Do You Believe 10 Years Return Data Is Just Noise?

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Topic Author
Random Walker
Posts: 3955
Joined: Fri Feb 23, 2007 8:21 pm

Do You Believe 10 Years Return Data Is Just Noise?

Post by Random Walker » Wed May 01, 2019 10:03 pm

Larry Swedroe has a line he says frequently, something to the effect “for most investors 3 years is a long time, 5 years a very long time, and 10 years an eternity; but financial economists know that 10 years is just noise in the financial data”

So many of us cite data looking back various amounts of time, it got me wondering how many appreciate this. I find timeframes interesting. Longer timeframes for data certain strengthen arguments for persistence, but go back too far and it seems natural to question relevance to present circumstance.

I like the concept of out of sample tests: different time periods, different geographic markets, and in some cases even different asset classes.

Dave

User avatar
gmaynardkrebs
Posts: 1621
Joined: Sun Feb 10, 2008 11:48 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by gmaynardkrebs » Wed May 01, 2019 10:40 pm

Random Walker wrote:
Wed May 01, 2019 10:03 pm
Larry Swedroe has a line he says frequently, something to the effect “for most investors 3 years is a long time, 5 years a very long time, and 10 years an eternity; but financial economists know that 10 years is just noise in the financial data”

So many of us cite data looking back various amounts of time, it got me wondering how many appreciate this. I find timeframes interesting. Longer timeframes for data certain strengthen arguments for persistence, but go back too far and it seems natural to question relevance to present circumstance.

I like the concept of out of sample tests: different time periods, different geographic markets, and in some cases even different asset classes.

Dave
Appreciate? 1%; pay lip service to? 99%.

User avatar
permport
Posts: 159
Joined: Sat Mar 31, 2018 11:20 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by permport » Wed May 01, 2019 11:11 pm

I agree that 10 years of return data is noise to the extent that it isn't enough to be indicative of what will happen over a lifetime of investing.

However, somewhat ironically, I think 10 years is a long enough time period in the future for forecasting purposes. Bogle himself has done this, in that he would forecast the 10 year return on stocks based on current valuations. He was remarkably consistent and his margin of error rather small.
Buy right and hold tight.

User avatar
JoMoney
Posts: 7118
Joined: Tue Jul 23, 2013 5:31 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by JoMoney » Wed May 01, 2019 11:17 pm

There's a tremendous gap between the stock returns for the ten years ending in March of 2009, relative to the ten years ending one year prior, or even one year later...
Is that noise?
"Forecasts" from ten years prior could have been any number of things all over the place from positive to negative returns and if you picked the right day/month as an example proved any of them correct.
"To achieve satisfactory investment results is easier than most people realize; to achieve superior results is harder than it looks." - Benjamin Graham

BJJ_GUY
Posts: 154
Joined: Wed Mar 13, 2019 7:45 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by BJJ_GUY » Wed May 01, 2019 11:19 pm

Understanding market cycles, and valuations/pricing is the importance generally referred to with the '10 year' references. Basically, 10 years is the most commonly used estimate of time for an economic and/or market cycle.

If ten years is simply a proxy for peak-to-peak and/or trough-to-trough time horizons, this definitely informs us with the estimated return data that answers the question of how prices will eventually reflect value -- but doesn't answer the when (despite the silly 10 year approximation in the title). Again, the timing element is not answered here, so don't attack!

User avatar
mrspock
Posts: 400
Joined: Tue Feb 13, 2018 2:49 am
Location: Vulcan

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by mrspock » Wed May 01, 2019 11:25 pm

Random Walker wrote:
Wed May 01, 2019 10:03 pm
Larry Swedroe has a line he says frequently, something to the effect “for most investors 3 years is a long time, 5 years a very long time, and 10 years an eternity; but financial economists know that 10 years is just noise in the financial data”

So many of us cite data looking back various amounts of time, it got me wondering how many appreciate this. I find timeframes interesting. Longer timeframes for data certain strengthen arguments for persistence, but go back too far and it seems natural to question relevance to present circumstance.

I like the concept of out of sample tests: different time periods, different geographic markets, and in some cases even different asset classes.

Dave
I think Larry's statement must be taken in context, and I'm not certain what the context was when he said this.

That said, I can say that Renaissance Technologies, Two Sigma and their AI powered Hedge Fund friends do not have large data warehouses because historical data is "noise", they train their (machine learning) models of historical information from a large diverse of data sets going back as many years as they can possible get. They put huge amounts of money into curation of these datasets (they are some of the most valuable IP in the world, next to the ML models themselves), and the further back they go the more valuable they are. In the world I'm in, it's similar.

What may seem like "noise" to humans, is most certainly not. Does the signal to noise ratio diminish if you go back too far? That depends on the thing you are building a model for. For the stock market though, its more likely case is it heavily correlates if you put the data in the right context, and this is what ML training tries to do. For situation X (fed rate rate, GDP, day of week, weather, earnings data, volumes, time of day, time of year etc), I predict Y (up/down) and therefore I do Z (sell/buy).

Anyways, I always laugh when people say "past performance doesn't predict the future (performance)". Really? That's why these guys are blowing billions on machine learning? I don't think so. :)

Dead Man Walking
Posts: 824
Joined: Wed Nov 07, 2007 6:51 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Dead Man Walking » Wed May 01, 2019 11:27 pm

Dave,

I’m a skeptic by nature; therefore, I don’t believe historical returns are relevant when predicting future returns. I believe that it may be different in the future. I agree with Larry’s idea that a decade is a short timeframe when predicting market returns. However, I believe that investors and the general public may suffer from “future shock.” The 21st century may be much different than any of the previous centuries. Globalization, technology, climate change, artificial intelligence, the art of war, terrorism, and central banking are some of the factors that may make this century different from the past. The disparity in the distribution of wealth may be the common factor between the future and the past. History has often been a battle between the haves and the have nots. How this may play out in the future is a mystery. As investors, we may be best served by the old adage that time in the market is our best option. Determining our asset allocation may be our greatest challenge. Of course, which factors will provide the greatest risk adjusted returns is another challenge for us. If investing was an exact science, we’d all be millionaires.

