Premium bonds for long term returns are absolutely terrible in respect to a stock index fund.

Exodi

david29dpo said:

Good for you
Over the last 4 years my return on PBs has been 7.8%

Good for you but it's not that relevant to anything? Going to a casino or buying lottery tickets doesn't become a good idea because somebody won.

Not that I'm defending the OP, especially with outrageous comments like "While with the stock market your chances are effectively 100%".

Personally I just invest what I can afford, and hope that the stock market returns more than savings would have - the rest is out of my control. I certainly wouldn't base my forecasting around a return of 7.7%.

(Plus the OP is looking at index returns, whereas in practice they'd be invested in a fund, paying things like platform fees, OCF, TC, etc - actual returns would be lower).

Hoenir

SneakySpectator said:

But I don't see the appeal of using premium bonds as a long term "investment" option.

It's readily available cash. As others have have said if I wanted to buy an apple I wouldn't chose an orange. Though it's interesting how much "long term investing" is now actually short term with a high degree of recency bias. 

aroominyork

boingy said:

SneakySpectator said:

I think premium bonds are fantastic for short periods where you need some ok returns on your money without committing it to something that could lose value in the short term. 

Premium bonds are terrible for short periods. You are not in the draw for at least a month when you buy them.

I used to think that but it is wrong. If you deposit on say 31 March, you will be in the 1 May draw and can withdraw your funds the same day. The draw operates on funds held during the previous month.

SneakySpectator

masonic said:

SneakySpectator said:

masonic said:

SneakySpectator said:

masonic said:

The probability of getting at least 7.7% is still about 50%, so your stated 60-70% chance of getting that return from Premium Bonds would still make PB the better investment if true.

If you can achieve 7.7% with 60-70% likelihood, the average return must be higher than 7.7%.

Either you're gaslighting me or I just don't get it, sorry.

The probability of the 20 year period averaging out to 7.7% per year is very high, close to 100%

Some years will be +10%, others -8%, others +14% but after a long enough period of time the probability of all those individual years averaging out to 7.7% is very high. If every year was 7.7% for 20 years, your average return would be smack bang on 7.7%. But since each year gives different returns, the probability of getting smack bang on 7.7% is low, but the probability of getting very close to that number is high...

I don't know where you're getting your reasoning from.

Maybe you can use a different example to help me understand?

I'm not gaslighting. You are fundamentally misunderstanding the nature of investment returns and probability. When the standard deviation is large, there is a large dispersion around the average.

Take this cFireSim simulation for 100% equities over 20 years using historical data: https://www.cfiresim.com/9bdf7cc4-f995-437e-9c24-b3fe4d5a27eb

This is in USD, but we'll have to live with that. I have set inflation to zero. Starting with $50,000, the median return (the minimum outcome for 50% of 20 year periods) is $183k - close to your average figure, but look at the dispersion: 10% of periods ended with $84k, the standard deviation is $111k.

The probability of getting at least the median return is around 50%, with a fair likelihood of achieving meaningfully less. You would have a 90% chance of achieving at least a 2.6% return based on this historical data, and only a 50% chance of achieving at least 6.7%.

I still don't understand but I do believe you. So hypothetically if I were to try and plan for the future and say I'm going to invest in an index fund that has historically returned on average 7.7% per year. What percentage annual return should I actually put in the interest rate section on this calculator? 

It doesn't normally make sense to use a figure that you only have a 50:50 chance of achieving, but equally, using a worst case scenario would be overly pessimistic. A common compromise is to use the lower quartile or quintile to give a return achieved in 75-80% of cases. It is then fairly unlikely you will be worse off than this. This would be in the ballpark of half the mean return for 100% equities, given the high dispersion around the mean.

But a compound interest calculator is not the right tool for predicting investment returns. Better to backtest and/or use Monte Carlo analysis. This is because the sequence of returns makes a big difference. Even then, it needs to be taken with a large pinch of salt.

Do you know of a website that has such a calculation tool because every investing type tool has similar calculator to the one I linked.

masonic

SneakySpectator said:

masonic said:

SneakySpectator said:

masonic said:

SneakySpectator said:

masonic said:

The probability of getting at least 7.7% is still about 50%, so your stated 60-70% chance of getting that return from Premium Bonds would still make PB the better investment if true.

If you can achieve 7.7% with 60-70% likelihood, the average return must be higher than 7.7%.

Either you're gaslighting me or I just don't get it, sorry.

The probability of the 20 year period averaging out to 7.7% per year is very high, close to 100%

Some years will be +10%, others -8%, others +14% but after a long enough period of time the probability of all those individual years averaging out to 7.7% is very high. If every year was 7.7% for 20 years, your average return would be smack bang on 7.7%. But since each year gives different returns, the probability of getting smack bang on 7.7% is low, but the probability of getting very close to that number is high...

I don't know where you're getting your reasoning from.

Maybe you can use a different example to help me understand?

I'm not gaslighting. You are fundamentally misunderstanding the nature of investment returns and probability. When the standard deviation is large, there is a large dispersion around the average.

