Back to Community
Should leveraged ETF's be banned from the Open?

From what I've read, there are two reasons to avoid leveraged etfs:

  1. Volatility Drift

  2. Fees

I might be wrong, but it seems that volatility drift is only a problem if you invest 100% of your money in leveraged etf's. However, using leveraged etf's does not necessitate fully investing in them. Here's an example:

Let's say you want 100% exposure to SPY. You could:
A: Invest 100% of your money in SPY
B: Invest 33% of your money in UPRO

In each case, volatility drift has the same effect on your returns. The difference between the two is that with case B, you can invest the other 66% of you money in different stocks or etf's. This gives quantopian users the ability to hedge and/or diversify their strategies more than if they could only put their money in SPY.

This is why, as far as I can see, volatility drift is not necessarily a downside of leveraged etf's. Algo writers can control the volatility of their strategies. Banning leveraged etf's just restricts our ability to write good algos for the Managers program.

I'm aware that fees are still a downside, but are they really so bad that they're not worth that extra 66% of your portfolio?

32 responses

If there's a flash crash then holding leveraged ETFs would be very very bad. They should not be allowed.

Hey fellow Jeff,
Are you saying that in a flash crash UPRO might go down more than 3x SPY? Or are you just saying that the super high betas you can get with leveraged etf's are bad?

Because if UPRO will only go down 3x as much as SPY, then all you need to do is have a strategy that only invests a small portion of your money in UPRO. As long as you do that your beta should stay minimal.

Well, to be honest, it's probably just because it's more complicated to calculate the margin requirements for them. If you want 100% exposure to SPY, you invest 100% in SPY and you still have 66% of your Quantopian 'leverage' left. You invest 33% of your money in UPRO and you'd still have 89% of your Quantopian 'leverage' left, but Quantopian themselves may only have 75% of their IB margin left (per http://ibkb.interactivebrokers.com/node/1124), less if they have to abide by Reg T overnight margin. Scale this up and you could see how unless Quantopian overhauled their leverage calculation to properly calculate broker margin requirements, this would lead to contest algos getting a margin call, which would be embarrassing.

So you think Quantopian can't mimic IB's leverage rules?

And even if they can't, aren't they only going to choose low-volatility strategies anyway?

I know a high-volatility strategy could potentially win the Open, but that seems unlikely considering the way the scoring system works.

I personally would go long a leveraged etf to hedge. Being long on a LETF for any substantial period of time is generally a bad idea. Your capital erodes pretty quickly with convexity and fees. I think you would actually be better off only partially hedging with 33% SPY in most cases

So you think Quantopian can't mimic IB's leverage rules?  

Of course they can, they just haven't. But it does skewer your main argument, which is that you want them for buying power, when they don't actually provide any buying power, given that 1/3 position in a 3x leveraged ETF uses almost as much margin as a 100% position in a 1x ETF.

So you're saying IB's leverage policy makes leveraged etf's obsolete?
If you look at this page: https://www.interactivebrokers.com/en/?f=margin&p=stk
you'll see that their margin requirements for leveraged etf's are: Minimum(50% * Leverage Factor, 100%)
That's for Reg T end-of-day.

So even if you put your money in a 3x leveraged etf, your initial margin will only be 100% of the position value.

As you can see, their model still allows for you to increase your leverage by using leveraged etf's.

So is my argument really skewered?

I know that leverage and restrictions for these etf's can get much more complicated, especially with short positions. But isn't it possible that they can be used correctly? We'll never know if there are good ways of using them if trying to use them is discouraged.

There's some confusion about whether Quantopian has to abide by Reg T leverage, but your example points out precisely that using leveraged ETFs in the Open would be counter-productive (or at least nil benefit), since Quantopian allows you 33% margin on everything else (which, I believe, is better than Reg T overnight). 100% of 3x LETF or 300% of 1x ETF. Same thing.

I think you misunderstood IB's policy. It says that for a 3x leveraged etf, you need initial margin of 100% the position value. That doesn't mean you can only put 100% of your money in a 3x etf. It means you can only put 150% of your money in a 3x etf. So 3x etfs can still increase your leverage.

Yes, if you sell it before the end of the day. Overnight margin will be maxed out at 100%, ie no margin at all.

There's still no benefit though, for getting leverage:

3x long LETF intraday with 25% x 3 margin = possible day trading leverage of 400%. 1x long ETF with 25% intraday margin = possible day trading leverage of 400%.