DMW

Thesaints
Posts: 2824
Joined: Tue Jun 20, 2017 12:25 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Thesaints » Thu May 02, 2019 12:27 am

mrspock wrote:
Wed May 01, 2019 11:25 pm
That said, I can say that Renaissance Technologies, Two Sigma and their AI powered Hedge Fund friends do not have large data warehouses because historical data is "noise", they train their (machine learning) models of historical information from a large diverse of data sets going back as many years as they can possible get. They put huge amounts of money into curation of these datasets (they are some of the most valuable IP in the world, next to the ML models themselves), and the further back they go the more valuable they are. In the world I'm in, it's similar.

What may seem like "noise" to humans, is most certainly not. Does the signal to noise ratio diminish if you go back too far? That depends on the thing you are building a model for. For the stock market though, its more likely case is it heavily correlates if you put the data in the right context, and this is what ML training tries to do. For situation X (fed rate rate, GDP, day of week, weather, earnings data, volumes, time of day, time of year etc), I predict Y (up/down) and therefore I do Z (sell/buy).

Anyways, I always laugh when people say "past performance doesn't predict the future (performance)". Really? That's why these guys are blowing billions on machine learning? I don't think so. :)
https://www.thriftbooks.com/w/the-signa ... iq=5622918

$4 well spent.

DonIce
Posts: 601
Joined: Thu Feb 21, 2019 6:44 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by DonIce » Thu May 02, 2019 12:47 am

Yes, a decade is a short timeframe. Just look at the dispersion of returns from a total stock market portfolio. The dispersion is very large for short time frames and continually becomes smaller for any longer timeframe that is analyzed. Historically, returns have always been positive over any 30 year period.

I would say 30 years is the minimum period for the data to not be "just noise".

User avatar
asset_chaos
Posts: 1556
Joined: Tue Feb 27, 2007 6:13 pm
Location: Melbourne

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by asset_chaos » Thu May 02, 2019 1:15 am

Somewhat different perspective is that with an average investing lifetime of around 60 years, 10 years is substantial fraction of the average investing lifetime.
Regards, | | Guy

rkhusky
Posts: 6747
Joined: Thu Aug 18, 2011 8:09 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by rkhusky » Thu May 02, 2019 6:47 am

It probably depends on what signal you are trying to extract. But, if there is any signal to extract, then 1 year returns are likely noise, but 10 years should be sufficient to pull out some signal.

3funder
Posts: 1004
Joined: Sun Oct 15, 2017 9:35 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by 3funder » Thu May 02, 2019 7:24 am

permport wrote:
Wed May 01, 2019 11:11 pm
I agree that 10 years of return data is noise to the extent that it isn't enough to be indicative of what will happen over a lifetime of investing.

However, somewhat ironically, I think 10 years is a long enough time period in the future for forecasting purposes. Bogle himself has done this, in that he would forecast the 10 year return on stocks based on current valuations. He was remarkably consistent and his margin of error rather small.
+1

larryswedroe
Posts: 15936
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by larryswedroe » Thu May 02, 2019 7:47 am

To help by clarifying
The context relates to RISK assets or sources of return. If there is a premium for risk then there must be the risk that the premium will be negative for a very long time, or there is no risk.
While people abandon strategies like value and small and emerging markets and reinsurance (as just examples) after few years of bad returns, they fail to consider that market beta has been negative in US for as long as 17 years (66-17) and 15 years (29-43) and 13 years (2000-13), and that totals 1/2 of the last 90 years!. And TSM strategies have almost all risks in that one factor (note in those three periods size and value premiums were fairly strong, providing diversification benefits).
That is why investing is simple, but not easy, takes great discipline to ignore noise.
Hope that is helpful
Larry

User avatar
jeffyscott
Posts: 8002
Joined: Tue Feb 27, 2007 9:12 am
Location: Wisconsin

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by jeffyscott » Thu May 02, 2019 8:08 am

JoMoney wrote:
Wed May 01, 2019 11:17 pm
There's a tremendous gap between the stock returns for the ten years ending in March of 2009, relative to the ten years ending one year prior, or even one year later...
Is that noise?
"Forecasts" from ten years prior could have been any number of things all over the place from positive to negative returns and if you picked the right day/month as an example proved any of them correct.
I think this can be addressed, at least to some extent, by this:
BJJ_GUY wrote:
Wed May 01, 2019 11:19 pm
Understanding market cycles, and valuations/pricing is the importance generally referred to with the '10 year' references. Basically, 10 years is the most commonly used estimate of time for an economic and/or market cycle.

If ten years is simply a proxy for peak-to-peak and/or trough-to-trough time horizons, this definitely informs us with the estimated return data that answers the question of how prices will eventually reflect value -- but doesn't answer the when (despite the silly 10 year approximation in the title). Again, the timing element is not answered here, so don't attack!
So, for example, you would want to look at a period from the low in (I think it was) 2002 to the low in March 2009. Or look at the period from the high in 2000 to the high in 2008, rather than some random 10 year interval.

While we don't know if we are at the next high, it's probably close enough that looking at the period from the 2008 high to today is a reasonable interval. But if instead you look at precisely 10 years, I think it is a biased sample as it would be going from a low to a high (or nearly so).
Time is your friend; impulse is your enemy. - John C. Bogle

Call_Me_Op
Posts: 7300
Joined: Mon Sep 07, 2009 2:57 pm
Location: Milky Way

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Call_Me_Op » Thu May 02, 2019 8:10 am

Yes, 10 years is essentially noise for the long-term investor.
Best regards, -Op | | "In the middle of difficulty lies opportunity." Einstein

CULater
Posts: 2044
Joined: Sun Nov 13, 2016 10:59 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by CULater » Thu May 02, 2019 8:21 am

The paradox is that "noise" in the data is an applicable concept only f the data represented a physical system governed by physical laws. But financial systems aren't the same because of uncertainty - meaning that totally unknown stuff pops up and invalidates your models and theories about how the system works. If we were to go out to longer backtesting periods such as 20 years, would backtesting results be less "noisy?" Not really, because 20 year results are even less meaningful as predictors of future results. It's random, you see. And "luck" plays a huge role in what the outcomes are. It ain't rocket science -- it isn't even science.
Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof. ~ John Kenneth Galbraith

