Macroeconomic Models: Imperfect but Indispensable

Milan Damjanović, Analysis and Research department

Macroeconomic models are necessarily simplified representations of the economy. Their value lies not in perfectly replicating reality, but in enabling economists to assess economic developments in a structured way, test the consistency of projections across different areas of the economy, and analyse risks and alternative scenarios. At Banka Slovenije, we therefore use a suite of different models that enter the projection process iteratively and complement one another in terms of economic structure and prediction accuracy.

Macroeconomists and the locus of control

Economists cannot control economic outcomes, but they can and should take responsibility for the process used to produce projections. This process should combine advanced analytical methods with expert judgement and lead to an outlook that is coherent, transparent and analytically well founded. At Banka Slovenije, we therefore follow an iterative projection process. Within this process, we reconcile the messages emerging from different areas of the projection, scrutinise model-based results and incorporate expert judgement to reflect additional information not captured by the models and to address remaining inconsistencies, until a sufficiently coherent and stable projection emerges (Figure 1).

As we can see, models play an important role in the projection process. Yet the economy is an open and constantly evolving system, continuously influenced by new factors, while the relevance of some previously important factors diminishes and established economic relationships are subject to instability and shocks. This naturally raises the question of why economists should entrust the analysis of such a complex and dynamic environment to models that are estimated on historical data and necessarily capture only a limited set of economic relationships.

Figure 1: Banka Slovenije's iterative projection process

Models as maps of the economic landscape

It is useful to think of models as maps of economic relationships. A map is never a perfect representation of the landscape in all its detail, yet it still shows us direction, distance and the connections between different points. Similarly, an economic model deliberately simplifies reality in order to shed clearer light on particular economic mechanisms, relationships and linkages.

At the same time, many central banks, including Banka Slovenije, use a suite of different models rather than relying on a single analytical framework (for an overview of projection models used at the ECB, see Ciccarelli et al., 2023). Different models are suited to answering different economic questions, embody different degrees of theoretical structure, and vary in their usefulness across forecasting horizons. Taken together, they allow economists to build a richer and more robust assessment of the economic outlook. Figure 2 presents a broad overview of Banka Slovenije’s modelling infrastructure.

In the short run, the data do most of the talking, while macroeconomic theory takes a back seat

At the shortest forecasting horizons, predictive accuracy is the main priority. Models for assessing current economic conditions, or “nowcasting”, can process large volumes of high-frequency data and estimate the current state of the economy before official national accounts or Harmonised Index of Consumer Prices (HICP) data are released. At the same time, they can provide useful signals about economic developments over the next few months or quarters.

Banka Slovenije’s platform for short-term model estimates of GDP growth combines a broad range of model specifications. These include dynamic factor models, which extract common signals from a large set of economic indicators; mixed-frequency models such as MIDAS, which combine data released at different frequencies; vector autoregressive models, which capture interactions among several economic variables; autoregressive distributed lag models, which trace the dynamics of these relationships over time; and bridge models, which link high-frequency indicators to quarterly GDP.

The models draw on more than one hundred time series at daily, weekly, monthly and quarterly frequencies. The data cover economic activity, business and consumer sentiment, the labour market, the balance of payments, financial markets, prices and economic developments in Slovenia’s main export markets. The range of model estimates provides information not only about the central estimate of current GDP growth, but also about the uncertainty surrounding it. Banka Slovenije’s automated tool for short-term GDP growth assessment is updated weekly and is publicly available on the Banka Slovenije website. The technical foundations of the short-term forecasting framework are presented in Radovan (2017) and Caka (2020).

Advances in short-term forecasting have also improved the ability to detect non-linear developments and turning points in the business cycle. Models that signal an increased probability of very weak or negative GDP growth can serve as early-warning indicators and point to the need for additional risk analysis or alternative scenarios. Banka Slovenije’s short-term forecasting apparatus therefore does not rely solely on central or median forecasts from linear models but utilizes also non-linear model techniques to monitor the business cycle, see Radovan (2023) for more detail.

