Challenging ShinyThing assumptions.

[Shiny Things]

Yup totally agree. How can you believe someone with zero contribution and has no banker. All I see is TCSS only in every thread.

http://forums.hardwarezone.com.sg/search.php?searchid=34265869&pp=25

Uh mate that search link doesn't work. Link directly to the post, maybe?

 

[Purplestars]

Well sometimes I feel if you've already sat out of 10 years of a bull run, you might as well sit it out all the way till the crash. So what if it takes 5 years to crash? It is more likely to crash sooner rather than later

 

[Shiny Things]

Your portfolio backtesting is based on average market performance. We all know that in the markets, there are extreme winners and majority are losers in long run.

How would you answer to Ole Peter testing?

I mean, my first answer, and it's to you rather than Mr Peter, is that the entire point of owning an index is that you don't have to try to pick the winners; you end up owning them automatically.

The thing I suspect that ST fails to verify is the cash to investment ratio. He is just using emergency fund as a heuristics. After investors had enough redundant cash, they will go overbetting.

This is why people do what's called Monte-Carlo testing. Instead of just backtesting the portfolio, they say "OK, what will the underlying assets probably do in the future?" They use a mean return and an expected volatility (which are, I stress, estimates; we're trying to predict the future here)

And then, using those inputs, they model the outcome across tens of thousands of possible "worlds", to use your term. That way, you can come up with an expected mean return, but you can also see what the tails of the distribution look like. What's the worst-case scenario?

And you can do this yourself - so, let's do it!

The question at issue is: "If 'investments' is the '110 minus your age' portfolio, and 'cash' is SGD cash, what ratio of cash to investments provides the best return?".

As a starting point, let's use the US market: allocating 80-20 between US stocks and US bonds, and then varying the cash portion between 0% and 50%. The handy-dandy tools at PortfolioVisualizer let us see how this portfolio would have performed between 1987 and today... and, no surprise, the answer is that the portfolios with less cash perform significantly better. There are higher drawdowns, obviously, but you get commensurately higher returns; and adding a higher cash weighting doesn't give you any better risk-adjusted returns (the "Sharpe Ratio" number).

So all other things being equal, it looks like you want to own more assets (stocks and bonds) and less cash.

Let's refine this, though. Your original complaint was that you thought I was only back-testing based on historical returns, and not looking at the possible worst-case outcomes. We can build our own monte-carlo simulator, plug in some expected returns and volatilities for the assets in the portfolio, and see what the tails of the distribution look like.

I'll write up the script for this tonight or tomorrow and post it on Github. (Disclaimer, I am not actually very good at Python, so my code is going to be awful. Feel free to laugh at it.)

The high-level flow looks a bit like this:

1) Initialise your model of the markets with your variables of choice. I'm going to assume that the returns of three assets - stocks, bonds, and cash - are normally distributed and uncorrelated, which is obviously not 100% right, but it'll give us an idea. (If anything, zero correlation understates how well the 100%-stock-and-bond portfolio will perform). I'm also going to assume continuous dividends and constant interest rates because I really can't be stuffed keeping track of discrete dividend payments (every six months or year) and variable interest rates; someone else is welcome to do this.

2) Generate a bunch of price trajectories for the assets based on that market model. 10,000 or 100,000 is a good number.

3) Find out what the ending value of the portfolio will be across those 10k or 100k trajectories.

4) Draw a nice chart of the distribution of the ending values, and put some summary statistics in there.

I think—and stop me if I'm wrong here—the key question that you'd like to know is: what does the lower tail of the distribution look like? How likely is it that a stock-and-bond portfolio will do worse than if you'd just held everything in cash?

Is that reasonable? I want to make sure I'm answering the question that you're actually asking.

Company earning of solid dependable companies did not disappear overnight even when the stock market plunged 50%. For example, you can check MacDonald's earnings during the GFC.

One Weird Thing that a few smart cookies realised in '08-'09: people slashed their spending on "large luxuries" (cars, luxury goods, things like that) but they increased their spending on "small luxuries". The hypothesis, I guess, is that people couldn't afford larger splurges, so they spent on smaller splurges instead to make themselves feel better.

