If I choose a fund with a solid track record, what are the chances that it would deliver in future? Does past performance matter? This is a question on the minds of most mutual fund investors. Let us try and answer this question using annual returns of all diversified equity funds from Value Research.
This post was first published on April 10th 2015. I have now included updated results. The text in brown has been added now.
Value Research provides annual returns for the last 12 years (as of this moment. It can be temperamental!). So I have calculated the CAGR of the first 6 years and plotted it against the CAGR of the next 6 years.
I recognize the results of such analysis could depend on the duration chosen (and it is!) . It can also depend on market movements.
Results based on annual returns from 2003-2014
Notice that the data points slope upwards from left to right. So the correlation between past and future is non-zero.
The rectangle represents all funds within one standard deviation, either side of the average for the first 6 years (length of the rectangle) and for the next 6 (breadth of the rectangle).
A good 48% of funds are inside the box (34 out of 71 funds). That is, if the first 6 year CAGR was within one standard deviation from the average, there is a 48% chance that the next 6 year CAGR will also be within one standard deviation from the average.
Or speaking in terms of star ratings, using VR online's methodology, a fund rated 3-star or 4-star after the first 6 years has 48% chances of being rated a 3-star or 4-star fund after the next 6 years.
It is also seen that 1-star or 2-star based on first 6 years performance ( points to the left of the rectangle) has a high probability of being rated the same after the next 6 years.
This means that past performance does have a significant influence on future performance for the data set studied.
The problem is only with 5-star funds. A fund rate 5-star after the first 6 years can just about be rated anything after the next 6-years.
Results based on annual returns from 2004-2015
Now 48 out of 95 funds are within the rectangle. That is ~ 50%. So the general conclusions made above are still valid. In fact, in this window some of the top performers over the first six years (to left of the box) have continued to do well in the next six.
Many average performent in the first six (funds above the box) have gone on to become top performents in the next six.
However, the boundaries of the box have changed greatly!
2003-2014
first six years: 18% - 31%
next six years: 18% -26%
They had a common lower bound which is purely accidental.
2004-2015
first six years: 16% - 27%
next six years: 7% -13% (that is huge downward shift!)
Yet another example of how volatile equity can be! Nonetheless, the central observation from this analysis is,
Funds with a 'reasonable' history of past performance have a pretty decent shot at reproducing that in future. Reasonable here refers to within one standard deviation of the average.
Or in other words, average performers have at least a 50% chance of remaining an average performer in future. Seems obvious when backed with some data! Now, this sounds terrible! However, for investors who want minimum maintenance portfolios, this is good news. While screening for mutual funds, they can cast a wide net (eg. average and above performance)
Investors who crave to be invested in the 'best' funds, will have to churn, perhaps frequently, in order to satisfy their craving.
Corollary: Expecting less from equity allows me (and Ashal Jauhari) to invest calmly with peace of mind and more importantly lower (not zero) portfolio management.
Read more: Understanding the nature of stock market returns
Register for the Hyderabad DIY Investor Workshop Nov 27th 2016
Check out the latest mutual fund returns listing
Buy our New Book!You Can Be Rich With Goal-based Investing A book by P V Subramanyam (subramoney.com) & M Pattabiraman. Hard bound. Price: Rs. 399/- (Rs. 359/- at Amazon.in). Read more about the book and pre-order now! |
Great article Pattu. I think your summary "reasonable past returns has a good chance of reasonable future returns" is something every investor should take to heart. The problem arises when "unreasonable returns" are chased and the basis of selecting funds is gut feel rather than rational analysis.
"Investors who crave to be invested in the ‘best’ funds, will have to churn, perhaps frequently, in order to satisfy their craving, for the period studied."
This reminded of scripbox and lo and behold, Sanjiv is the first one to comment 🙂
It will be interesting how scripbox accommodates/helps navigates taking above into consideration..
Thanks Sanjiv. Sorry for the late response.
Great article Pattu. I think your summary "reasonable past returns has a good chance of reasonable future returns" is something every investor should take to heart. The problem arises when "unreasonable returns" are chased and the basis of selecting funds is gut feel rather than rational analysis.
"Investors who crave to be invested in the ‘best’ funds, will have to churn, perhaps frequently, in order to satisfy their craving, for the period studied."
This reminded of scripbox and lo and behold, Sanjiv is the first one to comment 🙂
It will be interesting how scripbox accommodates/helps navigates taking above into consideration..