User avatar
gmaynardkrebs
Posts: 1621
Joined: Sun Feb 10, 2008 11:48 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by gmaynardkrebs » Thu May 02, 2019 8:36 am

DonIce wrote:
Thu May 02, 2019 12:47 am
Just look at the dispersion of returns from a total stock market portfolio. The dispersion is very large for short time frames and continually becomes smaller for any longer timeframe that is analyzed.
Not sure I get this. Don't the tails grow fatter with longer time frames?
CULater wrote:
Thu May 02, 2019 8:21 am
The paradox is that "noise" in the data is an applicable concept only f the data represented a physical system governed by physical laws. But financial systems aren't the same because of uncertainty - meaning that totally unknown stuff pops up and invalidates your models and theories about how the system works. If we were to go out to longer backtesting periods such as 20 years, would backtesting results be less "noisy?" Not really, because 20 year results are even less meaningful as predictors of future results. It's random, you see. And "luck" plays a huge role in what the outcomes are. It ain't rocket science -- it isn't even science.
+1
One can run as many Monte Carlos simulations as you wish, but it can't model a system that is fundamentally not a big roulette wheel (to the best of my knowledge).

CULater
Posts: 2044
Joined: Sun Nov 13, 2016 10:59 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by CULater » Thu May 02, 2019 8:43 am

It's worth taking another look at William Sharpe's observations about the reliability of financial data and it's usefulness for predicting even for the market as a whole and you have a half-century of data.
I ask students to set up a spreadsheet in which the market as a whole has, let’s say, a risk premium over cash of 5.5 percent and a standard deviation of 18 percent, so you know exactly what the risk premium of the market as a whole is—not just a piece of the market, but the whole thing. Then I ask them to generate a series of fifty years of data randomly from that distribution and compute the average historic premium. After repeating that step 1,000 times, they can see what a fifty-year record might look like in a world where the premium is truly 5.5 percent. Of course, as we all know, you can easily generate plenty of scenarios in which you observe 0 percent, or 2 percent, or 20 per- cent over fifty years.

The moral of the story is that even if the gods are kind and distributions never change— which is improbable—and even if you have lots of data, you can still be far off on the premium for the whole market. Anyone who thinks that looking at empirical data will produce a resolution to the question of whether the premium on small growth stocks is different from that of large value stocks with any degree of precision is just kidding himself. I’m very skeptical of the findings in historic data related to averages. As a result, I’m fairly dismissive of much of the anomaly literature. I like to think I know something about this topic, because I understand how frail empirical averages are as indicators of historic expectations and, a fortiori, current expectations.

So is your argument that returns are so “noisy” that it’s almost guaranteed that historical data will find what look like anomalies?
https://investmentsandwealth.org/getmed ... Sharpe.pdf
Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof. ~ John Kenneth Galbraith

Topic Author
Random Walker
Posts: 3955
Joined: Fri Feb 23, 2007 8:21 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Random Walker » Thu May 02, 2019 9:05 am

asset_chaos wrote:
Thu May 02, 2019 1:15 am
Somewhat different perspective is that with an average investing lifetime of around 60 years, 10 years is substantial fraction of the average investing lifetime.
I agree. And I think some of those 10 year fractions are more critical to the individual investor than others. The decade comprised of 5 years both sides of retirement is a significant period to be subject to the whims of noise. Especially here I think diversification across sources of risk/return is important.

Dave

dbr
Posts: 29419
Joined: Sun Mar 04, 2007 9:50 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by dbr » Thu May 02, 2019 9:12 am

That is an offhand comment that makes a point. Most people would probably agree that ten year return data is not useful for much of anything. I think there is less in that point than meets the eye. Maybe the comment is as noisy as the data it references.

The statistical analysis of financial data is a tough problem that has not been solved to great satisfaction by anyone. It may be a problem that will never be completely solved, if indeed any problem in statistical analysis of real world phenomena is ever solved.

I say this as a person who has invested portions of a career in trying to apply statistics to accomplishing useful things in the world whether in physical phenomena, human behavior, or mixtures of the two. Sometimes it works out pretty well and sometimes you really wonder if you have done anything.

User avatar
Dialectical Investor
Posts: 500
Joined: Mon Dec 03, 2018 11:41 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Dialectical Investor » Thu May 02, 2019 9:19 am

I think the statement is non-specific and unhelpful without context of what one is trying to demonstrate. Data is not really noise in itself. Comments about data may be noise. If the claim is, "Such and such outperformed the other over this 10-year time period, thus, it is a good idea to invest in such and such rather than the other," that is noise. If the claim is, "Such and such outperformed the other over this 10-year time period, thus, we can see that such relative performance is possible," is not noise.

Seasonal
Posts: 320
Joined: Sun May 21, 2017 1:49 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Seasonal » Thu May 02, 2019 9:29 am

mrspock wrote:
Wed May 01, 2019 11:25 pm
I think Larry's statement must be taken in context, and I'm not certain what the context was when he said this.

That said, I can say that Renaissance Technologies, Two Sigma and their AI powered Hedge Fund friends do not have large data warehouses because historical data is "noise", they train their (machine learning) models of historical information from a large diverse of data sets going back as many years as they can possible get. They put huge amounts of money into curation of these datasets (they are some of the most valuable IP in the world, next to the ML models themselves), and the further back they go the more valuable they are. In the world I'm in, it's similar.

What may seem like "noise" to humans, is most certainly not. Does the signal to noise ratio diminish if you go back too far? That depends on the thing you are building a model for. For the stock market though, its more likely case is it heavily correlates if you put the data in the right context, and this is what ML training tries to do. For situation X (fed rate rate, GDP, day of week, weather, earnings data, volumes, time of day, time of year etc), I predict Y (up/down) and therefore I do Z (sell/buy).

Anyways, I always laugh when people say "past performance doesn't predict the future (performance)". Really? That's why these guys are blowing billions on machine learning? I don't think so. :)
If have invested billions in machine learning and have a track record similar to Renaissance Technologies, then past data certainly seems useful to you. In that case, why are you posting here?

If not, past performance may well not be a very useful predictor for you. I'm going to guess most people who past here are in the latter category.

