Small Cap Value heads Rejoice !!!

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
Locked
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

klaus14 wrote: Mon Nov 30, 2020 2:38 am
Uncorrelated wrote: Mon Nov 30, 2020 2:31 am It is not a good diversifier in the same way roulette isn't: it's uncorrelated to the stock market, but you're not being compensated for that additional risk.

In practice all the return of VPU is explained by stock market and bond market factors. What you're left with is a highly concentrated, poorly diversified sector bet. Looking at correlations just doesn't work in a multi-factor world. The correlation between VPU and VTI is only low because VTI is over 50% long-term bonds. The correlation between tesla and the stock market is also low. Are you going to over-weight tesla as well just because it has a low correlation? Of course not, low correlation is not a sufficient condition for being a good investment.

See this research paper for more details: https://papers.ssrn.com/sol3/papers.cfm ... id=2965146 (the title says REIT's, but they also run regressions on utilities and the exact same mathematical methods are suitable for all sectors).
- Why not? It performed better than SP500 with lower volatility?

- Citation needed. Yes, regression shows term loading, but the highest R i was able to get was 50% in PV. This means its returns are not well explained by factors.
The paper you link has R=37% for UTIL. How can you say "all the return of VPU is explained by stock market and bond market factors" ?
Tesla also performed better than the S&P 500. It doesn't mean you should buy tesla. You should buy tesla if it increases diversification (obviously not true that buying tesla in excess of market weight improves diversification) or when tesla contains statistically significant positive risk factors.

Here's a citation:
Image

Here is one from an actual academic paper:
Image

And here with bond market factors:
Image

Low R^2 does not imply high diversification. It's the exact opposite: low R^2 usually means poor diversification, this can be easily observed by factor regressing individual stocks. You want a portfolio's with high R^2.
klaus14
Posts: 953
Joined: Sun Nov 25, 2018 6:43 pm

Re: 1,000 Small-Cap Value Posts

Post by klaus14 »

Uncorrelated wrote: Mon Nov 30, 2020 2:55 am
klaus14 wrote: Mon Nov 30, 2020 2:38 am
Uncorrelated wrote: Mon Nov 30, 2020 2:31 am It is not a good diversifier in the same way roulette isn't: it's uncorrelated to the stock market, but you're not being compensated for that additional risk.

In practice all the return of VPU is explained by stock market and bond market factors. What you're left with is a highly concentrated, poorly diversified sector bet. Looking at correlations just doesn't work in a multi-factor world. The correlation between VPU and VTI is only low because VTI is over 50% long-term bonds. The correlation between tesla and the stock market is also low. Are you going to over-weight tesla as well just because it has a low correlation? Of course not, low correlation is not a sufficient condition for being a good investment.

See this research paper for more details: https://papers.ssrn.com/sol3/papers.cfm ... id=2965146 (the title says REIT's, but they also run regressions on utilities and the exact same mathematical methods are suitable for all sectors).
- Why not? It performed better than SP500 with lower volatility?

- Citation needed. Yes, regression shows term loading, but the highest R i was able to get was 50% in PV. This means its returns are not well explained by factors.
The paper you link has R=37% for UTIL. How can you say "all the return of VPU is explained by stock market and bond market factors" ?
Tesla also performed better than the S&P 500. It doesn't mean you should buy tesla. You should buy tesla if it increases diversification (obviously not true that buying tesla in excess of market weight improves diversification) or when tesla contains statistically significant positive risk factors.

Here's a citation:
Image

Here is one from an actual academic paper:
Image

And here with bond market factors:
Image

Low R^2 does not imply high diversification. It's the exact opposite: low R^2 usually means poor diversification, this can be easily observed by factor regressing individual stocks. You want a portfolio's with high R^2.
Low R^2 means, regression doesn't fit well. From the paper you linked:
"We do note that the REITs regression
shows the third lowest R-squared (51 percent) of the industries considered, UTIL and ENRG
being the two lower, which indicates a relative deficiency in the ability for the factor model
to explain the variance in REIT returns."

You can also see with your eyes if you click on the "Rolling Returns" tab in PV you cited. Compare it with how well SLYV fits. (Link)

And can you explain why you circled t-stat of Alpha there?
My investment algorithm: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=351899&p=6112869#p6112869
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

klaus14 wrote: Mon Nov 30, 2020 3:03 am
And can you explain why you circled t-stat of Alpha there?
Investment decisions should be based on expected return and risk. A low R^2 does not necessarily indicate a deficiency in forecasting expected returns, that would be indicated by statistically significant alpha. But there is none.

A low R^2 is not a reason to invest in a particular asset class. Independent positive return factors or improved diversification are reasons to invest in specific asset classes. As far as I'm aware of there is no evidence that any sector provides this.
klaus14
Posts: 953
Joined: Sun Nov 25, 2018 6:43 pm

Re: 1,000 Small-Cap Value Posts

Post by klaus14 »

Uncorrelated wrote: Mon Nov 30, 2020 3:21 am
klaus14 wrote: Mon Nov 30, 2020 3:03 am
And can you explain why you circled t-stat of Alpha there?
Investment decisions should be based on expected return and risk. A low R^2 does not necessarily indicate a deficiency in forecasting expected returns, that would be indicated by statistically significant alpha. But there is none.

A low R^2 is not a reason to invest in a particular asset class. Independent positive return factors or improved diversification are reasons to invest in specific asset classes. As far as I'm aware of there is no evidence that any sector provides this.
Alpha there is meaningless since whole regression fit is meaningless due to low R^2.

Low R^2 means there is something more to this stock collection than 3 (or 5) factors, i guess you can call it "Independent positive return".
You can see in PV, adding VPU increases returns and lowers volatility of a US stock portfolio. I guess you can call it "improved diversification"

It's not TSLA, VPU includes 65 stocks, that is enough to erase idiosyncratic risk of any single company.

----

I found an older utilities fund to see more history: FKUTX
Results are same.
My investment algorithm: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=351899&p=6112869#p6112869
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

klaus14 wrote: Mon Nov 30, 2020 3:34 am
Uncorrelated wrote: Mon Nov 30, 2020 3:21 am
klaus14 wrote: Mon Nov 30, 2020 3:03 am
And can you explain why you circled t-stat of Alpha there?
Investment decisions should be based on expected return and risk. A low R^2 does not necessarily indicate a deficiency in forecasting expected returns, that would be indicated by statistically significant alpha. But there is none.

A low R^2 is not a reason to invest in a particular asset class. Independent positive return factors or improved diversification are reasons to invest in specific asset classes. As far as I'm aware of there is no evidence that any sector provides this.
Alpha there is meaningless since whole regression fit is meaningless due to low R^2.
So you believe the alpha on HmL is meaningless, since it has a low R^2 on CAPM?

Why do you think HmL is an accepted factor in academia, but VPU is not?

Low R^2 means there is something more to this stock collection than 3 (or 5) factors, i guess you can call it "Independent positive return".
You can see in PV, adding VPU increases returns and lowers volatility of a US stock portfolio. I guess you can call it "improved diversification"
My definition of diversification is reducing idiodyncratic risk. There is no evidence that VPU contains systematic risk other than the usual factors, therefore it cannot be claimed that over-weighting VPU improves diversification.
It's not TSLA, VPU includes 65 stocks, that is enough to erase idiosyncratic risk of any single company.
Not even 1000 stocks is sufficient. See How diversification impacts the reliability of outcomes.

