QSPIX - thoughts on interesting fund

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Johno
Posts: 1883
Joined: Sat May 24, 2014 4:14 pm

Re: QSPIX - thoughts on interesting fund

Post by Johno » Wed Aug 19, 2015 3:24 pm

afan wrote:Johno,

Probably my fault for trying to condense this down to forum posts. Think about what independent means in this context. Think about what out of sample means. Think about publication bias. Think about the differences in performance of hypothetical portfolio constructs and their behavior in the real world, where the assumption that you have accounted for all relevant factors may or may not be true. Think about the need for data from the real world on this particular management approach, not just whether a given factor exists. Read the papers.

Or don't. Just a suggestion. It is a free country.

An actually convincing argument from your POV would not just keep repeating 'read the (generic) paper'. Instead you would point to eg. which statistical tests in AQR's paper are inadequate and how (as I said, there was actually a useful discussion of this at one point here); you would point which aspects of 'academic' v 'real world' assumptions would have a significant impact in terms of what QSPIX actually does; you would compare and contrast various research unrelated to AQR which made generally similar findings about one or more of the basically same 'style premia'; you would again explain how the reality of past profitability of the currency carry trade (as one example) could possibly be refuted by reading a generic statistics paper. This would be the road map for a useful argument in keeping with your theme. But you're free to lack the interest to gain the familiarity to make such an argument.

But again, I would point to the difference between actual statistical issues with past data and/or formation of hypothesis on one hand, under the assumption of a stationary distribution, the only workable assumption from a stats analysis POV, v. the fact that the process of returns is not stationary in reality. That is, the world can change, and may even change in direct response to financial research findings. No result to any level of T stat etc confidence in the past can be taken literally because the returns process is non-stationary. This isn't a problem with statistical analysis per se (including 'out of sample', since the future is a sample from a new unknown distribution, always, compared to the past), or a real world assumptions problem, or a 'publication bias' problem. Nevertheless it's an uncertainty. But seeing how it plays out over a period of years adds no more certainty, because it can just change again in further periods. But this is common to any finding in finance, including again for example the hypothesis that there's a nicely remunerative equity risk premium over long periods.

needtosave
Posts: 16
Joined: Wed Aug 19, 2015 10:31 pm

Re: QSPIX - thoughts on interesting fund

Post by needtosave » Wed Aug 19, 2015 10:46 pm

Johno wrote:
afan wrote:Johno,

Probably my fault for trying to condense this down to forum posts. Think about what independent means in this context. Think about what out of sample means. Think about publication bias. Think about the differences in performance of hypothetical portfolio constructs and their behavior in the real world, where the assumption that you have accounted for all relevant factors may or may not be true. Think about the need for data from the real world on this particular management approach, not just whether a given factor exists. Read the papers.

Or don't. Just a suggestion. It is a free country.

An actually convincing argument from your POV would not just keep repeating 'read the (generic) paper'. Instead you would point to eg. which statistical tests in AQR's paper are inadequate and how (as I said, there was actually a useful discussion of this at one point here);


The argument, that I'm 95% sure afan is trying to make, is that anytime you test multiple hypothesis, you should do a statistical adjustment to control the family wise error rate, such as Bonferroni, https://en.wikipedia.org/wiki/Bonferroni_correction. The key point about multiple hypothesis testing is that the statistics don't care whether 1 phd student tests 10,000 factors or 10,000 phd each test 1 factor. I think nisiprius has written about this before.

User avatar
grap0013
Posts: 1861
Joined: Thu Mar 18, 2010 1:24 pm

Re: QSPIX - thoughts on interesting fund

Post by grap0013 » Thu Aug 20, 2015 11:18 am

^ I think you are getting way too far into the weeds on this one.

This fund basically has multiple factor tilts that have worked well in the past and I think are likely to work well into the future. They are robust. The big question with QSPIX is implementation and actual realized returns after trading costs and ER.

Also, this fund is not designed to protect in "flight to safety" scenarios such as 2008. Some posters are alluding to that being its goal. For true max drawdown protection you want treasury bonds. You can definitely see the value in choppy markets at reducing downside volatility with this fund. QSPIX is my second best performing fund YTD and it's my best performing fund for trailing 1 week, 3 mo, and 1 yr returns.
There are no guarantees, only probabilities.

afan
Posts: 2860
Joined: Sun Jul 25, 2010 4:01 pm

Re: QSPIX - thoughts on interesting fund

Post by afan » Thu Aug 20, 2015 11:48 am

Free country. Everyone gets to decide whether they want to read up on statistics before trying to interpret statistical arguments. It turns out that there are some excellent references specifically on point for financial data. But they are useful only to those who read them.

The wikipedia reference does introduce the problem. One could not actually apply this particular attempt to control this error. To use Bonferroni, you need to know how many tests were run, how many models were tried. Here, we have no idea. So the principles apply, but not the solution.

Even other approaches to controlling for the error depend on the data. We are trying to infer what would happen in a real world fund based on backtest results from a strategy without all the confounding messiness that happens in real life. Statistical concerns aside, only oos data FROM THE REAL WORLD IMPLEMENTATION will help with this problem.

As I said, free country. But many people find Campbell's work and the whole field interesting, even if they have no intention of chasing the latest hot investment idea.
We don't know how to beat the market on a risk-adjusted basis, and we don't know anyone that does know either | --Swedroe | We assume that markets are efficient, that prices are right | --Fama

lack_ey
Posts: 5556
Joined: Wed Nov 19, 2014 11:55 pm

Re: QSPIX - thoughts on interesting fund

Post by lack_ey » Thu Aug 20, 2015 12:15 pm

The multiple hypothesis testing and all the related statistical problems with "significance" levels are very well known (and plague many fields of study including finance and economics) and I think anybody with any exposure to statistics at least knows them and basic fixes (cludges?) like Bonferroni, the basic concepts.

But you keep failing to mention how specifically they apply here, to which styles, etc. What's more suspect and possibly or probably not significant, only reflecting randomness and not any real signal?

Which "confounding messiness that happens in real life" affects the results of the fund with live money relative to simulated backtest performance, and by how much?

Do you have similar questions and concerns about the equity risk premium? Anything else?

User avatar
nisiprius
Advisory Board
Posts: 33797
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: QSPIX - thoughts on interesting fund

Post by nisiprius » Thu Aug 20, 2015 1:12 pm

I certainly have questions about "the" equity risk premium, ever since Simba pointed out a paper with the amazing title and even more amazing abstract,
simba wrote:The Equity Premium in 150 Textbooks - Pablo Fernandez
Abstract:
I review 150 textbooks on corporate finance and valuation published between 1979 and 2009 by authors such as Brealey, Myers, Copeland, Damodaran, Merton, Ross, Bruner, Bodie, Penman, Arzac… and find that their recommendations regarding the equity premium range from 3% to 10%, and that 51 books use different equity premia in various pages. The 5-year moving average has declined from 8.4% in 1990 to 5.7% in 2008 and 2009.
Now, I'll agree that I personally invest on the assumption that there is a nonzero positive risk premium. The risk premium seems relatively robust and relatively well established.

It is sufficiently robust that when that author checked 150 textbooks, he came up with a value of "7% ± 3%" and that range doesn't include zero.

I do not believe you could find 150 textbooks that define or give values for the value, carry, momentum and defensive premiums, but in any case these premiums are much less robust and I think any honest range estimates for them would include zero.
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.

Johno
Posts: 1883
Joined: Sat May 24, 2014 4:14 pm

Re: QSPIX - thoughts on interesting fund

Post by Johno » Thu Aug 20, 2015 1:39 pm

afan wrote:Free country. Everyone gets to decide whether they want to read up on statistics before trying to interpret statistical arguments. It turns out that there are some excellent references specifically on point for financial data. But they are useful only to those who read them.

Again, you talk about a statistical argument but haven't ever made one specific to the topic we're discussing. Then you keep trying to deflect from your lack of an argument by claiming that others aren't familiar with statistics issues in finance research in general. I guess it's well past clear by now you aren't going to mount any specific argument about weakness in the statistical tests as it specifically relates to QSPIX or general research into value, carry etc premia (again some of which in some markets obviously have existed, another point you've always refused to address). But hey, you can always repeat 'it's a free country'. :D

Johno
Posts: 1883
Joined: Sat May 24, 2014 4:14 pm

Re: QSPIX - thoughts on interesting fund

Post by Johno » Thu Aug 20, 2015 2:15 pm

nisiprius wrote:I certainly have questions about "the" equity risk premium, ever since Simba pointed out a paper with the amazing title and even more amazing abstract,
simba wrote:The Equity Premium in 150 Textbooks - Pablo Fernandez
Abstract:
I review 150 textbooks on corporate finance and valuation published between 1979 and 2009 by authors such as Brealey, Myers, Copeland, Damodaran, Merton, Ross, Bruner, Bodie, Penman, Arzac… and find that their recommendations regarding the equity premium range from 3% to 10%, and that 51 books use different equity premia in various pages. The 5-year moving average has declined from 8.4% in 1990 to 5.7% in 2008 and 2009.
Now, I'll agree that I personally invest on the assumption that there is a nonzero positive risk premium. The risk premium seems relatively robust and relatively well established.