User avatar
gmaynardkrebs
Posts: 1621
Joined: Sun Feb 10, 2008 11:48 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by gmaynardkrebs » Thu May 02, 2019 9:47 am

Seasonal wrote:
Thu May 02, 2019 9:29 am
If have invested billions in machine learning and have a track record similar to Renaissance Technologies, then past data certainly seems useful to you. In that case, why are you posting here? If not, past performance may well not be a very useful predictor for you. I'm going to guess most people who past here are in the latter category.
I don't think BH would exist were past not seen as prologue.

User avatar
Tyler9000
Posts: 423
Joined: Fri Aug 21, 2015 11:57 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Tyler9000 » Thu May 02, 2019 9:56 am

Random Walker wrote:
Wed May 01, 2019 10:03 pm
Larry Swedroe has a line he says frequently, something to the effect “for most investors 3 years is a long time, 5 years a very long time, and 10 years an eternity; but financial economists know that 10 years is just noise in the financial data”

So many of us cite data looking back various amounts of time, it got me wondering how many appreciate this. I find timeframes interesting. Longer timeframes for data certain strengthen arguments for persistence, but go back too far and it seems natural to question relevance to present circumstance.
I see two interesting topics here. The first is Larry's observation that what normal people would consider a long-term investing timeframe is just a tiny blip in the normal ups and downs of a market lifetime. And the second is your point that the timeframe you choose to observe can be deceptive in a naturally noisy system.

I agree wholeheartedly with both points, but will add one caveat -- the core issue is generally not the length of the timeframe studied but the number of timeframes studied. The natural instinct to anchor any backtest by looking only at a single timeframe ending in the present day introduces a level of recency bias that distorts any meaningful discussion of uncertainty. As an example, just recently the rolling 10-year stock return finally cropped out the market crash in 2008/09, and the numbers look awesome! But look at the exact same 10-year average a few years ago, and you get a completely different result.

So IMHO, the key to understanding financial markets is to learn to embrace noise. It's a tricky process that requires thinking about ranges of outcomes rather than single averages, but once it sinks in you'll have a much greater appreciation for what "risk" really means.
Last edited by Tyler9000 on Thu May 02, 2019 9:58 am, edited 1 time in total.

Seasonal
Posts: 320
Joined: Sun May 21, 2017 1:49 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Seasonal » Thu May 02, 2019 9:57 am

gmaynardkrebs wrote:
Thu May 02, 2019 9:47 am
Seasonal wrote:
Thu May 02, 2019 9:29 am
If have invested billions in machine learning and have a track record similar to Renaissance Technologies, then past data certainly seems useful to you. In that case, why are you posting here? If not, past performance may well not be a very useful predictor for you. I'm going to guess most people who past here are in the latter category.
I don't think BH would exist were past not seen as prologue.
It's always difficult to gauge counter-factuals.

The main point is that just because a multibillion dollar hedge fund can do something does not necessarily have any useful implications for you.

Beliavsky
Posts: 927
Joined: Sun Jun 29, 2014 10:21 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Beliavsky » Thu May 02, 2019 10:10 am

Random Walker wrote:
Wed May 01, 2019 10:03 pm
Larry Swedroe has a line he says frequently, something to the effect “for most investors 3 years is a long time, 5 years a very long time, and 10 years an eternity; but financial economists know that 10 years is just noise in the financial data”
If you are trying to predict future values of a time series (such as the return of value stocks minus growth stocks) based on past values, one approach is to use the full sample average, which is correct if the mean is constant. Another approach is to use an exponential moving average, which will give more weight to more recent data, such as the last 10 years.

I would not call 10 years or even one month of returns "noise". I would call it more data to be added to the data set. The way you combine it with past data, for example using a simple or exponential moving average, depends on whether you think the properties of the time series are -- is the time series "stationary"?

CULater
Posts: 2044
Joined: Sun Nov 13, 2016 10:59 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by CULater » Thu May 02, 2019 10:32 am

It's all noise - 10 years, 50 years, 100 years. Back to William Sharpe's observations: perhaps the only thing we can reasonably be sure of is the ordinal ranking of very broad investment class returns. The historical data are insufficient to provide much more than that unless you're willing to overlook simple probability statistics:
To some extent, if you take stocks, bonds, and cash, there is at least an ordering of the long-term average returns. I think that part of the theory is reasonably secure.
Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof. ~ John Kenneth Galbraith

User avatar
Dialectical Investor
Posts: 500
Joined: Mon Dec 03, 2018 11:41 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Dialectical Investor » Thu May 02, 2019 10:41 am

CULater wrote:
Thu May 02, 2019 10:32 am

The historical data are insufficient to provide much more than that unless you're willing to overlook properly use simple probability statistics:
A minor correction. It's unfortunate so many practitioners make poor use of statistics, giving the field a bad name. We "know" to varying degrees of confidence a lot more than just ordinal rankings of long-term returns. If you truly thought all data was noise, you would not be invested at all, or you would be acting irrationally.

User avatar
willthrill81
Posts: 10751
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by willthrill81 » Thu May 02, 2019 10:47 am

It's only a matter of opinion. There isn't an objective way to measure what's 'noise' and what's not.

That being said, the historical evidence certainly indicates that the longer the period, the more stable returns (i.e. there has been far less volatility over a 10 year period than during a single year). That alone certainly seems to suggest that there is less 'noise' in 10 years of return data than in one.
“It's a dangerous business, Frodo, going out your door. You step onto the road, and if you don't keep your feet, there's no knowing where you might be swept off to.” J.R.R. Tolkien,The Lord of the Rings

User avatar
willthrill81
Posts: 10751
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by willthrill81 » Thu May 02, 2019 10:49 am

Dialectical Investor wrote:
Thu May 02, 2019 10:41 am
CULater wrote:
Thu May 02, 2019 10:32 am

The historical data are insufficient to provide much more than that unless you're willing to overlook properly use simple probability statistics:
A minor correction. It's unfortunate so many practitioners make poor use of statistics, giving the field a bad name. We "know" to varying degrees of confidence a lot more than just ordinal rankings of long-term returns. If you truly thought all data was noise, you would not be invested at all, or you would be acting irrationally.
Correct. Many here who use the phrase "nobody knows nothing" incorrectly apply it to every aspect of investing, which is just plain wrong.
“It's a dangerous business, Frodo, going out your door. You step onto the road, and if you don't keep your feet, there's no knowing where you might be swept off to.” J.R.R. Tolkien,The Lord of the Rings

CULater
Posts: 2044
Joined: Sun Nov 13, 2016 10:59 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by CULater » Thu May 02, 2019 11:09 am

willthrill81 wrote:
Thu May 02, 2019 10:47 am
It's only a matter of opinion. There isn't an objective way to measure what's 'noise' and what's not.