Thanks Sanjiv. Sorry for the late response.
In my experience mutual funds go up and down in a cycle pattern.Buying and selling pattern, stock selection, addition deletion, stock concentration diversification, political situation ,budget taxation, rules regulations, foreign investments social and health situation, managers, business pattern is changing everyday. Change in return is bound to occur.
It is very difficult to predict returns of mutual funds. Slight correlation 10-30% may not be significant in long run because up and high levels also influence future returns. One should be ready for change and progress in future.
Agreed. Change is certain. Investors need not act in response to each change.
In my experience mutual funds go up and down in a cycle pattern.Buying and selling pattern, stock selection, addition deletion, stock concentration diversification, political situation ,budget taxation, rules regulations, foreign investments social and health situation, managers, business pattern is changing everyday. Change in return is bound to occur.
It is very difficult to predict returns of mutual funds. Slight correlation 10-30% may not be significant in long run because up and high levels also influence future returns. One should be ready for change and progress in future.
Agreed. Change is certain. Investors need not act in response to each change.
Past performance does matter. The investor has to know the type of mutual fund. Some mutual fund families concentrate on long term performance, 10 to 20 years and beet the benchmarks over that period. Some mutual funds concentrate on momentum, trading or other short term strategies. The performance of such funds is more inconsistent.
Madhu Tahiliani
Past performance does matter. The investor has to know the type of mutual fund. Some mutual fund families concentrate on long term performance, 10 to 20 years and beet the benchmarks over that period. Some mutual funds concentrate on momentum, trading or other short term strategies. The performance of such funds is more inconsistent.
Madhu Tahiliani
Hi,
It looks like one of your post is published in Economic Times.
http://economictimes.indiatimes.com/why-investing-is-not-only-about-equity/editionlist/edition-633,artid-51955474.cms?intenttarget=no&utm_source=newsletter&utm_medium=email&utm_campaign=ETwealth&type=wealth&ncode=2fbce2a8298d4718b7c2f7a6726f3f0d
Yes, thanks.
What value did the =correl() formula give us for the two sets of arrays? Statistically speaking, if correlation coefficient isn't really upwards of 0.7 then for the data set chosen, it might not be the right thing to call it a positive correlation.
There would always be certain data points, if selected in silo, that might show a trend but correlation makes sense for the entire sample that one chooses, not really selecting a sample from a sample after a result is derived.
What are your thoughts?
The correlation is about 33%. Positive correlation is any value above 0. A value of 0.33 is, imo not bad at all.
Statistically speaking, a 0.33 value doesn't really qualify for a strong positive correlation i.e. something to base one's next action on. I'm a Six Sigma Black Belt hence was interested in knowing the exact value.
There are so many variables in play when it comes to Mutual Funds, that positive & negative correlations might be totally misleading. Hence, we shouldn't really bother with any value that is between -0.7 to 0.7
Not to mention, you would rarely end up with an absolute zero when calculating correlation coefficient between any two data sets which is why it is extremely important to note how to use the correlation coefficient and take the next step. Values such as 0.1 or 0.3 or 0.5 don't really mean anything, statistically speaking.
To add, values above 0.9 and below -0.9 is actually what makes such correlations really interesting.
0.7 to 0.9 is the window to get curious, 0.9 to 1 is the window when you get that wide grin on your face (same for negatives). For everything else, poker face as it doesn't really make any sense 🙂
Those are opinions and I have no comment on that. With reg. to this post,
1) It had a bit of history which has been edited out when I re-posted
2) The conclusions arrived at work well enough 0.33. The correl uses the sq of the std dev while I have used the std dev. I am comfortable with the no of funds within the stdev window to standby my conclusions.
I am not looking for a "strong correlation". My investment philosophy is a lot more generous and I am happy with 0.33 or rather 0.48 std dev window.
In my opinion obviously! past performance always matters especially when you have to invest your money in a mutual or hedge fund.
To interpret the performance of an investment fund its important to evaluate its history. Only those funds will get attraction of the investors whom are performing as per the expectation.
I think if one is blind to the 'details' of the data and is treating it with no knowledge of it, the r ~ 0.3 would be considered weak positive correlation and a significance test might be needed.
However, if one is very knowledgeable about the data, like it is in this case, it does make sense to draw from the knowledge-base to make meaningful conclusions even on weakly correlated systems.
Wonderful article. Will have to analyze it deeper...