It is sufficiently robust that when that author checked 150 textbooks, he came up with a value of "7% ± 3%" and that range doesn't include zero.

I do not believe you could find 150 textbooks that define or give values for the value, carry, momentum and defensive premiums, but in any case these premiums are much less robust and I think any honest range estimates for them would include zero.

I'm not sure number of textbooks covering a topic is a good measure of truth. It does demonstrate that perception of the ERP varies pretty widely, which is what you've used the same source before to demonstrate and I which I think is a logically stronger use of it than as a comparison to value, momentum, etc. premia.

That said I also rely a lot on the ERP, which is kind of the point of alternatives to begin with, as a way to reduce extreme reliance on the ERP, perhaps along with other forms of somewhat (but hoped to be not exactly) equity like risks as a source of return found in assets other than stocks (credit, real estate, etc).

But your last clause I think is useful to return to the point made several times but obfuscated by repeated 'read statistics papers' answers by somebody not interested enough to get familiar enough with the topic here to mount an actual argument. You say the 'honest' range of estimates of premia like value, momentum etc would have to include zero. Again, as it relates to the past, I think that would need to actually be demonstrated in specific terms of statistics of known past outcome for those premia. IOW I don't think that statement stands as self evident at all.

OTOH as to future returns, they could (will in some respects almost surely) come from a different underlying process of returns we don't know. It could even be affected by knowledge of the apparent existence of 'alternative' premia in the past. 'Statistics' in financial research always refers to past returns, and virtually always under the assumption there was a single underlying process of returns. Otherwise you couldn't make any objective statement about anything ever. But futures returns are not just 'out of sample', they will be the product of a different, perhaps significantly, underlying process.

In the latter context I'm quite willing to accept that value or momentum etc premia might be zero in the future. But I also have a lot of concern about the ERP's future, and again I rely on it much more heavily.
Last edited by Johno on Thu Aug 20, 2015 2:25 pm, edited 2 times in total.

User avatar
grap0013
Posts: 1861
Joined: Thu Mar 18, 2010 1:24 pm

Re: QSPIX - thoughts on interesting fund

Post by grap0013 » Thu Aug 20, 2015 2:17 pm

nisiprius wrote:I do not believe you could find 150 textbooks that define or give values for the value, carry, momentum and defensive premiums, but in any case these premiums are much less robust and I think any honest range estimates for them would include zero.


Nice point. Also, if you look at that graph buried somewhere in this thread that shows all those premiums concurrently if you follow each little squiggly line you do notice several 10 year flat periods with no premium. QSPIX is definitely not a lock for :moneybag :moneybag :moneybag by any means, but I'll put my :moneybag on it that it makes :moneybag.
There are no guarantees, only probabilities.

User avatar
matjen
Posts: 1671
Joined: Sun Nov 20, 2011 11:30 pm

Re: QSPIX - thoughts on interesting fund

Post by matjen » Thu Aug 20, 2015 3:08 pm

grap0013 wrote:
nisiprius wrote:I do not believe you could find 150 textbooks that define or give values for the value, carry, momentum and defensive premiums, but in any case these premiums are much less robust and I think any honest range estimates for them would include zero.


Nice point. Also, if you look at that graph buried somewhere in this thread that shows all those premiums concurrently if you follow each little squiggly line you do notice several 10 year flat periods with no premium. QSPIX is definitely not a lock for :moneybag :moneybag :moneybag by any means, but I'll put my :moneybag on it that it makes :moneybag.


Grap, Grap, Grap...what are we going to do with you? Everyone knows that the standard that QSPIX has to be judged against is the following:
1) Never, ever lose money for any period of time
2) In mixed markets do great (better than bonds and equities)
3) In bull markets do better than the slice of total equities that is doing the absolute best and/or the best actively managed funds for that period
4) In bear markets do better than Treasuries (and have their correlations in bear markets)
5) Be 100% transparent with your trading strategies so that lay persons can follow them
6) Never adjust your strategy
7) Cost less than .09 basis points
8) Be tax efficient

I don't think I have heard anyone ask for the returns to be inflation protected but I may have missed that. 8-)
A man is rich in proportion to the number of things he can afford to let alone.

lack_ey
Posts: 5556
Joined: Wed Nov 19, 2014 11:55 pm

Re: QSPIX - thoughts on interesting fund

Post by lack_ey » Thu Aug 20, 2015 3:20 pm

Johno wrote:
nisiprius wrote:I do not believe you could find 150 textbooks that define or give values for the value, carry, momentum and defensive premiums, but in any case these premiums are much less robust and I think any honest range estimates for them would include zero.


But your last clause I think is useful to return to the point made several times but obfuscated by repeated 'read statistics papers' answers by somebody not interested enough to get familiar enough with the topic here to mount an actual argument. You say the 'honest' range of estimates of premia like value, momentum etc would have to include zero. Again, as it relates to the past, I think that would need to actually be demonstrated in specific terms of statistics of known past outcome for those premia. IOW I don't think that statement stands as self evident at all.


I think I've seen some analysis over some period of many decades (my memory is great, huh? sorry, can't find the link) where momentum had a higher t-stat than market beta. That's by gross returns, of course, before transaction costs, where obviously momentum strategies incur more in costs.

In any case, I brought up the ERP in part because I am not 100% confident in it either, same for the four things here. I'm more confident in the ERP than in value, momentum, carry, and defensive, but that's just my personal interpretation. I give them all (including ERP) honest range estimates that include zero, sure, when it comes to the future.

User avatar
Maynard F. Speer
Posts: 2139
Joined: Wed Mar 18, 2015 10:31 am

Re: QSPIX - thoughts on interesting fund

Post by Maynard F. Speer » Thu Aug 20, 2015 3:20 pm

9) Absolutely zero beta - otherwise you're paying active fees for passive performance
10) Public manager floggings for any quarterly period the fund fails to beat its benchmark*

* - which is of course the S&P 500 + fund fee + compensation for any distress caused by this event
"Economics is a method rather than a doctrine, an apparatus of the mind, a technique of thinking, which helps its possessor to draw correct conclusions." - John Maynard Keynes

lack_ey
Posts: 5556
Joined: Wed Nov 19, 2014 11:55 pm

Re: QSPIX - thoughts on interesting fund

Post by lack_ey » Thu Aug 20, 2015 3:56 pm

^ re: (9)

On a more serious note, I think the majority of AQR alternatives-class mutual funds have pretty much zero beta except the long/short equity fund that specifically says it is "long and short global equities with moderate net long equity exposure." If Asness had his way, hedge funds in general would have zero beta, according to what he says and writes.

Anyway, more important than the amount of beta is the amount of alpha or non-beta content, seeing as that is what one is paying for (beta being very cheap to access elsewhere). The data seems to show average hedge fund alpha shrinking over time, so people are piling into strategies in many funds with net beta and not much to show for it.

This tangentially reminds me of a terrible article I once read about the IQ Hedge Multi-Strategy Tracker ETF (QAI), a somewhat popular hedge-fund-holdings-replicator fund with $1B AUM and an ER of 0.91%. This is one of the ETFs I said I looked at and found nothing of interest in:
http://www.etf.com/sections/features-an ... posure-qai

You have to see it to believe it. (If the graph there isn't obvious enough, check the holdings and it's 30% long stocks and sure enough has a beta around 0.3, with negative alpha so far. The article touts said fund as a replacement for bonds.)

User avatar
Maynard F. Speer
Posts: 2139
Joined: Wed Mar 18, 2015 10:31 am

Re: QSPIX - thoughts on interesting fund

Post by Maynard F. Speer » Thu Aug 20, 2015 5:00 pm

lack_ey wrote:^ re: (9)

On a more serious note, I think the majority of AQR alternatives-class mutual funds have pretty much zero beta except the long/short equity fund that specifically says it is "long and short global equities with moderate net long equity exposure." If Asness had his way, hedge funds in general would have zero beta, according to what he says and writes.

Anyway, more important than the amount of beta is the amount of alpha or non-beta content, seeing as that is what one is paying for (beta being very cheap to access elsewhere). The data seems to show average hedge fund alpha shrinking over time, so people are piling into strategies in many funds with net beta and not much to show for it.

This tangentially reminds me of a terrible article I once read about the IQ Hedge Multi-Strategy Tracker ETF (QAI), a somewhat popular hedge-fund-holdings-replicator fund with $1B AUM and an ER of 0.91%. This is one of the ETFs I said I looked at and found nothing of interest in:
http://www.etf.com/sections/features-an ... posure-qai

You have to see it to believe it. (If the graph there isn't obvious enough, check the holdings and it's 30% long stocks and sure enough has a beta around 0.3, with negative alpha so far. The article touts said fund as a replacement for bonds.)