The short-term analytical toolkit also includes indicators designed to monitor uncertainty in the economic environment. These draw on the behaviour of a broad set of macroeconomic and financial variables, the dispersion of forecasts and information contained in economic news. Because individual indicators capture different dimensions of uncertainty and respond to changes in the environment at different speeds, they can signal emerging risks before these become fully visible in conventional macroeconomic data, see Gabrovšek (2025). Newer tools based on artificial intelligence and natural-language processing can further broaden this information set by extracting signals on economic sentiment and perceptions of risk from large volumes of unstructured information.

Figure 2: Bird's eyeview of Banka Slovenije's modelling aparatus

An effective projection process must strike a balance between predictive accuracy and a credible interpretation of the economic outlook

The strength of the short-term forecasting apparatus lies in its ability to let the data speak to a large extent for themselves. Its limitation, however, is that the economic relationships underlying the results can be difficult to interpret and are not necessarily grounded directly in economic theory. As the projection horizon lengthens, uncertainty increases and empirically estimated statistical relationships become less reliable. Projections therefore require more structure. In economics, such structure brings greater coherence, discipline and consistency with economic theory to the medium-term outlook.

At Banka Slovenije, projections for individual GDP expenditure components are based on a so-called “thick modelling” approach, under which economists consider the results of several small-scale structural time-series models. These include error-correction models as well as standard and Bayesian vector autoregressive models. Labour market projections are based on equations reflecting established economic relationships, such as Okun’s law for employment and a wage Phillips curve for compensation per employee and labour costs. Inflation projections combine error-correction models for detailed HICP components with a Phillips-curve framework for assessing underlying price pressures.

Once model-based estimates for a particular area have been produced, they are complemented by expert judgement from the economists responsible for the respective parts of the projection. These area-specific projections cover private consumption, investment, government consumption, external trade, the labour market and wages, and inflation. Expert judgement allows information not captured by the models to be incorporated and adjustments to be made for known model limitations.

Small and medium-scale models combine forecasting power with economic structure, but each covers only a limited set of variables and economic relationships. In the final stage of each iteration of the projection process, projections for individual parts of the economy are fed into Banka Slovenije’s main macroeconomic model, which is used to assess the projection from a broader economy-wide, or general-equilibrium, perspective. The model makes it possible to test whether projected developments across different parts of the economy are mutually consistent and at the same time in line with economic theory.

The main model also allows revisions relative to the previous projection to be decomposed into the effects of new data, changes in assumptions and expert judgement. Such a decomposition is essential for understanding and explaining how and why the economic outlook has changed. Policymakers should be able to understand clearly how the projection moved from point A — the previous forecast — to point B — the new one — and which factors contributed most to that change. Making the contribution of expert judgement explicit and quantifying it also allows it to be scrutinised again in subsequent iterations, thereby strengthening transparency and allowing robust narrative to be developed.

The main macroeconomic model is also well suited to producing alternative projection scenarios, which have become a regular complement to baseline macroeconomic projections in an era of frequent shocks. Scenarios are used to examine how adverse risks might materialise, how alternative paths for key assumptions could affect the outlook, and how alternative transmission channels might operate, see, for example, Ciccarelli et al. (2025). In addition to analysing risks and alternative assumptions, scenarios are also used to assess appropriate policy response across different plausible paths for economic developments, see ECB, Monetary Policy Statement, June 2026

Macroeconomic projections as a basis for informed decision-making

Macroeconomic projections are integral to policy process because economic policy decisions must be taken before their full economic consequences are known. Policymakers therefore need a structured view of where the economy is heading, which forces are driving its development, and which risks could alter its course. By carefully combining model-based estimates with expert judgement within an iterative process, projections provide substantive analytical support for informed decision-making. At the same time, models make it possible to assess the effectiveness of such decisions more objectively and, where necessary, to adjust them in the future.