That being said, the historical evidence certainly indicates that the longer the period, the more stable returns (i.e. there has been far less volatility over a 10 year period than during a single year). That alone certainly seems to suggest that there is less 'noise' in 10 years of return data than in one.
A simple confusion of the arithmetic average of stock returns with the geometric average. The dispersion of the arithmetic average becomes smaller, but the dispersion of the geometric average becomes larger. If we equate noisiness with dispersion, then you arrive at different conclusions. Since geometric returns are what you actually end up with, that would seem to be the more relevant numeric.
Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof. ~ John Kenneth Galbraith

User avatar
willthrill81
Posts: 10751
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by willthrill81 » Thu May 02, 2019 11:23 am

CULater wrote:
Thu May 02, 2019 11:09 am
willthrill81 wrote:
Thu May 02, 2019 10:47 am
It's only a matter of opinion. There isn't an objective way to measure what's 'noise' and what's not.

That being said, the historical evidence certainly indicates that the longer the period, the more stable returns (i.e. there has been far less volatility over a 10 year period than during a single year). That alone certainly seems to suggest that there is less 'noise' in 10 years of return data than in one.
A simple confusion of the arithmetic average of stock returns with the geometric average. The dispersion of the arithmetic average becomes smaller, but the dispersion of the geometric average becomes larger. If we equate noisiness with dispersion, then you arrive at different conclusions. Since geometric returns are what you actually end up with, that would seem to be the more relevant numeric.
There's no confusion present on my part. Both metrics are potentially relevant. Many investors view risk as the likelihood that they will lose money in an investment, not volatility as measured by the dispersion of an investment's future value. Logically, this makes a lot of sense. If hypothetical investment A had a range of long-term returns of 5% - 25% while investment B had a range of long-term returns of 1-3%, the dispersion of returns would certainly be lower with investment B, but in the absence of other considerations, virtually everyone would choose investment A.

Historically, the likelihood of an inflation-adjusted loss in stocks was considerably less over a 10 year period (about 12%) than over a single year (about 33%).
“It's a dangerous business, Frodo, going out your door. You step onto the road, and if you don't keep your feet, there's no knowing where you might be swept off to.” J.R.R. Tolkien,The Lord of the Rings

CULater
Posts: 2044
Joined: Sun Nov 13, 2016 10:59 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by CULater » Thu May 02, 2019 11:50 am

willthrill81 wrote:
Thu May 02, 2019 11:23 am
CULater wrote:
Thu May 02, 2019 11:09 am
willthrill81 wrote:
Thu May 02, 2019 10:47 am
It's only a matter of opinion. There isn't an objective way to measure what's 'noise' and what's not.

That being said, the historical evidence certainly indicates that the longer the period, the more stable returns (i.e. there has been far less volatility over a 10 year period than during a single year). That alone certainly seems to suggest that there is less 'noise' in 10 years of return data than in one.
A simple confusion of the arithmetic average of stock returns with the geometric average. The dispersion of the arithmetic average becomes smaller, but the dispersion of the geometric average becomes larger. If we equate noisiness with dispersion, then you arrive at different conclusions. Since geometric returns are what you actually end up with, that would seem to be the more relevant numeric.
There's no confusion present on my part. Both metrics are potentially relevant. Many investors view risk as the likelihood that they will lose money in an investment, not volatility as measured by the dispersion of an investment's future value. Logically, this makes a lot of sense. If hypothetical investment A had a range of long-term returns of 5% - 25% while investment B had a range of long-term returns of 1-3%, the dispersion of returns would certainly be lower with investment B, but in the absence of other considerations, virtually everyone would choose investment A.

Historically, the likelihood of an inflation-adjusted loss in stocks was considerably less over a 10 year period (about 12%) than over a single year (about 33%).
I don't understand then why you would choose to make the argument that long term stock returns are less "noisy" based on the dispersion of arithmetic returns. It seems entirely arbitrary. I can arbitrarily assert that long term returns are actually noisier based on the fact that the dispersion of geometric returns increases with time. Both arguments implicitly assume that the dispersion of returns around their mean value represents "noise" and one argument is as good as the other.
Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof. ~ John Kenneth Galbraith

psteinx
Posts: 3044
Joined: Tue Mar 13, 2007 2:24 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by psteinx » Thu May 02, 2019 11:53 am

To me, it's pretty ridiculous to think that 10 years of returns data is just noise.

Of course, a lot depends on the context. Ten years of A, in isolation, or ten years of A, compared to B, and C, and D?

That said, for most specialist type investments (i.e. things outside of broad sweeping equity, bond, cash returns and the like), ten years represents a substantial portion of the time they've been readily investable.

Ten years is also a substantial portion of a typical investor's active investing life.

=====

But to go beyond the simple question, I think some folks are looking for a way to justify long periods of basically mediocre returns of certain quasi-exotic strategies they favor, compared to more mainstream strategies. OK, the last ten years have been bad, but the strategy was good! The problem is, many of the strategies in question were backtest darlings.

It's not unduly hard to find exotic strategies that backtest well, for 50 years or more. But how many of those are due to chance? Then, if you actually pursue the exotic strategy, and it underperforms for 10 years from the time you start pursuing it, do you shrug that off and say, well, 50 of the last 60 years it's been great*!

====

Another view is: it doesn't really hurt, ex-ante, to pursue the exotic strategy. If it was a real phenomenon, you benefit from it going forward, and if it was random noise, well, you're not really worse off than just following the mainstream strategy, right?

But, there are issues. Among them:

1) The exotic strategies almost always come with significantly higher fees, from the native provider (i.e. the fund, manager, or the like). Sometimes, it may be 0.45% versus 0.05% for the conventional strategy. But sometimes, the exotic strategy has fees well above 1%.