----

I found an older utilities fund to see more history: FKUTX
Results are same.
A factor regression shows that KFUTX is approximately 50% stocks and 50% long term bonds. Simply replacing KFUTX by long term bonds provides better results than KFUTX. Simply replacing KFUTX by walmart also provides better results. In the paper on REIT's, they use generic optimization methods to show that it is always possible to construct a portfolio with only small cap value and bonds that is better than a portfolio with REIT's. I suspect the same holds for portfolio's that over-weight utilities.

A backtest is not an appropriate tool for portfolio selection.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: 1,000 Small-Cap Value Posts

Post by vineviz »

klaus14 wrote: Mon Nov 30, 2020 3:34 am Low R^2 means there is something more to this stock collection than 3 (or 5) factors, i guess you can call it "Independent positive return".
That's right.

The sectors within the S&P 500 are not all equal in the degree to which the conventional risk factors explain their returns, and utility stocks are among the least well-explained using those factors. The low correlation of utility stocks with other stocks, their variance, and their low weight in most market cap indexes combine to make a utility sector fund (like VPU or XLU) a very reasonable and low-cost source of additional diversification for investor portfolios.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
klaus14
Posts: 953
Joined: Sun Nov 25, 2018 6:43 pm

Re: 1,000 Small-Cap Value Posts

Post by klaus14 »

vineviz wrote: Mon Nov 30, 2020 11:32 am
klaus14 wrote: Mon Nov 30, 2020 3:34 am Low R^2 means there is something more to this stock collection than 3 (or 5) factors, i guess you can call it "Independent positive return".
That's right.

The sectors within the S&P 500 are not all equal in the degree to which the conventional risk factors explain their returns, and utility stocks are among the least well-explained using those factors. The low correlation of utility stocks with other stocks, their variance, and their low weight in most market cap indexes combine to make a utility sector fund (like VPU or XLU) a very reasonable and low-cost source of additional diversification for investor portfolios.
Would you say the same for residential real estate (REZ) ?
My investment algorithm: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=351899&p=6112869#p6112869
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: 1,000 Small-Cap Value Posts

Post by vineviz »

klaus14 wrote: Mon Nov 30, 2020 2:03 pm
vineviz wrote: Mon Nov 30, 2020 11:32 am
klaus14 wrote: Mon Nov 30, 2020 3:34 am Low R^2 means there is something more to this stock collection than 3 (or 5) factors, i guess you can call it "Independent positive return".
That's right.

The sectors within the S&P 500 are not all equal in the degree to which the conventional risk factors explain their returns, and utility stocks are among the least well-explained using those factors. The low correlation of utility stocks with other stocks, their variance, and their low weight in most market cap indexes combine to make a utility sector fund (like VPU or XLU) a very reasonable and low-cost source of additional diversification for investor portfolios.
Would you say the same for residential real estate (REZ) ?
That fund has a shorter history, and the historical evidence on REITS in general isn’t as powerful. But in principle, yes: I’d say a dedicated allocation to REZ would generate some incremental diversification benefit.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
klaus14
Posts: 953
Joined: Sun Nov 25, 2018 6:43 pm

Re: 1,000 Small-Cap Value Posts

Post by klaus14 »

vineviz wrote: Mon Nov 30, 2020 2:44 pm
klaus14 wrote: Mon Nov 30, 2020 2:03 pm
vineviz wrote: Mon Nov 30, 2020 11:32 am
klaus14 wrote: Mon Nov 30, 2020 3:34 am Low R^2 means there is something more to this stock collection than 3 (or 5) factors, i guess you can call it "Independent positive return".
That's right.

The sectors within the S&P 500 are not all equal in the degree to which the conventional risk factors explain their returns, and utility stocks are among the least well-explained using those factors. The low correlation of utility stocks with other stocks, their variance, and their low weight in most market cap indexes combine to make a utility sector fund (like VPU or XLU) a very reasonable and low-cost source of additional diversification for investor portfolios.
Would you say the same for residential real estate (REZ) ?
That fund has a shorter history, and the historical evidence on REITS in general isn’t as powerful. But in principle, yes: I’d say a dedicated allocation to REZ would generate some incremental diversification benefit.
But i think expense ratio is not worth it unlike VPU.
My investment algorithm: https://www.bogleheads.org/forum/viewtopic.php?f=10&t=351899&p=6112869#p6112869
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

vineviz wrote: Mon Nov 30, 2020 11:32 am
klaus14 wrote: Mon Nov 30, 2020 3:34 am Low R^2 means there is something more to this stock collection than 3 (or 5) factors, i guess you can call it "Independent positive return".
That's right.

The sectors within the S&P 500 are not all equal in the degree to which the conventional risk factors explain their returns, and utility stocks are among the least well-explained using those factors. The low correlation of utility stocks with other stocks, their variance, and their low weight in most market cap indexes combine to make a utility sector fund (like VPU or XLU) a very reasonable and low-cost source of additional diversification for investor portfolios.
According to that definition, a random collection of 20-100 small cap stocks, or even a casino gambling round, is a better diversifier than VPU. This definition is clearly not useful in any way.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: 1,000 Small-Cap Value Posts

Post by vineviz »

klaus14 wrote: Mon Nov 30, 2020 2:46 pm But i think expense ratio is not worth it unlike VPU.
I agree. I don't think the benefits would be worth the hassle and expense.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: 1,000 Small-Cap Value Posts

Post by vineviz »

Uncorrelated wrote: Mon Nov 30, 2020 3:20 pm According to that definition, a random collection of 20-100 small cap stocks, or even a casino gambling round, is a better diversifier than VPU.
It's highly doubtful that a random collection of 20+ small cap stocks would be less correlated with a market cap weighted index of US stocks than VPU is. Such a large sample is, in fact, very likely to behave much like the small cap index from which the sample was drawn and therefore be about as powerful a diversifier as that index.

And unless this "random collection of 20-100 small cap stocks" happened by chance to be highly correlated with VPU, there's no particular reason the portfolio couldn't contain contain both VPU and the random collection.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
Pawpatrol
Posts: 198
Joined: Sun Sep 24, 2017 5:21 pm

Re: Small Cap Value heads Rejoice !!!

Post by Pawpatrol »

Well that was quick- i think today was the swing low (sp600 value). If not today it will be in the 130 area(viov)Get ready for next leg up and hopefully finally take out that 52 week high
james22
Posts: 1992
Joined: Tue Aug 21, 2007 2:22 pm

Re: 1,000 Small-Cap Value Posts

Post by james22 »

There's good reason Utilities are less correlated with the market. They are different than other businesses: the services they provide are essential, they are near (if not actual) monopolies, often their profits are regulated.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: 1,000 Small-Cap Value Posts

Post by willthrill81 »

Uncorrelated wrote: Mon Nov 30, 2020 4:24 am Not even 1000 stocks is sufficient.
I've heard this before, and it just doesn't pass the sniff test. Regardless of what someone claims the statistics involved to say, just look at how closely correlated the S&P 500 has been with the TSM. There hasn't been a meaningful difference in their long-term returns over multiple decades. Heck, even the DJIA has pretty closed matched the S&P 500 and done so with only 30 stocks.
Last edited by willthrill81 on Tue Dec 01, 2020 11:49 am, edited 1 time in total.
The Sensible Steward
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

willthrill81 wrote: Tue Dec 01, 2020 10:42 am
Uncorrelated wrote: Mon Nov 30, 2020 4:24 am Not even 1000 stocks is sufficient.
I've heard this before, and it just doesn't pass the sniff test. Regardless of what someone claims the statistics involved to say, just look at how closely correlated the S&P 500 has been with the TSM. There hasn't been a meaningful difference in their long-term returns over multiple decades. Heck, even the DJIA
I understand that sentiment. The problem is that you think in terms of 1st and 2nd moment (expected return and variance), and I think in terms of expected utility (all moments).