I saw another article not long ago suggesting you could recreate hedge-fund performance simply by holding an index tracker and 2/3rds cash .. You have got to be careful - certainly when a lot of people don't 'get' hedge funds (possibly sometimes even those who run them)

What I've found in the UK, at least, is if you want market neutral returns, with low volatility, you are typically looking at around 4% after fees .. There are long/short equity funds that do better, but they can be >80% long equities ... Which is fine if they can time when to be 80% short equities - but I've seen the mess they get in when the market correction doesn't come along

RenTech's Medallion Fund (I believe) is balanced to exactly 0% beta .. And Simons has said something I've tended to find, which is that these types of funds tend to do best with market volatility (RenTech returning >90% through 2008-09) .. Which obviously a simple defensive fund wouldn't achieve .... Then again, if QAI's alpha portion comes into its own through the next financial crisis, it may be worth its salt?
"Economics is a method rather than a doctrine, an apparatus of the mind, a technique of thinking, which helps its possessor to draw correct conclusions." - John Maynard Keynes

lack_ey
Posts: 5556
Joined: Wed Nov 19, 2014 11:55 pm

Re: QSPIX - thoughts on interesting fund

Post by lack_ey » Thu Aug 20, 2015 6:57 pm

Maynard F. Speer wrote:
lack_ey wrote:^ re: (9)

On a more serious note, I think the majority of AQR alternatives-class mutual funds have pretty much zero beta except the long/short equity fund that specifically says it is "long and short global equities with moderate net long equity exposure." If Asness had his way, hedge funds in general would have zero beta, according to what he says and writes.

Anyway, more important than the amount of beta is the amount of alpha or non-beta content, seeing as that is what one is paying for (beta being very cheap to access elsewhere). The data seems to show average hedge fund alpha shrinking over time, so people are piling into strategies in many funds with net beta and not much to show for it.

This tangentially reminds me of a terrible article I once read about the IQ Hedge Multi-Strategy Tracker ETF (QAI), a somewhat popular hedge-fund-holdings-replicator fund with $1B AUM and an ER of 0.91%. This is one of the ETFs I said I looked at and found nothing of interest in:
http://www.etf.com/sections/features-an ... posure-qai

You have to see it to believe it. (If the graph there isn't obvious enough, check the holdings and it's 30% long stocks and sure enough has a beta around 0.3, with negative alpha so far. The article touts said fund as a replacement for bonds.)


I saw another article not long ago suggesting you could recreate hedge-fund performance simply by holding an index tracker and 2/3rds cash .. You have got to be careful - certainly when a lot of people don't 'get' hedge funds (possibly sometimes even those who run them)

What I've found in the UK, at least, is if you want market neutral returns, with low volatility, you are typically looking at around 4% after fees .. There are long/short equity funds that do better, but they can be >80% long equities ... Which is fine if they can time when to be 80% short equities - but I've seen the mess they get in when the market correction doesn't come along

RenTech's Medallion Fund (I believe) is balanced to exactly 0% beta .. And Simons has said something I've tended to find, which is that these types of funds tend to do best with market volatility (RenTech returning >90% through 2008-09) .. Which obviously a simple defensive fund wouldn't achieve .... Then again, if QAI's alpha portion comes into its own through the next financial crisis, it may be worth its salt?

Nah, check the portfolio. It's just an ETF that invests in ETFs, mostly bond ones. The largest holdings now, comprising 40% of the fund, are Vanguard short-term bond and Vanguard total bond ETFs. It also has stocks and a bit of currency and commodity exposure, basically just long-only everything.

So okay, maybe they're switching the underlying holdings based on something over time (I don't quite get how it's theoretically tracking hedge funds), but the net effect is what looks like a conservative balanced index fund. Except they charge an ER of 0.91%. Or actually, their website says 0.75%, so let's just go with that.

I just think it's sad this is touted as a replacement for bonds and not as risky as stocks, based on the performance characteristics and underlying holdings. If you have something like this with a portfolio of stocks and bonds, you're not getting diversification or anything much new. You're just paying them extra fees for largely more of the same of what you already own.

Seriously, does this look like it's liable to do something much different in a downturn than VTINX (Vanguard target retirement income, which is 30% stocks) or any kind of conditions:
http://quotes.morningstar.com/chart/fun ... A%5B%5D%7D

There are a lot of bad funds out there, to be sure. Some, like this one, can apparently still gather $1B AUM on the strength of largely the market beta exposure dragging the fund returns up in a bull market. Maybe it will somehow copy or track hedge funds closely (and not with too much a time lag) and manage to properly market time to reduce equity exposure before a crash? In that case, then why not just watch its holdings rather than own the actual fund?


As for "these funds" or ones in general doing well in crises, it depends, right? If volatility is high you expect the magnitude of returns to be larger. If you're on the right side of the trades, you stand to gain a lot. If you're long the wrong assets, you lose a lot.

This just covers equity market neutral mutual funds, but you can see the average fund performance by Morningstar here, including through the financial crisis:
http://quotes.morningstar.com/chart/fun ... 2%3A955%7D

Vanguard's fund suffered in that time with the previous manager. Most funds did better. But the average doesn't exactly look spectacular.

User avatar
Maynard F. Speer
Posts: 2139
Joined: Wed Mar 18, 2015 10:31 am

Re: QSPIX - thoughts on interesting fund

Post by Maynard F. Speer » Thu Aug 20, 2015 8:23 pm

lack_ey wrote:
Maynard F. Speer wrote:
lack_ey wrote:^ re: (9)

On a more serious note, I think the majority of AQR alternatives-class mutual funds have pretty much zero beta except the long/short equity fund that specifically says it is "long and short global equities with moderate net long equity exposure." If Asness had his way, hedge funds in general would have zero beta, according to what he says and writes.

Anyway, more important than the amount of beta is the amount of alpha or non-beta content, seeing as that is what one is paying for (beta being very cheap to access elsewhere). The data seems to show average hedge fund alpha shrinking over time, so people are piling into strategies in many funds with net beta and not much to show for it.

This tangentially reminds me of a terrible article I once read about the IQ Hedge Multi-Strategy Tracker ETF (QAI), a somewhat popular hedge-fund-holdings-replicator fund with $1B AUM and an ER of 0.91%. This is one of the ETFs I said I looked at and found nothing of interest in:
http://www.etf.com/sections/features-an ... posure-qai

You have to see it to believe it. (If the graph there isn't obvious enough, check the holdings and it's 30% long stocks and sure enough has a beta around 0.3, with negative alpha so far. The article touts said fund as a replacement for bonds.)


I saw another article not long ago suggesting you could recreate hedge-fund performance simply by holding an index tracker and 2/3rds cash .. You have got to be careful - certainly when a lot of people don't 'get' hedge funds (possibly sometimes even those who run them)

What I've found in the UK, at least, is if you want market neutral returns, with low volatility, you are typically looking at around 4% after fees .. There are long/short equity funds that do better, but they can be >80% long equities ... Which is fine if they can time when to be 80% short equities - but I've seen the mess they get in when the market correction doesn't come along

RenTech's Medallion Fund (I believe) is balanced to exactly 0% beta .. And Simons has said something I've tended to find, which is that these types of funds tend to do best with market volatility (RenTech returning >90% through 2008-09) .. Which obviously a simple defensive fund wouldn't achieve .... Then again, if QAI's alpha portion comes into its own through the next financial crisis, it may be worth its salt?

Nah, check the portfolio. It's just an ETF that invests in ETFs, mostly bond ones. The largest holdings now, comprising 40% of the fund, are Vanguard short-term bond and Vanguard total bond ETFs. It also has stocks and a bit of currency and commodity exposure, basically just long-only everything.

So okay, maybe they're switching the underlying holdings based on something over time (I don't quite get how it's theoretically tracking hedge funds), but the net effect is what looks like a conservative balanced index fund. Except they charge an ER of 0.91%. Or actually, their website says 0.75%, so let's just go with that.

I just think it's sad this is touted as a replacement for bonds and not as risky as stocks, based on the performance characteristics and underlying holdings. If you have something like this with a portfolio of stocks and bonds, you're not getting diversification or anything much new. You're just paying them extra fees for largely more of the same of what you already own.

Seriously, does this look like it's liable to do something much different in a downturn than VTINX (Vanguard target retirement income, which is 30% stocks) or any kind of conditions:
http://quotes.morningstar.com/chart/fun ... A%5B%5D%7D

There are a lot of bad funds out there, to be sure. Some, like this one, can apparently still gather $1B AUM on the strength of largely the market beta exposure dragging the fund returns up in a bull market. Maybe it will somehow copy or track hedge funds closely (and not with too much a time lag) and manage to properly market time to reduce equity exposure before a crash? In that case, then why not just watch its holdings rather than own the actual fund?


As for "these funds" or ones in general doing well in crises, it depends, right? If volatility is high you expect the magnitude of returns to be larger. If you're on the right side of the trades, you stand to gain a lot. If you're long the wrong assets, you lose a lot.

This just covers equity market neutral mutual funds, but you can see the average fund performance by Morningstar here, including through the financial crisis:
http://quotes.morningstar.com/chart/fun ... 2%3A955%7D

Vanguard's fund suffered in that time with the previous manager. Most funds did better. But the average doesn't exactly look spectacular.


