While looking through the government’s official data website, I chanced upon data about the CPF minimum sum. Since CPF is related to FIRE, I got curious about what the numbers will look like when I turn 55.

I’m also curious when will the minimum sum hit 1 million.

##### CPF Retirement Sum Scheme

Here’s what happens in a nutshell. (More comprehensive information can be found on the CPF website.)

The Retirement Sum Scheme is set up so that Singaporeans and Permanent Residents (PRs) will have a monthly income enough for a basic retirement.

While working, Singaporeans and PRs contribute to our CPF, which is split between the Ordinary Account (OA), Special Account (SA) and Medisave Account (MA) (check out the exact split here).

When we turn 55, the Full Retirement Sum (FRS) is set aside from our OA and SA into our Retirement Account (RA). We also have the option to pledge a property we own and set aside the Basic Retirement Sum (BRS), which is half of the FRS. If we have more than the FRS, we can choose to set aside the Enhanced Retirement Sum (ERS), which is 1.5 times the FRS.

Note that the more money you set aside, the higher your monthly payouts. Someone who set aside the FRS will get a higher monthly payout than someone who set aside the BRS.

The FRS is set by the CPF and takes into account things like inflation and cost of living. This means that the FRS will increase every year.

##### Analysis (Regression)

It’s always a good idea to plot out your data points to get a sense of how the data looks like. From there, we can decide what to do next. X-axis (horizontal axis) dates are every two years, discrepancy is due to each year being 365.24 days long. Minor gridlines are half a year each.

Note a small gap in the points between 2015 and 2017. That was when CPF changed the effective date from 1st Jul to 1st Jan every year (makes more sense).

Looking at the scatter plot, it seems like a linear trend would fit pretty well. Let’s draw a line of best fit across the points (also known as a simple linear regression)

Turns out, it fits pretty well! If we extend the line, we will be able to get a good estimate of the future FRS according to this model.

Here’s the line extended 50 years into the future (till 2073) X-axis major gridlines are ten years, minor gridlines are one year each

For those interested, here are the more technical details:

Looking at the equation of the line, the coefficient of x is the slope of the line. Since the units of the x-axis is in days, a slope of 17.928 means a unit change in x results in a (positive) 17.928 unit change in y. If we put back the units (x in days, y in dollars), a change of 1 day means an increase of \$17.928.

A year has roughly 365.25 days. Thus, the model predicts that the minimum sum will increase by 365.25 * \$17.928 = \$6,548.20 every year.

The intercept is not valuable information since we are only interested in the future FRS.

For those who hate math, basically, the model predicts that the FRS increases by about \$18 every day, or roughly \$6,550 every year.

##### Analysis (Annual Increase)

Another way of looking at the increase in the FRS is through the underlying fundamentals, which are inflation and cost of living. Since the cost of living increases by a fixed percentage each year, we can also look at the percentage increase in the FRS. Here’s how the data looks like:

Between 2003 – 2013, the increase is roughly a flat 6%. Between 2013 – 2017, the rate of increase drops from 6% to 3%. From 2017 to 2020, the rate is almost a flat 3%. Note that the data point for 2017 is for 1.5 years, so the annual increase is actually lower.

How did inflation look like in the same period?

This figures are from the percentage increase in “MAS Core Inflation Measure” from the CPI figures in data.gov.sg.

Interestingly, there is not much similarities between the two plots (except the 2009 increase). Intuitively, this makes sense because CPF is looking at inflation in the future, not the current inflation.

Since we do not know what numbers CPF are using, it is very difficult to predict how much the FRS will increase by in the future. However, we can look at different projected rates into the future.

As a baseline, we can look at the average annualised rate of increase of the FRS. The FRS increased from \$80,000 to \$181,000 in 16.5 years. This means the compounded rate of increase is roughly 5.1% per year. For simplicity, let’s round it off to 5%.

Edit: “Sinkie” pointed out to me (refer to the comments section) that a comprehensive review of CPF adequacy in 2002/2003 found that the Minimum Sum/FRS was below the adequate level. This resulted in the annualised 6% increase in FRS.

An explanation by Tharman Shanmugaratnam in 2012 explained that their target MS/FRS in 2013 was \$120,000 in real terms (2003 dollars), but they pushed it back to 2015 after review.

Back of the envelop calculations with 2% inflation rate show that the nominal 2013 target was \$146,000 (vs \$148,000 actual), nominal target in 2015 was \$152,000 (vs \$161,000 actual) and ‘target’ in 2020 will be \$168,000 (vs \$181,000 actual). The increases have hit their targets, even exceeding it. In my opinion, it seems very likely that the increases going forward will only account for inflation, so 2% seems a safe bet.

##### Cheat Sheet

Interesting to note that results from the linear regression model is similar to an annual compounded increase of 2%. This is due to the shorter time horizon, as the 2% model is exponential whereas the linear model is, well, linear.

