Index | Recent Threads | Who's Online | Search

Posts: 23    Pages: 3    1 2 3 Next
Last Post
New Thread
This topic has been viewed 649 times and has 22 replies
yuvaltaylor
Do quant-based value models work out-of-sample?

If you want to have faith that quant-based value strategies still work, check out LSV Asset Management. This Chicago-based firm is run by Josef Lakonishok, a Ph.D. who had specialized in behavioral finance. In 1994, he realized that the key to investing success was quantitative value investing. He now has $112 billion under management, and his small cap and microcap value funds, all of which are quant-based, have been steadily beating the market, even over the last few years. The record is impressive: their US Small-Cap Value, US Microcap Value, Non-US Small-Cap Value, Emerging Markets Small-Cap Value, Japan Small-Cap Value, Europe Small-Cap Value, and Canada Small-Cap Value are ALL beating their benchmarks. Which is pretty remarkable for out-of-sample performance.

Here is their investment philosophy and practice, from https://www.lsvasset.com/about-lsv/


The fundamental premise on which our investment philosophy is based is that superior long-term results can be achieved by systematically exploiting the judgmental biases and behavioral weaknesses that influence the decisions of many investors. These include: the tendency to extrapolate the past too far into the future, to wrongly equate a good company with a good investment irrespective of price, to ignore statistical evidence and to develop a “mindset” about a company.

LSV uses a quantitative investment model to choose out-of-favor (undervalued) stocks in the marketplace at the time of purchase that have potential for near-term appreciation. LSV believes that these out-of-favor securities will produce superior future returns if their future growth exceeds the market’s low expectations.

LSV portfolios typically have a deep value orientation relative to the indices. Market timing is not part of the process and portfolios are fully invested (cash levels usually below 2%).

LSV uses quantitative techniques to select individual securities in what would be considered a bottom-up approach. The investment process is similar for each of our investment strategies but is segmented into different capitalization ranges or regions.

A proprietary investment model is used to rank a universe of stocks based on a variety of factors we believe to be predictive of future stock returns. The process is continuously refined and enhanced by our investment team although the basic philosophy has never changed – a combination of value and momentum factors. We then overlay strict risk controls that limit the over- or under-exposure of the portfolio to industry and sector concentrations. We also limit exposures in individual securities to ensure the portfolios are broadly diversified, further controlling risk.

The competitive strength of this strategy is that it avoids introducing the process to any judgmental biases and behavioral weaknesses that often influence investment decisions.

Portfolio turnover is approximately 30% for each strategy.


Also check out the bios of their investment team here: https://www.lsvasset.com/investment-team/. One of them (Han Qu) does nothing but backtesting and simulations.

The methods that Portfolio123 enable and encourage actually work. Portfolio123 was founded with a very similar philosophy to Lakonishak's: that judgmental biases and behavioral weaknesses often influence investment decisions for the worse, and that quant-based fundamentals investing based on a deep understanding of accounting measures together with rigorous and proper backtesting and simulations can lead an investor to systematically beat the market while minimizing risk.

Yuval Taylor
Product Manager, Portfolio123
invest(igations)
Any opinions or recommendations in this message are not opinions or recommendations of Portfolio123 Securities LLC.

Jan 15, 2020 9:50:05 AM       
philjoe
Re: Do quant-based value models work out-of-sample?

Looking at the performance, all their US strategies except micro (large, mid, small) have under-performed their benchmarks in the past 3 to 5 years... same for Japan and Canada large cap strategies. Europe strategies haven't been around long enough.

Jan 15, 2020 4:03:28 PM       
Edit 1 times, last edit by philjoe at Jan 15, 2020 4:04:06 PM
Jrinne
Re: Do quant-based value models work out-of-sample?

I was interested in what, exactly, their "quantitative methods" were.

I found this presentation by the Partner and Director or Research. Also a Professor of Finance. I have attached 2 slide images from his talk.

Looks like he is constructing a portfolio along the "efficient frontier" of modern portfolio theory in the second slide.

Obviously this just determines the weight of each position once they use their methods to determine the "return each stock is expected to earn" and each stock’s "variances and covariances."

I cannot find much else other than their use of behavioral finance.

