After personally suffering the costly effects of ECRI's notoriously ill-advised recession call in 2012, I embarked on the journey to create my own recession models. Since developing the initial models in late 2012, I have published the model forecasts shortly after the end of each month. As I learned more about the art of forecasting recessions (and market peaks and troughs), I have incorporated a few improvements to the models.
The structure of the models makes it very easy to incorporate new explanatory variables, so I am always on the lookout for new data series that are causally related to recessionary environments. I recently discovered two new data series, which will be incorporated into the Trader Edge recession models going forward.
The cornerstone of the recession models is the diffusion index. The Trader Edge diffusion index equals the percentage of independent variables indicating a recession. There were a total of 16 explanatory variables, each with a unique look-back period and recession threshold. With the addition of two new data series, there are now a total of 18 explanatory or independent variables. The resulting diffusion index and changes in the diffusion index are used by the probit, logit, and neural network models to forecast recessions.
The beauty of the diffusion index is that it is easy to add new data series at any time. The forecasting models do not use the data from each variable directly. Instead, the models only use the diffusion index values and changes in the diffusion index values. As a result, new variables can be added to the diffusion index without re-estimating the models.
In addition, it is even possible to add new variables to the diffusion index that do not have data for the complete historical time period. Granted, this would change the composition of the diffusion index, but the advantage of adding highly predictive variables to the diffusion index outweighs these concerns.
Recessions are similar, but they are never exactly the same. Therefore, the specific combination of explanatory variables indicating a recession is different from one recession to the next. Each independent variable has had explanatory power over many historical business cycles, but may or may not successfully predict any given recession.
The diffusion index eliminates the reliance on any individual variable. As long as a sufficient percentage of the explanatory variables accurately identify a weakening economic environment, the models will be successful. This makes the models more robust, especially when our primary concern is forecasting the next recession, which we know will not be exactly the same as past recessions.
Recession Slack Index
The recession slack index and changes in the recession slack index are also used as inputs into probit, logit, and neural network models. The Trader Edge recession slack index equals the median standardized deviation of the current value of the explanatory variables from their respective recession thresholds. The resulting value signifies the amount of slack or cushion relative to the recession threshold, expressed in terms of the number of standard deviations.
As was the case with the diffusion index, it is very easy to add new explanatory variables in the calculation of the recession slack index. Instead of calculating the median of 16 variables, the new recession slack index will calculate the median of 18 explanatory variables. The resulting change did not require any of the model coefficients to be re-estimated.
The Trader Edge data series and model coefficients are proprietary, but I do want to point out that each data series should have predictive value and the data series should cover a broad spectrum of economic and market data. Economic data is released with a delay and is also subject to revisions. This makes it challenging to get timely and accurate predictive estimates from economic data. While I use economic data extensively, I also prefer to include market data whenever possible.
The two new data series are both market data series. As a result, the new data is available with no lag and is not subject to revisions. Both of the new series offer early warning signals and are highly predictive. In addition, both new series reflect data that is not directly related to the existing market or economic variables.
The above changes were not designed to improve the fit of the historical data -- which was already quite good. Instead, the addition of two new market data series to the Trader Edge recession models were intended to make the models more robust and better able to identify future recessions as quickly and accurately as possible. Models are never foolproof, but the breadth of the economic and market data represented by the 18 explanatory variables effectively identify a wide range of recessionary environments.
Print and Kindle Versions of Brian Johnson's New Book are Now Both Available on Amazon (Fourteen 5-Star Reviews!)
Trader Edge Strategy E-Subscription Now Available: 20% ROR
The Trader Edge Asset Allocation Rotational (AAR) Strategy is a conservative, long-only, asset allocation strategy that rotates monthly among five large asset classes. The AAR strategy has generated annual returns of approximately 20% over the combined back and forward test period. Please use the above link to learn more about the AAR strategy.
Your comments, feedback, and questions are always welcome and appreciated. Please use the comment section at the bottom of this page or send me an email.
If you found the information on www.TraderEdge.Net helpful, please pass along the link to your friends and colleagues or share the link with your social or professional networks.
The "Share / Save" button below contains links to all major social and professional networks. If you do not see your network listed, use the down-arrow to access the entire list of networking sites.
Thank you for your support.
Copyright 2014 - Trading Insights, LLC - All Rights Reserved.