Skip to main content
rss feedour twitterour facebook page linkdin
If only investing had a 'system' - Eureka Report article

If only investing had a ‘system’

By Scott Francis
February 4, 2011

PORTFOLIO POINT: No matter how data is tortured, there is still no reliable way of predicting market tops and bottoms.

The Australian sharemarket has gone through four distinct phases in the past five years: an extended two-and-a-half year rally of about 69%, followed by a sharp 55% selloff over 14 months, before recovering with another year long rally of 63%. The fourth and final period has been one of largely inaction.

Given the magnitude of recent falls and the extraordinary rebound that followed, it is not hard to see the attraction in the idea of timing the market – or the strategy of buying shares at the bottom of the market before selling them at the top.

However, there is very little evidence to support the theory. Given the fame and fortune that awaits the first person to produce a reliable strategy that provides above-average sharemarket returns, this is unusual. The average returns from the Australian sharemarket, to November 2010, have been 8% over 10 years, 9.5% over 15 years and 10.8% over 20 years.

The first column from table below shows what $100,000 invested into the index would return over the same periods; the second column shows the end result of a mere 2% outperformance each year over the same period. Over 20 years, outperformance of this kind is the difference after tax is between $778,000 and $1.12 million.

-A little outperformance goes a long way
$100,000 investment earning average market return
Average market return
plus 2% a year 
10 Years
15 Years
20 Years

The moral of the story is that if we can find a good market timing strategy, it can potentially make a huge difference to our investment balance. There are, however, a number of issues I have in tracking down this millionaire-making strategy.

Concerns about tax

Central to the strategy of timing the market is the buying and selling of your market positions, which produces additional costs including brokerage, but by far the biggest of these will be tax.

The 20-year average sharemarket return to the end of November 2010 has been 10.8% a year and can be captured very tax effectively by an investor through buying and holding any of the following:
  • A diversified portfolio of shares.
  • An index fund.
  • A listed investment company (LIC).
  • An exchange traded fund (ETF).
In most cases there should be almost no capital gains tax from year to year, and for the average investor on the 31.5% tax rate, the franking credits from Australian share income will all but wipe out any income tax – and for superannuation and pension fund investors franking credits will increase their returns.

But if you choose to time the market – even if you only buy and sell once in a year – you are suddenly faced with capital gains tax.

To model this, I have assumed that an investor has the ability to earn a return, using a skillful market timing system, of 13.8% a year – an impressive 3% a year above the average market return. I have assumed that 4.5% of this return is from dividends, with the remaining 9.3% each year in capital gains. Because there is trading in the portfolio, tax has to be paid on this gain. At the average tax rate of 31.5%, this equates to an annual tax impost of 2.93% a year.

What this means is that even if a market timing system adds 3% to returns a year, it is basically wiped out by the capital gains tax it generates – assuming that it trades at least once in every 12 month period. So, for an investor on the average tax rate, the market timing system has to be better than 3% a year just to overcome the tax generated in the scenario here.

That 3% a year is a significant tax cost for a portfolio. This is less so for a superannuation fund in accumulation mode where it falls to 1.4%, and it must be said it is not a factor at all for a superannuation fund in pension mode.

Research and development

The largest study of market timing systems that I am aware of was carried up by John Graham at the University of Utah and Campbell Harvey at Duke University (click here). This study looked at 237 market timing systems over a 12 year period in the US and found that few of them were in operation long enough to gather any meaningful performance data. The author found “no evidence that (market timing) newsletters can time the market”.

Another study entitled The Behaviour and Performance of Investment Newsletter Analysts (click here) found that only one-third of the 329 market timing strategies that they considered over a 21-year period beat the average market return. Here, the study only considered before tax data; it is reasonable to assume that after tax the number of market timing strategies outperforming the average market return would be sustainably reduced.

Statistics and returns

As investors consider the opportunities put before them, they should also judge the value of statistics presented as evidence.

Many stock-picking and market timing systems misuse statistics to provide “evidence” for the efficacy of their system. This is usually done by finding a pattern in historical sharemarket data to build a strategy of how an investor might generate above-average returns – from either picking shares or timing markets. Some of the share-picking systems, in particular, imply potentially spectacular returns of 50% a year or more.

However, having found the pattern sharemarket data, the promoter of the investment scheme then uses the same data to demonstrate above-average market returns.

This is a significant statistical error. A pattern that is found in one set of data should not be tested on the same data. Any set of historical data will show a pattern.

For example, even long-term samples of historical sharemarket data will have shares that begin with one letter with a higher performance than all the others. The question is whether this pattern will hold in “out-of-sample” testing – perhaps using different time periods or data from different markets. This out-of-sample testing is what will prove the relationship to be a robust one, or simply one of chance.

The Journal of Investment published an article on data mining in 2007 that looked for coincidental relationships in data – to show how important it is to use statistics carefully. By way of example, the article showed that there was a strong relationship between Bangladesh butter production and US sharemarket returns. This is a totally bogus relationship, but one that was able to be proved using a sample of data.

Many investments are promoted on the basis of “modelled returns”. Any modelled return has a gap from what is experienced in real life, even if allowances are made for transaction costs, buy/sell spreads and taxes.

If these portfolios use real money to demonstrate returns, in my opinion, the financial services industry would be a safer place if only actual achieved investment returns were allowed to be advertised – as opposed to “modelled” or “backtested” returns.

Timing the market is the Holy Grail of investment strategy, the chance to add additional returns while missing the worst market downturns is an attractive proposition. But when I look at the evidence, think through some of the issues related to modelling and consider the taxes payable from the average investor, I am not convinced that the reasonable long-term returns from investing in the market are worth risking with a strategy that remains largely unproven.