Econometric election forecasting models contribute to election year commentary around the world. Andreas Graefe, Kesten Green, and Scott Armstrong tested whether applying three relevant conserative forecasting principles from the Golden Rule of Forecasting to modify established election forecasting models would increase the accuracy of out-of-sample forecasts. The short answer is, yes: errors were reduced by up to 43%.
The paper has been published in PLOS One with the long but descriptive title of "Accuracy gains from conservative forecasting: Tests using variations of 19 econometric models to predict 154 elections in 10 countries". It is available, Open Access, here.