Mass-affluent retailers like Whole Foods (a fancy-ass hippie organic supermarket over here; disclaimer, I shop at Whole Foods so I guess that makes me a fancy-ass hippie) did surprisingly well during the downturn.

 

[Shiny Things]

Well sometimes I feel if you've already sat out of 10 years of a bull run, you might as well sit it out all the way till the crash. So what if it takes 5 years to crash? It is more likely to crash sooner rather than later

Bull markets can last for a looooong time.

US stocks ran up in pretty much a straight line from 1983 to 2000. Even the 1987 crash only took the market back to flat for the year.

If you were sitting in cash in January 1993, with the SPX at 443 (after having quadrupled in the last ten years), and you thought "oh man, I'll wait until another crash happens, 1987 wasn't that long ago"... you'd have spent the next seven years waiting for a crash, and the SPX would have tripled again while you were waiting. Even in the depths of the 2002 and 2008 pukes, we never got close to 443 again.

 

[Purplestars]

Bull markets can last for a looooong time.

US stocks ran up in pretty much a straight line from 1983 to 2000. Even the 1987 crash only took the market back to flat for the year.

If you were sitting in cash in January 1993, with the SPX at 443 (after having quadrupled in the last ten years), and you thought "oh man, I'll wait until another crash happens, 1987 wasn't that long ago"... you'd have spent the next seven years waiting for a crash, and the SPX would have tripled again while you were waiting. Even in the depths of the 2002 and 2008 pukes, we never got close to 443 again.

Why would you wait out at 1993 just 6 years after the crash? If you are waiting out 10 years then you would have stayed out at 1997 which would have worked out pretty well since the crash happened 3 years after.

Yes there's always a risk of missing out on gains, but I think the risk is better than risking losses.

 

[Shiny Things]

Why would you wait out at 1993 just 6 years after the crash? If you are waiting out 10 years then you would have stayed out at 1997 which would have worked out pretty well since the crash happened 3 years after.

Yes there's always a risk of missing out on gains, but I think the risk is better than risking losses.

January '97, the S&P 500 was trading 790. It nearly doubled between 1997 and the peak.

If you look in other asset classes: US bonds went up in a nearly straight line from 1979 to today. Aussie housing has gone up and to the right for a good 20 years.

My point is that your 10-year cadence doesn't exist; you're seeing a pattern that isn't there. You're better off being invested and having a strategy to methodically buy and sell, instead of waiting for a "crash" and hoping that you'll be brave enough to actually hit the buy button when that dip comes.

 

[Purplestars]

January '97, the S&P 500 was trading 790. It nearly doubled between 1997 and the peak.

If you look in other asset classes: US bonds went up in a nearly straight line from 1979 to today. Aussie housing has gone up and to the right for a good 20 years.

My point is that your 10-year cadence doesn't exist; you're seeing a pattern that isn't there. You're better off being invested and having a strategy to methodically buy and sell, instead of waiting for a "crash" and hoping that you'll be brave enough to actually hit the buy button when that dip comes.

Well the 1997 spike to 2000 is a false peak. It came tumbling back down quickly afterwards.

As you said other asset classes went up and to the right at the same time. So if you put your money in bonds and property or other defensive investments during this time you won't miss out on too much.

Of course the 10 year cycle thing is by no means a science, it's just a sign that the economic cycle might be in its last legs. You use other indicators and fundamentals to check whether stocks are overvalued before pulling out.

So in 1997, do a p/e ratio check on the market or something since you suspect the business cycle is coming to an end. Still healthy? Ok continue to DCA. 1999-2000 when p/e ratios are going crazy? Time to pull out. Surely the business cycle has run its course by now.

 

[Shiny Things]

As you said other asset classes went up and to the right at the same time. So if you put your money in bonds and property or other defensive investments during this time you won't miss out on too much.

We agree on this. I absolutely don't think that anyone should be 100% in stocks; having some allocation to bonds is a great idea.

