Every mutual fund advisor has a list of best mutual funds.
The greedy ones will simply sell you the funds that offer highest commissions.
Honest (but not so sophisticated) ones will use a rating website to recommend funds.
Sophisticated ones might put in some research to back their recommendations.
But one thing is common across the spectrum – no one will tell you about the “methodology” used to arrive at the recommendations.
At best, you will get to hear statements like – “this is a 5-star rated fund and has delivered 19% return in last 5 years” or “our model suggests that this fund has been a top performer for the last 3 years”.
We, on our part, realise that it is important for our investors to know how we select mutual funds. Going by that philosophy, we came up with a framework for selecting the best mutual funds and put it up here – Looking for the Best Mutual Funds to Invest? It’s a bit trickier than you think!
It’s been more than a year since we launched this framework. Now, as we enter 2019, we are launching a much more sophisticated version of our previous framework.
And we have a name for it as well!
It’s called CRAFT – Categorize, Rank and Filter Framework.
Please note that in this blog, we will be focusing more on the framework rather than actual rankings. Having said that, we will go through some actual examples too.
For those who are impatient and want to jump directly to the results, you can check out this page – Best Mutual Funds to Invest in India in 2019
And for those of you who are up for this long and technical blog, let’s get started!
CRAFT framework for selecting the best mutual funds
Here is a quick snapshot of the CRAFT framework:
As evident from the picture above, CRAFT is a 3-step framework.
The first step is to categorize the funds into different categories based on their type. For example, large cap is one such category.
Once we have the categories, the next step is to rank the funds within each category. Ranking should be done for funds within a category and NOT across categories.
And finally, once we have ranked the funds, we should filter the funds based on qualitative assessment.
Everything appears to be a complete black box right now? Don’t worry! We will cover each step in detail.
The idea was to give you a quick flavour of the framework before we move on to the real slog.
And now we are ready.
Step 1: Categorize the funds into type buckets
The main idea here is that you want to compare funds within a category and NOT across categories. Something on the lines of making apple-to-apple comparison.
For example, a large cap fund will have a completely different risk and return profile vis-à-vis a small cap fund. Comparing them will not make much sense.
Instead, we should compare all small cap funds separately and all large cap funds separately.
In this framework, we will use SEBI’s new categorization mechanism (explained here – How SEBI’s recategorization changed Equity Mutual Fund categories?).
Also, for the purpose of this blog, we will only look at Equity Mutual Funds. But the beauty of this rank-based framework is that once you categorize the funds, you can apply the next 2 steps on any category (equity, debt or liquid).
Here are the categories that we will consider:
- Large Cap Funds
- Multi Cap Funds
- Small Cap Funds
- Mid Cap Funds
- Large & Mid Cap Funds
- Focused Funds
Once we have categorized the funds, we will apply the next 2 steps for each category separately.
Step 2: Rank based on performance
This is the heart of our framework.
Now that we have funds in distinct categories, our job is to be able to rank them in order of performance.
The guiding philosophy that this framework follows can be summed up in a single sentence – “Long-term track record of consistent performance”.
Following this philosophy, the entire ranking process can be broken down into 3 parts:
- Long-term performance with respect to peers and consistency of that performance
- Overall risk associated with the fund
- Recent performance
We will discuss each part in detail and then bring it all together.
But before that, we need to agree upon parameters that we will use to measure performance.
We use Rolling Percentile Rank instead of Sharpe Ratio
Sharpe ratio has been (and still is) one of the most commonly used metric to evaluate investment instruments including mutual funds.
The concept of rolling returns and standard deviation has been explained in detail in this blog here.
If you don’t feel like reading the whole theory, let me try to simplify Sharpe ratio for you – just think of it as a measure of “risk adjusted return”. Higher the Sharpe ratio, better is the fund performance.
Note that most of the Mutual Fund industry uses Sharpe ratio as one of the primary indicators of performance.
To that extent, even our last fund-selection framework relied largely on Sharpe Ratio.
Sharpe Ratio, however, has a big limitation:
It is highly sensitive to time period under consideration. This makes relative comparison between funds very difficult.
For example, consider 2 funds A and B. Fund A started in Jan 2004 and Fund B started in Jan 2011. If you look at their Sharpe ratios since inception, chances are that Fund A which started in 2004 will have a poorer ratio as it would have faced the entire brunt of 2008 crash.
Fund B, on the other hand, started in 2011 and never faced any such big crash.
One option would be to compare A and B for their performance only since Jan 2011.
This however limits our analysis as we are forced to ignore data for fund A from 2004 to 2011.
