Historical data on IB short lending fee and share availability

[feldy]

@marco

But what's the main use-case of this data? You want to use it in a seamless, all-in-one long/short portfolio strategy backtest, no? So shouldn't the L/S strategy backtest come first ?

We have started the L/S project. We will add separate buy/sell rules (similar to the screener), and support having two ranking systems. For the borrow rate we were planning to use some widely accepted formula.

That’s great to hear. A few weeks ago, I definitely would have agree that a combined L/S strategy is probably more important, though maybe less so now that simulated books support leverage. This bug with the Manage strategy performance for short positions I think is still present, and is a showstopper before I would reconsider re-enabling shorting.

The use case for adding per-stock IBKR borrow fees is to build more accurate simulations for short strategies. The current system only allows a single carry cost that is fixed across time and across assets. Unless you are restricting your universe to easy to borrow names, your broker may charge you 1% annualized to borrow one name, and 100+% to borrow the next. As-is, these short strategies may be tilting towards these high cost names as they are blind to the true transaction costs involved.

@korr123

@feldy what is the advantage of this set up versus P123 licensing the global borrow data from factset, building a parallel process from IB and then creating a proxy of IB data vs Factset to backfill pre 2015 (when factset borrow data starts)?

I can’t see enough detail on FactSet’s global borrow dataset to tell. Unless if gives you the granularity of borrow costs at per-broker, per-symbol lvel, then it wouldn’t be as useful for the modeling accurate borrow costs as these can vary so widely across executing brokers. Still could be useful for alpha/ranking system use though.