We can clearly observe the error by refering to DFA's How Diversification Impacts Investment Outcomes. In this study, they subsample portfolio's from their large-cap-value ETF with the same expected return as the original portfolio. You might be put-off by their use of the value portfolio, but that's not integral to their analysis.

Here, we can clearly see that 50 stocks is "enough" to reduce the standard deviation to acceptable levels.
Image

But that's only the standard deviation. Individual stocks are not normally distributed, but have significant left skew. This can clearly be observed if we plot the 25th percentile of portfolio returns:
Image

Here we can see that a portfolio of 50 value stocks has 25th percentile returns of -2.19% below the benchmark over a time horizon of 1 year. For a portfolio of 1000 stocks, the 25th percentile return is -0.18%.

Here is the same information, but plotted as probability that the value index (constructed with N stocks) will outperform the broad index:
Image


From this data, it is abundantly clear that 2000 stocks is a little bit better than 1000, and that 1000 stocks is much better than 50. But if all you're doing is toying around with correlations, mean return and standard deviation, you'll never see this.

This is also the exact reason I recommend against highly concentrated small cap funds.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: 1,000 Small-Cap Value Posts

Post by willthrill81 »

Uncorrelated wrote: Tue Dec 01, 2020 11:46 am
willthrill81 wrote: Tue Dec 01, 2020 10:42 am
Uncorrelated wrote: Mon Nov 30, 2020 4:24 am Not even 1000 stocks is sufficient.
I've heard this before, and it just doesn't pass the sniff test. Regardless of what someone claims the statistics involved to say, just look at how closely correlated the S&P 500 has been with the TSM. There hasn't been a meaningful difference in their long-term returns over multiple decades. Heck, even the DJIA
I understand that sentiment. The problem is that you think in terms of 1st and 2nd moment (expected return and variance), and I think in terms of expected utility (all moments).

We can clearly observe the error by refering to DFA's How Diversification Impacts Investment Outcomes. In this study, they subsample portfolio's from their large-cap-value ETF with the same expected return as the original portfolio. You might be put-off by their use of the value portfolio, but that's not integral to their analysis.

Here, we can clearly see that 50 stocks is "enough" to reduce the standard deviation to acceptable levels.
Image

But that's only the standard deviation. Individual stocks are not normally distributed, but have significant left skew. This can clearly be observed if we plot the 25th percentile of portfolio returns:
Image

Here we can see that a portfolio of 50 value stocks has 25th percentile returns of -2.19% below the benchmark over a time horizon of 1 year. For a portfolio of 1000 stocks, the 25th percentile return is -0.18%.

Here is the same information, but plotted as probability that the value index (constructed with N stocks) will outperform the broad index:
Image


From this data, it is abundantly clear that 2000 stocks is a little bit better than 1000, and that 1000 stocks is much better than 50. But if all you're doing is toying around with correlations, mean return and standard deviation, you'll never see this.

This is also the exact reason I recommend against highly concentrated small cap funds.
I care about returns and their volatility. You care about other aspects too. That's fine, but it doesn't mean that 'Not even 1000 stocks is sufficient'. Different investors have different goals.
The Sensible Steward
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: 1,000 Small-Cap Value Posts

Post by vineviz »

willthrill81 wrote: Tue Dec 01, 2020 11:51 am I care about returns and their volatility. You care about other aspects too. That's fine, but it doesn't mean that 'Not even 1000 stocks is sufficient'. Different investors have different goals.
The "you need 1,000 stocks" argument is based on people regurgitating data they don't understand. The majority of tracking error reduction happens well before you get up to 100 stocks in a portfolio. Sure there are incremental benefits to increasing the number of stocks, but those benefits are economically and statistically trivial.

Just take a look at the performance of the iShares Core S&P 500 ETF (IVV) compared to two concentrated analogs: iShares S&P 100 ETF (OEF) and Invesco S&P 500 Top 50 ETF (XLG).

Image

Not much daylight between them and there is even less if you mix them down into a portfolio with bonds, international stocks, etc. There is no risk measure (max drawdown, VaR, standard deviation, skewness, kurtosis, downside capture, etc.) where the XLG or OEF behave significantly worse than IVV. Granted these aren't random subsamples, but no one was seriously arguing for randomly subsampling 50 stock portfolios to begin with.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

willthrill81 wrote: Tue Dec 01, 2020 11:51 am
Uncorrelated wrote: Tue Dec 01, 2020 11:46 am
I care about returns and their volatility. You care about other aspects too. That's fine, but it doesn't mean that 'Not even 1000 stocks is sufficient'. Different investors have different goals.
That is a classic misunderstanding. You don't care about returns and volatility and neither do I. We care about hitting our financial goals. Return and volatility are imperfect proxies for that. If we construct a portfolio with the same expected return and stddev as the stock market but a significant skew, it is not a suitable investment for most individuals even though it looks indistinguishable when viewed through the lens of return and volatility.

And that's exactly what happens with under-diversified portfolio's, significant skew.

vineviz wrote: Tue Dec 01, 2020 12:45 pm
willthrill81 wrote: Tue Dec 01, 2020 11:51 am I care about returns and their volatility. You care about other aspects too. That's fine, but it doesn't mean that 'Not even 1000 stocks is sufficient'. Different investors have different goals.
The "you need 1,000 stocks" argument is based on people regurgitating data they don't understand. The majority of tracking error reduction happens well before you get up to 100 stocks in a portfolio. Sure there are incremental benefits to increasing the number of stocks, but those benefits are economically and statistically trivial.

Just take a look at the performance of the iShares Core S&P 500 ETF (IVV) compared to two concentrated analogs: iShares S&P 100 ETF (OEF) and Invesco S&P 500 Top 50 ETF (XLG).

Image

Not much daylight between them and there is even less if you mix them down into a portfolio with bonds, international stocks, etc. There is no risk measure (max drawdown, VaR, standard deviation, skewness, kurtosis, downside capture, etc.) where the XLG or OEF behave significantly worse than IVV. Granted these aren't random subsamples, but no one was seriously arguing for randomly subsampling 50 stock portfolios to begin with.
Describing an improvement of the 25th percentile outcome over a 1 year time horizon by 2.13% (exhibit 3) at no additional cost as "statistically and economically trivial" sounds completely delusional.


You're right that the discussion wasn't about subsampling 50 random stocks. The discussion was about subsampling 65 stocks from the same sector. As far as I'm concerned that is significantly worse than subsampling 50 random stocks, which are at least expected to have some amount of spread between the different sectors.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: 1,000 Small-Cap Value Posts

Post by vineviz »

Uncorrelated wrote: Wed Dec 02, 2020 6:07 am Describing an improvement of the 25th percentile outcome over a 1 year time horizon by 2.13% (exhibit 3) at no additional cost as "statistically and economically trivial" sounds completely delusional.
If the the total stock market has a 1-year loss of 5% (which is about a 25th percentile outcome) and a 50-stock sample of the stock market has a loss of 3% or 7%, then only the person who thinks that is a deal-breaker is delusional in my book. To each their own, I suppose.
Uncorrelated wrote: Wed Dec 02, 2020 6:07 amYou're right that the discussion wasn't about subsampling 50 random stocks. The discussion was about subsampling 65 stocks from the same sector. As far as I'm concerned that is significantly worse than subsampling 50 random stocks, which are at least expected to have some amount of spread between the different sectors.
I assumed that people would know what "subsampling" means. If there are 65 stocks with a characteristic you desire and you buy all 65 of those stocks (as a fund like VPU does), that isn't subsampling. That's just indexing.

