Back in 2018, I learned that CPF computes its interest using a method different from that of typical bank accounts, and shared this information as a footnote to a post then. In a nutshell:
- Bank account computes interest on a daily basis, and then pays you interest monthly (i.e. monthly compounding).
- CPF computes interest on a monthly basis, and then pays you interest annual (i.e. annual compounding).
- For the exact same amount, you will get slightly more interest when using monthly compounding as compared to annual compounding.
The CPF Board also explains this in layperson-friendly language on its website:
 |
| Source: CPF website, retrieved 28 June 2022. |
Based on this, I have been under the impression that the monthly interest is computed in this manner:
A post I came across at Fatty's Finance also interprets likewise:
 |
| via Fatty's Finance |
Interest is calculated on the "lowest amount in the month" - sounds correct, and matches with the wording from the CPF website, at least from a layperson's reading of the latter.
After doing my own checks, I believe that this interpretation, as well as my diagram above, is not precisely correct.
A comment ("It Is More Complicated Than You Think") within that post at Fatty's Finance made me curious whether my understanding is 100% correct. Hence, using my past CPF transactions, I back-tested the computation method shown in my diagram, to see if I was able to derive the exact same CPF interest figures that was paid to me at the end of that year.
The results surprised me. While I managed to obtain the correct figures for my Special Account and Medisave Account, I ended up a few dollars short of the actual interest figure for my Ordinary Account. Initially, I thought it might have been a data input error, so I checked through my inputs again, but they were all correct. The difference of a few dollars was also too large to be a result of rounding errors. At this point, my interest was piqued.
Since my SA and MA calculations were spot-on, and only the OA calculations were off, it meant that the issue with my method would likely have something to do with the nature of OA transactions. Notably, the OA is the only account from which I have withdrawals and contributions within the same month in varying order - meaning to say, withdrawals may precede contributions, or withdrawals may be after the contributions. Experimenting with different settings of my test model, I believe I have determined the precise method used by CPF to compute interest. It is in fact based on the "lowest balance during the month" but the nuance is that any withdrawals that take place after a contribution is made in the same month will first be deducted from the contribution (which we already know, will not earn interest that month). If the withdrawal amount exceeds the contribution amount, then the excess will be deducted from the starting balance for the month (in which case, some interest is foregone). To clarify, I call this the "lowest balance on rolling basis" although this term is not used widely. The following diagrams will help to clarify.
How it actually works
This is how Step 1 actually works:
This diagram should make it easier to understand:
The key implication here is that timing of withdrawals impact the interest earned for that month. See illustrative example #2, where changes from the previous example are marked in red.
In example #1, the withdrawal (for investment) is made on 20 Feb, after the contribution, hence interest will be computed based on $700 as the lowest balance for the month. In example #2, the withdrawal (for investment) is made on 5 Feb, before the contribution, hence interest will be computed based on $400 as the lowest balance for the month.
The later steps are more or less the same as my prior understanding.
Following this adjustment, the discrepancy in my back-test of OA interest amount became 1 cent. I believe this means that rounding of interest amounts (to nearest cent, see Footnote 1) occurs on the monthly figure rather than the yearly total, which seems logical since this would cater for edge cases where CPF Board pays only certain months' interest not amounting to the full year.
Conclusion
Holding all else equal, from the point of earning interest, it is preferable to withdraw funds from CPF accounts after that month's contributions are made. However, based on my understanding, CPF withdrawals for housing, i.e. for repayment of housing loan, takes place on the 1st day of every month, and this cannot be changed. Hence, the main area where CPF members should take note is in terms of other withdrawals, such as recurring investments under CPF-IS. If the member's monthly CPF contribution falls on the 15th of every one, then a recurring withdrawal for CPF-IS occurring on the 16th (or later) of each month would be preferable to one that occurs on the 14th (or earlier) of each month.
Another area where this could be relevant is when a CPF member decides to make a voluntary contribution (VC) to their CPF account. If the member plans to make a VC in the same month as a one-off investment withdrawal, then the optimal arrangement is to time the VC before the withdrawal is executed.
To be honest, for most circumstances, the difference between the two methods (as shown in example #1 versus example #2) may not add up to much, perhaps a delta of a few dollars over the course of a year.
For me, the process was a satisfying investigation. Firstly, I discovered that my previously-held assumptions were not entirely correct. Then, I investigated what might have been wrong with my prior beliefs, before successfully reverse-engineering the underlying process, and deriving the correct results.
=====
[Footnote 1 - It may feel pedantic to pay attention to rounding, but note that CPF Board itself has guidelines on rounding of CPF contributions: rounding the nearest dollar (i.e. up or down as applicable) for total contributions, and dropping the cents (i.e. down only) for the employee's share of contributions.]
No comments:
Post a Comment