Eurozone GDP Nowcasting

The ability to accurately understand underlying changes in economic conditions is an essential component of the decision-making process for a wide variety of groups. By building confidence and understanding the direction of the economy, policymakers, investors, and businesses can all make decisions with greater clarity and certainty, especially when events are moving quickly, such as during global financial crises, pandemics, or disruptions to regional and global supply chains.

Subsequently there is a demand to understand economic developments in near-real time rather than waiting for updates to official Gross Domestic Product (GDP) figures. To solve this challenge, in this short paper we introduce our preferred modelling approaches, updates for which will now be made available to clients with access to our Connect platform.

Case Study: First Quarter 2026

To illustrate how our models operate, Chart 1 shows the evolution of two nowcasts during the first quarter of 2026. These models are based on different econometric approaches, details for which are provided below.

We began estimating first-quarter Eurozone GDP on 23 January 2026, coinciding with the first significant economic release for the quarter—the Eurozone Composite PMI for January. Our preferred MIDAS model initially pointed to Q1 GDP growth of 0.23%, while an alternative Dynamic Factor Model (DFM) suggested growth of just 0.11%. The ECB projection was for growth of 0.3%. For the MIDAS model we include a measure of uncertainty around the estimate, allowing users to assess the confidence range associated with each nowcast (based on a one-standard-deviation estimate of typical forecasting error).

The nowcast models were updated weekly from 23 January 2026 onwards, enabling us to adjust the estimates in line with the data news flow from the 28 indicators we consider key to tracking Eurozone performance. Chart 1 illustrates this progression, culminating in our final GDP estimate around two weeks before Eurostat’s first official release.

Both models performed exceptionally well in predicting first quarter GDP growth. The MIDAS model suggested a quarter-on-quarter increase in GDP of 0.15%, whilst the DFM model predicted 0.14%. The official GDP estimate from Eurostat released some two weeks later came in at 0.14%.

This strong performance was broadly consistent with our modelling performance throughout much of 2025. Real-time testing for Q2, Q3 and Q4 2025 showed that the final MIDAS nowcasts differed from the initial Eurostat releases by just 0.04, 0.08 and 0.18 percentage points respectively. Over a longer sample period (since 2013 but excluding the pandemic-affected years of 2020 and 2021), the typical Root Mean Square Forecasting Error (RMSFE) for the MIDAS-based models has been around 0.21, indicative of strong, consistent and reliable performance.

The DFM models similarly also performed well in 2025, though tended to underestimate first estimates of GDP growth in 2025.

Nowcasting Frameworks

To nowcast Eurozone GDP, we have utilised the best of a rich academic literature that has developed over recent years (some examples are available here). This literature has provided an array of useful econometric methods that can translate, into a single figure estimate of economic growth, the signals from a vast array of indicators of varying quality, volatility, time frequency, and timeliness.

One approach is to create a set of individual MIDAS nowcast models that can easily and flexibly deal with the problem of misaligned time frequencies typically found in this type of econometric problem i.e. GDP data are of quarterly time frequency, but many economic indicators are released and provided in a monthly time format. Our first method to deal with these challenges has been to create a set of individual MIDAS nowcast models from 28 indicators relevant to Eurozone economic performance (see appendix for full list). These indicators reflect a balanced mix of “soft” and “hard” data.

Soft indicators include the globally renowned Purchasing Managers’ Indices (PMIs) produced by S&P Global Market Intelligence. The PMIs have been consistently found to be important in nowcast applications due to their high frequency, timeliness, non-revision, and strong correlation with GDP statistics. So-called ‘hard’ indicators generally mean important official datasets covering, for example, industrial production, retail sales, and trade.

For the soft indicators, individual MIDAS nowcasts are underpinned by an Almon function that uses a three-month lag of the high frequency data. Such a function allows us to impose a smooth polynomial weighting structure over a pre-determined lag period but importantly capturing the influence of the most recent data in forming an estimate of GDP growth. In contrast, high frequency hard data indicators are linked with GDP in an unrestricted manner (albeit also based on a three-month lag).

From here, two RMSFE (root-mean squared forecasting error) weighted estimates of current-quarter growth — based respectively on the sets of soft and hard indicators —are combined in a dynamically optimised way to produce the S&P Global MI EZ MIDAS Nowcast. By adopting a dynamic approach, the model can flexibly adjust the weights assigned to the soft and hard indicator estimates based on recent nowcasting performance.

While this “double-weighted” approach is our preferred method, we recognise the value of alternative nowcasting techniques, most notably the widely used Dynamic Factor Model (DFM). In this framework, a single time-series “common factor” is extracted from the 28-indicator dataset. This factor typically captures a substantial proportion of the co-movement observed across macroeconomic indicators. Once converted into a quarterly time series, the factor can be used in a simple OLS regression to predict current GDP growth. Alongside the MIDAS model, we also present the results of the DFM model.

Finally, to provide a “judgement-based” estimate of current-quarter growth, we also provide the most recent European Central Bank (ECB) Staff Projections. Note that these projections are updated only four times a year (March, June, September, and December).

Availability

Going forward, we will be trialling updates to our Eurozone GDP nowcast models on a weekly basis for those clients interested in tracking the evolution of growth with the economic news cycle (releases will be scheduled for Fridays). Updates can be found within the PMI landing page on the S&P Global Connect platform.

Feedback on our approach and suggestions for improvement are welcome. Please reach out to us using our email address at economics@spglobal.com