Quantopian's overnight margin of 33% is actually more than we can get as Reg T traders, which is max 50% for 1x products.

Again, the 25% refers to initial margin. IB isn't saying that you can only use 75% percent leverage for a 3x etf. Anyone who has used IB can confirm that they allow you to use at least 100% of your money in a 3x etf. Quantopian doesn't change that.

What?? Of course you can put 100% of your money in a 3x LETF. You have the cash, you buy the ETF. You can't hold more than 100% overnight, because the Reg T overnight margin will cap out 100% on leveraged ETFs. Your intraday maintenance margin is on a 3x is 75%, which means you could potentially have 133% invested in a 3x ETF during the day, as long as you get it down to 100% before 3:50pm.

25% margin means you can invest 400% of your account value, if it were only in that one thing. 50% margin means you can invest 200%. 75% margin means you can invest 133%.

Quantopian's 'leverage' constraint of 3x or 300% means they effectively have a margin requirement of 33%. This margin requirement is more favorable than what you can hold overnight as a Reg T trader with your own account at IB.

I think you are confusing "margin requirement" with "leverage percentage". Anyway, this topic has gone in circles for long enough.

All I'm saying is that leveraged etfs allow you to use more leverage. If you want to post quantopian backtests that show the same leverage whether or not you used leveraged etf's or not then I would understand your point.

Here you go. Benchmark is UPRO, a 3x long LETF, something I think we both agree IB will only let you hold overnight up to 100% of equity, ie: it has a margin requirement of 100%, ie. you cannot margin this.

The 'algo' is a constant 300% leveraged position in SPY, a 1x ETF that IB has an overnight margin requirement of 50% for, ie: in our IB accounts we could hold up to 200% of, but which the Quantopian contest allows you to hold up to 300% of.

Clone Algorithm
5
Loading...
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
def initialize(context):
    set_slippage(slippage.FixedSlippage(spread=0.0))
    set_commission(commission.PerShare(cost=0.0, min_trade_cost=0.0))
    # benchmark is UPRO, a 3x long LETF with an overnight margin requirement of 100%, so
    # we can only buy our full equity of it, no more
    set_benchmark(sid(38533))

def handle_data(context, data):
    # SPY is a 1x ETF with an overnight margin requirement of 50%.  In our Reg T IB accounts,
    # we'd be able to only hold 200% of this overnight, but the Quantopian contest lets us
    # leverage to 300%
    order_target_percent(sid(8554), 3.0)
    record( leverage=context.account.leverage )
    
There was a runtime error.

Right. That's without a leveraged etf. I'm saying that leveraged etf's allow you to use even more leverage. In this backtest it puts 300% in UPRO. You get more volatility but quantopian still considers it to be a leverage of 3, just like the algo above.

Clone Algorithm
8
Loading...
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
def initialize(context):
    set_slippage(slippage.FixedSlippage(spread=0.0))
    set_commission(commission.PerShare(cost=0.0, min_trade_cost=0.0))
    set_benchmark(sid(38533))

def handle_data(context, data):

    order_target_percent(sid(38533), 3)
    record( leverage=context.account.leverage )
    
There was a runtime error.

Yes, my point is you cannot put 300% in UPRO in real life. It's not allowed. It's Quantopian's error that they do not treat it differently in their simulations, which is likely why they have banned them from the contest.

In real life outside of the Quantopian contest, you can gain a little leverage using 100% of a 3x LETF vs being allowed to use 200% of a 1x ETF. In terms of the contest, where you are allowed 3x leverage, there's not much to be gained, leverage-wise, by using 100% of a 3x LETF vs 300% of a 1x ETF. Nowhere are you allowed to have 300% of a 3x LETF.

I think all this confusion is caused by the fact that we've been talking about two different leverage models: quantopian's and interactive brokers. They're very different since IB treats leveraged etf's differently and quantopian treats them the exact same way it treats every other stock and etf.

I'm gonna post two backtests which might explain my argument in a more effective way.

This backtest is a hypothetical strategy that the Quantopian Open currently allows.

It uses SPY to hedge against the market. The leverage stays around 2. You could live trade this algo with IB and everything would be fine.

This example is already really volatile. But a lot of properly hedged strategies aren't. We've all seen the low-beta strategies submitted to the Open that get annual returns less than 10%. Wouldn't Quantopian want to increase the leverage applied to those strategies?