-Jim

Attachment Author Information.png (116078 bytes) (Download count: 116)


Attachment Use of MT.png (165290 bytes) (Download count: 119)


From time to time you will encounter Luddites, who are beyond redemption.
--de Prado, Marcos López on the topic of machine learning for financial applications

Jan 15, 2020 5:50:20 PM       
Edit 1 times, last edit by Jrinne at Jan 15, 2020 6:01:18 PM
ustonapc
Re: Do quant-based value models work out-of-sample?

Looking at the performance, all their US strategies except micro (large, mid, small) have under-performed their benchmarks in the past 3 to 5 years... same for Japan and Canada large cap strategies. Europe strategies haven't been around long enough.


Philip,

You are right to point out that LSV has been underperformaning their benchmarks in the past 3-5 years.

Symmetric.io is a hedge fund tracking firm used by Bloomberg and they calculated the returns of hedge funds by comparing the performance of equities in the funds with that of a sector exchange-traded fund. The firm grades the stock-picking ability of nearly 1,000 hedge funds in its database with their proprietary indicator, Stockalpha. Below are the comparisons of LSV Asset Management vs Tiger Global/Lone Pine Capital.

Regards
James

Attachment LSV Asset Management.png (202930 bytes) (Download count: 98)


Attachment Tiger Global.png (203961 bytes) (Download count: 97)


Attachment Lone Pine.png (205204 bytes) (Download count: 96)


Jan 15, 2020 11:05:38 PM       
Schm1347
Re: Do quant-based value models work out-of-sample?

James,

That’s pretty cool information on hedge fund alpha. Also, Jim mentioned that LSV uses MPT based on variance of stock pricing and covariance between stock prices. If nobody was doing this it would be a great approach. However because the strategy of diversifying through historical correlation is commonly used in portfolio construction it has the unintended consequence of leading to higher security correlation. Perhaps that and indexing have led to higher correlation in stocks over the last few years. It is probably like all commonly used investment strategies that get broadly used that it is losing both its alpha generating ability and actually accumulating a big fat risk tail. When everybody is constructing MPT portfolios based on historical anti-correlation that collapses now everyone is exposed to the same risk which multiplies that risk to an avalanche scenario. By no means am I a finance PhD, but the herding but I do know that herding in markets is dangerous. MPT also concerns me because in its purest form you are selecting companies not on fundamentals but on historical price action. Correlations can quickly change despite history leaving a portfolio fully exposed.

Just some of my musing...

Jeff

Jan 16, 2020 12:08:45 AM       
Edit 1 times, last edit by Schm1347 at Jan 16, 2020 12:11:59 AM
RTNL
Re: Do quant-based value models work out-of-sample?

Thanks James - do you know who the top stockalpha funds are?

Jan 16, 2020 7:35:49 AM       
ustonapc
Re: Do quant-based value models work out-of-sample?

RT,

Here are are a few small hedge funds which has high scores. You can replicate their portfolios with their quarterly 13F disclosures.

Regards
James

Attachment SCGE Management.png (205527 bytes) (Download count: 77)


Attachment Whale Rock Capital.png (202352 bytes) (Download count: 76)


Attachment RA capital management.png (203754 bytes) (Download count: 77)


Jan 16, 2020 8:12:03 AM       
yuvaltaylor
Re: Do quant-based value models work out-of-sample?

It's a mistake to judge performance based on the last three years. It's simply far too short a sample, and everybody knows that 2019 was deeply aberrant. I'm talking about long-term success here.

Yuval Taylor
Product Manager, Portfolio123
invest(igations)
Any opinions or recommendations in this message are not opinions or recommendations of Portfolio123 Securities LLC.

Jan 16, 2020 8:50:31 AM       
ustonapc
Re: Do quant-based value models work out-of-sample?

Yuval,

We are all talking about alpha capture these days.

If a hedge fund manager underperformed the index consistently for the last 3 years (ie. -ve alpha), you will notice a lot of fund redepmtions. For instance, AQR did poorly in the last 2 years and their AUM dropped by more than 30% due to redemptions.

Regards
James

Jan 16, 2020 9:05:27 AM       
philjoe
Re: Do quant-based value models work out-of-sample?

Agreed, the psychological cost of underperforming the market for 3 -5 years outweighs the benefit of maybe marginally outperforming in the long-term. I would just invest in the market instead of LSV.

Jan 16, 2020 9:49:16 AM       
Posts: 23    Pages: 3    1 2 3 Next
 Last Post