6 ms·

They already did an offering earlier this year for $2B when it first started going crazy. You're right that they should do another. What I don't understand is

They already did an offering earlier this year for $2B when it first started going crazy. You're right that they should do another. What I don't understand is how all the S&P 500 index funds don't get soaked every time a stock enters or leaves the index. Everyone in the world can front-run those giant trades, just like we're seeing now. Do the index funds just absorb the losses every time or do they have enough flexibility to avoid it somehow?

dcolkitt 6y ago
> Everyone in the world can front-run those giant trades This is a perennial quant strategy. Buy the names likely to be included into the index, short those that are likely to be dropped. And the answer is that arbitrage has mostly wiped out the premium. The average gain (loss) on being added (removed) to the index has steadily decreased since the 1990s[1]. [1] https://repository.upenn.edu/cgi/viewcontent.cgi?article=1020&context=joseph_wharton_scholars https://repository.upenn.edu/cgi/viewcontent.cgi?article=102...
bsamuels 6y ago
Great question; index funds rarely carry the entire portfolio of 500 securities - it's very expensive to manage that much stock at scale and rebalance each one. What they do instead is hold a limited portfolio of 10, 50, 100 stocks from the S&P500 that best "represent" the index. Technically Tesla is already in the S&P500; the real question is whether index fund managers are going to decide to start using it in their representation of the index. The other commenter is also correct. There are many ways to hedge in favor of or against a stock issue becoming indexed by S&p500 funds in a way to dampen the impact of people trying to 'front-run' (although it's not actually frontrunning in the legal sense)
- CyberDildonics 6y ago
  Why is it expensive to manage something that can be automated?
  - pmiller2 6y ago
    It basically comes down to math and taxes. Here's a vastly oversimplified scenario. Think about how you'd automate this. It ain't easy, is it? Say you have, say, $10M under management, and the index is at $3,228.90, as it is right now according to Google. For simplicity, let's just make your share price equal to the exact value of the index (which it theoretically would be with no tracking error). If you do the division, you find out that you have to allocate 3097.02994828 shares. If you're an ETF, that's literally impossible. If you're a mutual fund, you can do it, but it introduces some accounting overhead to deal with fractional shares and the like. That's okay for the mutual fund, but not great. As far as taxes, one thing you want to do is hold at least some cash in your fund's portfolio so you don't have to be constantly buying and selling to fulfill purchases. Buying and selling generate capital gains, which you want to avoid as much as possible. However, this also introduces a bit of cash drag to the portfolio, which introduces some tracking error, so your share price will never exactly match the index. That right there blows our assumption that your share price is exactly the current value of the index. Again, because of taxes, you don't want to just own the exact amount of stock necessary to allocate the number of shares you have outstanding. You probably want a bit of a buffer in terms of shares, so you can deal with orders by just twiddling bits internally in your accounting system rather than trying to go to the market and buy shares "just in time" (which is, of course, literally impossible, because the index moves in sub-millisecond time steps due to HFT). So, given that you can't just own 1 share each of all 500 S&P 500 companies and say that's one share of your fund, what you do is go and pick a basket of stocks that, when added up together, correlate well to the index price. You can pick 10, 30, 50, 100, or whatever stocks, but you have to do a lot of number crunching to figure out precisely which stocks you want to include. You want to do this tax-efficiently (remember what I said about avoiding buying and selling), and you also want as low of a tracking error as possible. This is a hard problem. Software can aid in solving this problem, but you need to pay people to make sure that your software does something sensible and isn't going to drown you in taxes. And, then, there's the small matter of dividends, which complicates the hell out of things. If you don't own all 500 stocks and call that 1 share of your fund, your dividend will not match the sum of all the dividends of all 500 stocks in the index. That's fine, but you have to be aware of it, and you have to track dividend dates and amounts for however many stocks you hold. Did that convince you that running an index fund isn't as simple as it sounds? :) That's why all funds, including index funds, have a nonzero expense ratio. Ya gotta pay for all this processing. Oh, and I haven't even covered rebalancing and all the headaches that go with that at scale. And, because you're not Joe Six Pack investor, you're not trading on a $0 commission platform. You're paying for pro-level access to the market. TL;DR: Owning all 500 stocks and bundling them as 1 share is pretty impractical. Since every trade you make on the open market costs money and possibly generates tax liability, you want to avoid that as much as possible and fulfill orders via moving bits around in your internal accounting systems. These systems are nontrivial.