2) There is generally further cost to connect with the strategy provider. Sometimes, this is an AUM advisor, charging perhaps 1%/year (generally non-deductible now, IIUC, so requiring well above 1% excess return to offset it, for taxable investors). Sometimes, this might just be a large investment of time for average joe to read about the MANY MANY possible ways to beat the market, and choose and execute appropriately.

3) The exotic strategies tend to be less tax efficient - a potentially big hit for taxable investors.

4) The exotic strategies can be used to justify lower overall equity exposures, from a risk budget standpoint (i.e. a barbell strategy). But then, when the exotic strategy not only fails to beat the (equity) market, but actually underperforms it, joe investor may take an extra whack, because he had too much in safe, low-returning investments (cash and short/intermediate bonds), offsetting the supposedly higher risk of the exotic investment.


----

* More realistically, a strategy with a favorable backtest may be positive for, say, 38 out of 50 years, with an overall performance of, say, +3.2% relative to the baseline strategy. Then, a decade of real life implementation may produce 3 out of 10 good years, and a negative overall performance for the decade, but still over the 60 year period, it will have been positive for 41 of 60 years, for say, a +2.6% performance relative to the baseline.

Still looks good! Just an unlucky decade!

Or, maybe there was nothing there to begin with except random noise, or even if there was something there once, it's not there anymore.

(Made edits for punctuation, clarity, consistency and/or other issues.)
Last edited by psteinx on Thu May 02, 2019 12:34 pm, edited 3 times in total.

User avatar
willthrill81
Posts: 10751
Joined: Thu Jan 26, 2017 3:17 pm
Location: USA

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by willthrill81 » Thu May 02, 2019 11:56 am

CULater wrote:
Thu May 02, 2019 11:50 am
willthrill81 wrote:
Thu May 02, 2019 11:23 am
CULater wrote:
Thu May 02, 2019 11:09 am
willthrill81 wrote:
Thu May 02, 2019 10:47 am
It's only a matter of opinion. There isn't an objective way to measure what's 'noise' and what's not.

That being said, the historical evidence certainly indicates that the longer the period, the more stable returns (i.e. there has been far less volatility over a 10 year period than during a single year). That alone certainly seems to suggest that there is less 'noise' in 10 years of return data than in one.
A simple confusion of the arithmetic average of stock returns with the geometric average. The dispersion of the arithmetic average becomes smaller, but the dispersion of the geometric average becomes larger. If we equate noisiness with dispersion, then you arrive at different conclusions. Since geometric returns are what you actually end up with, that would seem to be the more relevant numeric.
There's no confusion present on my part. Both metrics are potentially relevant. Many investors view risk as the likelihood that they will lose money in an investment, not volatility as measured by the dispersion of an investment's future value. Logically, this makes a lot of sense. If hypothetical investment A had a range of long-term returns of 5% - 25% while investment B had a range of long-term returns of 1-3%, the dispersion of returns would certainly be lower with investment B, but in the absence of other considerations, virtually everyone would choose investment A.

Historically, the likelihood of an inflation-adjusted loss in stocks was considerably less over a 10 year period (about 12%) than over a single year (about 33%).
I don't understand then why you would choose to make the argument that long term stock returns are less "noisy" based on the dispersion of arithmetic returns. It seems entirely arbitrary. I can arbitrarily assert that long term returns are actually noisier based on the fact that the dispersion of geometric returns increases with time. Both arguments implicitly assume that the dispersion of returns around their mean value represents "noise" and one argument is as good as the other.
It all comes down to how you define "noise"; dispersion of returns is certainly a measure of volatility but not necessarily a good measure of risk. Whether an investor should be more concerned with one or the other is a matter of opinion, and there is no objectively right answer.
“It's a dangerous business, Frodo, going out your door. You step onto the road, and if you don't keep your feet, there's no knowing where you might be swept off to.” J.R.R. Tolkien,The Lord of the Rings

marcopolo
Posts: 2110
Joined: Sat Dec 03, 2016 10:22 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by marcopolo » Thu May 02, 2019 12:05 pm

Random Walker wrote:
Wed May 01, 2019 10:03 pm
Larry Swedroe has a line he says frequently, something to the effect “for most investors 3 years is a long time, 5 years a very long time, and 10 years an eternity; but financial economists know that 10 years is just noise in the financial data”

So many of us cite data looking back various amounts of time, it got me wondering how many appreciate this. I find timeframes interesting. Longer timeframes for data certain strengthen arguments for persistence, but go back too far and it seems natural to question relevance to present circumstance.

I like the concept of out of sample tests: different time periods, different geographic markets, and in some cases even different asset classes.

Dave
I think it depends on the point the one is trying to make.

If their favored investment style is doing poorly recently, one is likely to assert that the recent data is "noise", and insist people need to consider the longer term.

On the other hand, if the recent data is favorable to their investment style, then one is likely to imply more significance to shorter term results, sometime as short as 3 months. viewtopic.php?f=10&t=222780&p=4310205&h ... x#p4310205 :beer

To me, it seems that the only time frame that matters is your own personal timeline.
Once in a while you get shown the light, in the strangest of places if you look at it right.

psteinx
Posts: 3044
Joined: Tue Mar 13, 2007 2:24 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by psteinx » Thu May 02, 2019 12:12 pm

marcopolo wrote:
Thu May 02, 2019 12:05 pm
On the other hand, if the recent data is favorable to their investment style, then one is likely to imply more significance to shorter term results, sometime as short as 3 months. viewtopic.php?f=10&t=222780&p=4310205&h ... x#p4310205 :beer
Hey now! The linked bit covers a *4* month timespan. Not that short-termist 3 month thinking...

BTW, what's the relative performance of the funds in question over the most recent 4 months?

