TheMarketForIdeas

Economic forecasts rarely materialize exactly as expected, as economic reality is highly complex and full of surprises. Macroeconomic models used to produce forecasts are simplified representations of the economic system; the data they rely on are not always available in a timely manner, and economic developments are often affected by unforeseen shocks.

Inevitably, forecasts are based on a set of assumptions regarding how demand, supply, external conditions, and other factors will evolve. If the observed developments of these variables differ from expectations, the validity of the assumptions – and the accuracy of the forecasts – are reduced. However, analysing forecast errors helps identify the types of shocks that influenced the economy beyond initial projections. In this way, it becomes possible to understand why forecasts failed to materialize, which in turn contributes to improving forecasting methods.

For example, if errors in GDP and inflation move in the same direction – both being higher or both lower than forecast levels –, the unexpected factor is usually related to demand. Such deviations may result from higher or lower-than-expected private and/or public spending, fiscal and/or monetary policy impulses, or the transmission mechanisms of these policies (through fiscal multipliers, interest rates, exchange rates, etc.).

When forecast errors move in opposite directions – one variable being higher and the other lower than expected –, the shocks likely originated on the supply side. These may include fluctuations in energy and fuel prices, disruptions in supply chains, or changes in production capacity.

Simply mapping forecast errors can therefore help interpret economic developments without necessarily resorting to complex modelling.

A relevant illustration of this is the analysis of the forecasting accuracy of the National Bank of Romania (BNR). This analysis shows that in all eight successive forecast rounds conducted between February 2024 and November 2025, the annual inflation rate for December 2025 – 9.69% – was underestimated (Chart A – BNR, Raport asupra inflației/Inflation Report – Feb. 2026, p. 52). More recently, forecast errors have actually widened (Chart C – Raport asupra inflației/Inflation Report – Feb. 2026, p. 53). This trend continued into the first quarter of 2026, for which the NBR had forecast a decline in inflation to 9.2%, while the actual rate reached 9.9% in March.

This highlights the fact that forecasting inflation in Romania is an extremely challenging task, as the economy is a highly complex system in which small changes in certain macroeconomic variables can trigger significant adjustments in others.

The causes of the unexpected developments described above are multiple, and their transmission mechanisms differ. In the first part of 2025, forecast errors were mainly driven by the persistence of inflationary pressures from the previous year, due to developments in agri-food prices and increasing wage pressures, fuelled, among other factors, by successive minimum wage increases. In the second half of 2025 and in the first quarter of 2026, the underestimation of inflation was supported by the rise in indirect taxes introduced in August 2025. This led to larger and more widespread price increases than those corresponding to a simple full pass-through of higher VAT rates into final prices, also reflecting upward rounding of prices by retailers and delayed transmission effects for certain seasonal goods.

Moreover, the increase in indirect taxes significantly intensified inflation expectations. Pessimistic expectations among economic agents combined with a series of unforeseen supply-side shocks, particularly rising labour costs – especially in services and the food industry –, as well as the indirect effects of higher electricity prices for companies, which fed into final prices through production costs.

Finally, inflation was also influenced by a series of price increases in certain food products (such as coffee and sweets) and services (for example, radio-TV subscriptions), as well as by the effects of tensions in international commodity markets.

Thus, the inherent uncertainty surrounding inflation forecasts has been driven mainly by government fiscal and budget consolidation measures, developments in energy and utility costs, changes in food and services prices, and external factors. BNR forecasts either failed to anticipate or underestimated these elements, highlighting the difficulty of predicting inflation dynamics in Romania in recent years.

Nevertheless, ex post analysis of forecast errors remains essential. Interpreting these errors through the lens of demand- and supply-side forces allows for the early detection of structural changes and the identification of inherent risks. This approach helps ground baseline forecasts, design realistic risk scenarios, improve communication by linking forecast errors to identifiable economic factors, and better understand economic changes before misguided policy decisions become frequent and persistent.

The conclusion is that turning the analysis of forecast errors into a coherent analytical tool enhances the foundation, communication, and implementation of economic policy decisions. It enables policymakers to learn from the past, recognize structural shifts in the economy, and anticipate future risks – even in conditions of high uncertainty.

Photo source: PxHere.com.