It's a shame the data doesn't start before the financial crisis - because that would be the yardstick for me .. But I'm quite sure you're right with your reservations

I actually use a few (essentially) long-only hedge funds .. The two I use are top-rated UK funds (Ruffer and Newton Real Return - purple and orange on the chart)

And both of them are 30-40% equities, the rest in bonds and cash .. So they look very conservative when markets are rallying (half market returns if you're lucky .. very similar to a 30:70 retirement-style fund - in fact the volatility in the orange is premium/discount to NAV, as it's a closed-ended fund) ... But they're run by perma-bears, and as risk increases in markets, they tend to load up on options, currencies and gold, so what's been a drag on returns through a bull market can suddenly pay off - but you really are investing in the managers, and they really are *hedges*

Image

The pink line (2nd best since 2008) is the largest fund in the UK .. It's this new breed of quite transparent hedge fund - where you've got 1/3rd of the portfolio producing market returns, 1/3rd essentially doing pair trades (market neutral returns), and 1/3rd hedging against their own assumptions being wrong .. So it's not in the habit of disappointing so much, and it's very well risk-managed as a whole

And then the brown line (at the bottom) is what you get from a fund with low-risk, low-to-zero beta .. So some might argue that's the only market neutral fund to own, if you just combine it with beta, and in fact they recommend leveraging too (but it's also got the highest fees)
"Economics is a method rather than a doctrine, an apparatus of the mind, a technique of thinking, which helps its possessor to draw correct conclusions." - John Maynard Keynes

needtosave
Posts: 16
Joined: Wed Aug 19, 2015 10:31 pm

Re: QSPIX - thoughts on interesting fund

Post by needtosave » Thu Aug 20, 2015 8:35 pm

lack_ey wrote:The multiple hypothesis testing and all the related statistical problems with "significance" levels are very well known (and plague many fields of study including finance and economics) and I think anybody with any exposure to statistics at least knows them and basic fixes (cludges?) like Bonferroni, the basic concepts.


1. If Bonferroni is so well known, then it's a mystery why it or some similar statistical adjustment isn't used more in investing literature. Of all the fixes for multiple hypothesis testing, out of sample testing is probably the crudest tool because it doesn't take into account the number of hypotheses tested.
2. In machine learning in other fields, you train on a training data set(in sample), take the strategy that works the best, and then report the results of that strategy on the test data(out of sample). The performance of your algorithm is reported as its performance on the out of sample data. The investing literature seems to take the anomalous approach of using the out of sample data as some sort of pass/fail test and then quotes the in sample results as the best estimate of various factor premiums.

User avatar
grap0013
Posts: 1861
Joined: Thu Mar 18, 2010 1:24 pm

Re: QSPIX - thoughts on interesting fund

Post by grap0013 » Fri Aug 21, 2015 7:27 am

matjen wrote:
Grap, Grap, Grap...what are we going to do with you? Everyone knows that the standard that QSPIX has to be judged against is the following:
1) Never, ever lose money for any period of time
2) In mixed markets do great (better than bonds and equities)
3) In bull markets do better than the slice of total equities that is doing the absolute best and/or the best actively managed funds for that period
4) In bear markets do better than Treasuries (and have their correlations in bear markets)
5) Be 100% transparent with your trading strategies so that lay persons can follow them
6) Never adjust your strategy
7) Cost less than .09 basis points
8) Be tax efficient

I don't think I have heard anyone ask for the returns to be inflation protected but I may have missed that. 8-)


Reminds me of those crazy gold bug fixed false beliefs. We must be loco! :shock:

Did I mention QSPIX was up again yesterday while equities were down? :wink:
There are no guarantees, only probabilities.

User avatar
nisiprius
Advisory Board
Posts: 33797
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: QSPIX - thoughts on interesting fund

Post by nisiprius » Fri Aug 21, 2015 8:10 am

needtosave wrote:
lack_ey wrote:The multiple hypothesis testing and all the related statistical problems with "significance" levels are very well known (and plague many fields of study including finance and economics) and I think anybody with any exposure to statistics at least knows them and basic fixes (cludges?) like Bonferroni, the basic concepts.
1. If Bonferroni is so well known, then it's a mystery why it or some similar statistical adjustment isn't used more in investing literature...
Indeed.

lack_ey, I challenge you to cite an actual example of some investing literature you've read in the past month or so that shows that the Bonferroni adjustment or anything superior was being used.

(I don't think "kludge" is a fair description; the Bonferroni adjustment is completely valid and is an application of simple probability theory. The only objection to it can be that it is conservative. If something is significant after Bonferroni correction, it's significant. The only objection to it is that in some situations, where you have planned experiment to meet the preconditions for a more powerful test, the Bonferroni correction may not show significance where the more powerful test would).
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.

lack_ey
Posts: 5556
Joined: Wed Nov 19, 2014 11:55 pm

Re: QSPIX - thoughts on interesting fund

Post by lack_ey » Fri Aug 21, 2015 9:09 am

nisiprius wrote:
needtosave wrote:
lack_ey wrote:The multiple hypothesis testing and all the related statistical problems with "significance" levels are very well known (and plague many fields of study including finance and economics) and I think anybody with any exposure to statistics at least knows them and basic fixes (cludges?) like Bonferroni, the basic concepts.
1. If Bonferroni is so well known, then it's a mystery why it or some similar statistical adjustment isn't used more in investing literature...
Indeed.

lack_ey, I challenge you to cite an actual example of some investing literature you've read in the past month or so that shows that the Bonferroni adjustment or anything superior was being used.

(I don't think "kludge" is a fair description; the Bonferroni adjustment is completely valid and is an application of simple probability theory. The only objection to it can be that it is conservative. If something is significant after Bonferroni correction, it's significant. The only objection to it is that in some situations, where you have planned experiment to meet the preconditions for a more powerful test, the Bonferroni correction may not show significance where the more powerful test would).

I just mostly meant that Bonferroni is
(1) a very intuitive result, with the concept of uncorrected multiple tests producing too many false positives being common sense (except that I'll admit that somehow many practitioners don't seem to have this common sense—or rather, they prefer to see if they can get away with publishing weaker results)
(2) has unnecessarily low power for many applications
(3) still just relates concepts to arbitrarily chosen levels of significance and supposed 5% (or whichever number we please) false positive rates*

*except that, you know, when p values, t-stats, etc. are quoted they're often on distributions that obviously violate the underlying assumptions of the tests by having dependence, non-normality, or what have you.

Personally I'm not that big a fan of frequentist statistical inference for the real world, having a more subjective and Bayesian mindset. Though admittedly, I don't have much of a statistical background at all and don't use this stuff for my work, so I am not as familiar with the mechanics as I would like. I've had one statistics course in my life (several more dealing with probability, stochastic processes, and related analysis, though), and it definitely covered Bonferroni early on, and everybody else seemed to already know these things from previous classes. Maybe my perception of the use and abuse of statistics is way off the mark. I could easily believe that.

Here are a couple papers not from Campbell that at least address some related issues (I don't think I've read any investing paper in the last month, so I just searched, so this may be cheating):
http://papers.ssrn.com/sol3/papers.cfm? ... id=2184849
http://papers.ssrn.com/sol3/papers.cfm? ... _id=659941

My overall impression of the investing literature is that, like in every field, people generally don't admit when they have tested multiple things and only write about one of them, leading to publishing bias in that sense. In some cases they're not really running multiple tests on the same data, though, but the same test on multiple sets of data (that are not independent).

User avatar
Ketawa
Posts: 1841
Joined: Mon Aug 22, 2011 1:11 am
Location: Norfolk

Re: QSPIX - thoughts on interesting fund

Post by Ketawa » Fri Aug 21, 2015 9:14 am

nisiprius wrote:lack_ey, I challenge you to cite an actual example of some investing literature you've read in the past month or so that shows that the Bonferroni adjustment or anything superior was being used.


I was skimming the paper linked in this thread recently: "most [] research findings in financ[e] are likely false".

larryswedroe
Posts: 15597
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: QSPIX - thoughts on interesting fund

Post by larryswedroe » Fri Aug 21, 2015 9:17 am

Let's assume that value was based on data mining by Fama and French. Now you run the same test using the same metric in the pre sample period and you get the same basic finding. Now you run the same test in the post sample and get the same basic finding. Now you run the same test in say 30 different countries and get basically the same findings. Then you run the same type of test for value on commodities and currencies and bonds and get the same basic finding.

Now that's what has been done with all of the factors involved in the strategy here.

Each of the factors is not only persistent and pervasive but is implementable after costs and his supported by logical risk or behavioral explanations in the literature (where behavioral explanations there are logical explanations for why arbitrage limits another issues prevent the mispricing from being eliminated)

IMO this persistence and pervasiveness and the out of sample tests provides solid support for the factors and why they are likely to persist. IN other words, the data is highly robust, with the only debates being whether they are risk factors or behavioral which is very difficult to sometimes determine, though they are all highly variable (making them look like risk factors).

Larry

longinvest
Posts: 2447
Joined: Sat Aug 11, 2012 8:44 am

Re: QSPIX - thoughts on interesting fund

Post by longinvest » Fri Aug 21, 2015 9:36 am

larryswedroe wrote:Each of the factors is not only persistent and pervasive but is implementable after costs and his supported by logical risk or behavioral explanations in the literature (where behavioral explanations there are logical explanations for why arbitrage limits another issues prevent the mispricing from being eliminated)

Larry,

I think that you are simply predicting the future, and, I do not believe that anybody can predict the future, not even you.

In a previous post, I have reminded readers of your similar use of the present tense to describe an investment in 2008, a statement that turned out to be utterly false.

