Should We Go With 3% Rule?
I recently wrote a post about asset allocation in which a reader commented whether we should be using 3% safe withdrawal rate in Indian context instead of the 4% safe withdrawal rate that is generally accepted in the US (at least in the past). The commenter also provided a video supporting the 3% rule. I strongly encourage you to watch that first. It makes some good arguments. If you are interested, I’d also suggest you read the paper which contains all the details, assumptions, data and methodology used for the simulations. According to the research paper, it is recommended that one should go with a 3% safe withdrawal rate instead of the generally well known 4% rate.
I am glad that finally there is some research done with data available about investment vehicles in India to determine a safe withdrawal rate. So thanks to the anonymous commenter. While I watched the video, I did not read the paper completely. But from what I understand, the authors recommend a 3% safe withdrawal rate with an asset allocation of 60% in fixed income and 40% in equity. It is suggested that it is better to go with a 10% allocation into gold, 30% in equity and 60% in debt funds. This is quite the opposite allocation that I normally suggest which is an allocation of 30% fixed income and 70% equity for a safe withdrawal rate of 4%. How can it be?
Well, my recommendation is loosely based on the research done in the US. In that paper it was suggested that the allocation should be 60% in equity and 40% in debt. It was done in the US and it was a long time ago in 1994. The suggested safe withdrawal rate was 4% based on the data available from 1926 to 1992. As the equity returns changed over the years post the publication of that paper, the safe withdrawal rate was revised in future papers. But this particular paper was based on Indian context, so we should certainly give it more weight, although I am not very sure of the methodology used (because I did not read the paper completely).
The author indicates that having more allocation to risky investments like equity will actually decrease the safe withdrawal rate because it increases risk. I am assuming that in some of the simulations where the markets fall rapidly soon after retirement will severely affect the corpus and as a result may not last the 30 years that one expects it to last after retirement. Perhaps that’s why the safe withdrawal rate has to be reduced to avoid getting trapped in those unfortunate sequence of returns. It makes sense if you think about it. Lets understand with an example.
Sequence of Returns
For the rest of the examples, I will be making these assumptions
Annual expenses: 4L
Corpus required per 4% rule: 100L
Inflation: 5%
Equity returns: 12%
Fixed income returns: 6%
Asset allocation: 70% to equity and the rest in fixed income
With a 70:30 split between equity and fixed income asset allocation, your returns will be the average by weight of the the two asset classes:
Returns = 12% x 70% + 6% x 30% = 10.2%
If we take the simple approach of 10% annual return on investment every year, then yes, the corpus will last a long time as can be seen in the below table:
| Year | Returns | Corpus |
|---|---|---|
| 0 | 100 | |
| 1 | 10.2% | 106 |
| 2 | 10.2% | 111 |
| 3 | 10.2% | 118 |
| … | … | … |
| 28 | 10.2% | 516 |
| 29 | 10.2% | 550 |
| 30 | 10.2% | 587 |
Of course reality is far from this. Like the author mentioned, if we assume a bad sequence of returns in the early part of your retirement, how would the portfolio look like? We will make just minor change to the table. In the first year, lets say the stock market crashed by 50%. Assuming you have a 70:30 split between equity and fixed income asset allocation, your returns in the first year would be:
Returns = -50% x 70% + 6% x 30% = -33%
After that drop in returns in the first year, if we assume a 10% constant return on investments for the rest of retirement years, we will run out of money by year 25. So yes, sequence of returns matter a lot.
| Year | Returns | Corpus |
|---|---|---|
| 0 | 100 | |
| 1 | -33.2% | 64 |
| 2 | 10.2% | 66 |
| 3 | 10.2% | 67 |
| … | … | … |
| 24 | 10.2% | 12 |
| 25 | 10.2% | -2 |
Generally, but not always, after a crash, the stock market will return back to new highs with in a few years. We have seen this in 2002, 2008 and 2020. If we add in that information, the risk is not as bad as it looks. Here I take the example of Nifty 50’s 2008 crash (-50%) and 2009 (+75%) recovery as an example.
