Reversion to Mean

[marko]

Right up front: I'm far from a math wiz.

I keep reading about stocks "reverting to mean".
Yes, I'm not such a dope that I don't know what 'mean' means, but I'm wondering how the mean is defined.

It is a 3 year, 5 year, 10 year average? Over time a stock's mean should/could change considering the wild market swings over the past 15 years; I'd assume 2008 and 2020 would figure into a long term average to some extent.

Companies change their positions in the market over time as well (ex: GE), so how is the mean arrived at?

Per usual, I could be entirely in the weeds here so apologies in advance if so.

 

[pb4uski]

....I'm wondering how the mean is defined.

It is a 3 year, 5 year, 10 year average? Over time a stock's mean should/could change considering the wild market swings over the past 15 years; I'd assume 2008 and 2020 would figure into a long term average to some extent....

mean is average.... sum divided by the number of entries.

median is the middle value of a series of numbers in order

In terms of "reversion to the mean" it is just a concept.... that over time things will average out and that statistical outperformance will eventually be offset by statistical underperformance and vice versa.

Median	42
Mean	46.9
	8
	8
	10
	20
	24
	26
	26
	31
	33
	33
	35
	36
	37
	40
	42
	44
	45
	48
	48
	54
	56
	58
	73
	81
	83
	84
	89
	90
	98

 

[Which Roger]

The term always makes me think of the following quote: "In the short term, the stock market is a voting machine. In the long term, it's a weighing machine". In other words, long-term performance has a lot less noise than short-term.

 

[fosterscik]

I think it's just buzzy jargon people say to sound as if they know what's going to happen. Shorthand for "things look too pricey at the moment so its going to drop some". I don't think there's any agreed expected mean price. Any significant drop will be described by the talking heads as a mean reversion...

 

[audreyh1]

Right up front: I'm far from a math wiz.

I keep reading about stocks "reverting to mean".
Yes, I'm not such a dope that I don't know what 'mean' means, but I'm wondering how the mean is defined.

It is a 3 year, 5 year, 10 year average? Over time a stock's mean should/could change considering the wild market swings over the past 15 years; I'd assume 2008 and 2020 would figure into a long term average to some extent.

Companies change their positions in the market over time as well (ex: GE), so how is the mean arrived at?

Per usual, I could be entirely in the weeds here so apologies in advance if so.

It’s just a hackneyed phrase that’s often thrown out there, and the speaker rarely defines the mean, because that’s where people would start to argue.

And not every seemingly random phenomenon has the statistics of a bell curve.

I, personally, have never seen it happen (reversion to some magical mean).

Means nothing. [Yes, pun intended!]

 

[NW-Bound]

Stocks on the average go up. Inflation makes sure that they do, if nothing else.

Hence, when talking about reversion to the mean, nobody says a stock will go back to its average value of the last 3, 5, or 10 years, the same as nobody says house prices will go back to the average value of the last several years, or people's salary, or whatever have you.

What they are talking about is more about stock valuation, such as P/E ratio, book value, or the yearly gain. Stock prices cannot outrun the GDP growth forever. If they do, nobody has to work, and we can just sit at home selling stocks back and forth. Heck, we are almost doing that already, and in fact are selling back and forth not just stocks, but also just bits, like Bitcoin and its ilk and NFTs. And fewer and fewer people bother to work, as all the Help Wanted signs are telling us.

So, no more going back to the old days. No reversion anything.

Hurrah, it's a new paradigm. It's heaven on earth. Prosperity without work! What's not to like?

 

[GenXguy]

The correction is coming. You heard it here first.

 

[FREE866]

The correction is coming. You heard it here first.

Corrections are always coming. Corrections of 10-20% are part of the process. As are bear markets! That volatility is the emotional price we pay for outstanding long term returns of stocks. The historical numbers prove this over and over and over again, yet people still don't get it. They cling to the headlines of the day and consequentially make awful decisions that hurt their goals.

 

[Exchme]

People are just opining that they think stocks are expensive.

I'm sure if you go back a decade and look at posts on this site, you will see fear, gloom and doom shouted about back then too and it turned out to be a pretty good decade for stock returns.

It has to be this way, each and every day in the market has to be balanced where an equal amount of money is spent by buyers as paid to sellers at what both think is an acceptable price. So there can by definition never be a clear cut moment where you want to be in or out of the market.

 

[Markola]

It means,

What goes up must come down
Spinnin' wheel got to go 'round
Talkin' 'bout your troubles it's a cryin' sin
Ride a painted pony let the spinnin' wheel spin

 

[GravitySucks]

Always thought of it as running a linear regression best fit type of line of historical data and if the recent actual data is above that line it will soon have to be below that line.
Here's a chart showing you should have gotten out of the market in 2019 and missed out on the last 2 years growth.
Eventually we will be below the line, but nobody can tell us when.

 

[twaddle]

Mean-reversion doesn't apply to price. It applies to characteristics that are rooted to some fundamental value.

The classic example is height. The mean height is a function of biology (hormones, energy input, etc). So if you see a bunch of very tall people, it's likely that the next person you see won't be very tall, because you expect mean reversion for height.