Take this cFireSim simulation for 100% equities over 20 years using historical data: https://www.cfiresim.com/9bdf7cc4-f995-437e-9c24-b3fe4d5a27eb

This is in USD, but we'll have to live with that. I have set inflation to zero. Starting with $50,000, the median return (the minimum outcome for 50% of 20 year periods) is $183k - close to your average figure, but look at the dispersion: 10% of periods ended with $84k, the standard deviation is $111k.

The probability of getting at least the median return is around 50%, with a fair likelihood of achieving meaningfully less. You would have a 90% chance of achieving at least a 2.6% return based on this historical data, and only a 50% chance of achieving at least 6.7%.

I still don't understand but I do believe you. So hypothetically if I were to try and plan for the future and say I'm going to invest in an index fund that has historically returned on average 7.7% per year. What percentage annual return should I actually put in the interest rate section on this calculator? 

It doesn't normally make sense to use a figure that you only have a 50:50 chance of achieving, but equally, using a worst case scenario would be overly pessimistic. A common compromise is to use the lower quartile or quintile to give a return achieved in 75-80% of cases. It is then fairly unlikely you will be worse off than this. This would be in the ballpark of half the mean return for 100% equities, given the high dispersion around the mean.

But a compound interest calculator is not the right tool for predicting investment returns. Better to backtest and/or use Monte Carlo analysis. This is because the sequence of returns makes a big difference. Even then, it needs to be taken with a large pinch of salt.

Do you know of a website that has such a calculation tool because every investing type tool has similar calculator to the one I linked.

There is https://www.portfoliovisualizer.com

This can do backtesting and Monte Carlo, with some stress testing options, so can be used to build a fairly good picture of the spread of outcomes. You already have the cFireSim link from my earlier post.

Linton

SneakySpectator said:

masonic said:

SneakySpectator said:

I don't follow. If flipping a coin averages at 50% heads or tails. Then over a long period of flips the probability of getting 50% heads and 50% tails becomes very close to 100%.

Incorrect. Suppose you flip a coin 4000 times. The probability of getting exactly 2000 heads is just 1.26%. The probability of fewer than 2000 heads is 49.37% and the probability of greater than 2000 heads is 49.37%. This seems to be the root cause of your misunderstanding.

With investment returns, the standard deviation is quite large, so the range of possible, or even likely, outcomes over a 20 year period might surprise you.

Yes the probably of getting exactly 2000 heads and 2000 tails is low, but the probably of getting very close to those figures is extremely high.

I just did a simulated coin flip and first try got extremely close to 50% / 50%



I bet you could run this simulator all day and you'd get 50% / 50% give or take 1% either way. So applying that to the 7.7% return for 20 years would give you an expected multiple 5.46x your money. But in reality it's probably somewhere in the region of 4.5x - 6.6x. Getting smack bang on 5.46x would be really unlikely. 

With coin tosses there is the obvious underlying probability of 50% which will be approached as the number of tosses approaches infinity.

With a coin toss the number of possible outcomes is 2 - either heads or tails.   With investments the return could be anywhere between  -100% and plus some undefined large % . Is there an obvious underlying probablility of some particular investment return over an extended time period?  If so what time period would be meaningful in determining its value without the result being compromised by shorter term events?   We can see for example that 20 years is not sufficient. Is a lifetime sufficient? If not does the average matter?

 My second point is that a coin tossed today will have the same expected result in say 100 years time.  Can the investment returns of say 200 years ago be usefully compared with those of today? Or do changes in technology and the global economic environment affect average returns?  If they do there is no meaningful future average return.

So in summary it seems to me that in practice assuming any particular average return can be very misleading.  With investing you have to accept that the future is unknown.

Pollycat

Linton said:

SneakySpectator said:

masonic said:

SneakySpectator said:

I don't follow. If flipping a coin averages at 50% heads or tails. Then over a long period of flips the probability of getting 50% heads and 50% tails becomes very close to 100%.

Incorrect. Suppose you flip a coin 4000 times. The probability of getting exactly 2000 heads is just 1.26%. The probability of fewer than 2000 heads is 49.37% and the probability of greater than 2000 heads is 49.37%. This seems to be the root cause of your misunderstanding.

With investment returns, the standard deviation is quite large, so the range of possible, or even likely, outcomes over a 20 year period might surprise you.

Yes the probably of getting exactly 2000 heads and 2000 tails is low, but the probably of getting very close to those figures is extremely high.

I just did a simulated coin flip and first try got extremely close to 50% / 50%



I bet you could run this simulator all day and you'd get 50% / 50% give or take 1% either way. So applying that to the 7.7% return for 20 years would give you an expected multiple 5.46x your money. But in reality it's probably somewhere in the region of 4.5x - 6.6x. Getting smack bang on 5.46x would be really unlikely. 

With coin tosses there is the obvious underlying probability of 50% which will be approached as the number of tosses approaches infinity.