- currymj 6y ago
  wait, does Vanguard S&P 500 or SPY actually not own stock in all 500 companies? that is very surprising!
  - basch 6y ago
    SPY has 506 holdings, VOO has 510, some of which could be Class A and C stock for the same company, or cash.
- asaph 6y ago
  > index funds rarely carry the entire portfolio of 500 securities - it's very expensive to manage that much stock at scale and rebalance each one. This doesn't pass the sniff test. Funds calling themselves "S&P 500" funds but not actually holding the 500 companies in the index seems like false advertising at best and downright fraud at worst. And rebalancing a fund's holdings should be an automated process whose cost should be covered by the funds fees.
  - pmiller2 6y ago
    They're called "S&P 500 index funds" because they closely track the S&P 500. If you look at literally any such fund, you will see they always have some nonzero tracking error. That's mathematically unavoidable. The task of the fund manager is to minimize that error, among other things.
    - asaph 6y ago
      I thought the whole point of "index funds" was that they are not "actively managed" by a human fund manager. They should just automatically track the index (with some minimal tracking error).
      - pmiller2 6y ago
        Mostly. They aren't "actively managed" in the same sense as non-index funds. Humans need to make some decisions about what to include, and monitor to make sure things like tracking error are low and tax efficiency is high. This is not trivial, which is why index funds were not just some obvious concept that has existed forever, and why, when John Bogle articulated the concept in the 70s at Vanguard, tens of them didn't just immediately pop up. They're the type of object that looks very simple from the outside, and acts very simply, but, under the covers there's a good deal of stuff going on. It's kind of like how the steam engine is a very simple concept (in fact, the ancient Greeks invented a simple steam engine [0]), but, if you look at it at the very lowest level, you can't model how each individual water molecule contributes to the functioning of the whole. That's not to say index funds have emergent behaviors. Rather that they work internally in ways you, the investor, don't need to care about, because they're designed to have that fairly predictable and simple external behavior. Put another way: ever wonder why an index fund would need to have any management fee at all (i.e. why there's a nonzero expense ratio)? The reason is because they are managed, and you have to pay people to do that. --- [0]: https://en.wikipedia.org/wiki/Aeolipile https://en.wikipedia.org/wiki/Aeolipile
- pmiller2 6y ago
  Speaking of not carrying the full boat of 500, [0] is a good explanation of some of the mechanics behind why you might want to do that, and what difficulties it introduces (warning: auto-playing video). If you choose not to own all 500, that is going to introduce some tracking error. A casual search turned up [1], which, although old, has some interesting numbers on the tracking error, expense ratio, and alpha [2] of several funds. I found it interesting that some funds actually generated a small amount of alpha. I bet over time the mean alpha for a sampled-index strategy is 0 though, just by the law of large numbers. --- [0]: https://www.etf.com/etf-education-center/etf-basics/understanding-tracking-difference-and-tracking-error?nopaging=1 https://www.etf.com/etf-education-center/etf-basics/understa... [1]: https://seekingalpha.com/article/217707-keeping-track-of-the-s-and-p-500-index-trackers https://seekingalpha.com/article/217707-keeping-track-of-the... [2]: Alpha, intuitively speaking, is basically the amount of excess return over and above what you'd get if you theoretically just owned the whole index with 0 transaction fees. Pretty much, it encapsulates "how much did this strategy beat the market by." Predictably, alpha values for these funds are small and close to zero. Most are negative, meaning you're going to lag the market. Unless you have enough money to use a direct indexing strategy, you basically have to live with that.
T-hawk 6y ago
As the sibling comment to mine says, predictive arbitrage will already have accounted for the price shift by the time the index inclusion/exclusion actually happens. That arbitrage means the price change effectively gets smeared over time, as the probability of being included in the index rises. The price effect of index inclusion isn't a binary in-or-out, it's a probability cloud based on speculation. Consider a random company like say Dropbox, and take a wild guess that they might have a 2% chance of making the S&P 500 in the foreseeable future. That small chance is already built in to the current stock price, assuming efficient enough arbitrage. If that perceived chance were to rise, so would the stock price, and as it approached 100% would converge to the post-index inclusion value. Stated another way: By the time an index fund needs to acquire a particular stock, enough arbitrageurs and market-makers have already acquired it that they're willing to sell at the price the index fund will pay. There isn't really a loss for the index fund to absorb. If there is, it shows up as tracking error in the fund's value not perfectly replicating the index, which is a common occurrence for plenty of reasons.