Clone Algorithm
8
Loading...
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
def initialize(context):
    pass
def handle_data(context, data):

    order_target_percent(sid(43202), 1)
    order_target_percent(sid(8554), -1)
    record( leverage=context.account.leverage )
    
There was a runtime error.

This backtest uses UPRO. Even though it only puts 50% of your money into it, it hedges against the market just as well as SPY.

The important thing is that strategy is leveraged more. There a lot of strategies that shouldn't be leveraged more, but as long as there are strategies which should be leveraged more, these etf's have a purpose.

Clone Algorithm
8
Loading...
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
def initialize(context):
    pass
def handle_data(context, data):

    order_target_percent(sid(43202), 1.5)
    order_target_percent(sid(38533), -.5)
    record( leverage=context.account.leverage )
    
There was a runtime error.

Except the second strategy is impossible in real life! In a Reg T account, long 150% of a stock will use up 75% of your margin equity, short 50% of a LETF will use up 50%, and that's 125% > 100% of equity. The ratios which WOULD work are 120% long -> 60% margin and 40% short -> 40%. And that's because it's a Reg T account.

In Quantopian's case, though they disallow Leveraged ETFs, they do allow 300% leverage on vanilla stocks, so you could do 150% long and 150% short, which is more overall leverage than your Reg T 120%/40%.

Quantopian without LETFS > Reg T with LETFs > Reg T without LETFS. If what you want is Quantopian With LETFs, then we need to figure out how they themselves will get access to that margin, given that even IB's portfolio margin doesn't allow leveraged ETFs...

EDIT: I just realized that for a 3x, my ratios above would not be the correct hedge ratios. Sorry about that. Need to solve the equation where x + 3y = 0 and 0.5x + y = 1 I think? Something like that
EDIT2: The correct ratios are 1.2 of 1x and -0.4 of 3x. Doh! I've updated the post.

I just realized what I did wrong. You're right that the actual correct ratios are about 0.4 and 1.2. In this algo I put 46% in UPRO and 138% in stocks. So it's basically what you said. The leverage is less than 2 and as far as I can tell it would work with IB.

Is this right? Is this how you can increase your leverage in IB with leveraged etfs?

Clone Algorithm
8
Loading...
Total Returns
--
Alpha
--
Beta
--
Sharpe
--
Sortino
--
Max Drawdown
--
Benchmark Returns
--
Volatility
--
Returns 1 Month 3 Month 6 Month 12 Month
Alpha 1 Month 3 Month 6 Month 12 Month
Beta 1 Month 3 Month 6 Month 12 Month
Sharpe 1 Month 3 Month 6 Month 12 Month
Sortino 1 Month 3 Month 6 Month 12 Month
Volatility 1 Month 3 Month 6 Month 12 Month
Max Drawdown 1 Month 3 Month 6 Month 12 Month
def initialize(context):
    pass
def handle_data(context, data):

    order_target_percent(sid(43202), 9/6.5)
    order_target_percent(sid(38533), -3/6.5)
    record( leverage=context.account.leverage )
    
There was a runtime error.

I think that algo would still get a margin call, since the 50% overnight margin requirement for the PANW works out to ~69% and the 100% overnight margin requirement for UPRO works out to 46%, which sum up to ~115%, which is more than your equity.

Note that Quantopian's 'leverage' calculation has no bearing on when IB would give you a margin call, particularly when it comes to leveraged products, which I think is the root of why they don't allow them in the contest.

The reason why the margin requirement is over 100% is that I didn't use all of the equity. The 138% in PANW and 46% in UPRO adds up to 184%, so it isn't using all of the 200% leverage. That's why it's ok that the margin requirement is greater than 100%.

Your algo above uses more margin than you have access to in a Reg T account.

I have a paper trading account with IB and it let me open the exact same positions that my algo would have.

Is that still different?

Nope that should still be the same. Does it let you keep them overnight? Can you post a screenshot of your margin requirements for those two holdings at those percentages?

I'll have to wait to see what happens when the market closes. In the meantime, here's a screenshot:
http://i.imgur.com/dhCAXhz.png

Very interesting, thanks. If I had to guess, it's requiring 25% margin on the long equity and 90% margin on the short LETF. See what happens at 3:50pm.

Here it is after market close:
http://i.imgur.com/yRUX69Q.png

Very interesting, I wonder why that is - according to their documentation, it should be more. Perhaps it's out of date?