I think that your above statement is misleading, specially for new investors who do not understand that you probably don't mean that they are persistent and pervasive; instead, you mean that they have been so in some markets over some historical periods, an that you and some others, not everybody, believe that there is a good probability that they could remain so in the future. There is a world of difference, here.

If my understanding is false, please tell me so, and explain how you can guarantee what will happen in the future. I am all ears.

Regards,

longinvest
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

lack_ey
Posts: 5556
Joined: Wed Nov 19, 2014 11:55 pm

Re: QSPIX - thoughts on interesting fund

Post by lack_ey » Fri Aug 21, 2015 9:52 am

Is there really a need for clarification? If you read the last paragraph, Larry wrote "why they are likely to persist," which is pretty self explanatory.

Commodities have provided diversification from stocks and bonds since 2008 (too small a period for assessing asset classes, but we'll go with it). Just on the negative side. Keeping up with inflation could still well happen in the long run.

In any case, if one accepts some likelihood of robustness and legitimate existence in the past beyond noise, questioning if something will persist or not is just another prediction, same as predicting that something will continue on. You've got to have some kind of stance. If you don't know the future, then can you guarantee something won't happen? There's always a degree of uncertainty. It's just about placing bets on what seems worthwhile, taking all the risks, costs, benefits (not in isolation but in tandem with everything) into mind.

Johno
Posts: 1883
Joined: Sat May 24, 2014 4:14 pm

Re: QSPIX - thoughts on interesting fund

Post by Johno » Fri Aug 21, 2015 9:53 am

Ketawa wrote:
nisiprius wrote:lack_ey, I challenge you to cite an actual example of some investing literature you've read in the past month or so that shows that the Bonferroni adjustment or anything superior was being used.


I was skimming the paper linked in this thread recently: "most [] research findings in financ[e] are likely false".

This was (among) the actually good threads here recently about this issue, as opposed to 'there can be a problem with overfitting, go read about it, hence QSPIX is bogus, but don't ask me to provide any specific comment about the statistics of QSPIX, it's bogus so why should I?' :D

But note the comment in OP of that thread "It's less of a problem if you start with a theory and then test it than if you just run tests (aka data mining)." It still doesn't mean it can't be a problem, but again if we're going to actually talk about the concepts in this fund, the research as a whole is clearly more the first of those things than the second. As Larry Swedroe's post says, maybe you can say Fama/French original value findings were 'data mining', but a lot harder to say that about all subsequent attempts to verify or expand those findings. Likewise as mentioned a bunch of times now, it's absurd to treat the currency carry trade as a 'data mining' artifact. It's been a source of real profit for real traders on a significant scale for a long time. Does this mean it will persist indefinitely? No, a good paper posted in an earlier discussion here of QSPIX, by another investment outfit, gave an interesting argument for a theoretical basis of the currency carry trade and why it was less likely to work in the future (it is a real risk premium related to risk of suddenly different inflation rates among currencies, which risk has tended to decline). But nobody knows the future, again most importantly in that all statistical arguments are based on an implicit assumption of a stationary distribution, which isn't really the case. And yes that affects the future of the ERP as well.

User avatar
Maynard F. Speer
Posts: 2139
Joined: Wed Mar 18, 2015 10:31 am

Re: QSPIX - thoughts on interesting fund

Post by Maynard F. Speer » Fri Aug 21, 2015 10:13 am

longinvest wrote:
larryswedroe wrote:Each of the factors is not only persistent and pervasive but is implementable after costs and his supported by logical risk or behavioral explanations in the literature (where behavioral explanations there are logical explanations for why arbitrage limits another issues prevent the mispricing from being eliminated)

Larry,

I think that you are simply predicting the future, and, I do not believe that anybody can predict the future, not even you.


Estimating reasonable future outcomes ..

As Jim Simons says, all investing success comes down to "luck" ... It's just mathematicians prefer to use the term "probability" - which is something that can be studied, calculated and manipulated, rather than just putting all your money on Black and hoping the wind blows your way

As William Sharpe says, there is only *one* passive portfolio, and that's the global market portfolio - anything else involves prediction .. And if you owned the global market portfolio, you would have at least 3-5% in market neutral hedge funds, because the market does
"Economics is a method rather than a doctrine, an apparatus of the mind, a technique of thinking, which helps its possessor to draw correct conclusions." - John Maynard Keynes

larryswedroe
Posts: 15597
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: QSPIX - thoughts on interesting fund

Post by larryswedroe » Fri Aug 21, 2015 10:35 am

Few things
First we all have to "predict" the future to some degree. But IMO it isn't about luck as Simons is quoted as saying.
It's about putting the odds of success on your side by using the evidence we have. This is no different with the equity risk premium.
I'm confident Simon doesn't believe it is luck, just a cute saying. He is in fact betting on relationships with strong evidence to persist. He knows it isn't certain, so to that degree it's luck (but it's really about putting odds on your side by using the best available information, making informed decisions)

Having said that my crystal ball is cloudy and I don't know which factors will provide positive returns so I want to diversify across them as much as possible while considering the costs of implementation. And TSM portfolios have exposure to only beta factor as a systematic factor.

Best wishes
larry
Last edited by larryswedroe on Fri Aug 21, 2015 10:56 am, edited 1 time in total.

User avatar
Maynard F. Speer
Posts: 2139
Joined: Wed Mar 18, 2015 10:31 am

Re: QSPIX - thoughts on interesting fund

Post by Maynard F. Speer » Fri Aug 21, 2015 10:55 am

I absolutely agree it's just a cute saying ..

But I like that it bridges these two typically conflicting perspectives:
- When you roll a dice once, it's luck;
- When you know how many sides a dice has, and roll it 20 times, then you're talking about probability .. I think that's a great concept to start from
"Economics is a method rather than a doctrine, an apparatus of the mind, a technique of thinking, which helps its possessor to draw correct conclusions." - John Maynard Keynes

larryswedroe
Posts: 15597
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: QSPIX - thoughts on interesting fund

Post by larryswedroe » Fri Aug 21, 2015 11:05 am

Maynard
I knew you knew it was being cute, just bringing that to the attention of others.

Here's another interesting point most investors miss, with typical balanced portfolios the VAST majority of the risks are in stocks, which is behind the idea of risk parity strategies (which I'm not particularly a fan of though like the direction it takes you).

With that in mind from a recent paper. "From 1999 through 2013, the correlation of returns between a 60/40 portfolio and a 100% equity portfolio was 0.98. Even if an investor invests 30% in stocks and 70% in bonds,the resulting correlation with a 100% stock portfolio is 0.85. As a result, a long-only stock
and bond portfolio is not particularly well diversified."

So IMO it's prudent to think in non-traditional terms. Thinking of diversification in different ways than just "asset classes", but at same time making sure there is persistent and pervasive evidence and that there are logical risk based (preferable) explanations or behavioral ones (and that in these cases limits to arbitrage are likely to allow the mispricings to continue).

BTW--was on call with Fama today and he noted this of interest, Rolf Banz's famous paper on small stocks wasn't a discovery of the size premium but the discovery of the small value premium. The stocks he looked at were NYSE only which tend to be small value (risky) stocks, not the small growth stocks you see on the NASDAQ.

Larry

User avatar
Maynard F. Speer
Posts: 2139
Joined: Wed Mar 18, 2015 10:31 am

Re: QSPIX - thoughts on interesting fund

Post by Maynard F. Speer » Fri Aug 21, 2015 12:00 pm

I felt the same about stock and bond correlations - I believe I've read the correlations are higher in a low rate environment (which seems quite likely to stick around)

What's really held up for me recently is an allocation to P2P lending (which I think of as an unknown but containable risk), and actually active management - we've got a great fund manager over here, Giles Hargreave, who's been investing in very small companies since the 60s ... I think of it as a space where there's no real substitute for face-to-face meetings with the firms you're investing in (many being pre-profit, early stage), and the fact they conduct thousands of these every year means (to me) you're paying for something slightly inaccessible to many managers, and operating on different principles

I'd love to see AQR launch in the UK though - I think there'd be a lot of interest
"Economics is a method rather than a doctrine, an apparatus of the mind, a technique of thinking, which helps its possessor to draw correct conclusions." - John Maynard Keynes

larryswedroe
Posts: 15597
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: QSPIX - thoughts on interesting fund

Post by larryswedroe » Fri Aug 21, 2015 12:26 pm

Longinvest
Just simple question. The evidence on the equity premium is persistent and pervasive. And you are betting on it continuing. Are you predicting the future? If rational, and I assume you are, you are predicting a premium as the mean (expected) but with a very wide dispersion of potential outcomes also being considered and you are willing to live with those risks in return for the "expected" premium.

The other factors have the same type of persistence and pervasive evidence across time, economic regimes, countries, regimes, and even asset classes. The reasons to consider investing in these factors are the same for investing in beta, the historical evidence of persistent and pervasive premiums backed by logical risk/behavioral explanations, and implementable after costs.

Larry

longinvest
Posts: 2447
Joined: Sat Aug 11, 2012 8:44 am

Re: QSPIX - thoughts on interesting fund

Post by longinvest » Fri Aug 21, 2015 2:32 pm

Larry,

larryswedroe wrote:Longinvest
Just simple question. The evidence on the equity premium is persistent and pervasive.


