We use up to 10 years of available data for an investment to help determine its relationship to different economic levers. For statistical significance, we need a minimum 20 data points to compare between an investment and economic lever. For most levers that are daily/weekly, this is 4.5 months of data. For monthly levers, we need two years of data. If there isn't enough data for statistical significance, then we will use industry averages for the levers to back fill the correlations.
This allows us to handle funds that are recently created, stocks without a long history, and other instances of short performance history.
Once we determine what the correlations are, one can then begin stress testing. When you see S&P beta in financial analysis, you can roughly estimate how an investment would perform in different S&P movements. We do this same analysis, but for over 100 economic indicators.
If you want to see the accuracy of the model's past predictions, there is a news article and study HERE.