Of course the 10 year cycle thing is by no means a science, it's just a sign that the economic cycle might be in its last legs. You use other indicators and fundamentals to check whether stocks are overvalued before pulling out.

So in 1997, do a p/e ratio check on the market or something since you suspect the business cycle is coming to an end. Still healthy? Ok continue to DCA. 1999-2000 when p/e ratios are going crazy? Time to pull out. Surely the business cycle has run its course by now.

So hang on, you've just thrown your 10-year rule out the window there, and replaced it with monitoring P/E ratios (which makes a lot more sense). You've said that "if valuations are reasonable, then keep buying - whether it's eight, ten, or twelve years into the cycle".

My point is that there is no 10-year cycle. Markets turn down when people stop buying, and they don't start and stop buying on a regular 10-year cycle. As you correctly pointed out, valuation is a better measure of whether a market's overstretched or not.

 

[Purplestars]

We agree on this. I absolutely don't think that anyone should be 100% in stocks; having some allocation to bonds is a great idea.



So hang on, you've just thrown your 10-year rule out the window there, and replaced it with monitoring P/E ratios (which makes a lot more sense). You've said that "if valuations are reasonable, then keep buying - whether it's eight, ten, or twelve years into the cycle".

My point is that there is no 10-year cycle. Markets turn down when people stop buying, and they don't start and stop buying on a regular 10-year cycle. As you correctly pointed out, valuation is a better measure of whether a market's overstretched or not.

Eh no, I didn't set a 10-year rule. You set it.

All I said was if you have sat out a bull run for 10 years, you might as well sit out the rest of it, since the chances are the crash will come sooner rather than later. People have to stop buying eventually.

Even our transport minister believes that this theory is statistics, if our MRT has not broken down for a long while then it is due to break down soon.

There is no guarantee that this "strategy" will work better than continuously buying as no one can accurately time the market. But if you are going make a bet that the market will continuously go up, a bet that the market will soon go back down is equally valid since no one actually knows when the crash will come.

But most people I know believe in economic cycles though. It is taught in economics textbooks.

https://courses.lumenlearning.com/boundless-economics/chapter/key-topics-in-macroeconomics/

Time is probable the best indicator of cycle ending. The longer it goes the more likely it will end. Your strategy is to ignore that the slowdown as things will go back up eventually. That might be true, but meh. There are other things you can buy, and if you can live with missing out on gains why not?

 

[Shiny Things]

Eh no, I didn't set a 10-year rule. You set it.

All I said was if you have sat out a bull run for 10 years, you might as well sit out the rest of it, since the chances are the crash will come sooner rather than later. People have to stop buying eventually.

Ahhh, my bad. I thought you were giving 10 years as a hard-and-fast rule; I didn't realise you were saying "if you missed most of the run-up you might as well not chase it". (I think I got confused because a lot of people talk about 10-year cycles: they see 1998 and 2008 and assume that there's inevitably going to be something happening in 2018 as well. It drives me up the wall.)

But most people I know believe in economic cycles though. It is taught in economics textbooks.

https://courses.lumenlearning.com/boundless-economics/chapter/key-topics-in-macroeconomics/

You're giving me flashbacks to my economics degree! (I still have some of my old textbooks floating around somewhere, the ones that I couldn't sell back at the end of semester. Classic example of supply and demand right there; zero demand, lots of supply...)

Also, you'll want to be careful about conflating economic cycles and stock-market cycles; the two generally go together but not necessarily.

Time is probable the best indicator of cycle ending. The longer it goes the more likely it will end.

I think this is where we disagree, but at least we agree on what we disagree on. You think "the longer it goes, the closer it is to ending"; I think "bull markets don't die of old age". My general position is that markets can keep going up as long as there are buyers; markets don't just stop going up because the bull market's been going for a long time. There needs to be an actual trigger for the markets to go down.

 

[w1rbelw1nd]

And you can do this yourself - so, let's do it!

The question at issue is: "If 'investments' is the '110 minus your age' portfolio, and 'cash' is SGD cash, what ratio of cash to investments provides the best return?".