If I remember correctly, in the U.S. stock and bond markets of the 1800s there were very long stretches when bonds had higher returns than stocks. Doesn't this mean that there was no equity premium over these periods? What does it mean, really, for something to be "persistent and pervasive", if it doesn't happen during the investor's lifetime?

It is often forgotten that in the 1900s the gold standard was abandoned across the world. This has affected the value of bonds everywhere, as it caused unexpected inflation in all the affected countries. So, testing for a property around the world does not eliminate the fact that a very similar unexpected event happened everywhere.

Do you really know that stocks will beat bonds in the next 5, 10, 20, or 30 years? I do not think that anybody knows.

The law of supply and demand informs us that stocks must be priced often enough so that they will return more than short-term bonds, otherwise, investors wouldn't buy them and their price would drop. But, longer term bonds have risks, too. Which is more important: stock market risk over a 30 year period, or inflation risk for a 30-year nominal bond? Why shouldn't markets strive to price assets relative to their perceived risks?

Note that I refrain from using statistical terms such as "expected returns". I am not a statistician and I do not know the exact and complete definition of this term.

If I look at the past returns of Canadian stocks and bonds over the last 54 years (1961-2014), I get:

Bonds: annualized return 7.761%, Standard Deviation 7.150%
Stocks: annualized return 9.684%, Standard Deviation 16.282%
50/50: annualized return 9.104%, Standard Deviation 8.775%

We must be very careful, here. 1961-2014 is an arbitrary period; it just happens to be the longest period for which I have data. For an investor with a 30-year horizon, this is less than 2 data points!

Yet, in various sub-periods, bonds outperformed stocks, and in others, stocks outperformed bonds. Even for long periods. In the 30-year period 1979-2008, for example, bonds outperformed stocks. Was there an equity premium, over that period?

So, no, I do not predict that bonds or stock will outperform in the future. I only predict that they will continue to have different risks and behavior, and that investment-grade bonds are most likely to pay coupons and principal back*. I have trouble predicting future stock growth and dividends. I have no crystal ball.

* There is still a possibility that another country could invade Canada and all stocks and bonds loose their value.

larryswedroe wrote:And you are betting on it continuing. Are you predicting the future? If rational, and I assume you are, you are predicting a premium as the mean (expected) but with a very wide dispersion of potential outcomes also being considered and you are willing to live with those risks in return for the "expected" premium.


I do not give you permission to misrepresent what I do or think. You are wrong. I do not bet on the fact that stocks will return more than bonds. Actually, my portfolio is divided 50% in stocks and 50% in bonds. I could have added more stocks, knowing that it would affect the volatility of my portfolio and increase the probability of both higher and lower future returns. Stocks do not offer any specific guarantee of future performance.

If you can tell me which Canadian common stock certificate I could buy that contains an explicit future return promise, I would be very grateful.

I know that my stock ETFs will pay me regular distributions (based on dividends) and have a volatile NAV. I do also know that my bond ETF will pay me regular distributions (based on coupons) and have a less volatile* NAV than my stock ETFs. The fact that both my stocks and my bonds pay me "rent" for borrowing my money suits me. I do not like to lend my money without receiving rent for it. That's why I don't invest in commodities.

* Thanks to mathematics, I know that the volatility of my bond ETF is limited.

I can look at past data and learn about history (Bernstein's The Four Pillars of Investing), but I have no illusion about it: past data is helpful to improve my understanding of investment behavior and various risks, but it does not predict future return or outperformance.

larryswedroe wrote:The other factors have the same type of persistence and pervasive evidence across time, economic regimes, countries, regimes, and even asset classes. The reasons to consider investing in these factors are the same for investing in beta, the historical evidence of persistent and pervasive premiums backed by logical risk/behavioral explanations, and implementable after costs.


Amazingly, in a different thread, you explained to me something to the effect that the size factor did not exist in Canada and that it disappeared in the U.S. (My understanding could be inexact, here). How can you know that other factors won't appear and disappear, in various future investor lifetimes, too?

I do not wish to "bet" my money. I am willing to expose it to a reasonable amount of risk in both the bond and stock markets, while being paid a "rent" to do so. I rely on William Sharpe's theorem to simply aim at getting my fair share of the return of these markets* using low-cost total-market ETFs (there are no sufficiently low-cost equivalent mutual funds in Canada).

* I do not expect a "rebalancing bonus". I am a fan of investing new money into lagging assets (below target allocation) and only rebalance actively once a year or two, to control the volatility of the portfolio.

I do not think that all other investors should invest the way I do. If an investor believes in factors and that concentrating parts of his portfolio into them can improve the outcome, I think that this investor should invest accordingly. What I do object to, is presenting probabilistic future outcomes as sure future outcomes.

Best regards,

longinvest
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

larryswedroe
Posts: 15597
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: QSPIX - thoughts on interesting fund

Post by larryswedroe » Fri Aug 21, 2015 3:39 pm

longinvest
fWIW you are confusing many issues, not reading clearly what I have said.

First, by investing in stocks you are making a "bet" that they will outperform, otherwise you'd be very foolish to take the much greater risks of stocks. There must be an EX-ANTE risk premium to compensate investors for taking the greater risk. Whether the risks show up or not determines whether there was compensation for taking the risk or not. But EX ANTE, before we know the outcome we have an expectation of higher returns while accepting the risks that we MIGHT have lower returns. The same is true of other factors.

And that means you might have to wait even 40 years to be rewarded --long term bonds outperformed stocks from 1969-08, and Japanese investors in stocks may never be rewarded. That's why we don't take more risk than we have the ability, willingness or need to take and why it's prudent to diversify across different risks. But we have to make decisions in the presence of uncertainty, acknowledging we don't know what the future holds, but using the evidence we do have to put the odds most likely in our favor, giving us the best chance to succeed.

I fully acknowledge and could not have stated it any more clearly that I don't know what factors will provide the best returns and then explain that's why I diversify across them.

And yes I can ESTIMATE the expected return to Canadian or other stocks as we have tools to do that. But we acknowledge that there is great uncertainty in the forecasts. One has to add or subtract about 8% from the mean forecast to get the full range of 10-year outcomes we have experienced, and nothings says future outcomes cannot be wider. The way financial economists have found to be the best estimate of future returns is to take either current valuations or something like the CAPE 10 and use the earnings yield. You might want to read NYU professor Damodaran on the subject. He has a great website/blog.

Now you may not like the term "bet" but any investment we make is based on some type of bet, whether informed about risks or not.

And btw, rebalancing doesn't control volatility. If you rebalance and buy stocks you are increasing the volatility of the portfolio, not controlling it. What you have done by rebalancing is restoring the asset allocation you desired, not allowing the market to determine it for you.

Best wishes
Larry

longinvest
Posts: 2447
Joined: Sat Aug 11, 2012 8:44 am

Re: QSPIX - thoughts on interesting fund

Post by longinvest » Fri Aug 21, 2015 4:08 pm

Larry,

There are many definitions for the word bet; the most common is numbered 1 in The Free Dictionary:
http://www.thefreedictionary.com/bet
1. An agreement usually between two parties that the one who has made an incorrect prediction about an uncertain outcome will forfeit something stipulated to the other; a wager

This is not how I perceive investing into stocks and bonds.

Best regards,

longinvest
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

longinvest
Posts: 2447
Joined: Sat Aug 11, 2012 8:44 am

Re: QSPIX - thoughts on interesting fund

Post by longinvest » Fri Aug 21, 2015 4:19 pm

larryswedroe wrote:And btw, rebalancing doesn't control volatility. If you rebalance and buy stocks you are increasing the volatility of the portfolio, not controlling it. What you have done by rebalancing is restoring the asset allocation you desired, not allowing the market to determine it for you.


I agree with your statement.
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

larryswedroe
Posts: 15597
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: QSPIX - thoughts on interesting fund

Post by larryswedroe » Fri Aug 21, 2015 4:37 pm

longinvest
RE definition of bet.
That is precisely why I put a " " around the term. You can call it what you want but that's just semantics. You are betting on something, you are just calling it an investment. And I would not disagree of course. But the investment is based on a bet that the premium will be realized, otherwise you would not make the investment. If you didn't "expect" the premium to be realized there is no logic to investing in stocks.

But go back and see where you can find that I said that I knew which premiums would be realized as you seem to state that I did. I never made any such statement. My only point is that while you are "betting" on one equity premium, beta, and one or two bond premiums (term and credit) IMO there is substantial evidence of equal quality to justify "betting" on other premiums. Now you can come to a different conclusion on which premiums you want to "bet" on or invest in, but that doesn't change the fact that the evidence is strong for those other premiums, just as it is for the ones you are investing in. I just choose to diversify them more. Not saying one strategy is superior to another either. IN fact I believe that the best portfolio is the one the investor is most likely to stick to. And if that's TSM and US only that's the right one for them---though I might try to convince them that it isn't. (:-))

Best wishes
Larry

longinvest
Posts: 2447
Joined: Sat Aug 11, 2012 8:44 am

Re: QSPIX - thoughts on interesting fund

Post by longinvest » Fri Aug 21, 2015 4:54 pm

Larry,

larryswedroe wrote:longinvest
First, by investing in stocks you are making a "bet" that they will outperform, otherwise you'd be very foolish to take the much greater risks of stocks. There must be an EX-ANTE risk premium to compensate investors for taking the greater risk. Whether the risks show up or not determines whether there was compensation for taking the risk or not. But EX ANTE, before we know the outcome we have an expectation of higher returns while accepting the risks that we MIGHT have lower returns. The same is true of other factors.


