Scientific Paper about a Long-Short approach

[frommi]

I found this last week and think this is interesting.

 

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2312432

 

Happy reading!

[constructive]

I found this last week and think this is interesting.

 

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2312432

 

Happy reading!

 

The evidence is solid but the magnitude seems a little disappointing.

 

~6.7% annual, and the borrow rate on the short side would eat up a significant amount of that, especially in a normal interest rate environment. People generally want to short "junk" - I think short interest will have a higher correlation with this factor than other risk factors.

 

Of course if you change the strategy to long bias instead of market neutral, returns would be better.

 

Thanks for the link.

[frommi]

Mr. Asness has set up a hedge fund in 2013 using this approach that is 200% levered on each side and has delivered around 13.5% in the last 11 months, thats is nearly exactly what the paper suggests. Personally i would use this only when the broad market is running hot like currently and not levered up so just 100% long and beta adjusted 100% short. Imagine you could dodge most of the crashes by using this approach and even make money when you are not right on the timing. And then when the market has tanked you slowly get rid of the hedges.

 

Sounds like a perfect plan. :D

[randomep]

I found this last week and think this is interesting.

 

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2312432

 

Happy reading!

 

The evidence is solid but the magnitude seems a little disappointing.

 

~6.7% annual, and the borrow rate on the short side would eat up a significant amount of that,

 

What is the borrow rate? There is no borrowing cost to shorting, I thought.

 

[constructive]

I found this last week and think this is interesting.

 

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2312432

 

Happy reading!

 

The evidence is solid but the magnitude seems a little disappointing.

 

~6.7% annual, and the borrow rate on the short side would eat up a significant amount of that,

 

What is the borrow rate? There is no borrowing cost to shorting, I thought.

 

http://www.cornerofberkshireandfairfax.ca/forum/general-discussion/borrow-rates-lookup-please/msg163287/#msg163287

[constructive]

Mr. Asness has set up a hedge fund in 2013 using this approach that is 200% levered on each side and has delivered around 13.5% in the last 11 months, thats is nearly exactly what the paper suggests. Personally i would use this only when the broad market is running hot like currently and not levered up so just 100% long and beta adjusted 100% short. Imagine you could dodge most of the crashes by using this approach and even make money when you are not right on the timing. And then when the market has tanked you slowly get rid of the hedges.

 

Sounds like a perfect plan. :D

 

400% gross leverage seems unsafe. Hopefully they have a good plan for unwinding in a hurry.

 

I believe this paper assumes 50 / 50, not 100 / 100.  So that might mean the fund is losing half of the hypothetical return to transaction / margin costs.

[frommi]

Oh, i thought it was just 100%/100% in the paper, as i said i wouldn`t invest in their fund and wouldn`t lever it up that much.

You can reduce the borrowing cost if you use the Russel 2000 future, it costs around 1.5-2% per year and has enough junk in it. (at the moment)

[tng]

Mr. Asness has set up a hedge fund in 2013 using this approach that is 200% levered on each side and has delivered around 13.5% in the last 11 months, thats is nearly exactly what the paper suggests. Personally i would use this only when the broad market is running hot like currently and not levered up so just 100% long and beta adjusted 100% short. Imagine you could dodge most of the crashes by using this approach and even make money when you are not right on the timing. And then when the market has tanked you slowly get rid of the hedges.

 

Sounds like a perfect plan. :D

 

That timeframe has also been really good for long/short quant funds. I work for a quant fund and we have a small long/short fund that we are running to build up a track record. Made about 11% last year and about 6% so far this year (both after trading costs), going close to 100/100 without additional leverage. We were also pretty conservative and had really tight boundaries on the business, industry, and sector deltas, so we barely deviated from the index weightings in all of these categories and held 100+ positions on each side. We don't expect to keep posting numbers like this, a more realistic return based on our backtests would be in the 6-7% range or so, meaning going forward it may be lower as more and more people use quantitative techniques. Also, since we don't sell this product to clients yet (just managing internal money to build a track record), we are not charging performance/management fees which could bring these returns down significantly.

 

The attractiveness of such a product will likely depend on interest rates. I would much rather invest in a market neutral product instead of getting a few basis points on short term treasuries. In theory, you can leverage up these products or do something like buy S&P500 ETFs and use that as collateral to fund a market neutral strategy or buy an S&P500 future and hold a market neutral product. But there is tail risk to things blowing up in an unexpected way (which is why we are super conservative with our business/industry/sector deltas as well as our large number of positions for diversification). If the market were to become far more efficient than it is right now, there would be a lot of trading costs associated with market neutral strategies, and that drag can bring returns down very quickly.

[frommi]

Thanks for your feedback tng!

 

I know a lot of professionals scoff at saisonal statistics, but have you considered hedging only in the month from may to september, where the market has statistically a negative return? I would really like to see or hear from a backtest that does this.

You are probably right, in a high interest environment cash might have a better return than going long-short.

[tng]

Thanks for your feedback tng!

 

I know a lot of professionals scoff at saisonal statistics, but have you considered hedging only in the month from may to september, where the market has statistically a negative return? I would really like to see or hear from a backtest that does this.

You are probably right, in a high interest environment cash might have a better return than going long-short.

 

If you already know some statistical characteristic about the data, it is pretty much cheating to exploit that in a backtest because it is basically having access to future data. If there is no economic reason for the seasonal trend to exist, it is hard to have faith that it would continue to exist into the future. Financial datasets are actually very small, there is a severe lack of data before the 80s that doesn't have significant selection, survivorship, or other biases, so you expect some patterns to emerge in the data. If you keep flipping a series of 12 coins, and you do it 20 times, you might notice that the 5th coin came out heads 15/20 times. A lot of times when people see seasonal trends, it is just that.

 

In fact, half of the published research I read are nonsense because they fall into the trap of looking at biased data. The most common one is doing stuff that exclude financials or exclude a certain type of stock, or they have variables that heavily bias stocks into or out of one sector. You can imagine how avoiding tech during the tech bubble or financials during the financial crisis can easily add significant annualized returns over a 20-30 year backtest period. For example, if you are doing research in 2010 or 2011, and your dataset ends in 2009, anything that gets you outside of financials would look amazing. A lot of papers get published like this.

 

In terms of a short term hedge, you can't do an active hedge (as in picking out 100 statistically overvalued, low quality stocks and shorting them) because the trading cost of putting it on and taking it off 3 months later would likely eat up any gains, so you are limited to passive hedges like shorting an index during those months. But if you actually believe that it is possible to use statistics to generate outperformance, you can find stronger relationships that have economic rationale than just seasonal ones (ex: a weighting scheme based on P/E ratio beats out market cap weighting). Why hedge out a few months when there is evidence that you can go long/short the entire year or go long on cheap, high quality stocks only? You will probably make more money that way than looking at the seasonal trends (and that is assuming that the trends continue). I've done a few experiments with seasonal trends, but it never made sense to rack up the trading costs to modify our optimized portfolios for a few months and then undoing it.

[frommi]

The seasonal patterns exist for a very long time and are recorded for nearly all stock markets around the world. (Bouman and Jacobsen 2002, 36/37 markets)

 

http://seekingalpha.com/article/1183461-seasonal-patterns-in-stock-markets-319-years-of-evidence

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1697861

 

If we only consider the full sample (306 years) we find four monthly anomalies robust across different estimation methods (significantly positive returns for January and December and significantly below average returns for July and October) and also a positive Halloween or Sell in May effect.

 

There are lots of possible explanations like tax effects, holidays, hurricane season or that people look at the forward-pe after Halloween.