As a starting point, let's use the US market: allocating 80-20 between US stocks and US bonds, and then varying the cash portion between 0% and 50%. The handy-dandy tools at PortfolioVisualizer let us see how this portfolio would have performed between 1987 and today... and, no surprise, the answer is that the portfolios with less cash perform significantly better. There are higher drawdowns, obviously, but you get commensurately higher returns; and adding a higher cash weighting doesn't give you any better risk-adjusted returns (the "Sharpe Ratio" number).

So all other things being equal, it looks like you want to own more assets (stocks and bonds) and less cash.

Let's refine this, though. Your original complaint was that you thought I was only back-testing based on historical returns, and not looking at the possible worst-case outcomes. We can build our own monte-carlo simulator, plug in some expected returns and volatilities for the assets in the portfolio, and see what the tails of the distribution look like.

I'll write up the script for this tonight or tomorrow and post it on Github. (Disclaimer, I am not actually very good at Python, so my code is going to be awful. Feel free to laugh at it.)

The high-level flow looks a bit like this:

1) Initialise your model of the markets with your variables of choice. I'm going to assume that the returns of three assets - stocks, bonds, and cash - are normally distributed and uncorrelated, which is obviously not 100% right, but it'll give us an idea. (If anything, zero correlation understates how well the 100%-stock-and-bond portfolio will perform). I'm also going to assume continuous dividends and constant interest rates because I really can't be stuffed keeping track of discrete dividend payments (every six months or year) and variable interest rates; someone else is welcome to do this.

2) Generate a bunch of price trajectories for the assets based on that market model. 10,000 or 100,000 is a good number.

3) Find out what the ending value of the portfolio will be across those 10k or 100k trajectories.

4) Draw a nice chart of the distribution of the ending values, and put some summary statistics in there.

I think—and stop me if I'm wrong here—the key question that you'd like to know is: what does the lower tail of the distribution look like? How likely is it that a stock-and-bond portfolio will do worse than if you'd just held everything in cash?

Is that reasonable? I want to make sure I'm answering the question that you're actually asking.

You really shouldnt even bother spending your time crafting out such a well thought out answer when someone isnt even reciprocating with a half decent answer.

Till now, he is only criticising your position and shared nothing about his own cash %. Really a great contribution to the forum

 

[Shiny Things]

And you can do this yourself - so, let's do it!
[...]
I'll write up the script for this tonight or tomorrow and post it on Github. (Disclaimer, I am not actually very good at Python, so my code is going to be awful. Feel free to laugh at it.)


I think—and stop me if I'm wrong here—the key question that you'd like to know is: what does the lower tail of the distribution look like? How likely is it that a stock-and-bond portfolio will do worse than if you'd just held everything in cash?

Is that reasonable? I want to make sure I'm answering the question that you're actually asking.

So, because I keep my promises, here's my monte-carlo portfolio simulator in Python, up on my Github. It's nothing too fancy; just a quick two-hour job over a quiet beer at the brewery across the street from my house, and also I am bad at Python so be gentle.

All it does is ask: "how likely is it that an 80-20 stock-and-bond portfolio will outperform cash? No rebalancing, no glide path, just a straight 80-20."

I used the following assumptions:

• SGD cash yields one-and-an-eighth percent
• ES3 returns 6% total return with 10.5% vol;
• A35 returns 2% total return with 3.5% vol.

The return assumptions are conservative; the volatility assumptions are just the last three years' annualised volatility.

The upshot is that with these assumptions, over a 10-year period, an 80-20 stock-and-bond portfolio outperforms cash 90% of the time. And these assumptions are conservative: no rebalancing, no glide path, an aggressive allocation and deliberately low return assumptions.


The important paragraph for ExtremeWays to read: if you're less than 100% invested, you're just going to reduce your returns until they converge with the return on cash. So 90% of the time, having any allocation to cash at all will reduce your returns. Why would you take those odds?

 

[Purplestars]

So, because I keep my promises, here's my monte-carlo portfolio simulator in Python, up on my Github. It's nothing too fancy; just a quick two-hour job over a quiet beer at the brewery across the street from my house, and also I am bad at Python so be gentle.

