Visit our sponsor.
The question of past performance and it's use in making a fund selection is an area of great interest to all investors and fund managers, and as a result, there is a lot of information, but getting to the truth can be harder. The purpose of this article is to explore some of the underlying statistics, ( pictures only, no equations ! ), and the methods used by marketers to show their funds in the best light. It will not tell you which fund to invest in, but will help you make a rational decision with regard to your investment.
It is assumed that you have selected your overall approach, ( how much in cash, managed , equities, speculative etc. ), and simply seek to establish which funds within each sector to invest in. There is no argument as to the relative merits and demerits of different sectors. The real debate is whether or not there is any way to choose between similar funds that has any advantages over throwing darts at a list.
The fundamental question is " Does Past Performance act as a guide to FUTURE Performance?" The standard answer is that " past performance is not a guide to future performance", but this is often found at the base of a graph showing the amazing skill of the fund manager over the last decade or two. In other words, " our past performance is great, and we'd like you to invest in the belief that we will maintain it, but we've given you the formal warning so there is no come back if we fail".
Can the idea that past performance is a guide to the future be easily tested using the information provided by most surveys and companies? If it can be tested, and shows that there is a relationship, can that relationship be used to make money?
To explore this issue needs a little statistics, ( statisticians please bear with me, I am simplifying for clarity ). The graph below is known as a Bell Shaped Curve, ( for obvious reasons), or as a Normal Distribution. It appears when the criteria being measured is distributed randomly amongst the population being measured, but around an average. It could appear if you plotted the shoe size, or height, or lifespan of a few thousand randomly selected people, and took no other factors into account. The peak is the average.
General graph notes. These graphs are schematic in the interests of small file size and clarity.
Graph Notes. Normal Distribution Curve. Peak is the average, and the sharpness shows how close to the average most people are. The tails at low and high show that there is a broad spread of real results. If it was shoe size and 99% of people were size 7, then it would be a very thin spike. If peoples shoe size was evenly spread from 1-14 then the graph would be an horizontal line.
It defines the population, and can have other factors tested against it. For instance you might think that shoe size and height could be related. If you took a lot of people of a fixed height, say 6'2'', and plotted their shoe size, you would get another Bell Shaped Curve, and if it was different, perhaps with the peak at size 10, then you would have evidence that tall people tended to have bigger feet. No surprise there. But what about plotting against lifespan? This might show no difference at all, indicating that height did not have any impact on lifespan. ( In fact a plot of 6'2" people against lifespan would show a curve with a peak at a lower age than that for the population as a whole. Why? It could be sign that being tall is bad for your health. In fact it is simply because most people who are 6'2'' are men, and men die younger than women. Which just goes to show how careful you have to be when looking at statistics).
Graph notes. The two curves are not to the same scale. The General Population curve peaks later than the tall people curve, and this shows that, on average, tall people die younger, but the tails show that there is a lot of overlap and not all tall people die young.
This is the time to cover two of the other main misconceptions that surround statistics, applying them to the individual, and why a correlation is not evidence of cause.
First of the correlation question. There is a statistical correlation between smoking and early death. But this is not evidence that smoking causes early death, only additional studies have provided that concrete evidence of the link, ( although it was the statistics that rang the alarm bells). Statistical correlation's may be useful signals, but must be treated with care, and are not normally taken seriously unless a plausible hypothesis to account for the effect can be found.
Next, statistics say nothing useful about individuals.They only apply to groups. For example it is accepted that smoking is dangerous, and the life insurance premiums for smokers are higher than for non smokers. However this does not tell you that, if you smoke, you will die before your time, only that you are more likely to. Hence the tendency for people facing unpleasant facts, such as the fact that they smoke, to point to Aunt Flo who smoked like a chimney and was killed by a bus at the age of 88, so how about that then? The fact that there is a high proportion of early deaths among their smoking friends is overlooked.
Relating this to Past Performance is quite simple in theory, but not so easy in practice. You need to plot the performance of all the funds in a sector and see what appears. ( In practice sectors are so broad that a much tighter basis of selection would need to be used. The aim must be to compare like with like, and care is therefore needed in selecting the right groups of funds).
A Bell Shaped Curve , with the peak matching the average performance of the sector would imply that the position of any one fund is pretty well random, and that, all other things being equal, it is impossible to use Past Performance as a guide to future performance. The fact that a fund plots at point A is irrelevant. Next year it might plot at B.
