1. Market beating returns. The strategy aims to out-perform the JSE All-Share Index (ALSI) by at least 3-5 times over any rolling three year period.

2. Medium to long term performance. The strategy is designed for a 2-10 year investment horizon

3. Small portfolio sizes. On-line brokerage is still expensive in SA and thus a strategy is sought that performs best with 4-6 share portfolio sizes. This will require an excellent picker of "winners".

4. Performance maximised for annual re-balancing, to minimise maintenance and brokerage overhead.

5. Maximised risk-adjusted returns. This implies growth accompanied by low volatility. It requires annual portfolios with low constituent share volatility as well as consistency and low volatility of per annum portfolio growth rates over 3,5 and 10 year periods.

6. Proven performance. The strategy must have at its core, the foundations of a proven internationally tested system that has also been back-tested for at least a 10 year period on the JSE.

7. Liquidity. The strategy must allow constituent shares to be easily acquired and disposed of. STRATEGY MECHANICS

PowerShares-A has at its core, the Magic Formula Investing strategy that picks stocks with the best combination of low Price/Earnings (P/E) and high Return on Capital Employed (ROCE). However it has been modified for use on the JSE as described below:

Portfolios are 5 shares in size and are rebalanced annually. Screening is done with the most recent published financials, be they finals

__or__interims, to ensure the freshest fundamental data is used.

In addition, screening is targeted at a special TOP-100 JSE shares universe (a proprietary universe devised by us based on market capitalisation, liquidity, cash-flow and other variables) which provides for low constituent share and per annum growth volatility (see our previous research titled "Size Counts".). This universe consists of large liquid issues and has maximised sensitivity to PE and ROC.

Finally, only shares that pay dividends are considered (Dividend Yield > 0) since these provide for superior returns based on our tests, and with dividends re-invested, the strategy has a multiplier effect on performance. By default, the strategy picks low PE shares which implies high yields and with the market at all time lows, and dividend yields high, we feel this is a prudent component of the strategy that is fit with the times.

WHY P/E AND ROCE?

When designing PowerShares-A, we back-tested various single and double combinations of P/E, Price/Book, Dividend Yield and ROCE over the most recent 10 year period, and the Magic Formula method of combining PE and ROCE provided the most convincing and consistent risk adjusted returns. The diagram below shows the results of our tests conducted on the Top-100 JSE shares universe by market capitalisation. It depicts the average per-annum growth achieved over each of the 10 years together with standard deviation and Sharpe ratio (average divided by deviation), for various different 5-share portfolio strategies.

PE+ROC is a two-factor strategy based on the standard Magic Formula investment ranking method, whereas PWRSHR is the same method but limited to dividend yielding stocks only. DY is akin to the famous "Dogs of the Dow" theory, seeking the 5 highest yielding stocks from the Top-100 list. P2B+ROC is also a Magic Formula ranking method, but applied to Price/Book and ROCE. We see that the Price/Earnings Magic Formula derivatives significantly outperform all the other strategies both on an average growth and risk adjusted (Sharpe) basis.

WHY THE TOP-100?

We conducted "full monty" strategy backtests (as you have come to know on this site) on the Magic Formula method on ALL-STOCKS, MEDIUM STOCKS and LARGE STOCKS universes and the best risk adjusted returns and base rates that allowed for 5-share portfolios were delivered by the special PowerStocks TOP-100 universe over the 10 year test period. In today's terms, the Top-100 represents shares with market capitalisation in excess of R3.0Bn or approximately the top 25% of the JSE. This means even fund managers with heaps of investment cash can participate in this strategy. As we saw from a previous research paper of ours "

__Size Counts__" this universe displays low volatility in portfolio constitution.

WHY 5 SHARES?

We conducted backtests on various sized Magic Formula portfolios over the 10 year period and amazingly, PowerShares-A delivers higher risk adjusted returns the

*smaller*the portfolios! The returns peaked at 4 shares and below 4 and above 5 shares started regressing. We felt 4 shares was a little too small, even if it had better returns and 5 is a nice round number.

