Please Note: Blog posts are not selected, edited or screened by Seeking Alpha editors.

The REDS S&P 500 Prediction (RSP)

The REDS S&P 500 Prediction (NYSEARCA:RSP) (June 16, 2013)

My article ("The RED Spread: A Market-Breadth Barometer - Can It Predict Black Swan?") introduced the RED Spread (REDS) (with the RED) as a Market Condition Monitor. The RED Spread is performing one-month-ahead forecasts of the S&P 500 index. The predictions follow.



DATE S&P 500 DATE S&P 500 RANGE % (ABS Mean Error)
2/24/2013 1,420.40 3/22/2013 1,556.89 5/2/12-2/22/13 9.1 8.7
3/3/2013 1,421.09 3/29/2013 1,569.19 5/2/12-3/1/13 9.9 8.8
3/10/2013 1,428.79 4/5/2013 1,553.28 5/2/12-3/8/13 8.3 8.6
3/17/2013 1,431.12 4/12/2013 1,588.85 5/2/12-3/15/13 10.4 9.0
3/24/2013 1,428.24 4/22/2013 1,562.50 5/2/12-3/22/13 9.0 8.8
3/31/2013 1,429.61 4/29/2013 1,593.61 5/2/12-3/29/13 10.8 8.6
4/7/2013 1,425.70 5/6/2013 1,617.50 5/2/12-4/5/13 12.6 9.6
4/14/2013 1,433.87 5/13/2013 1,633.77 5/2/12-4/12/13 13.1 10.8
4/21/2013 1,426.25 5/20/2013 1,666.29 5/2/12-4/19/13 15.5 12.7
4/28/2013 1,432.37 5/28/2013 1,660.06 5/2/12-4/26/13 14.7 13.7
5/5/2013 1,440.94 6/3/2013 1,640.42 5/2/12-5/3/13 12.9 13.2
5/12/2013 1,424.34 6/10/2013 1,642.81 5/9/12-5/10/13 14.2 13.1
5/20/2013 1,407.06 6/17/2013 ? 5/16/12-5/17/13 ? ?
5/28/2013 1,438.23 6/24/2013 ? 5/23/12-5/24/13 ? ?
6/3/2013 1,410.32 7/1/2013 ? 5/30/12-5/31/13 ? ?
6/10/2013 1,439.78 7/8/2013 ? 5/6/12-5/7/13 ? ?
6/16/2013 1,438.26 7/8/2013 ? 5/6/12-5/7/13 ? ?
Forecast Error Percentage (NASDAQ:FEP) = 2*(A-F)/(A+F) (A=Actual F=Forecast)
Absolute Mean Error (NYSE:AME) = (sum(ABs of 7-day errors)/7)/A (A=Actual F=Forecast)

The evaluation of the predictions made on 2/24/2013, 3/3/2013, 3/10/2013, 3/17/2013, 3/24/2013, 3/31/13, 4/7/2013 4/14, 4/21, 4/28, and 5/5/2014 is available: The Forecast Error Percentages were 9.1, 9.0, 8.3,10.4, 9.0, 10.8, 12.0, 13.3, 15.5, 14.7, and 12.9, respectively; and the Mean Absolute Errors were 8.7 8.8, 8.6, 9.0, 8.8, 8.6, 9.6, 10.8, 12.7, 13.7, and 13.2; respectively. They were under-forecasts. The error were significant because the data range (5/2/2012 to 4/12/2013) don't cover a full year. From the 5/5/2013 prediction the data have a full year, as shown in the above Table..

It was also because the budget negotiations have been worked out this time, surrounding the sequester (March 1) and the smooth passage of the spending bill or a continuing resolution (NYSE:CR): The market not only didn't respond negatively on across-the-board spending cuts even though White House exaggerated the impact of the sequester, but the new CR keeps government open through September, averting shutting down government on March 27 when all stopgaps spending measurements will be expired.

All these timely good movements of Congress boosted the market more much than expected one month earlier.

