slack, some of which (such as participation) have not fully recovered yet (for the record, the forecasts using job openings as a measure of slack are above the mean/median of the distributions shown below in each year as this measure of slack implies more upper pressure on wages than other).
Note: the figure shows the distribution of point forecasts of a large set of wage Phillips curve models with “anchored” expectations. Total number of models is about 500. This set of models restricts the coefficient of the long-term expectations (pi*) to sum up to one with the lags of the endogenous variable. Whiskers show the 25th, 75th, 10th, and 90th percentile of the distribution. The red line and the light blue dot in each whisker show the median and the mean of the distribution, respectively.
Takeaways and implication for the Fed staff
Takeaways. No matter what model is used, the evidence suggests that wage growth in 2022 should remain above (or well above) the “fundamentals” (trend inflation, trend productivity, and slack). According to the “main” wage model, the “fundamentals” in 2022 are estimated a bit below 3.5%. Importantly, all models suggest some persistence of wage inflation. According to the models, it should take at least another 3-4 quarters before wage growth moderates significantly.
Implications for the Fed staff. In our view, the Fed staff forecasts wage inflation using a calibrated model similar to the one presented above (with the unemployment gap as a measure of slack). In our view, the Fed staff is currently forecasting ECI wage growth around 4.5% in 2022 before returning to the fundamentals (3.5%) in 2023 and 2024. As previously discussed, in our view the evolution of wage growth is crucial for the Fed staff because the staff needs to consider the general equilibrium of the economy when assessing the level of pi* (in other words, wage growth must confirm the staff prior in the price “decomposition”). The issue is that the evidence of the models is now against the Fed staff because it signals either that wage growth will prove more persistent than what the Fed staff seems to assume or that the level of pi* currently assumed in the price decomposition (a touch above 2%) is too low to explain wage growth. In other words, in our view, either the Fed staff is in trouble because wage growth will put more upper pressure on consumers’ prices than expected or (even worse) because the level of underlying inflation is already well above target.
The Fed staff and the FOMC can get lucky going forward. But conditional on the March SEP path of the FF rates, it seems that the probabilities are not in their favor. For this reason, we continue to believe that the actual path of the FF rates will be steeper than the one implied by the March dots plot.