The Top-100 universe is the only universe that displayed this characteristic. The mathematics behind how the Magic Formula ranking works on this universe typically means the top-ranked 4 or 5 shares have tiny combined ranking scores and above this the scores jump radically in size and are all bunched closely together in an ever increasing curve. The larger or smaller universes display less of this characteristic. Since the size of the combined rank score is indicative of the "cheapness and efficiency" of the share being ranked, this obviously has a significant bearing on overall returns and would explain the declining returns ratio the larger the portfolios (since larger portfolios are offering less "differentiation" between the shares). Another amazing characteristic is that as portfolio size increases, Sharpe decreases radically. This counter-intuitive characteristic is merely testament to the fact the system is an excellent picker of out-performing stocks.

TOTAL RETURNS

The performance delivered by various strategies back-tested with 5-share portfolios is depicted below. Returns include share appreciation only and not dividends (we cover dividends later.) The PowerShares-A, and its cousin, the traditional Magic Formula (PE+ROC) leave all other strategies in the dust. Remember this is a 10-year test period and a strategy must consistently out-perform to deliver this kind of gap (it is not attributable to one or two years of over-performance.)

In the 10-year investment period from 30 March 1999 to 30 March 2009, PowerShares grew some 1,680% or 33.4% compounded annually, versus an equivalent ALSI strategy that delivered only 216% growth, or 12.21% compounded annually. That is a massive 8 times out-performance. In fact if you look at the progress of some of the strategies depicted below over the test period, you can see the ALSI was well and truly trampled in the dust on a CONSISTENT basis.

We will show later that the incorporation of dividends into the back-tests, increases PowerShares' performance to 3,322% or 42% CAGR, but for now we keep focussing on capital appreciation/returns only.

RESILIENCE IN BIG BEAR MARKETS

Now the 2008 test period covers 30 March 2008 through to 30 March 2009 and you can see that the ALSI fell some 33% in this period. The ONLY two strategies tested that NEVER WENT DOWN in this latest vicious bear market were PE+ROC (Magic Formula) and PowerShares, which both went UP 20% These two strategies never faltered ONE BIT in their upward trendlines. It's almost as if the bear market never existed for them! The same happened in the big 2002 bear market (you may have to squint a bit in the above picture, but its there to see!) where the ALSI lost 29% and Powershares went UP 15%.

The diagram below shows how Powershares-A fared against the ALSI in the last 2008-2009 test period that covered the recent ugly market crash. The shares in the portfolio were TRU-Truworths , TBS-Tiger Brands, LEW-Lewis Stores, MPC-Mr Price and FOS-Foschini Group. The 3 months from inception to JUL, PowerShares dipped 18% and was under-performing the ALSI. As with the 2002 bear market, the inception date in this period was rather unfortunate - 2 months before the peak of the bull run came to an end - generally accepted as the worst time to be getting into the market! But PowerShares ensures you NEVER OVERPAY for stocks and even at the worst possible time to be buying, the strategy made handsome profits. Imagine the investors that never stuck by the strategy and bailed in Jul as they lost their nerve and faith. Bad news was coming out the press, sub-prime blisters were bursting and the strategy seemed to be dropping faster than the JSE itself. But after JUL, the JSE kept going down for a hefty overall loss of 43% whilst PowerShares rose 45% from JUL, after the investing public cottoned on to the value of the underlying QUALITY shares.

The identical scenario played out in the big May 2002 market crash. The poor PowerShares portfolio got in at the top on 29 March 2002 and spent 10 out of 12 months of its existence in an ugly protracted 29% decline in the JSE. The portfolio shrugged off the doom and gloom and went on to post a 14% rise for the year! The stocks in that 2002 portfolio were IMP-Impala Platinum, MVG-Mvelavanda Group, TBS-Tiger Brands, ASR-ASSORE and MSM-Massmart.

This is an important lesson, as generally the PowerShares are

*neglected or oversold quality high dividend-yielding stocks*and it can take 2-3 months for them to revert back to true value after portfolio inception. In our backtests, the underlying shares of the portfolio dropped by 5-20% in the first 3 months, in 6 out of the 10 years, but then powered up strongly in the last 75% of the year. In 4 out of the 10 years the portfolios powered up strongly straight after inception and then growth slowed in the 2nd half. This is obviously affected by portfolio inception date and timing of general market events but we advice AGAINST trying to time the market. Pick your inception date, see what shares qualify on that date and just go straight in as the qualifying shares have the best value-to-quality ratio on the JSE at that time and you are likely to erode overall gains (and miss the high dividend yields) trying to time when the market realises they are undervalued. If you feel compelled to set stop-losses then set them at 20%. Alternatively use the PowerShares Dollar-Cost Averaging investment method to spread your investment over 3-4 separate PowerShares portfolios spread 2-3 months apart to reduce risk (more on this in later research notes)

