Governments routinely collect huge amounts of data that they often present in extended time series - giving figures for populations, gross domestic product, employment etc. Collect a reasonably long set of data, and see how well standard forecasting methods work. How can the forecasts be improved?