In the context of stocks, it's usually applied to annual returns. If you believe that returns are rooted in fundamentals, like earnings growth, GDP growth, etc., then you'd expect returns to be mean-reverting.

 

[NW-Bound]

Always thought of it as running a linear regression best fit type of line of historical data and if the recent actual data is above that line it will soon have to be below that line.
Here's a chart showing you should have gotten out of the market in 2019 and missed out on the last 2 years growth.
Eventually we will be below the line, but nobody can tell us when.View attachment 39283

Interesting chart.

I noticed that it starts from 2009. If the linear regression starts from an earlier year, say 2000 or 1990, I wonder what the chart looks like. I would think it is harder to linear fit cataclysmic crashes of 2003 and 2009, or the dot-com mania of 2000. These events are way out of the norm.

 

[donheff]

Here is one that goes all the way back.

 

[NW-Bound]

Thanks.

Note that the straight-line fit is on a log chart, hence it's an exponential regression, and not a linear regression.

The deviations off the straight-line are huge. And the market can stay above trend or below trend for decades.


PS. Caveat: This chart shows the stock price. It does not include the gain from dividends. Dividend yield is so low now, but back in the past it was often higher than 5%.

 

[The Cosmic Avenger]

Mean-reversion doesn't apply to price. It applies to characteristics that are rooted to some fundamental value.

The classic example is height. The mean height is a function of biology (hormones, energy input, etc). So if you see a bunch of very tall people, it's likely that the next person you see won't be very tall, because you expect mean reversion for height.

In the context of stocks, it's usually applied to annual returns. If you believe that returns are rooted in fundamentals, like earnings growth, GDP growth, etc., then you'd expect returns to be mean-reverting.

Except that this is a good example of the Gambler's Fallacy. When flipping coins, the chance of 100 heads in a row is...vanishingly small, but after 99 heads, what's the chance that the next flip is heads?

It's always 1/2, regardless of the past.

This is just another way of saying that past performance cannot predict future returns.

 

[Lsbcal]

Mean reversion refers to the market being a nasty brute. Last week the market was very mean to me!

 

[NW-Bound]

Except that this is a good example of the Gambler's Fallacy. When flipping coins, the chance of 100 heads in a row is...vanishingly small, but after 99 heads, what's the chance that the next flip is heads?

It's always 1/2, regardless of the past.

This is just another way of saying that past performance cannot predict future returns.

A coin toss is a random walk. Each coin toss is independent of the past tosses.

Market movement is not a true random walk. After a long string of gains, people are edgy and want to book profits. Conversely, after a large decline, more people are willing to bid the price back up. The coin does not have memory. People have memory.

All this does not mean that the market automatically goes up after exactly so many days of decline, as if on a timer. It's just that the probability of an up day is higher and higher after more and more days of decline. Of course, the market often turns back down after a single up day. I have seen this enough. That's why market timing is so difficult.

I make short-term trades based on prices via option writing (out-of-the-money calls after a string of up days, out-of-the-money puts after many down days). But the stocks I write options on, I hold them long-term based on their fundamentals. No betting on no-earning stocks, let alone companies without even any sales. Ditto on bits of any kind. There are no metrics to tell what their reasonable values should be. Pure gambling to me.

 

[audreyh1]

Mean reversion refers to the market being a nasty brute. Last week the market was very mean to me!

Last week was a tiny drop in the bucket!!!

 

[twaddle]

I think that Bogle had the best way to look at it.

Expected return = dividend yield + earnings growth rate + change in P/E.

Sometimes P/E expands (as it has recently) and that fuels most of the return.

Sometimes P/E contracts.

Each year that P/E expansion fuels returns, that adds to the "gravitational pull" towards lower returns in subsequent years. Unless you believe that P/E can grow forever.

 

[Lsbcal]

Last week was a tiny drop in the bucket!!!

But that drop hit me in the eye.

 

[REWahoo]

But that drop hit me in the eye.

Can't hit you in the eye if you don't look.

 

[GravitySucks]

Here is one that goes all the way back.

Thank you.

 

[NW-Bound]

I think that Bogle had the best way to look at it.

Expected return = dividend yield + earnings growth rate + change in P/E.

Sometimes P/E expands (as it has recently) and that fuels most of the return.

Sometimes P/E contracts.

Each year that P/E expansion fuels returns, that adds to the "gravitational pull" towards lower returns in subsequent years. Unless you believe that P/E can grow forever.

The late Bogle talked about P/E reversion, and used it to set his forecast. However, he used that to temper people's expectations, and never claimed he knew when that P/E reversion would happen or how it would happen. It may happen in a hurry, or spread out over many years. Hence, he stressed that his forecast of future returns was for long periods of a decade or more, not for next year.


PS. Forgot to say, P/E is somewhat tied to the interest rate. When the interest rises, P/E goes down. Interest down, P/E up. It's because stocks have to compete with other investment instruments.

Can interest keep on going down? To negative?

 

[NW-Bound]

Last week was a tiny drop in the bucket!!!

I tried to catch these drops, and they added up.