User avatar
HomerJ
Posts: 13019
Joined: Fri Jun 06, 2008 12:50 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by HomerJ » Thu May 02, 2019 12:16 pm

Duplicate
Last edited by HomerJ on Thu May 02, 2019 12:58 pm, edited 1 time in total.
The J stands for Jay

User avatar
HomerJ
Posts: 13019
Joined: Fri Jun 06, 2008 12:50 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by HomerJ » Thu May 02, 2019 12:22 pm

psteinx wrote:
Thu May 02, 2019 11:53 am
But to go beyond the simple question, I think some folks are looking for a way to justify long periods of basically mediocre returns of certain quasi-exotic strategies they favor, compared to more mainstream strategies. OK, the last ten years have been bad, but the strategy was good! The problem is, many of the strategies in question were backtest darlings.
This seems likely.

However, I agree somewhat that 10 years is not long enough to say "This idea has failed".

Stock bear markets have lasted longer than 10 years in the past.

But stocks have a much longer history.

I'm always VERY suspicious of a new back-tested theory that fails fairly quickly going forward. 10 years doesn't prove the theory wrong, but it sure doesn't prove the theory correct or give one confidence in it.
Last edited by HomerJ on Thu May 02, 2019 12:58 pm, edited 3 times in total.
The J stands for Jay

asif408
Posts: 1694
Joined: Sun Mar 02, 2014 8:34 am
Location: Florida

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by asif408 » Thu May 02, 2019 12:32 pm

Dave,

I think 10 years is somewhat useful, but not in the way most people look at it. For instance, I look at the last 10 years performance of, for instance, foreign developed and emerging markets large value stocks, gold stocks and global energy stocks (losers over the last decade) and project better performance in the future because of their poorer relative performance to, say, technology stocks or US small growth stocks (winners over the last decade). This is not to say it is a guarantee, or that I know exactly when and how the change will take place if it does, how much it will outperform by, etc. But I think the concept of mean reversion can be somewhat helpful over a decade period of time between different parts of the global stock markets, at least that has been the case historically. More often than not last decade's best performer becomes one of the next decade's worst performers. I wouldn't apply this logic to individual stock picking, newer areas of investing (e.g., bitcoin) where data is limited, but at least from a high level (e.g., foreign, vs US, growth vs value, small vs large), where there is a decent amount of historical data, I think it is useful.

Of course, like most things in investing, it still requires patience and discipline to stick with it, because it could take another decade before things turn around, if they do at all. I'm willing to take something of a bet that mean reversion is not dead, others may not be comfortable with that.

CULater
Posts: 2044
Joined: Sun Nov 13, 2016 10:59 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by CULater » Thu May 02, 2019 12:42 pm

willthrill81 wrote:
Thu May 02, 2019 11:56 am
CULater wrote:
Thu May 02, 2019 11:50 am
willthrill81 wrote:
Thu May 02, 2019 11:23 am
CULater wrote:
Thu May 02, 2019 11:09 am
willthrill81 wrote:
Thu May 02, 2019 10:47 am
It's only a matter of opinion. There isn't an objective way to measure what's 'noise' and what's not.

That being said, the historical evidence certainly indicates that the longer the period, the more stable returns (i.e. there has been far less volatility over a 10 year period than during a single year). That alone certainly seems to suggest that there is less 'noise' in 10 years of return data than in one.
A simple confusion of the arithmetic average of stock returns with the geometric average. The dispersion of the arithmetic average becomes smaller, but the dispersion of the geometric average becomes larger. If we equate noisiness with dispersion, then you arrive at different conclusions. Since geometric returns are what you actually end up with, that would seem to be the more relevant numeric.
There's no confusion present on my part. Both metrics are potentially relevant. Many investors view risk as the likelihood that they will lose money in an investment, not volatility as measured by the dispersion of an investment's future value. Logically, this makes a lot of sense. If hypothetical investment A had a range of long-term returns of 5% - 25% while investment B had a range of long-term returns of 1-3%, the dispersion of returns would certainly be lower with investment B, but in the absence of other considerations, virtually everyone would choose investment A.

Historically, the likelihood of an inflation-adjusted loss in stocks was considerably less over a 10 year period (about 12%) than over a single year (about 33%).
I don't understand then why you would choose to make the argument that long term stock returns are less "noisy" based on the dispersion of arithmetic returns. It seems entirely arbitrary. I can arbitrarily assert that long term returns are actually noisier based on the fact that the dispersion of geometric returns increases with time. Both arguments implicitly assume that the dispersion of returns around their mean value represents "noise" and one argument is as good as the other.
It all comes down to how you define "noise"; dispersion of returns is certainly a measure of volatility but not necessarily a good measure of risk. Whether an investor should be more concerned with one or the other is a matter of opinion, and there is no objectively right answer.
Agree completely. Maybe we should cycle back to the original post and determine exactly how "noise" is defined and measured.
Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof. ~ John Kenneth Galbraith

dkturner
Posts: 1493
Joined: Sun Feb 25, 2007 7:58 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by dkturner » Thu May 02, 2019 1:16 pm

If you can spend it, it’s more than “just noise”. :D

UpperNwGuy
Posts: 2111
Joined: Sun Oct 08, 2017 7:16 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by UpperNwGuy » Thu May 02, 2019 1:30 pm

If I could only see a single ten year period, I would want it to be the most recent ten years.

User avatar
AtlasShrugged?
Posts: 689
Joined: Wed Jul 15, 2015 6:08 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by AtlasShrugged? » Fri May 03, 2019 6:34 am

RandomWalker....No, 10 years of data is not simply noise, to the individual investor.

Example: First 10 years of retirement, Years 1-5 have a negative real return of 3% annually, then 0% real for years 6-10. My retirement is screwed because of sequence of return risk. That ain't just noise....it is deafening noise.

The individual investor maybe invests 50-60 years. That 10-year slice matters a lot; it is 17% to 20% of their entire investing life.

Mr. Swedroe is right on the mark (BTW, I was personally very happy to see his post here) that a) context matters, and b) they call it market risk for a reason - sometimes you wind up on the wrong side of risk and get burned.

Mr. Spock also nailed an important point. There are a few very successful quants out there. I personally think as digital data proliferates, the models will become more predictively (not a real word, but it fits) accurate.

So what to do? Have your IPS, develop your plan, follow your plan. Keep things simple. Be adaptable to changes in the investing environment. A case in point...I myself started a thread on the wisdom of holding 45% bonds in retirement. I am very seriously considering having a 70/30 allocation (instead of 55/45) because total bond returns won't even be close to their historical averages. In all likelihood, neither will stocks if CAPE 10 is truly predictive of forward returns. Yes, I know. CAPE 10 can only tell you the likely dispersion of outcomes, with a higher CAPE 10 suggesting that returns will be much more to the left tail than right tail - I get that.