We both agree that there is a relation between risk and potential return. This was clear in my previous text. I wrote it in a non-statistician vocabulary, saying that stocks must be priced often enough as to return more than short-term bonds. The problem is that we don't know "ex-ante" if stocks are currently priced to deliver higher returns.

If your "equity premium" is some sort of statistical term, like "expected returns", then I can say nothing about it, because I do not fully understand the concept, and I have funnier things to do in life than reading statistics books. :)

So, everytime I decide to buy a stock of a great business, like the Royal Bank of Canada, I am taking some risk, but I also become part owner of the biggest Canadian bank, whose employees work hard to make sure it survives and continues to pay their salaries. I am not making a bet with my neighbor.

If I decide to buy a 30-year Government of Canada nominal bond, I am also taking the risk that the bond and its coupons will fail to even match future inflation.

In both cases, I am taking risk. Which one is riskier? Is it much riskier? How does one quantify this? Yes, the Royal Bank could go bankrupt, but it also is likely to continue growing its dividends in the future. So, shouldn't I discount the Government of Canada bond in proportion of its perceived risk of under-performance?

I am not so sure that premiums are stable. I am actually pretty sure that they change, based on which market was measured and which periods were measured.

There is a real possibility for investors to cherish businesses they like and bid up their prices to what could be viewed, after the fact (at the end of an investing period), as irrational. They could similarly develop an irrational aversion to long-term nominal bonds.

So, whether the statistical concepts of "expected returns" and "equity premiums" hold or not, for me, a single investor with one specific investing horizon, they do not contain enough precision to tell me, with certainty, which of bonds or stocks will outperform over my horizon. (I should say "horizons", to be more exact, as I make successive investments, and one day, hopefully, I will make successive withdrawals).

So, instead of selecting an asset allocation by trying to estimate my future returns, and "bet" more money on the highest rewarding asset, I instead focus on the risks. I invest in minimal-cost total market index funds so as to virtually eliminate the risk of under-preforming each market I invest into. I invest both in stocks and bonds to diversify and limit exposure to the risks of each asset. I invest both in the domestic and international stock markets for the same reasons. That might not look very sophisticated, but it works for me, and it is rational.

Best regards,

longinvest
Bogleheads investment philosophy | Lifelong Portfolio: 25% each of (domestic/international)stocks/(nominal/inflation-indexed)bonds | VCN/VXC/VAB/ZRR

larryswedroe
Posts: 15597
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: QSPIX - thoughts on interesting fund

Post by larryswedroe » Fri Aug 21, 2015 5:10 pm

Longinvest
Equity risk premium is simply the historical return of stocks, annual average, minus the annual average return on t-bills (for US).

We do know that stocks are more risky than bonds in your example, that's the evidence in any conventional sense of the word. Now we don't know which will prove more risky after we make the investment, but ex ante we know which is.

And of course, as I stated there is no precision. That's an illusion and don't know where you got the idea that anyone thinks there is precision.

Nor do I know where you got the idea that premiums are stable. We know that in fact they are not. Otherwise there would be no risk. Premiums are in fact many times more volatile than the size of the premium generally--same true for stocks (beta) in general. The SD has been about 2.5 x the premium, that's risk

So if you invest in TSM you are betting on beta. Period. That's all. It's not good or bad. Rational or irrational. It just is what you are investing in. A bet or investment in a single factor that you have decided to globally diversify (a good thing) but it's still an investment in one single factor. With bonds it's investing in one or two factors.

Nothing more can add. Hope it is helpful

Best wishes
Larry

needtosave
Posts: 16
Joined: Wed Aug 19, 2015 10:31 pm

Re: QSPIX - thoughts on interesting fund

Post by needtosave » Fri Aug 21, 2015 7:11 pm

Johno wrote:But note the comment in OP of that thread "It's less of a problem if you start with a theory and then test it than if you just run tests (aka data mining)." It still doesn't mean it can't be a problem, but again if we're going to actually talk about the concepts in this fund, the research as a whole is clearly more the first of those things than the second.


Now, you're getting into circular reasoning. If you only include tests of "reasonable theories", then the Bonferroni adjustment required is much less. That begs the question of who is the arbiter of what is a reasonable theory in investing. In a field, such as medicine, there is some existing body of knowledge that all or nearly all medical researchers agree to and a theory, can be judged as to how it fits into that existing body of knowledge. I doubt there's an academic consensus on much of anything in investing and to the extent there is, I would guess academics treat the market as being at least weak efficient.

User avatar
Yesterdaysnews
Posts: 424
Joined: Sun Sep 14, 2014 1:25 pm

Re: QSPIX - thoughts on interesting fund

Post by Yesterdaysnews » Fri Aug 21, 2015 7:36 pm

down only -0.78% today.

larryswedroe
Posts: 15597
Joined: Thu Feb 22, 2007 8:28 am
Location: St Louis MO

Re: QSPIX - thoughts on interesting fund

Post by larryswedroe » Fri Aug 21, 2015 7:51 pm

yesterdaysnews, and was up yesterday, but that's just a couple of days. There have been years when both stocks and QSPIX both down, the correlation isn't negative 1 so should expect that to on occasion happen.


Larry

Johno
Posts: 1883
Joined: Sat May 24, 2014 4:14 pm

Re: QSPIX - thoughts on interesting fund

Post by Johno » Sat Aug 22, 2015 8:38 am

needtosave wrote:
Johno wrote:But note the comment in OP of that thread "It's less of a problem if you start with a theory and then test it than if you just run tests (aka data mining)." It still doesn't mean it can't be a problem, but again if we're going to actually talk about the concepts in this fund, the research as a whole is clearly more the first of those things than the second.


Now, you're getting into circular reasoning. If you only include tests of "reasonable theories", then the Bonferroni adjustment required is much less. That begs the question of who is the arbiter of what is a reasonable theory in investing. In a field, such as medicine, there is some existing body of knowledge that all or nearly all medical researchers agree to and a theory, can be judged as to how it fits into that existing body of knowledge. I doubt there's an academic consensus on much of anything in investing and to the extent there is, I would guess academics treat the market as being at least weak efficient.

Again for example (maybe fourth, fifth time mentioned and not answered?) carry as a way to make money in currency trading is common knowledge. Whether or not it 'should' make money, it often has. It's beyond any reasonable question that it has. Therefore, researchers evaluating carry in currency markets, or evaluating its existence in other markets are IMO obviously not 'data mining' in the sense of throwing up numbers against the wall to see what sticks.

And I think in general your response falls in the same category as afan's. "Here's a general issue. I'm not going to get into how it applies to 'style premia' in a fund like QSPIX, because I'm naturally skeptical of any such instrument or approach, and thus have no interest in gaining the familiarity with it I'd need to make a specific argument".

needtosave
Posts: 16
Joined: Wed Aug 19, 2015 10:31 pm

Re: QSPIX - thoughts on interesting fund

Post by needtosave » Sat Aug 22, 2015 3:28 pm

Johno wrote:Again for example (maybe fourth, fifth time mentioned and not answered?) carry as a way to make money in currency trading is common knowledge.
The fact that something is "common knowledge" that it has made money in the past is an indication that it was tested years ago. That might tell you whether a factor can persist after it is well known or will be arbitraged away, but that has nothing to do with whether or not it is the result of data mining. No competent statistician would consider a result as more statistically significant just because someone else ran the same data tests years ago. Otherwise, you would end up in a bizarre world where factors that were lucky enough to be tested first are considered not the result of data mining while other factors tested later are considered data mining, just because of the order in which they were tested.

Whether or not it 'should' make money, it often has. It's beyond any reasonable question that it has.
The fact that something has often made money in the past addresses the issue of real world implementation costs, but that has absolutely nothing to do with whether something is or is not data mining
Therefore, researchers evaluating carry in currency markets, or evaluating its existence in other markets are IMO obviously not 'data mining' in the sense of throwing up numbers against the wall to see what sticks.


You seem to have an extremely narrow definition of data mining and then arguing from that definition, it doesn't apply to AQR factors. Perhaps alternative terminology(false positive rates in the presence of collective multiple hypothesis testing) would be a more accurate, if more cumbersome description, of these issues, but I think they do apply, regardless of what you want to label them.

Nothing more I can add.

Johno
Posts: 1883
Joined: Sat May 24, 2014 4:14 pm

Re: QSPIX - thoughts on interesting fund

Post by Johno » Sat Aug 22, 2015 3:46 pm

needtosave wrote:
Johno wrote:Again for example (maybe fourth, fifth time mentioned and not answered?) carry as a way to make money in currency trading is common knowledge.
The fact that something is "common knowledge" that it has made money in the past is an indication that it was tested years ago. That might tell you whether a factor can persist after it is well known or will be arbitraged away, but that has nothing to do with whether or not it is the result of data mining. No competent statistician would consider a result as more statistically significant just because someone else ran the same data tests years ago. Otherwise, you would end up in a bizarre world where factors that were lucky enough to be tested first are considered not the result of data mining while other factors tested later are considered data mining, just because of the order in which they were tested.

