Preliminary data releases can diverge significantly from subsequently revised figures, complicating economic policy decisions made in real time. This column shows how real-time GDP releases understated the strength of the post-pandemic recovery in both the euro area and the United States. This distorted view painted a more pessimistic picture than subsequently revised data revealed, masking demand pressures.
Preliminary data releases can diverge significantly from subsequently revised figures, complicating economic policy decisions made in real time. The Great Inflation of the 1970s offers a clear illustration of this point. Most retrospective accounts conclude that the Federal Reserve was excessively accommodative during that period, thereby contributing to persistently high inflation. The work of Orphanides (2001, 2004) showed that a key driver of this accommodative stance was the data available at the time: real-time estimates systematically understated economic activity and overstated economic slack. In Giannone and Primiceri (2025), we document some analogies between the 1970s and the post-COVID inflation episode in the US and the euro area – particularly in the divergence between real-time and revised data, and how this may have influenced policy decisions.
Forecasts versus outcomes: ECB and SPF projections
To assess whether policymakers in the euro area may have misperceived the state of the economy in real time – and whether this misperception shaped their policy stance – we examine the ECB’s macroeconomic projections for HICP inflation and real GDP since 2020. We compare these forecasts with both the initial data releases from Eurostat and the most recent revised data. For the US, we cannot use the Federal Reserve’s Greenbook forecasts due to a five-year embargo. As a proxy, we rely on projections from the Survey of Professional Forecasters (SPF), focusing on expectations for CPI inflation and real GDP.
These projections are represented by the dashed lines in Figure 1. The diamonds indicate nowcasts of GDP and inflation, as the ECB and SPF projections are typically finalised in the middle month of each quarter – when forecasters generally have access only to the preliminary estimate of GDP for the previous quarter and the price index for the first month of the current quarter. The solid lines in the figure show the latest vintage of realized data that are available to us. In our paper, we provide data sources and further details on the computations underlying Figure 1.
Figure 1 Inflation and GDP in the US and the euro area: Projections vs currently available data
Sources: Bureau of Economic Analysis, Bureau of Labor Statistics, European Central Bank, Eurostat, and Federal Reserve Bank of Philadelphia; data accessed via Haver Analytics, the website of the European Central Bank and the Federal Reserve Bank of Philadelphia; computations by authors.
Let us begin by focusing on the second row of Figure 1, which displays the inflation projections for the US and the euro area. It is evident that both the SPF and the ECB repeatedly underestimated the duration of the inflation surge in 2021 and 2022, anticipating that inflation would subside fairly quickly. This pattern serves as a useful sanity check: it aligns with the narrative that many central bankers initially viewed the rise of inflation as largely transitory, revising this assessment only later (e.g. Lagarde 2023, Powell 2024).
Turning to the first row of Figure 1, we see that both the SPF and the ECB not only under-predicted inflation, but they also underestimated the strength of the recovery. Their GDP forecasts were consistently below realised outcomes, indicating that forecasters expected more economic slack than actually materialised. This finding suggests that policymakers operating in real time perceived the recovery to be weaker than it actually was.
The perception of economic conditions real time
What accounts for the substantial and systematic under-prediction of real GDP shown in Figure 1? Figure 2 provides a clue. It replicates the projections from Figure 1, but compares them to the preliminary GDP releases available at the time – the “advance estimates” from the Bureau of Economic Analysis (BEA) for the US and Eurostat’s “flash estimates” for the euro area – instead of the most recent revised data. The contrast is striking: the forecasts from the SPF and the ECB align much more closely with the preliminary data than with the revised data. In other words, by comparing the solid lines in Figures 1 and 2, it becomes clear that the BEA and Eurostat substantially revised their GDP estimates upward for much of the 2020–2024 period.
Figure 2 GDP in the US and the euro area: Projections vs real-time data
Sources: European Central Bank, Eurostat, and Federal Reserve Bank of Philadelphia; data accessed via the website of the European Central Bank, Eurostat, and the Federal Reserve Bank of Philadelphia; computations by authors.
Implications for our understanding of the sources of post-pandemic inflation
As we have shown, preliminary data painted a more pessimistic picture of the post-pandemic recovery than what later emerged. In real time, economic activity appeared broadly in line with prevailing forecasts. Subsequent data revisions, however, reveal that both the SPF in the US and the ECB in the euro area substantially underpredicted GDP. Inflation was also systematically underpredicted in both regions during its surge, as is well known.
In other words, when forecast errors in 2021-2022 are evaluated using revised data, they exhibit a much stronger positive correlation between inflation and real activity than when benchmarked against real-time releases. Therefore, this shift in the pattern of forecast errors implies a greater role for demand forces than the real-time data initially suggested. Put differently, once data revisions are taken into account, the case for a demand-driven interpretation of the inflation surge becomes stronger than with real time data—a view supported by a growing body of work (e.g. Giannone and Primiceri 2024, 2025, Attinasi and Di Casola 2024, Ascari et al. 2023, Bergholt et al. 2024, Bocola et al. 2024; Faria e Castro 2024, Gagliardone and Gertler 2024, Garcia-Revelo et al. 2024, Mori 2025).
An additional implication of our findings is that policymakers operating in real time – perceiving a weaker recovery than what ultimately materialised – may have been more willing to tolerate higher inflation in order to support economic activity. To be clear, we do not question the ex-post optimality of those decisions. Allowing inflation to exceed typical bounds may well have been a reasonable trade-off to support a faster recovery. Our point is simply that real-time misperceptions of economic slack likely played a role in shaping those policy choices.
Concluding remarks
From a methodological perspective, our analysis underscores the enduring importance of research on real-time data. Assessing the state of the economy is especially challenging during periods of heightened uncertainty and volatility, and these limitations can have first-order consequences for how economic conditions are interpreted, how policy is designed, and how both are evaluated in hindsight. The parallel with the 1970s is instructive. It took more than two decades to fully appreciate how limitations in real-time data contributed to the overly accommodative monetary policy during the Great Inflation (Orphanides 2001, 2004). Today, we are better equipped, thanks in part to the research agenda that this episode helped to launch. Real-time databases are now routinely tracked and published by researchers and policy institutions (Croushore and Stark 2003, Giannone et al. 2012), and the literature on real-time data remains active and growing (see Croushore 2011 for a survey). Regular forums – such as the Real-Time Data Conference, hosted next month at the Bundesbank and celebrating its 25th anniversary next year at the Philadelphia Fed – continue to foster exchange between academics and policymakers.
Source : VOXeu