Let’s look at an actual example:
The example above shows that in order to make any meaningful comparison, we will have to consider all funds from June 2012. By doing this, we lose out on precious 2.5 years of performance data for other 4 funds.
And this is just for a 4-fund universe. Imagine the entire universe where some funds started as early as 1995 and others that started in 2017.
Moral of the story – we have to look beyond Sharpe ratio to measure and compare mutual fund performances. There are other ratios like Sortino, Treynor etc. However, none of them solve the time-period sensitivity issue that we discussed above.
This limitation of Sharpe ratio is one of the primary reason for us to launch our new framework. In our previous framework, we were forced to work with data beyond 2010 and had to ignore data before that. Even then, funds that started after 2012 ended up distorting the picture.
So, what’s the solution?
Introducing the concept of Rolling Percentile Rank.
What does it mean?
It is fairly simple. We pick a start date (Jan 2004 in our framework) and look at every 3-year period starting Jan 2004 all the way up to Dec 2018.
So, 3rd Jan 2004 to 3rd Jan 2007 is one such period. Similarly, 18th Sep 2012 to 18th Sep 2015 is another such period and 31st Dec 2015 to 31st Dec 2018 will be the last 3-year period. In total, there are roughly 3,000 such periods.
Now, for each 3-year period, we rank the funds based on the returns delivered during that 3-year period.
We don’t use absolute rank but percentile rank. For example, if there were 17 funds during a period, and a fund is ranked 4th, it’s percentile rank would be 23.5%.
The logic for using percentile rank instead of absolute rank is as follows – suppose there were only 7 funds in initial periods. On the other hand, during later periods, there were 25 funds. Being 5th in a 7-fund universe is way different than being 5th in a 25-fund universe. Using percentile ranking solves the problem.
Now we apply this concept to get the percentile rank of each fund in every period.
Confused? Perhaps a pictorial representation would help.
Now that we are equipped with the concept of percentile rank, we are ready to start ranking the funds based on our 3-part process.
1. Long-term performance with respect to peers and consistency of that performance
This is what we have done so far:
- Bucketed all the funds into separate categories
- Within each category, we have calculated the percentile rank of each fund for every 3-year period starting Jan 2004.
- Excluded funds that started post Jan 2012 (This shall be explained soon).
To measure long-term performance, all we need to do is to calculate the average percentile rank of each fund.
Remember, for each fund, we have its percentile rank for every 3-year period. To get a sense of the fund’s “Long-term performance with respect to peers”, the best way is to calculate the average rank over all periods.
While averaging gives us a sense of long-term performance, it doesn’t tell us much about the consistency of the fund. Suppose we have 10 years of performance data for a fund. For the first 7 years, the fund was consistently ranked in top 10%. However, for the last 3 years, it is ranked in the bottom 10%.
Average ranking might still look good. But the performance cannot be termed as “consistent”.
To measure consistency, we introduce another metric – percentage of times the fund has been in top 25%. Let’s call it Top25 percentage.
Suppose a fund has seen 2,000 3-year periods since Jan 2004. Of these, the fund was in top 25% in 1,400 periods. Top25 percentage would then be 70%.
That’s it! We now have Long-term average percentile rank to measure long-term performance and Top25 percentage to measure consistency of that performance.
These 2 will be the highest weightage parameters in our ranking framework.
For this whole exercise, we have used 3-year returns instead of 1-year or 5-year returns. The reason is that 3-year is the sweet spot. 1 year is too short a time frame to evaluate a fund’s performance. On the other hand, 5 years is too long (you would not want to stick around with an underperforming mutual fund for 5 years).
Also, we have not considered funds that started post Jan 2012. This is because we want at least 4 years of performance data. Since we are considering 3-year returns, any fund that started after Jan 2012 will not give us 4 years of performance data. Our period of assessment should be long enough to showcase track record and consistency of performance.
Now let’s revisit the example above (of Axis Focused 25) and see how the picture changes.
2. Overall risk associated with the fund
Part 2 of our 3-part process is to measure the overall risk associated with a fund.
The most commonly used metric for measuring risk is standard deviation (or volatility) of the fund. To know more about standard deviation (and how to calculate it), refer this blog here.
Standard Deviation, just like Sharpe ratio, has 2 big limitations:
- It is highly sensitive to time period under assessment. Funds that started before 2005 would have faced the whole brunt of the crash of 2008 and their standard deviation will be higher. It would be unfair to compare their standard deviation with those funds that started in 2010 or later.
- Standard Deviation places equal weightage to both upside and downside deviation. What this means is that if your expected return is 15%, then an actual return of 10% as well as an actual return of 20% will both increase the standard deviation (and the perceived risk). As an investor, anything above 15% should not be a risk.
To solve for time sensitivity we will take a cutoff date (1st Jan 2010 in our case) and evaluate all funds from this cut off.
To solve for equal weightage problem, we use a modified version of standard deviation called Downside Deviation.
Will not go into the technicalities of calculating Downside Deviation. It’s very similar to how we calculate Standard Deviation except that we ignore all instances with positive deviations (where actual returns are higher than average return).
Choice of Jan 2010 as cutoff is not arbitrary. By taking Jan 2010 as cutoff, we remove the impact of the crash of 2008 from our analysis.
3. How has the fund been performing recently
So far, we have been looking at the long-term performance of the fund – over a time frame of 15 years.
It is also important to look at recent performance of the fund. If a fund has been in the top 10% all along but has had a drastic fall recently, it should be a red flag and might require deeper investigation – is the recent underperformance a temporary blip or something fundamental has changed?
We will use the same concept of average percentile rank to measure recent performance as well. However, instead of using 3-year returns, we will use 1-year return and consider the performance from 1st Dec 2016 to 31st Dec 2017. This way, the last 1-year period ends on 31st Dec 2018.
There you go! We now have all the 4 parameters required to rank our funds. Let’s quickly summarise:
- Long-term average percentile rank to measure long-term performance
- Top25 percentage to measure consistency of that performance
- Downside Deviation since Jan 2010 to measure risk
- Near-term average percentile rank to measure recent performance
Next obvious step is to ascribe weights to each parameter. For our framework, we give the highest weight to #1, next is #2 and roughly the same weight to #3 and #4.
Using these weights, we arrive at the final rank of each fund within a category.
Time for an actual example.
Note that the ranking shown in the example above is still an interim ranking. The list will now go through the thrird and final stage of a subjective evaluation.
As we will shortly see, once we apply the final filter step, the rankings can change. And at times, quite drastically!
Step 3: Filter based on subjective parameters
So far in Step 2, we have dealt with purely objective (quantitative) parameters. All these parameters can be simply plugged into a formula and we will have our output ranking.
Unfortunately, life isn’t that simple.
When evaluating Mutual Funds, one must have a final stage of subjective evaluation based on qualitative parameters.
Here’s a list of qualitative parameters that one should account for:
1. Recent change in fund manager
Suppose your framework throws up an absolute rock-star fund. Unfortunately, you find out that the old fund manager has recently quit and the fund is now being managed by a new manager. The rock-star performance of the fund was actually attributable to the old fund manager.
You now have to take a subjective call on the new manager. Is she as good as the old one? What has been her track record with other funds?
For that matter, it could also be the other way around as well. An underperforming fund suddenly finds itself with a new manager with a great track record and starts performing. Again, subjective call.
2. Total size of the fund
Bigger is not necessarily better. Total Asset Under Management (AUM) of a fund can have a big impact on the potential returns it can deliver.
As a fund grows in popularity, more investors start investing in it. With all this new money, the total size of the fund increases.
As fund size rise, the number of new stock prospects shrink and it becomes difficult for the fund manager to stick to his/her investment style. End of the day, there are limited number of stocks that a manger can buy.
Asset size problem becomes much more acute as your move down the market cap category.
For a large cap fund that deals with highly liquid large cap stocks, a large fund size is still manageable.
As you move towards smaller market cap stocks, it becomes extremely difficult to find new stocks that are both attractive and liquid enough to buy and sell. For this reason, it is quite common for small cap funds to close down new subscriptions once they hit a particular size.
What size is the right size? This is again a very subjective question and has to be answered in conjunction with:
- The type of fund
- Ability and track record of the fund manager in managing funds of larger sizes
In my view, given the current context of Indian markets, the sweet spot could look something like this:
3. Crash score of the fund
One very interesting parameter that we should look at is the performance of the fund during a steep stock market crash. In our case, we can simply look at the returns delivered by the funds during 2008-2009 crash period.
Since not all funds in our universe will have a track record of performance in 2008-2009 period, this is again a subjective assessment.
Funds that do have a track record and a good one at that should get some added bonus points.
Navigating through crashes like 2008 can be a great learning experience. As a general rule-of-thumb, you should prefer fund managers who were actively managing funds during the 2008 crash period.
So, that in essence was the CRAFT framework for selecting the best mutual funds.
A quick pictorial summary of the framework and then we will close out our large cap mutual fund example.
Finally, let’s wrap this up by completing our list of top funds in the large cap category.
To see the application of this framework to pick the best mutual funds to invest in 2019, visit Best Mutual Funds to Invest in India in 2019