I’ve been talking to managers at long-short funds and they get this from brokers. Some have gotten it in the past from Interactive Brokers. Can’t really run short sim’s without this, unless they hold 100 positions or something.
I agree with you. When, based on a short sim, I screen for the recommended stocks at IB, I see borrowing rates like 46% (ONVO in this case). At the moment I am unable to run backtests that take this into account, let alone check if a stock is shortable.
Maybe other P123 users have creative ideas on how to circumvent this problem, but the only thing that I come up with so far is shorting an ETF like IWM and even then I think this ETF will have high lending costs when we are in the midst of a crisis. Another good example are the lending rates of current treasury ETF’s. Shorting EDV will cost you 7% on interest at the moment. Is it worth it? I do not know and I am unable to backtest these scenario’s.
Anyway, this is an ‘old’ topic, see this thread posted by Marco. Maybe time to reopen the discussion?
Agree it’s an old topic. Just feel that stock shorting systems are close to worthless without it. If you want to short 100 stocks, there are ways to assume a blended borrowing cost. Short of that, I don’t think so.
Yes this is VERY important. Surprised people do not ask for this more aggressively. This is crucial for selling short. IB does provide this, but its not like you could run a simulation.
: (
Problem is that this data varies from broker to broker…
Data explorers. Good stuff…used it in the past and you can build high short alpha systems on the various items that are very elite. But not cheap.
I’d like to implement a long-short strategy, and while the book simulations are not perfect (since they don’t allow for leverage), I’m comfortable with it for now directionally.
But, it would be great to be able to exclude stocks with very high cost to borrow, or no availability. This information is essential for simulations and portfolio “buy” rules. While the stocks with the highest cost to borrow are not necessarily the companies with the most downside potential, we wouldn’t know for sure unless it is tested.
I didn’t see this on the “to do” list that Marco posted at the beginning of the year, but is this an official feature request?
The only workaround I can see would be to download the information in Excel each week from Interactive Brokers and manually reject any stock that a portfolio recommends with either no availability or very high cost to borrow. I noted in one of my 20-stock short sims that only 3 stocks are too costly to short right now. This approach may be better than nothing, but I’d still be flying somewhat blind, and would much prefer to access the information through a portfolio “buy” rule.
Does anyone here run a true long-short stock portfolio? Are there any other approaches that others use?
The borrow can be an issue and having stock called away is another externality. ETFs and futures are my preference on the short side now as I like to have 2x the number of shorts as longs. I will take a look at the IB list and see how it might be incorporated.
Hold long is your average holding period for the short stocks?
It’s a big project and I’m not sure it’s worth it.
Assuming we have all historical borrow rates for every name point-in-time… Does a short simulation of 10 positions tell you anything at all? So many things can go wrong! One single difference in the simulated past would produce an entirely different result. The dreaded word “curve-fit” comes to mind.
The example above is true for long sims too, but with shorting it’s much easier to get in trouble. To short “safely” (so you can sleep at night) you need to short a lot of names, say 100, which is not possible for regular investors. And if you can do 100 names (that are liquid and not micro-caps), you should be ok with estimating a borrow rate and use P123.
In other words, adding this hard to get (and expensive) data, hoping that it will allow small investors to create a robust 10-20 short positions portfolio, that is a dream.
The historical borrow rates are the crux. Unless someone has been caching them for years then there is no point-in-time backtests. The issue is important as some academics have found that shorting is a lot less profitable than their academic studies would indicate when borrow costs are incorporated, (much less ability to borrow in size).
For example, Viggle Inc, VGGL. IB indicates borrow cost of 116.85 % p.a. and only 200 shares available. Short interest ramped up last Oct and the Short Interest ratio has doubled in past month as volume has waned. You may want to short this but the cost of borrow and ability to borrow your size would constrain your results.
The historical borrow cost could be estimated using observed borrow rates and building a model with market cap, price, short interest days, short interest / float, beta, price change, analyst sentiment, dividend, percent insider holding, percent institutional holding, trends in volume, etc. You would have to do the same for the shares available to borrow. Both issues are quite specific to a company’s balance sheet, capital structure and investor base so there would be relatively low confidence in the estimate.
The lack of robust historical data for cost and ability to short shares makes any short system likely to have over estimated returns. This is worst in small/microcaps, but is even an issue, though much smaller, is large cap systems, especially when there have been shorting restrictions as with the financial stocks in 2008/2009.
On higher volume stocks and ETFs with liquid options, one could also consider a synthetic short using options (short call + long put). In those cases one wouldn’t need to borrow stock, and the margin requirements should be more favorable.
http://www.theoptionsguide.com/synthetic-short-stock.aspx
–Tom C
Thanks for the color, Marco. As you mentioned, a long simulation of 10 holdings can have similar issues, but I take your point that shorting is a little bit trickier. Though I’m not sure if 100 names is required to be diversified on the short side. Perhaps 40 or 50 would be sufficient, but I’m still testing it out in simulations.
It looks like short borrowing cost data is neither robust, nor cheap, so I’ll rely on diversification with more holdings as you suggested. I’m hoping that a small allocation to a short strategy (e.g., 10-20% of a portfolio), and perhaps not using leverage at first, may help mitigate risk if I decide to try it out.