Subsampling would be taking an index like the S&P 600 Value index, which contains approximately 450 stocks and represents one possible sampling of small-cap value stocks, and choosing a smaller number (e.g. 50) of those.

A random subsample of 50 stocks from the S&P 600 Value index will almost certainly track the actual index pretty well. If ANY measure of risk (e.g. variance, skewness, kurtosis, max drawdown) for this random subsample differed from the sample in a statistically significant way I'd be shocked. And I'd want an actual test of statistical significance to verify it, not just a picture from a white paper.

But in the real world there'd be no need to choose a random subsample. There are much more powerful ways to choose the subsample, many of which are trivially easy to execute, in such a way that 40 or 50 stocks would be MORE than enough to replicate the performance of the primary sample.

Just look at chart I posted earlier of the real-world performance of the 50-stock and 100-stock versions of the S&P indexes for an example: no significant differences in either skewness or kurtosis despite having 1/10th or 1/5th the number of stocks.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

vineviz wrote: Wed Dec 02, 2020 8:46 am
Uncorrelated wrote: Wed Dec 02, 2020 6:07 am Describing an improvement of the 25th percentile outcome over a 1 year time horizon by 2.13% (exhibit 3) at no additional cost as "statistically and economically trivial" sounds completely delusional.
If the the total stock market has a 1-year loss of 5% (which is about a 25th percentile outcome) and a 50-stock sample of the stock market has a loss of 3% or 7%, then only the person who thinks that is a deal-breaker is delusional in my book. To each their own, I suppose.
Not a loss of "3% or 7%". A loss of 7%.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: 1,000 Small-Cap Value Posts

Post by willthrill81 »

Uncorrelated wrote: Wed Dec 02, 2020 6:07 am
willthrill81 wrote: Tue Dec 01, 2020 11:51 am
Uncorrelated wrote: Tue Dec 01, 2020 11:46 am
I care about returns and their volatility. You care about other aspects too. That's fine, but it doesn't mean that 'Not even 1000 stocks is sufficient'. Different investors have different goals.
That is a classic misunderstanding. You don't care about returns and volatility and neither do I. We care about hitting our financial goals. Return and volatility are imperfect proxies for that. If we construct a portfolio with the same expected return and stddev as the stock market but a significant skew, it is not a suitable investment for most individuals even though it looks indistinguishable when viewed through the lens of return and volatility.

And that's exactly what happens with under-diversified portfolio's, significant skew.
I'll definitely grant you that I'm misunderstood. :P

From the charts you've posted, the lion's share of the improvement occurs when moving from 50 stocks to 200. Everything else is marginal, at best. VBR, one of Vanguard's SCV ETFs, has 909 stocks. That's more than enough to reduce idiosyncratic risk for me. Further, the spread of possible outcomes shrinks over time. I'm not investing in stocks for a year or even three or five, but perhaps you are, and if so, then I can see why you would want to own absolutely everything.
The Sensible Steward
User avatar
Anon9001
Posts: 1884
Joined: Fri Dec 20, 2019 8:28 am
Location: بھارت

Re: Small Cap Value heads Rejoice !!!

Post by Anon9001 »

Interesting that his chart his showing significant improvements in 25th percentile for 50 stock portfolio the longer you hold it in terms of excess return compared to the other number of stock portfolios. This might be why there is no difference between the 50-100 and 500 stock S&P portfolio that vineviz showcased as he used long term returns from 2005 to 2020.
Land/Real Estate:89.4% (Land/RE is Inheritance which will be recieved in 10-20 years) Equities:7.6% Fixed Income:1.7% Gold:0.8% Cryptocurrency:0.5%
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: 1,000 Small-Cap Value Posts

Post by vineviz »

Uncorrelated wrote: Wed Dec 02, 2020 9:47 am Not a loss of "3% or 7%". A loss of 7%.
As I said earlier, it's important to understand something before you repeat it. You're making an assumption that the data don't support.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
langlands
Posts: 1093
Joined: Wed Apr 03, 2019 10:05 pm

Re: 1,000 Small-Cap Value Posts

Post by langlands »

Uncorrelated wrote: Wed Dec 02, 2020 9:47 am
vineviz wrote: Wed Dec 02, 2020 8:46 am
Uncorrelated wrote: Wed Dec 02, 2020 6:07 am Describing an improvement of the 25th percentile outcome over a 1 year time horizon by 2.13% (exhibit 3) at no additional cost as "statistically and economically trivial" sounds completely delusional.
If the the total stock market has a 1-year loss of 5% (which is about a 25th percentile outcome) and a 50-stock sample of the stock market has a loss of 3% or 7%, then only the person who thinks that is a deal-breaker is delusional in my book. To each their own, I suppose.
Not a loss of "3% or 7%". A loss of 7%.
Why does the dimensional paper only show how much the undiversified portfolio underperforms the full portfolio at the 25% percentile? Why don't they show the full comparison across all percentiles (as a competent researcher would)? Is it possibly because the undiversified portfolio actually outperforms at the higher percentiles (which we know must be the case since the expected value of a subsampled portfolio must be the same as the whole portfolio)? Perhaps this crossover point occurs as soon as the 80-85%. They should have just plotted the full probability distributions in a large clear graph instead of giving us only the part of the elephant they want us to see.

I'll note that the word "skew" literally doesn't appear once in the whole paper, so if that was supposed to be the take home message, the authors did a spectacularly poor job of presenting the phenomenon in a clear and complete way.
Last edited by langlands on Wed Dec 02, 2020 3:31 pm, edited 2 times in total.
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

willthrill81 wrote: Wed Dec 02, 2020 9:54 am
Uncorrelated wrote: Wed Dec 02, 2020 6:07 am
willthrill81 wrote: Tue Dec 01, 2020 11:51 am
Uncorrelated wrote: Tue Dec 01, 2020 11:46 am
I care about returns and their volatility. You care about other aspects too. That's fine, but it doesn't mean that 'Not even 1000 stocks is sufficient'. Different investors have different goals.
That is a classic misunderstanding. You don't care about returns and volatility and neither do I. We care about hitting our financial goals. Return and volatility are imperfect proxies for that. If we construct a portfolio with the same expected return and stddev as the stock market but a significant skew, it is not a suitable investment for most individuals even though it looks indistinguishable when viewed through the lens of return and volatility.

And that's exactly what happens with under-diversified portfolio's, significant skew.
I'll definitely grant you that I'm misunderstood. :P

From the charts you've posted, the lion's share of the improvement occurs when moving from 50 stocks to 200. Everything else is marginal, at best. VBR, one of Vanguard's SCV ETFs, has 909 stocks. That's more than enough to reduce idiosyncratic risk for me. Further, the spread of possible outcomes shrinks over time. I'm not investing in stocks for a year or even three or five, but perhaps you are, and if so, then I can see why you would want to own absolutely everything.
Investment decision is almost never related to investment horizon, and that's also true in this case.

I believe the graph shows the annualized numbers. Although the annualized tracking error decreases as the time horizon decreases, the dispersion in ending balance increases as the time horizon increases.

