Don’t be fooled by selective use of information

A picture is worth a thousand lies, when it is a graph.

It seems that every fund manager’s presentation I have ever seen, over the decades, has included a graph that seemed to demonstrate its performance was superior.

Although this tempts me to conclude that the laws of mathematics do not apply to investment funds, and that they are all above average, there is an alternative explanation.

There are certain tricks-of-the-trade in drawing graphs that are widely practised to create any desired impression. Once you know them you can look at any graph and instantly recognise if it uses them. It is also possible to draw graphs differently to eliminate these distortions.

Graph 1 ‘proves’ that Maple-Brown Abbot (MBA) has better performance in Australian equities than Merrill Lynch (ML).

Graph 2 ‘proves’ the opposite. Together they prove that graphs can be quite misleading and can be used to support any case you like.

There is only one difference between the two graphs — their starting points. Graph 1 begins in March 1994 and Graph 2 in February 1996. By careful selection of start dates, it is possible to produce a misleading visual impression of either superior or inferior returns for almost any investment or asset class.

The key to recognising whether data manipulation is occurring is to look at the left-hand side of a graph. Does the investment being promoted immediately begin to produce a better return? If so, you are probably looking at a case of selectivity of start dates that has occurred in the two graphs used in this example.

As a topical and hotly debated example, take the relative, long-term performance of asset classes.

Graph 3 demonstrates that, from November 1980, listed property trusts have produced a far better return than shares. The All Ords did not catch the property index at any point over this period. Yet Graph 4 demonstrates almost the opposite from December 1982!

These two graphs have nearly the last 19 years in common and only differ in start date by approximately two years, yet produce startlingly different impressions. This highlights how significant starting point selection can be, even for long term data.

All of these graphs reveal their start date bias near their left-hand edge, with the preferred investment jumping to a quick performance lead.

How can start date bias be removed?

One way is to adopt the practice of drawing backwards graphs. These do not follow the usual form of showing the various investments with the same start point but rather with the same end point. A conventional graph represents the amount that an initial investment of, say, $1,000 might be worth at various points in time.

A backwards graph shows how much would need to be invested at various points in time to have an investment worth $1,000 today.

Naturally it is better to be able to invest a smaller amount to achieve the same result.

Thus, at any point in time, the superior performing investment will be nearer the bottom of the graph and rising more steeply to the final date.

This is the opposite of a conventional graph where being near the top represents the best return.

Graph 5 is a backwards graph comparing MBA and ML since the launch of the latter in January 1993.

You can see that each has had periods of being the better performer as each has periods of being lower down on the graph. MBA has been best from any starting point since around the middle of 1998. ML was better from any starting point from around the middle of 1994 to the middle of 1997. Prior to this MBA again held the edge. This gives an unbiased representation of their relative returns.

Graph 6 is a backwards graph of property versus shares since the start of 1980 (this date chosen as the earliest Morningstar data for both). It highlights that the returns from the two have been reasonably close over many time frames.

As it happens shares have been better, that is, are lower in the graph, over only 17 per cent of these time frames (which are all periods from one month to 21 years and nine months) and property better over 83 per cent of the times. Most of the start dates from which shares did best were from early 1982 to early 1985.

Graph 7 is a backwards graph of bonds versus shares also since January 1, 1980. Ten-year bonds have produced a superior return to equities over 82 per cent of these time frames, which may surprise many people. Several factors should be noted in regard to this.

The period reported has seen a strong decline in interest rates due to the demise of inflation. As the 10-year bond index has a much longer maturity profile than a typical bond fund, it will usually produce a superior return whenever rates fall. Thus most bond fund investors will not have achieved results as spectacular as this index suggests.

Further, it is highly improbable that bonds can continue to produce the returns they have shown over recent decades as this would require rates to fall to near zero.

This highlights the validity of warnings that past performance, even over long time frames, does not guarantee future performance.

At the same time it is relevant for the debate about diversification versus 100 per cent equity portfolios, that shares have only produced a better return than both property and bonds over five per cent of timeframes since January 1, 1980.

It is hard to see evidence in this that no-one should ever have anything other than equities in long term portfolios.

Equally, I am not suggesting that the relativity of performance of these assets will be repeated over the next decade or so.

Backwards graphs are effective at preventing product promoters from presenting a misleading view of the relative performance of their products. Consequently, they may not become the accepted industry practice.

Thus it will remain necessary to scrutinise graphs carefully. A very interesting question to ask is: “Why did you pick that particular start date?”