Empirical Forecasting Models

Empirical forecasting grew alongside modern economic measurement. The hub below keeps that lineage practical by arranging the work around forecasting jobs, data choice, diagnostics, and comparison.

Section Notes

All Models Forecasting One Series Interacting Series Economy-wide Release Window State Shifts

Route notes

One route follows a single target series, another keeps a small interacting system together, and a third sketches a simplified economy-wide platform.

Diagnostics stay close to the route itself, where assumptions and forecast paths remain visible.

The live build currently covers one-series and interacting-series forecasting, with nowcasting, regime, and scenario layers nearby.

Empirical forecasting

Economic measurement turned macro forecasting into a practical workspace.

As governments started publishing national income, prices, labor, and spending data regularly, economists moved from describing macro cycles in words to estimating them from time series. On this site, forecasting stays organized by job: one target series, a small interacting system, or a simplified economy-wide platform.

Each route keeps data selection, transformations, diagnostics, and forecast comparison in one place.

Forecasting jobs

Baseline path, interacting system, or scenario platform.

The forecasting family is organized around the question at hand. Each route carries the same loop forward: data, preparation, model, diagnostics, forecast, and comparison.

The live build now includes one-series and interacting-series routes. Scenario layers keep the same logic once the system grows beyond those first forecasting jobs.

Live route

One Target Series

Available today

One macro series stays in view from target definition and benchmarks through live models, diagnostics, and holdout comparison.

Best for: Baseline projections and live monitoring for one macro variable at a time.

Workflow: Data → Preparation → Model → Diagnostics → Forecast → Compare

Diagnostics: Time plot, Rolling behavior, ACF / PACF, ADF and KPSS, Ljung-Box, RMSE / MAE / MAPE / MASE

Benchmarks run first. ETS and ARIMA are the live candidate models today, while AR, MA, and ARMA remain part of the learning scaffold behind them.

Live workspace

Live route

A Few Interacting Series

Available today

Small macro systems let inflation, labor, output, rates, or housing variables move together inside one shared read.

Best for: Joint dynamics across a small macro system.

Workflow: Data → Alignment → Transform → Model → Diagnostics → Forecast

Diagnostics: ADF / KPSS, Johansen cointegration, Lag order, Stability roots, IRFs / FEVD

The live lab keeps the pack small on purpose so lag choice, long-run links, and spillovers still mean something.

Live workspace

Guided route

A Practical Whole-Economy Platform

Guided route

A simplified scenario platform keeps linked blocks, accounting identities, exogenous assumptions, and baseline-versus-shock comparison in view.

Best for: Baseline and shock comparison across sector blocks.

Workflow: Map blocks → Assign equations → Link identities → Run baseline → Apply shock → Evaluate fit

Diagnostics: Identity consistency, Residual fit by block, Parameter stability, Scenario sensitivity

It stays simplified on purpose. The aim is to make economy-wide forecasting legible before institution-scale infrastructure enters the picture.

Guided route

Route comparison

Forecasting jobs shape the route.

Route	Best for	Typical models	Output
One Target Series	Baseline forecast for one macro variable	Benchmarks, ETS, ARIMA / SARIMA	Point and interval forecasts
Interacting Series	Joint macro dynamics	VAR, VECM, BVAR	Joint forecasts, IRFs, decomposition
Economy-Wide Platform	Scenario and policy-style projection	Semi-structural / macroeconometric blocks	Baseline plus shock comparison

Adjacent forecasting jobs

Nowcasting and regime risk sit beside the baseline routes.

Release-Window Nowcasting

Bridge equationsDynamic factors

Current-quarter and current-month updates belong here once mixed-frequency inputs and release timing matter more than the baseline path itself. Mixed-frequency updates become a different forecasting question once release timing outweighs the baseline path.

State Shifts and Recession Risk

Probit systemsMarkov switching

State-shift work watches whether the economy may be changing state, which differs from projecting one clean point path. Probability signals matter most when transition risk carries more weight than the next print.