Factor investing is one area where reducing diversification is actually rational: there is a trade-off between the factor tilt (expected return) and diversification. There are a few papers that claim some factors might be a compensation for skew risk, but I'm not sure how credible the evidence is.
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: 1,000 Small-Cap Value Posts

Post by willthrill81 »

Uncorrelated wrote: Wed Dec 02, 2020 3:20 pm
willthrill81 wrote: Wed Dec 02, 2020 9:54 am
Uncorrelated wrote: Wed Dec 02, 2020 6:07 am
willthrill81 wrote: Tue Dec 01, 2020 11:51 am
Uncorrelated wrote: Tue Dec 01, 2020 11:46 am
I care about returns and their volatility. You care about other aspects too. That's fine, but it doesn't mean that 'Not even 1000 stocks is sufficient'. Different investors have different goals.
That is a classic misunderstanding. You don't care about returns and volatility and neither do I. We care about hitting our financial goals. Return and volatility are imperfect proxies for that. If we construct a portfolio with the same expected return and stddev as the stock market but a significant skew, it is not a suitable investment for most individuals even though it looks indistinguishable when viewed through the lens of return and volatility.

And that's exactly what happens with under-diversified portfolio's, significant skew.
I'll definitely grant you that I'm misunderstood. :P

From the charts you've posted, the lion's share of the improvement occurs when moving from 50 stocks to 200. Everything else is marginal, at best. VBR, one of Vanguard's SCV ETFs, has 909 stocks. That's more than enough to reduce idiosyncratic risk for me. Further, the spread of possible outcomes shrinks over time. I'm not investing in stocks for a year or even three or five, but perhaps you are, and if so, then I can see why you would want to own absolutely everything.
Investment decision is almost never related to investment horizon, and that's also true in this case.
Perhaps it shouldn't, but virtually every investor I know does this.
Uncorrelated wrote: Wed Dec 02, 2020 3:20 pm I believe the graph shows the annualized numbers. Although the annualized tracking error decreases as the time horizon decreases, the dispersion in ending balance increases as the time horizon increases.
Except that I've already shown that in at least one instance, it didn't at all (i.e. S&P 500 vs. TSM).
Uncorrelated wrote: Wed Dec 02, 2020 3:20 pm Factor investing is one area where reducing diversification is actually rational: there is a trade-off between the factor tilt (expected return) and diversification. There are a few papers that claim some factors might be a compensation for skew risk, but I'm not sure how credible the evidence is.
I'll grant you that factor investing is rational.
The Sensible Steward
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

langlands wrote: Wed Dec 02, 2020 3:15 pm
Uncorrelated wrote: Wed Dec 02, 2020 9:47 am
vineviz wrote: Wed Dec 02, 2020 8:46 am
Uncorrelated wrote: Wed Dec 02, 2020 6:07 am Describing an improvement of the 25th percentile outcome over a 1 year time horizon by 2.13% (exhibit 3) at no additional cost as "statistically and economically trivial" sounds completely delusional.
If the the total stock market has a 1-year loss of 5% (which is about a 25th percentile outcome) and a 50-stock sample of the stock market has a loss of 3% or 7%, then only the person who thinks that is a deal-breaker is delusional in my book. To each their own, I suppose.
Not a loss of "3% or 7%". A loss of 7%.
Why does the dimensional paper only show how much the undiversified portfolio underperforms the full portfolio at the 25% percentile? Why don't they show the full comparison across all percentiles (as a competent researcher would)? Is it possibly because the undiversified portfolio actually outperforms at the higher percentiles (which we know must be the case since the expected value of a subsampled portfolio must be the same as the whole portfolio)? Perhaps this crossover point occurs as soon as the 80-85%. They should have just plotted the full probability distributions in a large clear graph instead of giving us only the part of the elephant they want us to see.

I'll note that the word "skew" literally doesn't appear once in the whole paper, so if that was supposed to be the take home message, the authors did a spectacularly poor job of presenting the phenomenon in a clear and complete way.
I agree the authors did a poor job. This isn't the full paper, I can't find the full paper. If you have a better paper I would like to read it, but this is the best I can find on the subject of diversification.

Although the paper contains a narrow view, I don't see any reasons to doubt their numbers.
langlands
Posts: 1093
Joined: Wed Apr 03, 2019 10:05 pm

Re: 1,000 Small-Cap Value Posts

Post by langlands »

Uncorrelated wrote: Wed Dec 02, 2020 3:27 pm
langlands wrote: Wed Dec 02, 2020 3:15 pm
Uncorrelated wrote: Wed Dec 02, 2020 9:47 am
vineviz wrote: Wed Dec 02, 2020 8:46 am
Uncorrelated wrote: Wed Dec 02, 2020 6:07 am Describing an improvement of the 25th percentile outcome over a 1 year time horizon by 2.13% (exhibit 3) at no additional cost as "statistically and economically trivial" sounds completely delusional.
If the the total stock market has a 1-year loss of 5% (which is about a 25th percentile outcome) and a 50-stock sample of the stock market has a loss of 3% or 7%, then only the person who thinks that is a deal-breaker is delusional in my book. To each their own, I suppose.
Not a loss of "3% or 7%". A loss of 7%.
Why does the dimensional paper only show how much the undiversified portfolio underperforms the full portfolio at the 25% percentile? Why don't they show the full comparison across all percentiles (as a competent researcher would)? Is it possibly because the undiversified portfolio actually outperforms at the higher percentiles (which we know must be the case since the expected value of a subsampled portfolio must be the same as the whole portfolio)? Perhaps this crossover point occurs as soon as the 80-85%. They should have just plotted the full probability distributions in a large clear graph instead of giving us only the part of the elephant they want us to see.

I'll note that the word "skew" literally doesn't appear once in the whole paper, so if that was supposed to be the take home message, the authors did a spectacularly poor job of presenting the phenomenon in a clear and complete way.
I agree the authors did a poor job. This isn't the full paper, I can't find the full paper. If you have a better paper I would like to read it, but this is the best I can find on the subject of diversification.

Although the paper contains a narrow view, I don't see any reasons to doubt their numbers.
I've tried searching, but it seems this research topic isn't worthy of most finance academics. I will note that the S&P 500 contains 500 stocks (approximately) whereas the TSM contains more like 2500. And the performance between them is almost exactly 0. So that is a strong empirical datapoint in my view that shows 500 stocks is definitely enough to be 99.999% diversified. The fact that the Dow (30 stocks) has outperformed the TSM over a 40 year period from 1980-2020 to me also complicates the narrative that a suitably chosen subset of 30 stocks is a disastrous mistake.

I don't doubt their numbers either, but I think I found a problem with their analysis that invalidates the core conclusion. In Exhibit 4, they plot the volatilities of the 50 stock, 100 stock, etc. portfolios and seem to be showing that they are all comparable. The implication being it must be the higher moments (skew) explaining the different distribution in outcomes. But this is very misleading. The plot of the 25% percentile outcome in Exhibit 3 is the excess return over the MSCI index benchmark, not the absolute return (say over risk free rate). If the whole point is measuring this excess return, they also need to plot the volatilities of the difference portfolios e.g. (50 stock portfolio) - (MSCI index portfolio). I think it's obvious these volatilities will show much more variation as you go from 50 to 100 to 1000 etc. than what's shown in Exhibit 4 (in fact, this is largely what the tracking error histograms show). And this means that most of the underperformance of the undiversified portfolio at the 25% percentile excess return over MSCI is explained by the 2nd moment, which in turn implies that in all likelihood the undiversified portfolio at the 75% percentile would outperform the full portfolio.

After writing all that, I really think it's impossible that the researchers didn't plot the 75% percentile results themselves. They clearly didn't like what they saw (that it was largely a reversal of the 25% percentile plots) and decided not to include it.

In short, I largely agree with vineviz that it's "3%" or "7%". There is a small skew effect, but the reality is much more something like "3.5%" or "7%" than a definitive loss of 7%.
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

willthrill81 wrote: Wed Dec 02, 2020 3:24 pm
Uncorrelated wrote: Wed Dec 02, 2020 3:20 pm Investment decision is almost never related to investment horizon, and that's also true in this case.
Perhaps it shouldn't, but virtually every investor I know does this.
Yes, there are many good reasons why investors would do this. But almost none of them are related to investment horizon. Here is a simple paper that explains the fallacy of time diversification. This problem was studied by many of the great minds of economics, Samuelson, Merton, Bodie, ...