Graph Notes. The graph plotted from real data is a Normal Distribution, therefore place on the graph is random, so the past is not a guide to the future.
If the result is anything other than a Bell Shaped Curve then you would know that there was something interesting going on. With luck you will be on the track of skilled managers, ( but in practice it might simply indicate that your data selection was poorly judged and that you were not comparing like with like). Unfortunately this will not help you because there are major problems with the interpretation of such data.
First of all the good performers will be a combination of managers there by chance and those who are there by skill. Since all fund managers believe that they know what they are doing, ( if they don't, they go Tracking ), it is no good asking them " is this due to chance, or do you have something valuable here ". The other key complication is that systems that may have served the manager well in the past may not continue to work in a changed environment in the future. ( The obvious way to overcome this is to only deal with managers with a 20 year record, which implies that they have known how to do it since age 25, and are now 45 plus and still an actual manager....)
This point becomes more obvious if you imagine that there is a great debate about how to ensure a long life, and one side says "Smoke , it kills all the germs in your lungs", and the other side says " Don't smoke, it clogs up your lungs". Imagine that no one actually knows the real danger of smoking. You are wondering whether or not to smoke, so you go to a lot of old people, and ask them how they have kept their health. Some will swear that this is due to their smoking, and others will tell you that is because they don't smoke. Without additional information it is impossible to select the correct approach. You face the same problem with fund managers. Because they believe that they have the answer their information is useless.
Even if there was clear evidence that a certain technique had worked, ( i.e. you could not find any old smokers, so non smoking appeared to be the answer), when applied to fund management the situation is not as clear because even if a system is effective for a time, it may not be effective in the future.
For instance one could imagine a situation in a boom where a number of managers may weight towards smaller manufacturing companies, and a study of techniques shows that their performance plot is as below, with an average better than the sector. This would be evidence that such an investment profile was effective in outperforming the market. ( Actually it is a correlation between investment approach and performance, but the link is causally plausible, and so can be considered evidence. Apologies for splitting hairs, but care is needed). To continue in the Cassandra mode, it doesn't take too much imagination to predict the outcome for anyone who bought into such a fund, on the basis of such research, just before a recession. That fact that you have a correlation, an explanation for the correlation, and proof therefore of management ability, does not mean that it will be wise to invest.
Graph notes. Both populations show a normal distribution, but one plots around a higher average performance than the main group. If you could be sure that your fund manager was in this small group then you would be onto a winner. The problem is to decide if your manager is in this group, and, if he is, will the modus operandi of the group continue to work in the future?
Have these curves been plotted? I don't know. But I am sure that if they have, and they provided solid evidence for the superior expertise of certain companies,management groups, or systems, we would have been bombarded with them by the companies concerned, ( who would leap on it as evidence that their past performance DOES indicate future performance). So I suspect that past performance, in the form normally shown, does not provide a statistically useful indicator of future performance.
The next area of interest is to assess how , and why, companies actually present their past performance in an effort to persuade you to invest. All of their efforts can be split into two types of appeal. The first is to say " look at our past performance", the second is to say " this is our strategy, you know it makes sense". Firms often combine both styles.
In this article we are not commenting on the " our strategy " approach, but will look at the methods by which past performance can be displayed, and the care with which such displays must be interpreted.
Rule number one is to beware data and presentations put together by fund managers or company marketers. It is the job of the marketing department to put the best gloss on whatever information they have. To help them in this they use statistics. The statistics themselves will be accurate, but their impact depends not so much on the numbers themselves, as the presentation. Effective presentation can be, ( for companies with little to be proud of ), primarily a matter of careful selection of data and the omission of unhelpful data.
Most information is presented as simple numbers or as graphs. The numerical approach is normally of the " Best fund in sector over past year" or " 12%pa for last ten years" type. It is usually a flag waving exercise and not to be taken too seriously, though it might indicate something worth following up. The graphical approach is more important as it is meant to be taken seriously, and often all the information available is presented as one form of graph or another.
Graphs are wonderful for displaying data in an easy to understand manner, but when you know that the compiler of a graph has an axe to grind you have to be cautious. In many cases it is the information that is left out of the graph may be as important, or more important, than the information in the graph itself. Always ask yourself what they are trying to show, and what parts of the picture might be hidden.