With a coin toss the number of possible outcomes is 2 - either heads or tails.   With investments the return could be anywhere between  -100% and plus some undefined large % . Is there an obvious underlying probablility of some particular investment return over an extended time period?  If so what time period would be meaningful in determining its value without the result being compromised by shorter term events?   We can see for example that 20 years is not sufficient. Is a lifetime sufficient? If not does the average matter?

 My second point is that a coin tossed today will have the same expected result in say 100 years time.  Can the investment returns of say 200 years ago be usefully compared with those of today? Or do changes in technology and the global economic environment affect average returns?  If they do there is no meaningful future average return.

So in summary it seems to me that in practice assuming any particular average return can be very misleading.  With investing you have to accept that the future is unknown.

Even though the OP says:

SneakySpectator said:

While with the stock market your chances are effectively 100%, not guaranteed of course but statistically very high probability.

Hmmm.

Exodi

aroominyork said:

boingy said:

SneakySpectator said:

I think premium bonds are fantastic for short periods where you need some ok returns on your money without committing it to something that could lose value in the short term. 

Premium bonds are terrible for short periods. You are not in the draw for at least a month when you buy them.

I used to think that but it is wrong. If you deposit on say 31 March, you will be in the 1 May draw and can withdraw your funds the same day. The draw operates on funds held during the previous month.

I think this may be a tad pedantic.

Draw is the action of selecting winners, and to that end @boingy's comment is completely correct that you are "not in the draw for at least a month". You may argue that the funds can be eligible or counted towards the next prize draw almost straight away, but it doesn't change the fact that you will not be included in a draw for at least a month.

eskbanker

Exodi said:

aroominyork said:

boingy said:

SneakySpectator said:

I think premium bonds are fantastic for short periods where you need some ok returns on your money without committing it to something that could lose value in the short term. 

Premium bonds are terrible for short periods. You are not in the draw for at least a month when you buy them.

I used to think that but it is wrong. If you deposit on say 31 March, you will be in the 1 May draw and can withdraw your funds the same day. The draw operates on funds held during the previous month.

I think this may be a tad pedantic.

Draw is the action of selecting winners, and to that end @boingy's comment is completely correct that you are "not in the draw for at least a month". You may argue that the funds can be eligible or counted towards the next prize draw almost straight away, but it doesn't change the fact that you will not be included in a draw for at least a month.

While it's undoubtedly factual that purchases aren't in the draw for at least a full calendar month, the point is that in itself this doesn't disadvantage PB holders versus those holding standard savings accounts, who also won't receive interest other than in arrears, so it's not the killer argument that some perceive it to be!

Granted, the comparison can be skewed away from PBs if buying and/or selling at inopportune times, but if buying at the end of one month and selling straight after the first draw, then the return from that one draw can legitimately be measured against the one month's interest that would have been earned from an equivalent savings account.

SneakySpectator

Linton said:

SneakySpectator said:

masonic said:

SneakySpectator said:

I don't follow. If flipping a coin averages at 50% heads or tails. Then over a long period of flips the probability of getting 50% heads and 50% tails becomes very close to 100%.

Incorrect. Suppose you flip a coin 4000 times. The probability of getting exactly 2000 heads is just 1.26%. The probability of fewer than 2000 heads is 49.37% and the probability of greater than 2000 heads is 49.37%. This seems to be the root cause of your misunderstanding.

With investment returns, the standard deviation is quite large, so the range of possible, or even likely, outcomes over a 20 year period might surprise you.

Yes the probably of getting exactly 2000 heads and 2000 tails is low, but the probably of getting very close to those figures is extremely high.

I just did a simulated coin flip and first try got extremely close to 50% / 50%



I bet you could run this simulator all day and you'd get 50% / 50% give or take 1% either way. So applying that to the 7.7% return for 20 years would give you an expected multiple 5.46x your money. But in reality it's probably somewhere in the region of 4.5x - 6.6x. Getting smack bang on 5.46x would be really unlikely. 

With coin tosses there is the obvious underlying probability of 50% which will be approached as the number of tosses approaches infinity.

With a coin toss the number of possible outcomes is 2 - either heads or tails.   With investments the return could be anywhere between  -100% and plus some undefined large % . Is there an obvious underlying probablility of some particular investment return over an extended time period?  If so what time period would be meaningful in determining its value without the result being compromised by shorter term events?   We can see for example that 20 years is not sufficient. Is a lifetime sufficient? If not does the average matter?

 My second point is that a coin tossed today will have the same expected result in say 100 years time.  Can the investment returns of say 200 years ago be usefully compared with those of today? Or do changes in technology and the global economic environment affect average returns?  If they do there is no meaningful future average return.

So in summary it seems to me that in practice assuming any particular average return can be very misleading.  With investing you have to accept that the future is unknown.

I see what you're getting at but for this hypothetical, we have to use existing past data. So you look back through the life time of the index fund, in the case of MSCI world index and take the absolute worst year and the absolute best year and make those your limits. So a -100% return has a 0% chance because it's never happened. 

Having a quick look at MSCI returns by year I can see that the best year was 42.8% and the worst year was -40.33% so we use those as the limits. 

How does that affect the overall probability? Maybe it doesn't I dunno I'm totally lost at this point haha.