Uncorrelated wrote: Wed Dec 02, 2020 3:20 pm I believe the graph shows the annualized numbers. Although the annualized tracking error decreases as the time horizon decreases, the dispersion in ending balance increases as the time horizon increases.
Except that I've already shown that in at least one instance, it didn't at all (i.e. S&P 500 vs. TSM).
TSM and S&P500 contain 90% of the same market capitalization, so it's no surprise that it's difficult to see the difference in an N=1 backtest... That is hardly an argument to avoid diversification.

If we have two ETF's, let's say growth stocks and value stocks. Growth stocks have an expected return of 5% and stddev of 20%. Value stocks have exactly the same risk factors as growth stocks and a tracking error of 1%. It's pretty clear that:

The annualized tracking difference scales by 1/sqrt(n), and therefore decreases every year.
The total tracking difference scales by sqrt(n), and therefore increases over long time horizons.

Although it looks like (in a backtest) that the S&P500 and TSM converge over long time horizons, they are actually expected to diverge. This can be more clearly seen when plotting small cap vs large cap, or large cap vs mid cap.
MotoTrojan
Posts: 11259
Joined: Wed Feb 01, 2017 7:39 pm

Re: Small Cap Value heads Rejoice !!!

Post by MotoTrojan »

kolder wrote: Thu Nov 19, 2020 4:30 pm
MotoTrojan wrote: Tue Sep 15, 2020 4:20 pm
manlymatt83 wrote: Tue Sep 15, 2020 1:25 pm
Massdriver wrote: Tue Sep 15, 2020 1:22 pm
manlymatt83 wrote: Tue Sep 15, 2020 12:33 pm Did anyone else replace their SLYV/DGS holdings with AVUV/AVDV? Just curious. And if so, did you dump your entire SLYV/DGS?
During the sell off, I sold 2/3 of my VSS position for AVDV. I sold VBR and bought SLYV for my SCV domestic (I used to have VBR+IJS for commission free reasons), but wish I would have also TLH my IJS for AVUV.
Interesting that your actions were due to TLH. If you were starting from scratch today, what would you buy between SLYV/IJS, VSS, VBR, DLS, DGS, AVUV, ADVD, and ADEM?
It really did open my eyes up recently when I realized the differences between these are pretty subtle once you adjust for factor exposure. You cannot really answer this question without first deciding how much of a tilt you want. For example, DFA's DFSVX is one of the more potent factor loaders (I would wager AVUV will fall in this realm). Let's say you were comfortable with 75/25 S&P500 (VFINX)/DFSVX; we can then use Portfolio Visualizer's Match Factor Exposure tool to find an equivalent portfolio of VFINX/VBR (VBR is a much less aggressive tilt, but also cheaper fund). We find that to match exposures the tilt increases to a whopping 65/35 VFINX/VBR, a 40% relative increase in small-value tilt. Even more powerful though, you can see the portfolio with VBR actually outperformed the one with DFSVX, in part because VBR's expense ratio is far lower.

https://www.portfoliovisualizer.com/mat ... tion2_1=75

The biggest takeaway here is that you may not actually need these more expensive funds unless you are constrained by available funds (most of funds in 401k in broad index, and only a small amount of IRA space for factor exposure), or want a more aggressive tilt than 100% VBR.
I was thinking about this idea recently and tried to dig a bit deeper myself. If you run the factor exposure match on VISVX, which goes back a bit further than VBR, it is actually the DFA fund that has slightly higher performance, despite the higher ER.

https://www.portfoliovisualizer.com/mat ... tion2_1=75

You get the same result if you do FF5 regression as well.

I wonder what exactly could be causing the differences. My best guess is that DFSVX has more consistent factor exposure on a day-day basis due to daily rebalancing, which might generate alpha that is more sensitive to the actual performance of the small/value factors if you are running a monthly regression. This would imply that during times when factors provide positive returns, you might expect the DFA fund to have more positive alpha than a SCV index fund, and when factor returns are negative, you might expect there to be more negative alpha than a SCV index fund.

This somewhat appears to be a reasonable hypothesis as you can see in the graph that through 2000-2007 (a period of highly positive factor returns) the DFA fund portfolio had slightly higher returns, despite having the same factor exposure. The higher returns for the DFA portfolio are mostly lost during 2017-present, as SCV did poorly.

This also appears to show up in the 60 month rolling regression for DFSVX, where its alpha seems to hit a bottom in 2000 (right as value was turning around) and peaks in 2005, as you might expect. You can also see that alpha has gone deeply negative in the last few years, as SCV has had very large negative returns over this time period.

Image

That being said, VISVX had deeply negative alpha during periods of good performance for SCV, but much higher alpha during periods of poor performance for SCV. It does seem that over the last few years it has had declining alpha though, which does not fit my hypothesis.

Image

(Note the date ranges are different for the above two images so they don't line up vertically)

Any thoughts? I could just be grasping for straws or overlooking something obvious here lol. Of course it's not too much data to go on and alpha can be quite volatile/random it seems.
VBR changed indices in 2013. The CRSP index it now uses obliterated all other SCV options since 2001 inception.
User avatar
vineviz
Posts: 14921
Joined: Tue May 15, 2018 1:55 pm
Location: Baltimore, MD

Re: 1,000 Small-Cap Value Posts

Post by vineviz »

langlands wrote: Wed Dec 02, 2020 3:40 pm After writing all that, I really think it's impossible that the researchers didn't plot the 75% percentile results themselves. They clearly didn't like what they saw (that it was largely a reversal of the 25% percentile plots) and decided not to include it.

In short, I largely agree with vineviz that it's "3%" or "7%". There is a small skew effect, but the reality is much more something like "3.5%" or "7%" than a definitive loss of 7%.
I think it's clear that DFA weren't looking at skewness at all. At first I was puzzled by the claim that portfolios with small numbers of stocks had negative skew, since every test I ran produced the opposite result (small portfolios have more POSITIVE skew, not more NEGATIVE skew).

This is likely an effect of equal-weighting the subsampled portfolios, and is one reason to prefer an equal-weighted portfolio IF (and it is a big if) transaction costs can be well-managed.

I suspect what DFA were examining is actually kurtosis, since low-n portfolios do tend to exhibit higher levels of kurtosis than large-n portfolios. But, again, this effect typically is not very big: for portfolios with n>25, neither skewness nor kurtosis is different from n>1,000 to a statistically significant degree.
"Far more money has been lost by investors preparing for corrections than has been lost in corrections themselves." ~~ Peter Lynch
langlands
Posts: 1093
Joined: Wed Apr 03, 2019 10:05 pm

Re: 1,000 Small-Cap Value Posts

Post by langlands »

vineviz wrote: Wed Dec 02, 2020 5:15 pm
langlands wrote: Wed Dec 02, 2020 3:40 pm After writing all that, I really think it's impossible that the researchers didn't plot the 75% percentile results themselves. They clearly didn't like what they saw (that it was largely a reversal of the 25% percentile plots) and decided not to include it.

In short, I largely agree with vineviz that it's "3%" or "7%". There is a small skew effect, but the reality is much more something like "3.5%" or "7%" than a definitive loss of 7%.
I think it's clear that DFA weren't looking at skewness at all. At first I was puzzled by the claim that portfolios with small numbers of stocks had negative skew, since every test I ran produced the opposite result (small portfolios have more POSITIVE skew, not more NEGATIVE skew).