To answer the million dollar question: it depends on which model you use.

The 7% model predicts 2046, just 27 years away. This is an aggressive estimate, but gives you an idea of what to expect if CPF suddenly expects future cost of living to increase.

The 6% model predicts 2050, 31 years away. This is also a relatively aggressive estimate. However, this rate of increase was seen between 2003 – 2013, so it is still a possibility (but unlikely).

The 5% model predicts 2056, 37 years away. So far, the FRS has increased by about 5.1% every year, so this seems quite possible.

The 4% model predicts 2064, 45 years away. Considering that the FRS has increased at about 3% recently, this is also quite possible.

The 3% model predicts 2078 (not shown in cheat sheet), 59 years away. Similar to the 4% case, this might be possible if the 3% rate of increase remains. Most of us will not be alive by then.

The 2% model predicts 2107 (not shown). This is a conservative estimate. However, an inflation rate of 2% in developed countries is expected, so this might be the case. Looking at the inflation rate graph, Singapore’s inflation rate has been below 2% for most years. Edit: As per previous edit, this seems the likely scenario (to me).

The regression model predicts 2146 (not shown). Ultimately, this is a linear model, and fundamentally does not make sense. CPF is meant to provide a reasonable payout for retirement, and inflation will cause prices to increase exponentially.

##### Warnings, Assumptions and Disclaimers

The future is never certain. We can only make an educated guess about what happens next. This analysis requires certain assumptions to be true. If any one of these assumptions fail, the results will no longer hold.

I did not include confidence intervals in my forecast (for regression). It is very unlikely that the numbers predicted will zhun zhun be listed the cheat sheet, but will roughly be in the ballpark if assumptions hold.

Predicting the future far in advance is very risky. Any small mistake will be compounded and result in very large errors (2.7 million FRS, anyone?).

I have assumed that the FRS deduction age continues to stay at 55, if this age changes, the results will change. Given that the life expectancy of Singaporeans and PRs are increasing, it seems logical that the age will be pushed back eventually. Again, no one knows the future.

Edit 2 (for those familiar with statistical methods): Additionally, not all assumptions required by the linear regression holds (eg in this case, homoscedasticity and independence of error terms). Simple linear regression also does not reflect the underlying reason behind increasing the FRS. It turns out that the good fit is due to chance, ie the decreasing rate of increase because of government policy. The significance level has also been omitted.

##### Conclusion

This was an interesting exercise in predicting the CPF FRS. I have provided two simple approaches with different underlying methodologies to forecast the FRS. Only time will tell which model is more accurate. Personally, I would use the linear regression for the first ten years, then the 4% model afterwards. Edit: Following the previous edits, I now think the 2% model will be more accurate.

The different models give us a sense of how different scenarios will pan out. However, what we know for sure is the FRS will increase. Hopefully, this gives you a better idea of what to expect in the future.

Categorised in: CPF, Finances

## 2 thoughts on “Predicting Your CPF Minimum Sum”

• Sinkie says:

The FRS increase from 2003-2015 was exaggerated i.e. way above inflation, as govt consciously ramped up the Minimum Sum (as it was known then), following a comprehensive 2002/2003 review of CPF adequacy. The 2003 review determined that the Minimum Sum at that time was at least 50% below the adequate level(!). That’s why the 6% annual increase in FRS, with everybody KPKB.

These few years from 2016-2020 is more of normalization as they use a generic 3% annual increase. CPF had said it will do another review in 2020. The basic premise is that future growth of FRS will largely follow the general consumer inflation rate e.g. CPI.

The actual CPI increase over the last 10-15 years have been roughly 1.7% annually. So if the trend continues, we may see CPF use 2% as the annual increase. We’ll know more in 2020.

• Ben says:

Thanks for your insightful comments! This explains the discrepancy between the inflation rate and rate of FRS increase. I was wondering why the FRS was increasing faster than inflation.

Found an explanation by Mr Tharman on the revision of CPF Minimum Sum. He mentioned that the target MS was to hit \$120k in real terms (2003 dollars) in 2013, but pushed back to 2015.

Using a 2% inflation rate: in 2013, the target was \$146k (vs \$148k actual). In 2015, the target was \$152k (vs \$161k actual). In 2020, the ‘target’ will be \$168k (vs \$181k actual). It seems like they have hit the target, even exceeded it. Personally, it seems very likely that the increase will be mostly adjusting for inflation, so 2% seems like a logical choice.

Another interesting (not-as-relevant-anymore) question regarding MS/FRS is: Given the need for retirement adequacy vs the average SG/PR’s financial capability, how much should the MS be raised? If the MS is raised too quickly, the average SG/PR cannot catch up and the number is meaningless. If too slow, the payouts will (probably) be inadequate for retirement (as at Feb 2019: Average CPF member’s monthly payout is \$355 – Josephine Teo).