Whether or not it 'should' make money, it often has. It's beyond any reasonable question that it has.
The fact that something has often made money in the past addresses the issue of real world implementation costs, but that has absolutely nothing to do with whether something is or is not data mining
Therefore, researchers evaluating carry in currency markets, or evaluating its existence in other markets are IMO obviously not 'data mining' in the sense of throwing up numbers against the wall to see what sticks.


You seem to have an extremely narrow definition of data mining and then arguing from that definition, it doesn't apply to AQR factors. Perhaps alternative terminology(false positive rates in the presence of collective multiple hypothesis testing) would be a more accurate, if more cumbersome description, of these issues, but I think they do apply.

Nothing more I can add.

You could add *anything at all* actually specific to the case at hand. :D

And even in terms of the vague generalities you can't seem to proceed beyond, it would seem you've in contrast to my 'too narrow' definition, defined just about all statistical study of past financial results 'data mining'. It's 'data mining' if researchers start with a theory based on commonly known results (eg currency carry trade makes money, not something 'tested years ago' but ongoing common knowledge of currency trading) seeking to evaluate its existence in a broader set of markets, not a 'later test on the same data'. It's 'data mining' if researchers search for factors found in previous academic studies (eg value in US stocks) and evaluate in a broader set of markets (many foreign stock markets, bonds, currencies, short term rates) using, perforce, different data. It's hard to see in context of the actual history of study of such factors what wouldn't be 'data mining'.

But, I don't think it's possible to have a rational discussion of this when one side just sticks to generalities, 'there can be problems with statistical analysis, but I've no familiarity with this particular case so I can't get into any specifics of why I think it undermines the results here it just does' which though I don't suppose you agree, pretty much summarizes your argument and that of afan previously IMO.

lack_ey
Posts: 5556
Joined: Wed Nov 19, 2014 11:55 pm

Re: QSPIX - thoughts on interesting fund

Post by lack_ey » Sat Aug 22, 2015 4:01 pm

So focusing on what we actually care about (future results), if we want to stick with generalities because nobody is talking specifics for whatever reasons...

What do you think is the probability of
(a) a strategy discovered today based on historical data producing positive returns in the future
relative to
(b) a strategy discovered decades ago based on historical data then, that has been implemented successfully with live money for years and tests today with similar "significance" levels as in (a), producing positive returns in the future?

User avatar
nedsaid
Posts: 8289
Joined: Fri Nov 23, 2012 12:33 pm

Re: QSPIX - thoughts on interesting fund

Post by nedsaid » Sat Aug 22, 2015 4:07 pm

larryswedroe wrote:Few things
First we all have to "predict" the future to some degree. But IMO it isn't about luck as Simons is quoted as saying.
It's about putting the odds of success on your side by using the evidence we have. This is no different with the equity risk premium.
I'm confident Simon doesn't believe it is luck, just a cute saying. He is in fact betting on relationships with strong evidence to persist. He knows it isn't certain, so to that degree it's luck (but it's really about putting odds on your side by using the best available information, making informed decisions)

Having said that my crystal ball is cloudy and I don't know which factors will provide positive returns so I want to diversify across them as much as possible while considering the costs of implementation. And TSM portfolios have exposure to only beta factor as a systematic factor.

Best wishes
larry


Larry this is a very good post. I have often said that successful investing is putting the odds in your favor the best you can. You have added to this thought by talking about using the evidence that we have. There are relationships between asset classes and sub-classes that tend to persist over time. The equity premium is one of them. I have also liked your thinking of diversification as across factors rather than just across asset classes. I have done this for years without realizing it.
A fool and his money are good for business.

needtosave
Posts: 16
Joined: Wed Aug 19, 2015 10:31 pm

Re: QSPIX - thoughts on interesting fund

Post by needtosave » Sat Aug 22, 2015 5:20 pm

Johno wrote:You could add *anything at all* actually specific to the case at hand. :D
AQR could use a Bonferroni correction in all their papers, although I thought I mentioned that in an earlier post.
And even in terms of the vague generalities you can't seem to proceed beyond, it would seem you've in contrast to my 'too narrow' definition, defined just about all statistical study of past financial results 'data mining'. It's 'data mining' if researchers start with a theory based on commonly known results (eg currency carry trade makes money, not something 'tested years ago' but ongoing common knowledge of currency trading) seeking to evaluate its existence in a broader set of markets, not a 'later test on the same data'. It's 'data mining' if researchers search for factors found in previous academic studies (eg value in US stocks) and evaluate in a broader set of markets (many foreign stock markets, bonds, currencies, short term rates) using, perforce, different data. It's hard to see in context of the actual history of study of such factors what wouldn't be 'data mining'.
Any paper that includes Bonferroni, Holm, or even BHY as a correction, which AQR does not do in their literature.
But, I don't think it's possible to have a rational discussion of this when one side just sticks to generalities, 'there can be problems with statistical analysis, but I've no familiarity with this particular case so I can't get into any specifics of why I think it undermines the results here it just does' which though I don't suppose you agree, pretty much summarizes your argument and that of afan previously IMO.

lack_ey
Posts: 5556
Joined: Wed Nov 19, 2014 11:55 pm

Re: QSPIX - thoughts on interesting fund

Post by lack_ey » Sat Aug 22, 2015 5:36 pm

Okay, so specifically in this paper:
http://pages.stern.nyu.edu/~lpederse/pa ... ywhere.pdf

Which corrections would you apply where? How many hypotheses do you consider to have been tested?

More importantly, in the end, what are the broad statistical concerns? What probability would you assign to each value+asset class combination being a fluke of the data (with the actual value being zero), and each momentum+asset class combination? How about a composite, as in the style premia fund? Then, as a separate question from the matter of data mining, what is your idea of the probability of the composite performance being positive in the future?

User avatar
grap0013
Posts: 1861
Joined: Thu Mar 18, 2010 1:24 pm

Re: QSPIX - thoughts on interesting fund

Post by grap0013 » Sun Aug 23, 2015 7:05 am

Opening up this thread from a link in my email.

Right above this email is an email from Papa John's Pizza stating: "A great deal, ANY way you slice it!"

How fitting it's next to the QSPIX thread. ROTFLMAO. If Papa John says it's a great deal, I'm buying!
There are no guarantees, only probabilities.

afan
Posts: 2860
Joined: Sun Jul 25, 2010 4:01 pm

Re: QSPIX - thoughts on interesting fund

Post by afan » Sun Aug 23, 2015 8:55 am

Why doesn't the academic finance literature talk, much, about multiple hypothesis testing?

My theory: the authors get much more credit for identifying effects, developing data sets, and characterizing relationships than they do for checking for false positives.

The intended audience is other academic finance types who are thoroughly familiar with the issues, so they are not mislead by papers that ignore them.

The audience understands how hard it is to generate oos data. They realize that testing a sample from, say, 1960 to 2000 does not render pre-1960 data oos if one knew the phenomenon was present in the earlier time period. The only oos are those that are truly unknown- entirely different and unstudied markets, or future returns.

Some of the discussion in this thread has conflated the behavior of a given factor with the effectiveness of a particular mix of factors. Two different issues. The latter always raises multiple testing concerns. The former might or might not depending on how the factor has been tested and supported in the past.

Some of the discussion has also ignored concerns about the definition of performance. If the goal is higher expected risk adjusted returns, then one must define risk. Some of the discussion has assumed risk is captured by variance alone. For other comments it is not clear what is meant by risk. I think it is clear that real world investors care about at least skewness and perhaps higher moments. Analysis of the effect on an overall portfolio should take this into consideration.

The concern about running a fund with a mix of strategies, a mix of factor bets, is that there may be covariances not accounted for when the factors are treated as independent. Even if one models some level of interaction, the risk remains. That was what happened to LTCM. They thought they had modeled the covariances, only to discover that in an extreme event, their bets were much more highly correlated than they imagined. Given the Nobel - level IQ's involved, these events say that getting the matrix right is hard. Really hard. Probably too hard for yhe smartest people in the world. Unless the collective knowledge is much greater now than then, a difficult argument to make, it implies that one should be skeptical of claims to have solved this problem.

Remember that once in a century events happen all the time. The odds on any one of them are low, but around the world, with ?millions? of possible events, the chances of many of them occurring approaches certainty.

This stuff is fascinating to discuss, but I would hesitate to invest my money in what is, for the academics, a game of thought experiments, with nothing more at stake than getting a paper in JOF or a less prestigious place.

I am not trying to be a naysayer, or to come off as arrogant, although I worry that is how my posts may sound. These issues do not go away because one decides to ignore them.


As for asking to point out, in painful detail, how all this applies to each particular factor. "It applies to all factors, all the time, everywhere." It is like asking which specific pages in a statistics book to read to address it. Start at page one. Read to the end. Then get another book.

Or don't. It is a free country.
We don't know how to beat the market on a risk-adjusted basis, and we don't know anyone that does know either | --Swedroe | We assume that markets are efficient, that prices are right | --Fama

Post Reply