Economic Models and Forecasting

Economic models provide formal representations of the interactions among agents, resources and institutions. At the macro level, models range from structural equation systems and dynamic stochastic general equilibrium modules to input–output frameworks and system-dynamics simulations. At the micro level, discrete-choice, regression and machine-learning models analyse firm- and household-level decisions. Forecasting applies these models and statistical methods to infer future trajectories of key variables such as output, inflation, interest rates and emissions. Reliable forecasts underpin fiscal and monetary planning, infrastructure investments and corporate strategy. Advances in computational power and data availability have extended forecasting from traditional time-series approaches to hybrid and high-dimensional schemes that integrate survey, administrative and real-time sensor data. Across all approaches, model validation and error-measurement techniques are essential to gauge confidence and guide interventions. The global reach of model-based forecasts is evident in disaster-response planning, debt-sustainability analysis, environmental policy and supply-chain logistics. Continuous innovation in model form, estimation algorithms and scenario generation seeks to balance academic rigour with practical relevance for policymakers and business leaders.

Research from Nature Portfolio

An empirical study of 136 countries assessed how digitalisation affects carbon productivity (CP) between 2000 and 2020. By combining fixed-effects panel regressions and mediation analysis, it showed that technological innovation and income-inequality mitigation are key channels through which digital tools boost CP. Quantile regressions revealed that the upward impact of digitalisation strengthens in economies with already high CP, while heterogeneity analysis highlighted especially strong effects in high-income, high-human-capital regions.

Another investigation forecasted carbon emissions in the power sector of Gansu Province, China, under dual-carbon targets. Using the IPCC inventory approach, ridge regression and an STIRPAT model, it quantified the influence of population, urbanisation, affluence, energy mix and technology. Scenario analysis projected emissions peaks and declines under baseline, low-carbon and economic-growth pathways, offering policymakers clear pathways to meet provincial and national climate commitments.

Economic Models and Forecasting publication trend

The graph below shows the total number of articles in economic models and forecasting across all publications each year (not limited to Nature Index journals).

Technical terms

Carbon productivity: Ratio of economic output to CO₂ emissions, indicating decarbonisation efficiency.

Fixed-effects model: A panel regression technique that controls for time-invariant entity characteristics.

Mediation analysis: Statistical method to decompose the effect of an independent variable into direct and indirect pathways.

STIRPAT model: An empirical variant of the IPAT identity, used to estimate environmental impacts based on population, affluence and technology.

Quantile regression: A technique that estimates conditional quantiles of the response variable, capturing heterogeneous effects across the distribution.

Ridge regression: A penalised linear regression that shrinks coefficients to mitigate multicollinearity.

Rolling-window validation: A time-series model assessment approach that uses sequential subsets of data for training and testing.

Mean-absolute percentage error (MAPE): A measure of forecast accuracy expressed as the average absolute percentage difference between forecasts and actuals.

References

1. Impact of digitization on carbon productivity: an empirical analysis of 136 countries. Scientific Reports (2024).
2. Research on driving mechanism and prediction of electric power carbon emission in Gansu Province under dual-carbon target. Scientific Reports (2024).
3. Macroeconomic Theory and Forecasting.