If you want to see a real live testament to what a juggernaut this strategy is in bear markets, then go and look at the live performance of our MF-5SH-1 created on 20 Nov 2008. This is a cousin of PowerShares-A, basically a Magic Formula strategy on the JSE TOP-40 universe, and it has delivered a whopping 60% plus CAGR since inception in the Nov 20 trough. We created a sister, MF-5SH-2 (we are creating one sister every quarter to measure real-world performance across a range of inception dates) in March and it's also steam-rolling ahead as if no bear market exists. You can track these two Top-40 strategies at the Magic Formula Blog or the Strategy ScoreBoard in the main menu.

WHAT ABOUT BASE RATES (WIN RATES?)

PowerShares-A out-performed the ALSI 8 out of 10 years (80% win rate). A high win rate of at least 67% is important for a strategy as it does not tempt the investor to make an emotional decision and switch out of the strategy. Switching is the main cause of under-performance by most private and even many professional fund managers. The two years that PowerShares-A underperformed the ALSI, were 2004 (it delivered 16% versus the ALSI's 23%) and 2006 (it delivered 18% versus the ALSI's 32%) These were hardly calamities and were not adjacent periods, so chances are a wise investor would have held his ground. The win-rates of the various strategies over the 10 year test period are below:

AND WHAT ABOUT VOLATILITY?

We measure two volatility numbers at PowerStocks, namely Portfolio (or Share) Volatility (the volatility of the growth of the various shares that make up the portfolio) and Annual Growth Volatility (the volatility of the annual growth rate of the portfolio as a whole over a given test period.) The volatility, represented by a Standard Deviation statistical function is divided into the average growth of the portfolio to derive a Sharpe ratio for each of Share and Annual growth to give a risk adjusted return measurement.

Portfolio, or Share Volatility of itself is not necessarily a bad thing. You won't mind if a portfolio is 200% up because a few shares in the portfolio completely sky-rocketed, thus driving Share Volatility up. On the other hand if growth of the whole portfolio is strongly up and Share Volatility is very low (leading to a high Sharpe ratio or high risk adjusted return), then you're going to be REALLY happy as this is an indication of the strategies ability to consistently pick winners! Share Volatility is a good correlation measurement for a strategies winning picking characteristics.

Annual Growth Volatility on the other hand should be low. We cant have a strategy swinging wildly from pillar to post from year to year. These sorts of gut wrenching ups and downs will scare of the investor, especially after 1 or 2 good "down" years, EVEN IF he is up on the whole. Annual year-on-year growth must be consistent and preferably the annual growth of a strategy must exceed its annual standard deviation, leading to a Sharpe of > 1.

The Share Volatility of the various strategies are depicted below. PowerShares exhibits a very acceptable average Portfolio Sharpe of 1.17 (most strategies are below 1). In 6 out of the 10 years, PowerShares exhibited a Portfolio Sharpe of >1

The Annual Growth Volatility study for the various strategies we tested is below. PowerShares and the Magic Formula produced stunning Sharpe's of 1.72 and 1.6 respectively. An average annual growth rate of 35% over 10 years against a backdrop of a mere 20% standard deviation of growth is every investors' dream, as most investors are content with 0.9 to 1.

DETAILS OF THE TEST PERIOD

The annual growths achieved by each strategy in the backtests are shown below:

The total returns from each portfolio assuming an initial R1 investment on 30 March 1999 and annual rebalancing and re-investment are shown below. The below table uses the growth rates in the previous table to compute values. Note that dividends are excluded from calculations, as are brokers fees.

WHAT ABOUT DIVIDENDS?

Most international research on various index strategies do not take dividends into account. For the tests we have conducted above, no amount of dividends would allow any of the competing strategies to catch up with PowerShares and PE+ROC, so strictly speaking we don't need to include dividends to arrive at the conclusion as to which is the superior strategy.

