Am picking apart some definitive technical analysis as applied to one of arguably few reputable economic indicators, the ECRI Weekly Leading Index (WLI). Leaving aside all and any assessment of its application to any purported economic reality, this exploration is purely technical analysis of the publicly available WLI data series, and provides a summary of where it sits amongst the myriad of purported recoveries or otherwise as regurgitated daily by the main stream media.

I cannot suggest if this data series has either a linear or power series (compounded, polynomial) relationship to the broader economic state of play, and merely present several methods of analysis that indicate the possibility that all is not what it seems.

Whole of data series linear regression showing +/- StdDev bands

The most obvious 2 observations are a) the rate of increases into 2007, broken by b) massive correction to the underside in 2008/09. This parallels with the observed outcomes of the GFC in 2008. Note however the WLI did not foresee the 06May2010 flash crash, yet the WLI responded to this event very abruptly. There is a 6 standard deviation correction to the March 2009 low, and we currently reside -3.5SD’s under the linear regression average. Obviously a weak recovery to date.

Note that this data series started during a period of 8 years indicating economic stagnation from 1967 to 1975. This is fortuitous since it allows a direct comparison to what might be a similar period as indicated by the ECRI WLI currently in 2011, from the June 2007 high.

Same whole of series, added moving averages (short, medium and long term)

– moving averages most clearly show momentum, and comparison to previous trend movements

When analysing many instruments and asset trends, there is a distinctive increase in rates of change and cyclic volatility from 1996 onwards, attributable to a marked and definite increase in electronic analysis and trading, and easing of market access coupled with excitement and fervor over the technology sector. This euphoria would eventually end in the very public bursting of the dot.com bubble and reveal massive stock price manipulation in what was to become the new black in IPO’s spinning and laddering. For this exercise, I am using the point that it last departed above the current linear regression average (being 1995) and using January 1995 so as to commence with a whole year.

Closeup of linear channels 1995 onwards

Applying some simple moving averages to the data series further highlights that we are yet to clear the obstacles of the GFC of 2008. As can be clearly seen in the following chart, with the addition of some Fibonacci levels

– showing major trend lines overlayed also

Combining some RSI and MACD analysis (not shown) of the ECRI WLI analysis, we can see a prevailing weakness in the forward movement of this Index indicating upwards direction being weak and limited, sideways trending being very likely and a possibility of a further decline into a dip and recovery within the next 12 months to Dec2012.

Low order dynamic bands applied to 1995 portion

CONCLUSION: Current situation indicates clearly we are still in a period of sustained weakness showing no immediate recovery. Momentum is negative (downwards) while a rise is not impossible, it is merely rallying in a down trend (a bear rally). Confirmation of a clear break of this downward trend is an ECRI WLI reading above 125. I would mark a recovery to resumed growth as confirmed only above 129.

Using a Dec1971 value of 60, and applying 2% CAGR to the ECRI WLI, we project a value of 132.5 for Dec2011. Clearly we are below this. Given that we are also below linear regression of 138 by some 3.5 standard deviations, it does suggest that the universe is conspiring aginst mean reversion, and that upwards progress is imminent. Mean reversion and regression to trends are poetry when you are placed on the right side of the move.

Very simply, upwards is a good thing. It’s a matter of getting it there, and keeping it there. Eventually further growth will become the most obvious certainty. But timing isn’t everything, it is the only thing. The downside is that we are in a period fo 6-8 years that is sideways to down still. I respect the lucid and thorough Kyle Bass who suggests large corrections to housing markets last anywhere between 6-9 years. Using 6 years aligns with the previous history of the ECRI low in 1975, placing another possible cyclic low still to come in Dec2012.

Hence I would put a floor of 120 not to break (must hold), and a ceiling of 129 that it needs to break (confirm recovery and growth) – buy the dip, sell the fade in the mean time.

Regards,

I would like permission to post your work on my new website: http://www.tedbits.com

Thank you for your time and consideration. sincerely

Ty Andros

Ty, have posted an email to you. Regards.

Pingback: ECRI WLI – quick update | atradersrant

I stumbled across your blog and found this post quite interesting. One question, speaking as an Excel amateur, how did you create the outer bands of the polynomial regression in the last chart?

Thanks,

Charles

Hi Charles, having done quite a few of these now, there’s two ways to go about it. They are a) the proper way (via the Solver addin), or b) the quick and dirty way (much quicker, more flexible)

a) using the solver addin, quite labourious but worthwhile to do a few properly (refer to a reputable Excel manual/online search)

b) quick and (not so) dirty – much quicker

Excel creates it’s own chart based trendlines (Add Trendline->polynomial) much faster than setting up a Solver system each time. (You soon get sick of playing with solver parameters). So instead –

Create the Excel chart, and plot a trendline on the data series (e.g. polynomial, order 2, 3, 4, 5 or 6). Your choice, but high powers make high frequency rolloff at the ends of the data sets. Sometimes 6 is a better fit, typically 3, 4 or 5 depending on duration (number of data elements).

Create a curve shape (Insert->Shape->Curve) and trace the polyline ‘exactly’. Once done, hold the shift key and move the trace vertically (care here, do not offset left or right) to approximate by eye a 95% capture of data offset (either above or below). You will begin to notice what looks right having done a few. Copy+Shift and create another copy vertically shifted but ‘offset the exact same’ on the opposite side of the original Excel polyline. Use other shapes to help measure this to maintain accuracy (you don’t have to be too rough, you lose the value in it).

2xSD offset is 95% boundary of data on one side of the central polyline regression average (Excel’s polyline). So you can use +/-1SD bands, or +/-2SD bands (4xSD total width). I use +/-2SD as this is easier to visualise, and gives better extremes in price movements with lower risk trades – which result in less frequency of trading. (my preference, you can suit yourself).

To change and experiment this way is much quicker and more flexible once you get an eye for the proper curve fit. Beware of using high power polynomials however. The rolloff is not always what you want to see.

Happy experimenting. You can do quick linear regression channels this way also. A carefully selected regression system can prove very useful (and profitable).

Regards,