For many years now we have published a quarterly data file as described in this note.
The old quarterly data file was extremely popular among institutions, particularly those that were using it for standard recession forecasting ensemble implementations or even custom ensembles to fit their investment approach. Also many clients were using the file as a single consolidated view or dashboard of all 15 of our econometric models. This popularity drove request for more timely information and we announced last month we would be updating the file on a monthly as opposed to quarterly basis.
The file is now available in the Excel section of the PRO menu:
There have been significant enhancements to functionality, such as ability to select pre populated subsets or popular, high-performing and robust ensemble models:
Additionally every one of our 15 models and even your custom selection now displays performance statistics covering average lead times to recession, standard deviations and coefficient of variance (standard deviation divided by average lead).
We display performance data for your custom selection based on the formation of a composite index or a recession ensemble with your chosen trigger level. You can also select to display and calculate your ensemble as a model count or percentage of its models:
Here is the RFE-14, a high performance collection of 14 of our best models, with a recession trigger of 3 as we selected above, available from the ENSEMBLE tab. You will note individual recession lead times associated with the selected trigger are displayed as well as performance statistics:
If you prefer building and using composite indices you can go and look at the COMPOSITE tab. Again individual leads to recession and performance statistics are displayed:
A new PERFORMANCE tab allows you to compare your selection to the 20 popular pre programmed options we have:
The exploded views for the composite indices of your selected models now automatically display lead times in months to each recession, in blue numbers on the top of the chart:
As we are now publishing monthly, some data may not be available which is highlighted in green. For the missing months we just assume the value of the prior valid print for computation of the composites and RFE models: