The Ultimate Comprehensive Analysis of The 11-Year Returns of the STI ETF: Part 2 – Rolling Internal Rate of Returns

This is Part 2 of a 4 part series. I started off trying to write a single article but the amount of analysis grew too large to fit into a single post, so I’ve decided to split it up into the following 4 parts:

• Part 1 : The 11-year performance of Dollar Cost Averaging into the STI ETF.
• Part 2 : Multi-Period (5-Year to 10-Year) Rolling IRR if you DCA into the STI ETF.
• Part 3 : Multi-Period (5-Year to 10-Year) Rolling IRR for Lump Sum Investing into the STI ETF.
• Part 4 : Comparing the performance of DCA vs Lump Sum.

After I posted the analysis of the 11-year returns of DCA last week, there was a lot of responses asking for a deeper analysis into the returns.

People were concerned about whether such a high return – 5.63% real return and 7.15% non-inflation adjusted return – was just a result of the 2008 financial crisis. If you need a reminder, here’s the price chart of the SPDR Straits Times Index ETF (ES3) from 2008 to 2018.

The market really bottomed out at the end of 2008 and early 2009 which could mean that if you purchased your shares around that period, that could constitute the majority of your returns. Although the price does not stay low for long and should not greatly impacts long-term returns, it still makes you wonder if the 7.15% non-inflation adjusted return was just luck.

Well it’s time to try and put that to rest with rolling returns analysis. *applause*

TL;DR: Summary of Findings:

1. In the last 11 years, no matter the rolling window, the STI ETF provided more than 0% return (non-inflation adjusted) more than 99% of the time.
2. Based on the average inflation rate of 1.52%, the STI ETF returned more than 0% real return more than 98% of the time.
3. The longer you stay invested, the lower variance in return – and thus lower risk – you will have.

What is this “Rolling Returns Analysis” you speak of?

A rolling returns analysis looks at the returns of a particular asset during a fixed window (i.e. 5-year or 6-year window) over time. For example, let’s look at a few 5-year rolling windows:

This way we can see the returns based on a different start date and end date to see how timing affects returns. I did this for the entire dataset (Dollar Cost Averaging and Reinvesting Dividends) and this is what it looks like:

So now we can see the returns we would have gotten if we only DCA into the STI ETF for 5-years given all of these start dates.

What kind of insight can we get from this chart?

Good question, so let’s take a look. Based on the chart above, here are a few observations:

1. The window that provided the best returns was if you invested from 10-May-2008 to 10-May-2013 with an IRR of 12.77%. Start investing at the bottom of the recession gives you the best returns? Who woulda thunk it?!
2. The worst window was from 10-Feb-2011 to 10-Feb-2016 with an IRR of -1.76%.
3. Out of the 73 windows, only 1 provided negative returns.
4. The average returns sits between 4% and 9%.

That’s if your investment window is exactly 5 years long.

“Hold up!” you might say. “That’s all well and good but 5 years is pretty short and doesn’t really tell us what happens to rates of returns the longer we invest. Can you do something about that?”

Well yes… yes I can. The better question would be, are you ready for it?

Beware, here be dragons: All the rolling returns

Here are the rest of the rolling returns, all of them: 5-year, 6-year, 7-year, 8-year, 9-year and 10-year rolling returns (DCA with dividends reinvested):

When mapped out, the longest window (10-Year) only has 12 starting dates (in 2008) as we only have data until 2018 (and January 2019). This is why it ends abruptly in Jan-2009. Similarly, each of the other rolling windows will also end when it reaches the starting point with an ending date that corresponds to January 2019.

However, this format isn’t so easy to read and get much information from. Let’s take a look at the scatter plot version instead:

The patterns are much clearer here. Let’s take a look at what it tells us:

What can we learn from combining all of this data?

1. The best time to start investing was in 2008 and early 2009. Again, that seems the best time to have started investing, no matter your time-horizon. Since that’s the bottom of the recession, it’s obvious and it’s great to see that the rolling returns reflects that.
2. Out of all the investment starting periods for all the rolling windows (all 125 of them) there were only 2 starting points that would end with a negative IRR, that is the 6-Year window from 10-Feb-2010 to 10-Feb-2016 and the 5-Year window from 10-Feb-2011 to 10-Feb-2016. You can also see that the periods ending on 10-Feb-2016 are the worst performing of all the rolling windows, that coincides with the lowest STI ETF price since 2012, and no rolling window ends in 2012 (you’d have to start in 2007 with a 5-year rolling window to end on a date in 2012), so that’s probably why.
3. The longer your investment horizon, the less variance in returns that you will see. As you observe the chart, you will see that the 5-Year IRR has a bigger spread between the lowest and highest returns and as the length increases to 6, 7, and finally 10 years, the difference between the highest return and lowest return becomes much lower. This suggests that the longer your time horizon the returns become more stable.

In order to help you see #3 better, I’ve added trend lines to the chart:

You can clearly see the variance in dots get smaller as the window length get longer from 5-year to 10-year.

Cool, so what kinds of returns was I likely to get based on each investment horizon for the last 11 years?

Well, just like in my last STI analysis article, let’s take a look at where the IRR for each starting period and each rolling window lands on a histogram to see the frequency of occurrence:

Here you can see the difference in variance really clearly.

1. The 5-Year rolling window has the highest variance with the lowest return of less than -1% but also the highest return at almost 13%. The majority of the return (over 17% of the time) is between 8-9% and is more than 6% return more than 72% of the time.
2. The 10-Year rolling window has the lowest variance with all of the returns between 7-10%. That’s 100% of the time the returns more than 7%! Of course, all the starting dates for the 10-year rolling windows started in 2008 which was near the bottom of a recession so this isn’t the most fair.

Let’s take out the lowest points in the market

To remove any bias due to starting at the bottom of the recession, let’s remove the data points from 2008 and 2009. This is what the histogram looks like now, with 2 years with the lowest prices removed as the starting points:

Frequency of occurrence of returns for 5 to 8-year rolling windows starting from Jan-2010 to Jan-2014.

With the lowest points of the market removed as starting points, the crazy returns on the top end are gone.

1. The 9 and 10-Year rolling window is removed as all the starting dates are in 2008 and 2009.
2. The variance pattern holds true here as well. The shorter the investment window, the more varied your return will be. The longer the investment window, the lower the variance.
3. For the 5-year and 6-year window, more than 59% of the time your returns will be more than 6%.
4. For the 7-year window, 96% of the time your returns will be more than 6%.
5. For the 8-year window, you’d make more than 6% return 100% of the time if you invested between Jan-2010 and Jan-2011.
6. Based on the average inflation rate of 1.52%, you’d still get greater than 0% real return more than 97% of the time even when the bottom of the market was taken out.

In any case, the returns of the STI ETF within the last 11 years (between Jan-2008 and Jan-2019), no matter what the rolling window is and no matter when you start, was greater than 0% more than 99% of the time! The result is the same between Jan-2010 and Jan-2019, removing the 2 years where the market bottomed.

Parting notes and conclusion

That certainly looks great doesn’t it? Personally based on my previous post on the 11-year DCA for the STI ETF and this one, I am very confident in continuing to maintain my regular savings plan to buy the STI ETF for S$1,000 each month.

However, before you take this data to mean that the STI ETF will definitely give you this type of return going into the future, I have to caveat that past performance does not guarantee future returns. Anything can happen in the market, we can face another 2008 style financial crisis and another recession could hit and the market could tank. This could wipe out our returns for long periods, but based on my knowledge of history, the market always recovers eventually and then the returns will go back to hover around 5-7%.

What I wish I could add to this analysis

The only thing I wish I could do to add to this analysis to make it more robust would be to have more market data to go further back to 2002 when the SPDR STI ETF was just setup. That will allow me to do more rolling return windows as well as provide more periods to analyse. This will give us even more data points to cover a longer period of Singapore’s development. If anybody has the pricing data going back that far (Yahoo Finance only has prices till 2008) please send it to me and I will add to this analysis.

So that’s it! The rolling returns analysis for DCA into the STI ETF for the investing periods in the last 11 years. I hope this was interesting for you. Please do post your comments, feedback and suggestions on what analysis you’d like me to do next in the comments section below.

You can also follow me on twitter @firepathlion to interact with me directly! I’d love to hear from you.

If you enjoyed this post, it would make my day if you could help share this article with your friends who might also find it interesting. Thank you!

Until next time!

FPL