Fortunately, I have a couple of years before I need to make a hard and fast decision on changing my glidepath.
“If you don't know, the thing to do is not to get scared, but to learn.”

RadAudit
Posts: 3421
Joined: Mon May 26, 2008 10:20 am
Location: Second star on the right and straight on 'til morning

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by RadAudit » Fri May 03, 2019 7:21 am

AtlasShrugged? wrote:
Fri May 03, 2019 6:34 am
The individual investor maybe invests 50-60 years. That 10-year slice matters a lot; it is 17% to 20% of their entire investing life.
And, if you're in the last ten (twenty?) or so years of your investing life, it's a lot bigger chunk of that remaining investing life. Maybe I should of known more about Mr. Swedroe's factor investing about 40 years ago.
FI is the best revenge. LBYM. Invest the rest. Stay the course. - PS: The cavalry isn't coming, kids. You are on your own.

Topic Author
Random Walker
Posts: 3955
Joined: Fri Feb 23, 2007 8:21 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Random Walker » Fri May 03, 2019 7:51 am

AtlasShrugged? wrote:
Fri May 03, 2019 6:34 am
RandomWalker....No, 10 years of data is not simply noise, to the individual investor.

Example: First 10 years of retirement, Years 1-5 have a negative real return of 3% annually, then 0% real for years 6-10. My retirement is screwed because of sequence of return risk. That ain't just noise....it is deafening noise.

The individual investor maybe invests 50-60 years. That 10-year slice matters a lot; it is 17% to 20% of their entire investing life.

Mr. Swedroe is right on the mark (BTW, I was personally very happy to see his post here) that a) context matters, and b) they call it market risk for a reason - sometimes you wind up on the wrong side of risk and get burned.

Mr. Spock also nailed an important point. There are a few very successful quants out there. I personally think as digital data proliferates, the models will become more predictively (not a real word, but it fits) accurate.

So what to do? Have your IPS, develop your plan, follow your plan. Keep things simple. Be adaptable to changes in the investing environment. A case in point...I myself started a thread on the wisdom of holding 45% bonds in retirement. I am very seriously considering having a 70/30 allocation (instead of 55/45) because total bond returns won't even be close to their historical averages. In all likelihood, neither will stocks if CAPE 10 is truly predictive of forward returns. Yes, I know. CAPE 10 can only tell you the likely dispersion of outcomes, with a higher CAPE 10 suggesting that returns will be much more to the left tail than right tail - I get that.

Fortunately, I have a couple of years before I need to make a hard and fast decision on changing my glidepath.
I agree with your points. It seems ironic, but it’s logical. If expected returns are lower, one might have to increase equity allocation to meet goals. That is very risky proposition in those last five years of working and first five years of retirement. Maybe better to modify goals.

Dave

MnD
Posts: 4287
Joined: Mon Jan 14, 2008 12:41 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by MnD » Fri May 03, 2019 8:07 am

30+ years in to investing and maybe ~30 to go and I agree completely, 10 years is just noise. It rarely contains a full business cycle or two so typically looks especially good or bad. I like all the fear, greed and and importance many investors attach to the past 10 years and 10-year "projections", CAPE10 and other nonsense. Their behavioral errors provides nice opportunities for the long-term rational investors.

CULater
Posts: 2044
Joined: Sun Nov 13, 2016 10:59 am

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by CULater » Fri May 03, 2019 8:30 am

Isn't Larry putting forth the "noise" argument in order to justify buying and holding an investment thesis for at least a decade in order for it to pay off? In other words, it is the argument that you can't reject theories since returns are "noisy" at least over time periods shorter than 10+ years. In this case, factor investing. First, I'd argue that if you're not prepared to do that, then maybe you're better off just holding a total market index fund. Second, I'd argue that just about any other investment thesis is likely to look like a winner randomly from time to time but does that prove that it "works" or is it just random luck? Sharpe says it's probably all noise, even over much longer periods than 10 years.
Last edited by CULater on Fri May 03, 2019 9:06 am, edited 1 time in total.
Faced with the choice between changing one’s mind and proving that there is no need to do so, almost everyone gets busy on the proof. ~ John Kenneth Galbraith

prairieman
Posts: 127
Joined: Thu Mar 01, 2018 3:17 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by prairieman » Fri May 03, 2019 8:50 am

I think what this means is simply that there is a chance you could experience a negative return over a ten year time period, but the market trend will still be up if viewed from a lifetime perspective.
Whatever the case, I feel extremely fortunate to have experienced this last ten year growth period, and hope it is not just noise. The ten year period before that apparently was noise, even though it did not feel like it at the time.

User avatar
Sandtrap
Posts: 7447
Joined: Sat Nov 26, 2016 6:32 pm

Re: Do You Believe 10 Years Return Data Is Just Noise?

Post by Sandtrap » Fri May 03, 2019 8:58 am

Random Walker wrote:
Wed May 01, 2019 10:03 pm
Larry Swedroe has a line he says frequently, something to the effect “for most investors 3 years is a long time, 5 years a very long time, and 10 years an eternity; but financial economists know that 10 years is just noise in the financial data”

So many of us cite data looking back various amounts of time, it got me wondering how many appreciate this. I find timeframes interesting. Longer timeframes for data certain strengthen arguments for persistence, but go back too far and it seems natural to question relevance to present circumstance.

I like the concept of out of sample tests: different time periods, different geographic markets, and in some cases even different asset classes.

Dave
Good points.
Thanks.

Return data and projections are "noise" to the extend that the time frame is applicable or relevant to the investor.
For example: if a retiree is 85, it's not as relevant given his diminishing time frame to apply the data or projections based on that data.

Sometimes, an argument or case is presented quoting 50 years of data. But,(to my simple brain) it seems a bit irrelevant to a 70 year old retiree with an annuity and a sound pension and SS. True?

So, I also question the relevance to present circumstance, and individual applicability.
Wiki Bogleheads Wiki: Everything You Need to Know

Post Reply