dcolkitt 6y ago

> Everyone in the world can front-run those giant trades This is a perennial quant strategy. Buy the names likely to be included into the index, short those that are likely to be dropped. And the answer is that arbitrage has mostly wiped out the premium. The average gain (loss) on being added (removed) to the index has steadily decreased since the 1990s[1]. [1] https://repository.upenn.edu/cgi/viewcontent.cgi?article=1020&context=joseph_wharton_scholars https://repository.upenn.edu/cgi/viewcontent.cgi?article=102...

bsamuels 6y ago

Great question; index funds rarely carry the entire portfolio of 500 securities - it's very expensive to manage that much stock at scale and rebalance each one. What they do instead is hold a limited portfolio of 10, 50, 100 stocks from the S&P500 that best "represent" the index. Technically Tesla is already in the S&P500; the real question is whether index fund managers are going to decide to start using it in their representation of the index. The other commenter is also correct. There are many ways to hedge in favor of or against a stock issue becoming indexed by S&p500 funds in a way to dampen the impact of people trying to 'front-run' (although it's not actually frontrunning in the legal sense)

CyberDildonics 6y ago
Why is it expensive to manage something that can be automated?
- pmiller2 6y ago
  It basically comes down to math and taxes. Here's a vastly oversimplified scenario. Think about how you'd automate this. It ain't easy, is it? Say you have, say, $10M under management, and the index is at $3,228.90, as it is right now according to Google. For simplicity, let's just make your share price equal to the exact value of the index (which it theoretically would be with no tracking error). If you do the division, you find out that you have to allocate 3097.02994828 shares. If you're an ETF, that's literally impossible. If you're a mutual fund, you can do it, but it introduces some accounting overhead to deal with fractional shares and the like. That's okay for the mutual fund, but not great. As far as taxes, one thing you want to do is hold at least some cash in your fund's portfolio so you don't have to be constantly buying and selling to fulfill purchases. Buying and selling generate capital gains, which you want to avoid as much as possible. However, this also introduces a bit of cash drag to the portfolio, which introduces some tracking error, so your share price will never exactly match the index. That right there blows our assumption that your share price is exactly the current value of the index. Again, because of taxes, you don't want to just own the exact amount of stock necessary to allocate the number of shares you have outstanding. You probably want a bit of a buffer in terms of shares, so you can deal with orders by just twiddling bits internally in your accounting system rather than trying to go to the market and buy shares "just in time" (which is, of course, literally impossible, because the index moves in sub-millisecond time steps due to HFT). So, given that you can't just own 1 share each of all 500 S&P 500 companies and say that's one share of your fund, what you do is go and pick a basket of stocks that, when added up together, correlate well to the index price. You can pick 10, 30, 50, 100, or whatever stocks, but you have to do a lot of number crunching to figure out precisely which stocks you want to include. You want to do this tax-efficiently (remember what I said about avoiding buying and selling), and you also want as low of a tracking error as possible. This is a hard problem. Software can aid in solving this problem, but you need to pay people to make sure that your software does something sensible and isn't going to drown you in taxes. And, then, there's the small matter of dividends, which complicates the hell out of things. If you don't own all 500 stocks and call that 1 share of your fund, your dividend will not match the sum of all the dividends of all 500 stocks in the index. That's fine, but you have to be aware of it, and you have to track dividend dates and amounts for however many stocks you hold. Did that convince you that running an index fund isn't as simple as it sounds? :) That's why all funds, including index funds, have a nonzero expense ratio. Ya gotta pay for all this processing. Oh, and I haven't even covered rebalancing and all the headaches that go with that at scale. And, because you're not Joe Six Pack investor, you're not trading on a $0 commission platform. You're paying for pro-level access to the market. TL;DR: Owning all 500 stocks and bundling them as 1 share is pretty impractical. Since every trade you make on the open market costs money and possibly generates tax liability, you want to avoid that as much as possible and fulfill orders via moving bits around in your internal accounting systems. These systems are nontrivial.
currymj 6y ago
wait, does Vanguard S&P 500 or SPY actually not own stock in all 500 companies? that is very surprising!
- basch 6y ago
  SPY has 506 holdings, VOO has 510, some of which could be Class A and C stock for the same company, or cash.