| Year | Returns | Corpus |
|---|---|---|
| 0 | 100 | |
| 1 | -33.2% | 64 |
| 2 | 54.3% | 92 |
| 3 | 10.2% | 96 |
| … | … | … |
| 28 | 10.2% | 272 |
| 29 | 10.2% | 281 |
| 30 | 10.2% | 291 |
But we cannot assume that markets will go back up every time or as soon. Take the example of Japan which every one often reminds us about. I present here the table with the real data of Nikkei from 1990 which is when the major crash happened and the markets were not able to get back to the previous high until very recently. Suppose India goes into this kind of decline and someone retired at that time, then this is how the table will look like assuming a 70% asset allocation to equity and debt returns remain unchanged at 6% throughout the period:
| Year | Nikkei | Year | Returns | Corpus |
|---|---|---|---|---|
| 0 | 100 | |||
| 1990 | -38.7% | 1 | -25.3% | 72 |
| 1991 | -3.6% | 2 | -0.7% | 67 |
| 1992 | -26.4% | 3 | -16.7% | 52 |
| 1993 | 2.9% | 4 | 3.8% | 49 |
| 1994 | 13.2% | 5 | 11.1% | 48 |
| 1995 | 0.7% | 6 | 2.3% | 44 |
| 1996 | -2.6% | 7 | 0.0% | 38 |
| 1997 | -21.2% | 8 | -13.0% | 28 |
| 1998 | -9.3% | 9 | -4.7% | 21 |
| 1999 | 36.8% | 10 | 27.6% | 18 |
| 2000 | -27.2% | 11 | -17.2% | 10 |
| 2001 | -23.5% | 12 | -14.7% | 2 |
| 2002 | -18.6% | 13 | -11.2% | -5 |
As you can see, this person will be wiped out by year 13 with that kind of sequence of returns. Now, you wouldn’t want to take that kind of gamble would you?
Asset Allocation
Now what about asset allocation? The authors suggest that going with an asset allocation of 40% in equity and the rest in fixed income reduces the risk. Lets work out a few examples. For the rest of the examples, I will be making these assumptions
Annual expenses: 4L
Corpus required per 4% rule: 100L
Inflation: 5%
Equity returns: 12%
Fixed income returns: 6%
Asset allocation: 40% to equity and the rest in fixed income
With a 40:60 split between equity and fixed income asset allocation, your returns will be:
Returns = 12% x 40% + 6% x 60% = 8.4%
I won’t work out the ideal case where the returns are constant throughout the retirement period because the corpus will certainly last the 30 years in retirement. Now coming to the one year bad returns case, the returns will be:
Returns = -50% x 40% + 6% x 60% = -16.4%
And the table will look like this:
| Year | Returns | Corpus |
|---|---|---|
| 0 | 100 | |
| 1 | -16.4% | 80 |
| 2 | 8.4% | 82 |
| 3 | 8.4% | 84 |
| … | … | … |
| 26 | 8.4% | 22 |
| 27 | 8.4% | 7 |
| 28 | 8.4% | -9 |
Again, your corpus will not last for 30 years. In the 70:30 allocation example above, the corpus got exhausted in 24 years and in this case it exhausted in 27 years. So no matter what your asset allocation is, if your luck is bad and there is a bad first year then you will not be able to recover.