All it does is ask: "how likely is it that an 80-20 stock-and-bond portfolio will outperform cash? No rebalancing, no glide path, just a straight 80-20."

I used the following assumptions:

• SGD cash yields one-and-an-eighth percent
• ES3 returns 6% total return with 10.5% vol;
• A35 returns 2% total return with 3.5% vol.

The return assumptions are conservative; the volatility assumptions are just the last three years' annualised volatility.

The upshot is that with these assumptions, over a 10-year period, an 80-20 stock-and-bond portfolio outperforms cash 90% of the time. And these assumptions are conservative: no rebalancing, no glide path, an aggressive allocation and deliberately low return assumptions.


The important paragraph for ExtremeWays to read: if you're less than 100% invested, you're just going to reduce your returns until they converge with the return on cash. So 90% of the time, having any allocation to cash at all will reduce your returns. Why would you take those odds?

Is it possible to do a monte carlo scenario of someone only starting to buy right after a crash, holding it for 10 years, selling it and going 100% bonds until the next crash and the market has gone up 10%?

 

[unhinged_loon]

Well sometimes I feel if you've already sat out of 10 years of a bull run, you might as well sit it out all the way till the crash. So what if it takes 5 years to crash? It is more likely to crash sooner rather than later

Just to note: this subscribes to the gambler's fallacy.

http://www.investopedia.com/university/behavioral_finance/behavioral7.asp

 

[IronMac]

Is it possible to do a monte carlo scenario of someone only starting to buy right after a crash, holding it for 10 years, selling it and going 100% bonds until the next crash and the market has gone up 10%?

This is one of the reasons why I decided to stop participating in the Financial forum. This is a timewaster of a question.

 

[IronMac]

We agree on this. I absolutely don't think that anyone should be 100% in stocks; having some allocation to bonds is a great idea.

I think it's a terrible idea to have any sort of allocation in bonds but I made my view clear here a couple of years ago.

The equity market historically has always gone up to the right. You get some bumps here and there but if you ignore that you are good to go. If you can't take a 20% hit to your overall valuation then you might as well stay out.

My last 12 months and YTD returns in the US market has been about 25% each just off the top of my head. Not bad. My Canadian dividend stocks have returned 16% in the past 12 months. Better than expected.

Since 2007 my portfolio has grown 8x. I expect it to slow down some but who knows?

 

[Mecisteus]

I think it's a terrible idea to have any sort of allocation in bonds but I made my view clear here a couple of years ago.

The equity market historically has always gone up to the right. You get some bumps here and there but if you ignore that you are good to go. If you can't take a 20% hit to your overall valuation then you might as well stay out.

0% bonds allocation is valid to me as long you are a high risk taker and you have a longer term view. And of course assuming the stocks portion is soundly selected.

Personally, I don't like bonds too because I can stomach the big volatilities. I will just adjust my cash and stocks allocation.

 

[frenchboxers]

Well sometimes I feel if you've already sat out of 10 years of a bull run, you might as well sit it out all the way till the crash. So what if it takes 5 years to crash? It is more likely to crash sooner rather than later

Why not use 50/50 allocation,that way u still get to enjoy some of the growth while waiting.

 

[Purplestars]

Just to note: this subscribes to the gambler's fallacy.

http://www.investopedia.com/university/behavioral_finance/behavioral7.asp

Tell that to Khaw Boon Wah

 

[Purplestars]

Why not use 50/50 allocation,that way u still get to enjoy some of the growth while waiting.

I'm on about 50/50 now, but sometimes it feels pointless.

I'd agree with Ironmac, of you think stocks are right then you should go stocks all the way. If you think now is not a good time to buy stocks, you should stay fully away. Why go halfway neither here nor there?

If you can't take the 20%-50% drop, you stay out. Especially for newbie investors who have missed the bulk of the bull run already.

The only reason to go bond allocation is when you are ageing and need to keep a big portion of your money safe for retirement.