For instance imagine two companies, Biglife and Nice Assurance, who are competing for business on the basis of past performance. Both make the claim " Our Managed Fund has done better than theirs, and we have the graphs to prove it". Common sense would suggest that one is better than the other, but in fact things are more complex than that.
Biglife
Nice Assurance
In actual fact only the terms are different. Multilife have taken performance from t1 ( perhaps past 10 years ) whereas Nice Assurance think that only the past 5 years should be considered (t2). Their graphs also indicate that marketers always follow the sage's advice to " choose your enemies with care ", as the graph below shows. Multilife can be thought of as a smaller company than the other two, and therefore will be ignored by them. If you ask questions you will probably be told that " Multilife are no doubt a fine company, but not really in our league". Absolutely.
The careful selection of start point and competitors ensures that any marketer worth his pension can make almost any fund look good, or at least reasonable. Of course there might be a total wipe-out. You might be the man with the worst performing fund since the Great Depression. In this case it is time to think laterally. Publish the information for a different fund, ( say a much more respectable Equity Fund ) in the hope that investors will assume that a good equity fund implies a good Managed or With Profits fund.
Given the growing tendency for every company to offer an ever increasing range of funds, there is plenty of scope for creative graphing. Everyone can show a success.
The next cause for concern is the trick of comparing a fund against its index, ( FTSE and the like ). At first glance this seems foolproof, but again care must be taken. The first, rarely answered , question is how relevant is the index to the fund? The closer the fund is to the index, the less it can depart, ( either in out or underperforming ), from it. In other words any fund that clearly outperforms the relevant index probably has little in common with it, making the comparison dubious at best. It does not mean that the managers are better than their opponents, only that their approach is different. To try to compare them is to try and compare apples and pears.
The final technique used by marketers is the " once upon a time " story. It runs like this, " Once upon a time in 1973 a far sighted investor put £10,000 in our Magic Beanstalk Fund, and now it has grown into a fortune far beyond his wildest dreams. If you want to be rich, give us your money now". There are normally two key points to these stories, the first is that the date of initial investment often represents a time of panic or recession, and the second is that in many cases the big gains occurred as the market lifted out of recession. Such adverts simply highlight the wisdom of , in the words of Kipling," keeping your head when all around are losing theirs" , and investing in low markets, but should be treated cautiously when considering investment. ( If the graph plots against an index trace them and rezero to indicate the effect of investing in , say 1975 or 1986. Is it still so impressive ?)
Having examined the statistics of fund performance, and raised questions about the information provided, how can you use this to try and select the right fund.
This is not an easy area, but the following guidelines offer as good guide as any:-
1) Ensure that you understand how the managers are running the fund. Are they going to follow the index; are they going to try and second guess the economy by buying stocks ahead of the cycle; are they going to concentrate on big companies, or do they like smaller companies; will they trade actively, or do they buy and hold? Whatever their approach, does it make sense to you? Do you like it? If not, it is probably best to avoid it, no matter how good it sounds.
2) Ask if these criteria are recent, or have they been in place for a long time.
3) Has the fund price been volatile ( A ) or more stable ( B ). ( Prefer stability unless short term volatility shows long term performance, in which case consider how much downside you can live with). In the tables at the back of this magazine a consistent position is better than fluctuations between top and bottom quartiles. There is an argument that states that any fund which was run by a truly gifted manager, who could produce long term better than average performance, would show lower than average, ( for it's sector ), volatility. According to some sources there is empirical evidence that approximately 60 such managers exist in the UK Unit Trust industry, but more work needs to be done. ( Personally I'm waiting for the graphs that show non Normal Distribution before accepting this as proven).
4) Look at more recent years performance in preference to older years. In graphical terms (a) is better than (b).
So to return to the original question, it does not appear that past performance is a guide to future performance, and that systems that show an advantage will tend to fail after a while due to changing market conditions. It is also clear that many sources of information are biased, and that neutral sources, ( such as simple fund data in magazines or software packages ), need a little care in their interpretation.
At the end of the day the use of past performance is a triumph of hope over experience. It is probably better to simply recognise that whatever fund you invest in, it will probably give a return typical of it's type and sector, and most of the research effort should be aimed at establishing that the aims and objectives of the fund are suitable, perhaps using past performance to help in the final decision. After all , if two funds are otherwise equal the better performing one does at least offer the hope that it has a little hidden talent!