Generally, but not always, after a crash, the stock market will return back to new highs with in a few years. We have seen this in 2002, 2008 and 2020. If we add in that information, the risk is not as bad as it looks. Here I take the example of Nifty 50’s 2008 crash (-50%) and 2009 (+75%) recovery as an example.
| Year | Returns | Corpus |
|---|---|---|
| 0 | 100 | |
| 1 | -16.4% | 80 |
| 2 | 33.6% | 101 |
| 3 | 8.4% | 105 |
| … | … | … |
| 29 | 8.4% | 141 |
| 30 | 8.4% | 134 |
If you use the Japan’s Nikkei returns from 1990 as an example, the situation slightly improves lasting 4 more years when compared to the 70:30 allocation, but it still does not last the full 30 years in retirement.
| Year | Nikkei | Year | Returns | Corpus |
|---|---|---|---|---|
| 0 | 100 | |||
| 1990 | -38.7% 1 | -11.9% | 84 | |
| 1991 | -3.6% | 2 | 2.1% | 82 |
| 1992 | -26.4% | 3 | -6.9% | 72 |
| 1993 | 2.9% | 4 | 4.8% | 70 |
| … | … | … | … | … |
| 2003 | 24.5% | 14 | 13.4% | 17 |
| 2004 | 7.6% | 15 | 6.6% | 9 |
| 2005 | 40.2% | 16 | 19.7% | 0 |
The lower allocation to equity does reduce the risk like the authors mention.
Safe Withdrawal Rate
Now we come to the most important point made in the video, which is the safe withdrawal rate (SWR). While it is generally accepted that a 4% SWR is sufficient using US data, the authors in the research paper suggest a lower 3.5% or even 3% SWR. So lets workout some examples with the following assumptions
Annual expenses: 4L
Corpus required per 3% rule: 134L
Inflation: 5%
Equity returns: 12%
Fixed income returns: 6%
Asset allocation: 40% to equity and the rest in fixed income
The ideal case where the returns are constant throughout the retirement period works out well, so no need to check that case. Where things change drastically is in the case of first year negative returns.
| Year | Returns | Corpus |
|---|---|---|
| 0 | 134 | |
| 1 | -16.4% | 109 |
| 2 | 8.4% | 113 |
| 3 | 8.4% | 117 |
| … | … | … |
| 28 | 8.4% | 242 |
| 29 | 8.4% | 244 |
| 30 | 8.4% | 246 |
The corpus easily outlasts the 30 years in retirement. If this worked out well then there is no point looking at the case where the returns bounce back after the first year loss. So the final test is of course the Nikkei returns and this is how it performs:
| Year | Nikkei | Year | Returns | Corpus |
|---|---|---|---|---|
| 0 | 134 | |||
| 1990 | -38.7% | 1 | -11.9% | 114 |
| 1991 | -3.6% | 2 | 2.1% | 112 |
| 1992 | -26.4% | 3 | -6.9% | 100 |
| 1993 | 2.9% | 4 | 4.8% | 100 |
| … | … | … | … | … |
| 2008 | -42.1% | 19 | -13.2% | 18 |
| 2009 | 19.0% | 20 | 11.2% | 8 |
| 2010 | -3.0% | 21 | 2.4% | -3 |
Unfortunately, it meets the same fate. The corpus will only last for 20 years in retirement if the sequence of returns are anything like Nikkei in the past. However, the corpus lasted 4 years longer when compared to previous case. I even tried 3% SWR with 70% equity allocation and the performance was far worse during bad times and far better during good times. You can play with the numbers yourself by making a copy of my spreadsheet and changing the numbers.
What all these examples teach us is that the asset allocation and SWR depends on our luck and individual risk appetite. I would still go with 70:30 asset allocation and 4% SWR for myself based on the analysis that I did earlier. I have trust in my analysis and have the risk taking ability. If you trust the research paper and have low risk profile then you should certainly follow 3% rule or even 2% SWR with 40% allocation to equity. There is no one right answer here. After all, personal finance is very personal. No one can predict the future and you should make decisions based on your situation. Perhaps the old adage – “be safe than sorry” is applicable more than ever.
Disclaimer: In all these calculations, I have not done any rebalancing. I assumed you would start with a certain asset allocation and then you are never going to rebalance and let the market forces dictate the value of the corpus. Which is probably not what one would do. Perhaps I will do another analysis with rebalancing.