The Global Predictions model uses a mix of state-of-the-art Machine Learning techniques, statistical forecasting, and classic economic models combined to make both accurate, dependable, and generalized
forecasts. This mix avoids overfitting and the breadth allows for prediction of non-linear extreme or fat tailed scenarios.
Rather than trying to tell the most compelling narrative, we are focused on forecasting accuracy to a level rarely seen in the economic sector. We have strong fundamental hypotheses about how the world works, but always rigorously validate our models to
see if those hypotheses bear out quantitatively. All forecasts are back-tested, and measured against real-world data, across 10,000s of indicator time series.
The world is complex, therefore the models need to be complex. Modeling real estate or GDP requires an understanding of macroeconomics, technology trends, public health, financial markets, and even politics. The Global Predictions model uses a large
ensemble model approach to maintain a widespread understanding of what is happening in the macro economy at all times, predicted out 2-12 months into the future.
Lower error represents higher accuracy
The relative percentage error is calculated as the absolute difference divided by the baseline of the last year economic data was available. Benchmarks retrieved from: International Monetary Fund World Economic Outlook Database, April 2015 to October 2020.