But PowerShares-A by its nature selects cheap stocks (low PE) with high productivity (high ROC). A by-product of the low PE is high Dividend Yields (since most "cheap" stocks generally have high dividend yields). A by-product of a high ROC is..you guessed it..high dividends! So we simply could not resist the temptation to incorporate dividends into the above backtest for PowerShares, since we suspected they could significantly turbo-charge returns.

This is not a task that is easily automated and so we had to spend a few days manually sifting through the annual results of 50 shares over the last 10 years to calculate all the various dividends, special dividends and share premiums. Most shares made two payouts per year, namely the interim and the final dividend. We calculated the cash value of said payouts for each years' portfolios and used these proceeds in the end-of-year rebalancing exercise to repurchase shares for the following years' portfolio. The results are below:

We see stunning affirmation of the ability of PowerShares to seek out stocks that provide for capital appreciation as well as high dividend returns. When incorporating dividends into the total return calculation we see that the R1 that became a "respectable" R17.80 through share price appreciation gets boosted to an incredible R33.27 with annually re-invested dividends! That equates to a 3,227% accumulated return, or 42% compound growth.

The average dividend payout (finals+interims+specials) in the 12-month life span of a PowerShares R50,000 portfolio was R4,276 or 8.55% yield on the initial investment. In 2005, midway through the bull market, profits were bursting at the seams of these high ROC companies and all manner of special dividends by management to shed excess cash

*coupled with the low price*we paid for the shares boosted the yield to a stunning 21%!

The compounding effect of re-investing the dividends is clear to see in the total returns chart above, but the chart below shows just how more important a role dividends played in total returns as time marched on, ending up contributing 45% of total returns over the 10 year period. This becomes an important factor later on in terms of the tax-efficiency of PowerShares, since dividends are tax-free. We will discuss taxes later on.

WHY 12 MONTHS?

Currently the best total return is achieved with annual rebalancing. If you do not like paying tax to the tax-man, you could conceivably only rebalance every 3 years as our tests show that 3 year portfolios also perform well, but are sub-par to 12 month portfolios (we have scheduled detailed backtests for 3 year "tax efficient" PowerShares portfolios, but this will take time). Also, we showed that re-invested dividends (which are tax-free in your hands) made up 40% of the returns over 10 years, so the capital gains tax issue only affects 60% of the returns.

Also, 3 year PowerShares portfolios are not as immune to bear markets as the 12 month portfolios, as after 2 years the PowerShares stocks become "popular" or "glamour" issues that have been "bid up" considerably and even though you paid a cheap price in year one, when the market turns down the older PowerShares get punished due to their high PE's and market-caps (read "Anatomy of a Crash" to see which shares got punished the most in the last correction). For now, we suggest you take your average 35% capital appreciation, pay your taxes and re-invest your dividends on rebalancing - you will still be knocking the socks off the ALSI!

Update : Our research into tax implications and the effect on annualy rebalanced portfolios is nearly complete. Our examination of the SARS guidelines with regard to capital gains indicates to us there may be an option to avoid taxes on the re-invested capital gains. We are in consultation with tax experts on this issue and will advise accordingly, but this is merely a bonus - do not let this hold you up as even with taxes the PowerShares performance is mind boggling.

I WANT TO BUILD A PORTFOLIO NOW!

We will publish monthly lists of the 5 shares that qualify for inclusion in the PowerShares-A portfolio (starting with PSA-MAR09 commencing 30 March 2009, basically starting where the backtests stopped) over at the Magic Formula Blog.

We will also be tracking live performance of the portfolios we create each month to stress test the strategy over 2009/2010. Dividends will be incorporated as they are paid to show total returns. You need to be a PowerStocks Research subscriber to access the Magic Formula Blog. You can read about our services and prices and how to apply for a membership account over HERE.

Optimum portfolio sizes are R60,000 (R12,000 per share) to minimise brokerage charges of 0.6%, but nothing stops you starting smaller. We do not recommend anything smaller than R30,000. If you want to spread disposable income over time in your investment, then if your portfolio is older than 2-3 months, you are advised to create

*new portfolios*rather, to avoid over-paying for shares. This is a very good strategy called "Dollar cost averaging" which we cover in detail later, as it is actually shown to deliver far higher returns than investing in an older portfolio that may now have expensive shares in it, and uses the condition of the market to your favour (when the market is expensive you buy less and when its cheap you buy more.)