This is likely an effect of equal-weighting the subsampled portfolios, and is one reason to prefer an equal-weighted portfolio IF (and it is a big if) transaction costs can be well-managed.

I suspect what DFA were examining is actually kurtosis, since low-n portfolios do tend to exhibit higher levels of kurtosis than large-n portfolios. But, again, this effect typically is not very big: for portfolios with n>25, neither skewness nor kurtosis is different from n>1,000 to a statistically significant degree.
Yes, the relative performance of small (number of stock) portfolios to full portfolios tend to have positive skew. Maybe I'm wrong, but I think Uncorrelated knows this. It's this positive skew that causes the median outcome to trail the expected outcome, which is the commonly mentioned "volatility drag." Say you buy a random stock (the most undiversified portfolio). After 10 years, you will trail the market with 90%+ probability, but in the remaining < 10% of the time, you will significantly outperform the market. So in expectation, any single stock has the same return as the market, but with high probability it will underperform. Also, the skewness gets worse as the investment time horizon increases. You can see by comparing the 1 year return in the first article and the 20 year return in the second article.

https://klementoninvesting.substack.com ... ock-market

https://www.stlouistrust.com/why-most-s ... orm-index/

So I do think DFA was implicitly looking at skew, even if they were very unclear about it.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

I know a lot of people are turned off by the 0.63% TER of DGS, WisdomTree Small Cap Dividend Emerging Markets.

I was looking through the annual report. It says it has 3,430,960 from securities lending income. This is 0.21% of total assets. From what I've read this reduces the actual TER, so ~0.41% would be closer to the real costs. Is this correct?

https://connect.rightprospectus.com/wis ... lTAID=True
User avatar
Uncorrelated
Posts: 1237
Joined: Sun Oct 13, 2019 3:16 pm

Re: 1,000 Small-Cap Value Posts

Post by Uncorrelated »

langlands wrote: Wed Dec 02, 2020 5:41 pm
vineviz wrote: Wed Dec 02, 2020 5:15 pm
langlands wrote: Wed Dec 02, 2020 3:40 pm After writing all that, I really think it's impossible that the researchers didn't plot the 75% percentile results themselves. They clearly didn't like what they saw (that it was largely a reversal of the 25% percentile plots) and decided not to include it.

In short, I largely agree with vineviz that it's "3%" or "7%". There is a small skew effect, but the reality is much more something like "3.5%" or "7%" than a definitive loss of 7%.
I think it's clear that DFA weren't looking at skewness at all. At first I was puzzled by the claim that portfolios with small numbers of stocks had negative skew, since every test I ran produced the opposite result (small portfolios have more POSITIVE skew, not more NEGATIVE skew).

This is likely an effect of equal-weighting the subsampled portfolios, and is one reason to prefer an equal-weighted portfolio IF (and it is a big if) transaction costs can be well-managed.

I suspect what DFA were examining is actually kurtosis, since low-n portfolios do tend to exhibit higher levels of kurtosis than large-n portfolios. But, again, this effect typically is not very big: for portfolios with n>25, neither skewness nor kurtosis is different from n>1,000 to a statistically significant degree.
Yes, the relative performance of small (number of stock) portfolios to full portfolios tend to have positive skew. Maybe I'm wrong, but I think Uncorrelated knows this.
I think I meant positive skew, but accidentally typed negative skew.

I'm not great at interpreting skew and kurtosis. My approach is to calculate the expected utility, since that's all the investor should care about. If I see two portfolio's with identical expected return and variance, but one portfolio has a worse 25th percentile outcome, then that's sufficient evidence for me to conclude the expected utility is worse. I would like to know how much worse exactly, but I simply can't find that information in the literature. Meh.
Hallman
Posts: 147
Joined: Fri May 17, 2013 4:25 pm

Re: Small Cap Value heads Rejoice !!!

Post by Hallman »

YRT70 wrote: Thu Dec 03, 2020 10:57 am I know a lot of people are turned off by the 0.63% TER of DGS, WisdomTree Small Cap Dividend Emerging Markets.

I was looking through the annual report. It says it has 3,430,960 from securities lending income. This is 0.21% of total assets. From what I've read this reduces the actual TER, so ~0.41% would be closer to the real costs. Is this correct?

https://connect.rightprospectus.com/wis ... lTAID=True
The securities lending figure sounds correct.

Considering it tracks an index, it seems easier to me to calculate total cost by looking at how well it tracks that. Wisdomtree uses an index net of withholding tax. I calculated it a few years ago and came to a total cost of about 1.7% for a tax sheltered account (IIRC it lagged the index by about 1.1% and the index was net of about 0.6% dividend withholding taxes). I assume it's not much different today.
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

Hallman wrote: Thu Dec 03, 2020 11:18 am
YRT70 wrote: Thu Dec 03, 2020 10:57 am I know a lot of people are turned off by the 0.63% TER of DGS, WisdomTree Small Cap Dividend Emerging Markets.

I was looking through the annual report. It says it has 3,430,960 from securities lending income. This is 0.21% of total assets. From what I've read this reduces the actual TER, so ~0.41% would be closer to the real costs. Is this correct?

https://connect.rightprospectus.com/wis ... lTAID=True
The securities lending figure sounds correct.

Considering it tracks an index, it seems easier to me to calculate total cost by looking at how well it tracks that. Wisdomtree uses an index net of withholding tax. I calculated it a few years ago and came to a total cost of about 1.7% for a tax sheltered account (IIRC it lagged the index by about 1.1% and the index was net of about 0.6% dividend withholding taxes). I assume it's not much different today.
Interesting. Dividend withholding taxes are still 0.6% so you could be right.

Can American investors get that back? I can't (afaik), I'm from The Netherlands.
Nicolas
Posts: 4923
Joined: Wed Aug 22, 2012 7:41 am

Re: Small Cap Value heads Rejoice !!!

Post by Nicolas »

YRT70 wrote: Thu Dec 03, 2020 11:34 am
Hallman wrote: Thu Dec 03, 2020 11:18 am
YRT70 wrote: Thu Dec 03, 2020 10:57 am I know a lot of people are turned off by the 0.63% TER of DGS, WisdomTree Small Cap Dividend Emerging Markets.

I was looking through the annual report. It says it has 3,430,960 from securities lending income. This is 0.21% of total assets. From what I've read this reduces the actual TER, so ~0.41% would be closer to the real costs. Is this correct?

https://connect.rightprospectus.com/wis ... lTAID=True
The securities lending figure sounds correct.

Considering it tracks an index, it seems easier to me to calculate total cost by looking at how well it tracks that. Wisdomtree uses an index net of withholding tax. I calculated it a few years ago and came to a total cost of about 1.7% for a tax sheltered account (IIRC it lagged the index by about 1.1% and the index was net of about 0.6% dividend withholding taxes). I assume it's not much different today.
Interesting. Dividend withholding taxes are still 0.6% so you could be right.

Can American investors get that back? I can't (afaik), I'm from The Netherlands.
Yes, American investors can claim a tax credit for foreign taxes paid, as long as the investment was in a taxable account.
bog007
Posts: 539
Joined: Fri Feb 28, 2020 2:27 am

Re: Small Cap Value heads Rejoice !!!

Post by bog007 »

:sharebeer
Don’t let anyone else ruin your portfolio. It’s your portfolio. Ruin it yourself!!!
Pawpatrol
Posts: 198
Joined: Sun Sep 24, 2017 5:21 pm

Re: Small Cap Value heads Rejoice !!!