asaph 6y ago
> index funds rarely carry the entire portfolio of 500 securities - it's very expensive to manage that much stock at scale and rebalance each one. This doesn't pass the sniff test. Funds calling themselves "S&P 500" funds but not actually holding the 500 companies in the index seems like false advertising at best and downright fraud at worst. And rebalancing a fund's holdings should be an automated process whose cost should be covered by the funds fees.
- pmiller2 6y ago
  They're called "S&P 500 index funds" because they closely track the S&P 500. If you look at literally any such fund, you will see they always have some nonzero tracking error. That's mathematically unavoidable. The task of the fund manager is to minimize that error, among other things.
  - asaph 6y ago
    I thought the whole point of "index funds" was that they are not "actively managed" by a human fund manager. They should just automatically track the index (with some minimal tracking error).
    - pmiller2 6y ago
      Mostly. They aren't "actively managed" in the same sense as non-index funds. Humans need to make some decisions about what to include, and monitor to make sure things like tracking error are low and tax efficiency is high. This is not trivial, which is why index funds were not just some obvious concept that has existed forever, and why, when John Bogle articulated the concept in the 70s at Vanguard, tens of them didn't just immediately pop up. They're the type of object that looks very simple from the outside, and acts very simply, but, under the covers there's a good deal of stuff going on. It's kind of like how the steam engine is a very simple concept (in fact, the ancient Greeks invented a simple steam engine [0]), but, if you look at it at the very lowest level, you can't model how each individual water molecule contributes to the functioning of the whole. That's not to say index funds have emergent behaviors. Rather that they work internally in ways you, the investor, don't need to care about, because they're designed to have that fairly predictable and simple external behavior. Put another way: ever wonder why an index fund would need to have any management fee at all (i.e. why there's a nonzero expense ratio)? The reason is because they are managed, and you have to pay people to do that. --- [0]: https://en.wikipedia.org/wiki/Aeolipile https://en.wikipedia.org/wiki/Aeolipile
pmiller2 6y ago
Speaking of not carrying the full boat of 500, [0] is a good explanation of some of the mechanics behind why you might want to do that, and what difficulties it introduces (warning: auto-playing video). If you choose not to own all 500, that is going to introduce some tracking error. A casual search turned up [1], which, although old, has some interesting numbers on the tracking error, expense ratio, and alpha [2] of several funds. I found it interesting that some funds actually generated a small amount of alpha. I bet over time the mean alpha for a sampled-index strategy is 0 though, just by the law of large numbers. --- [0]: https://www.etf.com/etf-education-center/etf-basics/understanding-tracking-difference-and-tracking-error?nopaging=1 https://www.etf.com/etf-education-center/etf-basics/understa... [1]: https://seekingalpha.com/article/217707-keeping-track-of-the-s-and-p-500-index-trackers https://seekingalpha.com/article/217707-keeping-track-of-the... [2]: Alpha, intuitively speaking, is basically the amount of excess return over and above what you'd get if you theoretically just owned the whole index with 0 transaction fees. Pretty much, it encapsulates "how much did this strategy beat the market by." Predictably, alpha values for these funds are small and close to zero. Most are negative, meaning you're going to lag the market. Unless you have enough money to use a direct indexing strategy, you basically have to live with that.

T-hawk 6y ago

As the sibling comment to mine says, predictive arbitrage will already have accounted for the price shift by the time the index inclusion/exclusion actually happens. That arbitrage means the price change effectively gets smeared over time, as the probability of being included in the index rises. The price effect of index inclusion isn't a binary in-or-out, it's a probability cloud based on speculation. Consider a random company like say Dropbox, and take a wild guess that they might have a 2% chance of making the S&P 500 in the foreseeable future. That small chance is already built in to the current stock price, assuming efficient enough arbitrage. If that perceived chance were to rise, so would the stock price, and as it approached 100% would converge to the post-index inclusion value. Stated another way: By the time an index fund needs to acquire a particular stock, enough arbitrageurs and market-makers have already acquired it that they're willing to sell at the price the index fund will pay. There isn't really a loss for the index fund to absorb. If there is, it shows up as tracking error in the fund's value not perfectly replicating the index, which is a common occurrence for plenty of reasons.