Central banks rarely devote prominent space to analysing their own forecast errors. The National Bank of Serbia has done exactly that in its August Inflation Report, and the result is more revealing than a conventional retrospective.
For GDP, the NBS finds that its short-term forecasts in the post-pandemic period have generally been too cautious rather than too optimistic. Between 2021 and the end of 2024, only one quarter — Q3 2022 — recorded a downside forecast error in the sense that realised growth was weaker than the one-quarter-ahead projection. In most quarters, actual growth exceeded expectations.
The central bank interprets this as evidence of a conservative approach to GDP forecasting. That conclusion is relevant to its current 3.2% growth forecast for 2026 and 4.5% for 2027, both of which are above several external projections.
The pattern changed during 2025. GDP undershot NBS forecasts as global uncertainty increased, the operating situation around NIS remained unresolved, and domestic protests and blockades affected investment and consumer confidence. The report explicitly links some of the forecast misses to those factors.
Since the start of 2026, the direction has reversed again. Actual GDP growth has been stronger than forecast, especially in the second quarter, where the NBS estimates annual growth at 3.6%.
Inflation forecasting followed a different history. The largest one-year-ahead misses occurred in 2022 and early 2023, when realised inflation far exceeded forecasts because of the surge in global energy prices, other primary commodities and international cost pressures.
As those shocks faded, forecast accuracy improved. From mid-2023 through mid-2026, one-year-ahead inflation errors returned to a range broadly comparable with pre-pandemic experience.
The NBS uses this to argue that the exceptional misses of the energy crisis reflected shocks rather than persistent weaknesses in its forecasting model.
That distinction is more than institutional self-defence. Monetary policy depends on forecasts rather than current inflation alone. A central bank that systematically underestimates inflation would risk setting policy too loosely. One that systematically underestimates growth might repeatedly assume more spare capacity than actually exists.
The NBS’s review suggests neither bias has been stable across the full period. Inflation errors surged during a once-in-a-generation global commodity shock and then normalised. GDP forecasts were generally cautious, except during the unusual confidence and geopolitical disturbances of 2025.
The margin-control episode provides another example of why forecasting can be difficult even without global shocks. Inflation came in below projections in the third quarter of 2025 after the government introduced retail-margin restrictions. In Q1 2026, inflation again undershot expectations because the rebound after the regulation expired was smaller than anticipated. The NBS treats these as specific policy-related deviations rather than a systematic tendency to overpredict inflation.
The current forecast contains similar uncertainties. The central bank is projecting inflation to rise from 1.9% in July to slightly above 4% from September, largely because of base effects and energy costs. Food prices, however, have repeatedly surprised on the downside as the agricultural season proved stronger than expected.
GDP could surprise in either direction. Construction, agriculture and automotive production have outperformed earlier assumptions, but the economy remains exposed to Middle East energy risks, weak European demand and project-execution delays.
The bank’s alternative scenarios demonstrate that a severe energy shock alone could subtract 0.8 percentage points from 2027 GDP growth. That is large enough to move the economy materially below the 4.5% central forecast.
At the other extreme, faster energy normalisation could add 0.3 points to 2027 growth.
The report’s self-assessment concludes that forecast deviations over recent years were largely caused by hard-to-predict international events rather than systematic weaknesses in the macroeconomic models. After the 2022 shock, forecast precision gradually returned towards pre-pandemic norms.
Investors do not need to accept that judgement uncritically to find the exercise useful. What matters is the pattern of errors.
The NBS has tended to be conservative about near-term real growth during normal or favourable periods. That history provides some context for a 2026 forecast that is already higher than the IMF and European Commission. Conversely, the inflation experience of 2022 demonstrates the limits of model-based forecasting when global commodity markets move outside historical ranges.
The August report therefore invites a more nuanced reading of the central bank’s projections. The 3.2% growth forecast is not an isolated optimistic call from an institution with a history of overforecasting activity. Recent history more often shows the opposite. The 4%-plus inflation path, meanwhile, remains particularly sensitive to food and energy shocks that have repeatedly produced the largest forecasting errors.
The NBS’s most important forecast may consequently be not the midpoint but the width of the uncertainty around it.
