Inflation reflects forces operating at different cycles, from short-lived shocks to slow-moving structural trends. Yet most forecasting models treat inflation as a single aggregate process. This column introduces a ‘sum-of-the-cycles’ method, which decomposes inflation into cyclical components, forecasts each with models suited to its persistence, and recombines them into an aggregate projection. Applied to US data, the method delivers sizable gains over standard benchmarks, especially at one- and two-year horizons, and by linking forecast improvements to specific cycles and predictors, it offers policymakers a clearer diagnosis of whether inflation surprises reflect temporary disturbances or more durable forces.
Accurately assessing inflation pressures remains a central challenge for monetary policy. Inflation movements arise from heterogeneous forces operating at different cycles: transitory shocks at high frequencies, demand- and supply-driven fluctuations at business cycle frequencies, and slow-moving structural forces that shape long-run inflation (Yellen 2016). Despite this well-understood structure, most forecasting models compress inflation into a single aggregate process. This mismatch helps explain why simple benchmarks, such as the Atkeson–Ohanian random walk model, remain difficult to outperform, a point already emphasised in several works on inflation forecasting (e.g. Faust and Wright 2013, Rossi 2025).
In a recent paper (Verona 2026), I propose a forecasting method that takes the multi-frequency nature of inflation seriously. The sum-of-the-cycles method forecasts inflation by decomposing it into cyclical components, modelling each component separately, and then recombining the forecasts. The key insight is that predictability is frequency-specific: different predictors and models are informative at different horizons, and aggregate models fail to exploit this structure.
The logic of the sum-of-the-cycles method is deliberately simple. High-frequency inflation components require models that respond quickly to new information, such as financial and commodity indicators. Medium-frequency components are best captured by Phillips curve forces, including slack and inflation expectations. Low-frequency components, which dominate medium- and long-horizon forecasts, benefit from persistent financial and structural indicators.
The sum-of-the-cycles method implements this logic systematically. Inflation is decomposed into a small number of frequency bands using wavelet methods. For each band, a broad set of standard forecasting models is evaluated out of sample, and the specification that performs best at that frequency is selected. The aggregate inflation forecast is then obtained by summing the frequency-specific forecasts.
This bottom-up structure mirrors how policymakers already reason about inflation – distinguishing temporary shocks from cyclical pressures and persistent trends – but embeds this reasoning in a transparent and disciplined forecasting framework.
Applied to US inflation since the early 2000s, the method delivers substantial and robust improvements over leading benchmarks. Forecast errors fall by roughly 25% at short horizons (h=1) and by about 40–50% at one- and two-year horizons (h=4 and h=8, respectively). These gains are economically meaningful and statistically significant. Figure 1 illustrates these gains for Personal Consumption Expenditures (PCE) inflation across three forecasting horizons. Results are quantitatively similar for CPI, core CPI, and the GDP deflator.
Figure 1 Forecast accuracy for PCE inflation across horizons
The post-pandemic inflation surge provides a demanding stress test for inflation forecasting models. Many approaches – including sophisticated machine-learning methods – were slow to recognise both the magnitude and the persistence of the shock. In contrast, the sum-of-the-cycles method performed markedly better.
A central reason is its ability to incorporate information on supply constraints at the frequencies where it matters most. News-based shortage indicators improve forecasts of business cycle and medium-frequency inflation components, allowing the method to anticipate both the build-up and the persistence of inflation pressures earlier. When the same indicators are included in conventional time-series models, the gains are much smaller, underscoring that the improvement stems from the frequency-domain structure rather than from any individual predictor.
Figure 2 illustrates this point by comparing real-time forecasts after 2020 with realized inflation (black lines). Across forecast horizons of one quarter, one year, and two years, the optimised sum-of-the-cycles method (blue lines) signals elevated inflation pressures earlier and tracks their persistence more closely than the best-performing time-series model (red lines), which systematically underpredicts the surge.
Figure 2 Real-time inflation forecasts after 2020
The sum-of-the-cycles method improves forecasts for two related reasons. First, it reduces bias arising from forcing heterogeneous inflation dynamics into a single aggregate process. Second, it enhances interpretability. Forecast revisions can be traced to specific predictors operating at specific cycles: financial variables dominate high-frequency movements, Phillips-curve forces shape medium-run dynamics, and structural indicators anchor long-run inflation.
For central banks, the primary implication of the sum-of-the-cycles method is improved forecast accuracy at policy-relevant horizons. Relative to standard benchmarks, the method delivers systematically lower forecast errors, particularly at one- and two-year horizons, which are central for monetary policy decisions.
These gains are especially relevant at turning points. During the onset of the 2020–21 inflation surge, using sum of the cycles would have provided earlier and more persistent signals of rising inflation pressures than conventional aggregate models, reducing the risk of interpreting the shock as predominantly transitory. By explicitly accounting for the frequency at which different predictors are informative, the method improves not only average forecast performance but also the timeliness of signals when inflation dynamics change rapidly.
Sum of the cycles should therefore be viewed as a complementary forecasting tool. It does not replace DSGE models, judgemental assessments, or professional surveys. Rather, it offers a quantitatively more accurate cross-check while simultaneously enhancing interpretability, by linking forecast revisions to specific cycles and predictors. In this respect, the method complements recent policy analysis emphasising how distinct cyclical and persistent components shape inflation dynamics across horizons (e.g. Hasenzagl et al. 2018).
The broader message is simple. Inflation is the sum of distinct cycles, and forecasts that respect this structure outperform those based on the aggregate alone. In this sense, the sum of the cycles truly outperforms the whole.
Source : VOXeu
Experts say $200bln bond-buying effort unlikely to significantly lower housing costs. There's scant evidence so…
That has helped at least to put a floor under euro zone bond prices. Euro…
Bank profitability will remain strong this year despite lower interest rates, says S&P. Saudi banks…
The 2026 review of the EU ETS must be anchored in facts and focus on…
Federal Open Market Committee statements typically sound unanimous, but the Committee’s internal debates rarely are.…
Local responses to gender-based violence, with femicide as its most extreme form, remain uneven across…