The Model

Y (t) = X (t-1) + e,

Where Y (t) = one-month average of daily changes in the S&P 500 at t, X (t-1) = one-month average of daily changes in REDGs at t-1, and e = errors.

(1) The Data

The RED Spread or REDS from 4/3/2012 to date.


Market forecasts refer to market perspectives of unknown events, whether in the past, present, or future. It is the future that attracts most attention of investors because the practical purpose of the forecasts is to help make and improve investment decisions, which are forward looking.

Large econometric models which have many variables and various data are available for market forecasts, but a naïve single-variable model is introduced here. For stock-market forecasts, a simple model with a well-defined set of data can compete large model because stock prices and bond yields (and bond prices in the opposite direction), all of which are highly volatile, responsive to a continuous flow of rumors and headlines that affects expectations of investors. As a result, it is almost impossible predict the future movements of the stock market in a relatively short term, say, a time horizon of one to three months which are my primary target.

The purpose of this column is not to expect to provide any acceptable results of near-term market forecasts, but rather to attempt to test a naïve autoregressive equation with seven-day moving averages of changes in both the S&P 500 index and the RED Spread or the REDS. The REDS reflects all market activities - large and small - at the closing every day. They are unbiased and directly observed. Hence I have a strong confidence on this data set.

The REDS series just starts from April 3, 2012. After making changes and moving averages, the final input series begin May 2, 2012. Changes and moving averages made the original series less noisy, but still a year-on-year forecasting is better to eliminate further any remaining cyclical or seasonal components.

Hence, weekly one-month-ahead predictions will continue to be performed using the same starting data, May 2, 2012, until May 3, 2013. After that, the starting data will change to cover just one year. As the series lengthen, two-month and three-month predictions will be performed in the future.

The evaluation of predictions will be made by two criteria: (a) The Forecast Error Percentage (FEP) and (b) The Absolute Mean Error (AME). The FEP is for at one point at the target day while the AME is for 7-day forecasts.

At this time, I am not certain about the likely outcomes of this experiment. The only hope is this model with this innovative data set would shed some light in a very changeable area of predictions of the equity market with an explicit link with the bond market.


The origin of The REDS went back to early 1980s when a paper was submitted to the American Economic Association [AEA] meeting in Montego Bay, Jamaica. The paper was to predict stock markets (the level of the S&P 500) by employing the NBER composite indexes which were compiled at the Center for International Business Cycles Research [CIBCR]. Since the S&P 500 is one of the components of the Composite Leading Index, the paper used the Composite Lagging Index inverted: The inverted Lagging index normally leads the Leading Index.

Last summer a science-oriented reader and I exchanged our views on whether the REDS can predict the market or not, as follows:

Reader - "Here are my predictions for today at closing based on your chart above RED REDS S&P 500 10Y T Yield 9/28/2012 52.8 5.0% 1445.99 1.665."

Author - "It is pretty difficult to grasp the relevance of your comment which perhaps is based upon a presumption that the RED and the RED Spread can forecast the closing prices of the S&P 500 and 10-year Treasuries. I wish your comment works, but unfortunately it cannot because economics (finance) is not an exact science."

Reader - "Still, I look forward to your post on Monday to see how far off I am and in what direction."

Author -- Yesterday (September 28) numbers were: RED = 52.*% RED Spread = 4.6% S&P 500 = 1,440.69 10-Y Treasury Yield = 1.64%...[A]ny economic forecast should be done with an explicit model…[W]e should build two equations as:

Y1 = a1X1 + b1X2 (1)

Y2 = a2X1 + b2X2 (2)

Where Y1 = the S&P 500, Y2 = the 10-Y Treasury yield, X1 = the RED, and X2 = the RED Spread. A further discussion along this line of thought is beyond of the scope of this article."

(From our comments on "The RED Spread: A Market-Breadth Barometer - Can It Predict Black Swan?".)

The experimental work is to show my analysis to the reader (quoted above) and all other readers who might be interested in a short-term market forecast which is not seen on Seeking Alpha.