Post by Pawpatrol »

Pawpatrol wrote: Mon Nov 23, 2020 8:29 pm
stocknoob4111 wrote: Mon Nov 23, 2020 7:37 pm Small Caps are shining... S&P 600 is within shouting distance of it's 2018 peak and R2000 futures are strongly up again, it's been a long wait (August 2018) but it does look like Small Caps are finally going to regain leadership...
This has been fun to watch. SCV are overbought so tomorrow might be a good time to rebalance. We will get those 2018 highs but we likely need to consolidate some after this great run.
Pawpatrol wrote: Fri Nov 27, 2020 8:05 pm
aristotelian wrote: Tue Nov 24, 2020 9:07 am
Pawpatrol wrote: Mon Nov 23, 2020 8:29 pm
stocknoob4111 wrote: Mon Nov 23, 2020 7:37 pm Small Caps are shining... S&P 600 is within shouting distance of it's 2018 peak and R2000 futures are strongly up again, it's been a long wait (August 2018) but it does look like Small Caps are finally going to regain leadership...
This has been fun to watch. SCV are overbought so tomorrow might be a good time to rebalance. We will get those 2018 highs but we likely need to consolidate some after this great run.
Another big day today. SLYV up 24% in November. What a month!
The swing high was on tuesday but healthy consolidation the last two trading days. It might be healthier for the longrun if we are choppy and range bound through the next month or so but there are still a ton of small cap value bears that have not capitulated. Prior to the 11/9 gap, we would retrace 50% of the this move up but this time is different given the breakout. My guess is we go parabolic upwards in an epic short squeeze in the coming week or two. For those wanting to get out of scv after such a huge run here is a good article about performance of the russell 2000 after 10%+ months.

https://awealthofcommonsense.com/2020/ ... thly-gain/
Pawpatrol wrote: Mon Nov 30, 2020 3:58 pm Well that was quick- i think today was the swing low (sp600 value). If not today it will be in the 130 area(viov)Get ready for next leg up and hopefully finally take out that 52 week high
Feel pretty good, called previous high on sp600, the swing low and now made a new high. (Made a good chunk of change with my play money on it).

I think im done with predictions here as its cloudy. My gut says we go parabolic given the amount of put buying and short interest in the russell 2000 etf leading to more short squeeze. We also should make an all time high like IJR VB VBR and russell 2000. But we could easily retrace some of this since we made a new high today and are overextended.

Im def rejoicing!
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Small Cap Value heads Rejoice !!!

Post by willthrill81 »

Really good day for SCV. VBR was up 2.18%.
The Sensible Steward
stan1
Posts: 14246
Joined: Mon Oct 08, 2007 4:35 pm

Re: Small Cap Value heads Rejoice !!!

Post by stan1 »

willthrill81 wrote: Fri Dec 04, 2020 4:59 pm Really good day for SCV. VBR was up 2.18%.
VIOV 2.73%
corp_sharecropper
Posts: 590
Joined: Thu Nov 07, 2013 1:36 pm

Re: Small Cap Value heads Rejoice !!!

Post by corp_sharecropper »

stan1 wrote: Fri Dec 04, 2020 5:02 pm
willthrill81 wrote: Fri Dec 04, 2020 4:59 pm Really good day for SCV. VBR was up 2.18%.
VIOV 2.73%
AVUV 2.89%
absolute zero
Posts: 1244
Joined: Thu Dec 29, 2016 3:59 pm

Re: Small Cap Value heads Rejoice !!!

Post by absolute zero »

corp_sharecropper wrote: Fri Dec 04, 2020 5:23 pm
stan1 wrote: Fri Dec 04, 2020 5:02 pm
willthrill81 wrote: Fri Dec 04, 2020 4:59 pm Really good day for SCV. VBR was up 2.18%.
VIOV 2.73%
AVUV 2.89%
RZV 3.48%
absolute zero
Posts: 1244
Joined: Thu Dec 29, 2016 3:59 pm

Re: Small Cap Value heads Rejoice !!!

Post by absolute zero »

I went from 0% to 20% VIOV in August in one fell swoop. Turned out good timing. I’m planning to hold steady at 20% (hopefully forever), but this recent SCV outperformance is feeling a little less exciting than I would’ve imagined. Part of me was hoping that SCV stocks would lay in the trenches until I am about 5 years away from retirement. And then overnight they’d shoot to the moon and turn that final 5 years into only 5 months instead. Oh well. Still enjoying the rejoicing that’s happening in the markets and in this thread.
:beer
csmath
Posts: 826
Joined: Sat Oct 13, 2018 11:32 am

Re: Small Cap Value heads Rejoice !!!

Post by csmath »

absolute zero wrote: Fri Dec 04, 2020 5:25 pm
corp_sharecropper wrote: Fri Dec 04, 2020 5:23 pm
stan1 wrote: Fri Dec 04, 2020 5:02 pm
willthrill81 wrote: Fri Dec 04, 2020 4:59 pm Really good day for SCV. VBR was up 2.18%.
VIOV 2.73%
AVUV 2.89%
RZV 3.48%
RESP & ECT 100%?

🥃
YRT70
Posts: 1289
Joined: Sat Apr 27, 2019 8:51 am

Re: Small Cap Value heads Rejoice !!!

Post by YRT70 »

absolute zero wrote: Fri Dec 04, 2020 5:25 pm
corp_sharecropper wrote: Fri Dec 04, 2020 5:23 pm
stan1 wrote: Fri Dec 04, 2020 5:02 pm
willthrill81 wrote: Fri Dec 04, 2020 4:59 pm Really good day for SCV. VBR was up 2.18%.
VIOV 2.73%
AVUV 2.89%
RZV 3.48%
That was funny.

3 month, YTD, 1 year returns (M*).
VBR 22.25 3.89 7.54
VIOV 23.68 1.06 5.06
AVUV 25.48 5.27 10.42
RZV 27.82 -1.78 3.19
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Small Cap Value heads Rejoice !!!

Post by willthrill81 »

YRT70 wrote: Sat Dec 05, 2020 4:12 am
absolute zero wrote: Fri Dec 04, 2020 5:25 pm
corp_sharecropper wrote: Fri Dec 04, 2020 5:23 pm
stan1 wrote: Fri Dec 04, 2020 5:02 pm
willthrill81 wrote: Fri Dec 04, 2020 4:59 pm Really good day for SCV. VBR was up 2.18%.
VIOV 2.73%
AVUV 2.89%
RZV 3.48%
That was funny.

3 month, YTD, 1 year returns (M*).
VBR 22.25 3.89 7.54
VIOV 23.68 1.06 5.06
AVUV 25.48 5.27 10.42
RZV 27.82 -1.78 3.19
For most of the year, VBR has had the lead among those four funds. It's only been about a month since AVUV moved into the lead.

Image
The Sensible Steward
Random Walker
Posts: 5561
Joined: Fri Feb 23, 2007 7:21 pm

Re: Small Cap Value heads Rejoice !!!

Post by Random Walker »

AVUV seems to be heavier on market beta and size than some of its peers, and not quite as heavy on value.
https://www.portfoliovisualizer.com/fac ... sion=false

Dave
User avatar
willthrill81
Posts: 32250
Joined: Thu Jan 26, 2017 2:17 pm
Location: USA
Contact:

Re: Small Cap Value heads Rejoice !!!

Post by willthrill81 »

Random Walker wrote: Sat Dec 05, 2020 9:31 am AVUV seems to be heavier on market beta and size than some of its peers, and not quite as heavy on value.
https://www.portfoliovisualizer.com/fac ... sion=false

Dave
Which do you prefer, VBR or VIOV?
The Sensible Steward
Locked