36

I saw a tweet from the Chief Investment Strategist of Charles Schwab that said the following:

Since 1993 from @bespokeinvest :

Buying SPY at open & selling at close every day (ie, only holding it during trading day): -13.9% return

Buying SPY at close & selling at next open (ie, only holding it after-hours): +634.2%

It included the following graphic:

enter image description here

Why has the S&P 500 generated such a massive gain since 1993 during the overnight hours (4pm to 9:30am), and has an overall loss during normal trading hours (9:30am to 4pm)?

Community
  • 1
7529
  • 3,753
  • 3
  • 24
  • 37
  • 21
    All of the existing answers are getting at the "fundamentals" that ultimately cause this pattern. But none of them address the natural sub-question "Why haven't any investors noticed this non-random pattern, executed an automatic investment strategy to exploit it, and neutralized it?" – tparker May 14 '20 at 02:54
  • 3
    @tparker - No one addressed that sub-question in their answers because it wasn't asked. – Bob Baerker May 14 '20 at 03:47
  • 16
    @BobBaerker It seems to me that that question is implicit in any apparently exploitable regularity in stock market behavior, but I guess that's a matter of opinion. – tparker May 14 '20 at 03:56
  • 3
    I will also point out that the cumulative nature of this chart hides the real truth - perhaps on most days, overnight results exceed intra-day results, but on many days, the reverse is true. You can see this in the chart when the difference between the two lines tightens, but it is not intuitive, and makes it look more like a 'sure thing' that overnight investing always wins. This downplays the risk of trading on this as a general strategy. – Grade 'Eh' Bacon May 14 '20 at 13:39
  • @tparker - If I was Carnac The Magnificent, I would put the card up against my magical ESP hat and I would know that the OP was implicitly thinking about the question that he did not ask. And as Grade 'Eh' Bacon points out, the long term performance belies the risk of trading an indicator such as this. In years 01, 02, 08, 12, and 15 there were sustained down moves in the overnight strategy and the draw downs would have been severe. No savvy trader would stuck with such a losing strategy in those periods. – Bob Baerker May 14 '20 at 14:25
  • If the chart starts from 2009 instead, would the question become "Why do most of the stock market's gains occur intraday"? – base64 May 14 '20 at 14:28
  • 5
    @tparker My answer does address that, because I point out that the effect is to be expected in equilibrium, on the basis of the risk and interest cost of holding overnight. It is not an "exploitable regularity" because attempting to exploit it (buy buying at the close and selling at the open) would earn only fair compensation for that risk/cost (also transaction costs). That is, it's consistent with an efficient-market equilibrium of risk and return. It's like asking whether the well-known long-term return advantage of riskier stocks over safer bonds is an "exploitable regularity". – nanoman May 14 '20 at 23:39
  • 2
    Also, it should be noted that holding it overnight only likely would not have produced better return than just holding it and never selling. By holding it and never selling you get those 531.9% minus one spread and twice trading fees. By selling every morning and buying every night for 27 years (assuming 250 business days a year) you pay 6750 spreads and 13500 trading fees... this will be more than those 100 percentage points... – Demosthenes May 15 '20 at 10:08
  • 1
    @Demosthenes Is that still true with $0 trades? –  May 15 '20 at 21:31
  • @Michael you still have the spread at the least – Dan May 16 '20 at 01:08
  • @tparker Because it is a market and 7529 is not the first person noticing this pattern there must be a hidden cost to reaping its windfall profit, and the difference between 530% and 630% is the cumulated amount of that hidden cost. Can be the spread, trading fees, taxes, labor, whatever. – Peter - Reinstate Monica May 16 '20 at 05:35
  • Before explaining the possible effect the first thing to check is whether this effect is real or whether the data has been cherry picked. All current answers might be trying to explain a non-existent effect. Could someone check that other market indices show the same effect? – Kvothe Jun 22 '21 at 16:24

3 Answers3

38

A couple of possible reasons:

  1. A disproportionate amount of stock-market risk (e.g., scheduled economic and earnings releases) happens outside market hours, with the goal of avoiding destabilization of the market. (Extended-hours and futures trading are typically occurring, and show the immediate impact, but the regular-hours SPY does not.) Thus, overnight returns should be commensurate with the risk.

  2. Interest (either on positive cash balances or on margin debt) is calculated daily on overnight balances regardless of intraday fluctuations. Thus, day trades have zero cost of capital and their return is lower to compensate.

I once read a suggestion that a countervailing mechanism explains the suppression of this tendency in parts of the 1990s (i.e., somewhat weaker overnight and stronger intraday returns): As Japan endured an extended bubble-deflating bear market, US morning futures and opening markets were often depressed in sympathy with the overnight Nikkei but then rebounded during the trading hours of bullish US investors.

nanoman
  • 29,638
  • 3
  • 69
  • 89
  • I like your reason #1. I am not sure how interest would result in much greater overnight returns though. – 7529 May 13 '20 at 00:36
  • 4
    @7529 Not a big factor right now when rates are close to zero, but when rates were ~5%/yr it would add to the temptation for traders to sell their positions by the close, earn more (or pay less) interest overnight, and buy again tomorrow. In the resulting equilibrium, overnight returns had to become more lucrative to justify the cost of capital. – nanoman May 13 '20 at 00:48
  • Most people hold stocks without margin though, so no interest is involved. If they are buying on margin, their return will be more lucrative just from the fact that they have invested more than 100% of their cash. – 7529 May 13 '20 at 18:24
  • @7529 Even without margin, brokers often pay interest on cash (e.g., by sweeping into a money market fund) -- granted, at lower rates than they charge for margin. So either way, holding a stock overnight means losing on overnight interest (this is what I mean by "cost of capital"). When you say "their return will be more lucrative", the issue is the return from an incremental dollar investment. No matter what position size they already have, holding an additional dollar overnight in stock instead of cash gains them the stock return on the dollar and costs them the interest on the dollar. – nanoman May 13 '20 at 20:09
  • 4
    My two cents: The number of non-business hours (night, weekend, holiday) is like 4× the number of trading hours (09:30~16:00), so it's a reasonable first assumption that more risk happens during those hours. – Nayuki May 13 '20 at 22:43
  • @Nayuki But by itself that doesn't explain why not just 80%, but 100% or more of net returns happen overnight. – nanoman May 14 '20 at 00:36
  • @nanoman To add some data to this idea: A gain of 634% over the 26 years depicted in the graph corresponds to an interest rate of ~8% – Hagen von Eitzen May 14 '20 at 15:34
24

Per the request above:

The market reacts to overnight news which includes after hours earnings announcements, economic reports, overseas trading, etc. (very, very few companies announce earnings during regular hours trading). Because of these, the market tends to gap up or down in the morning. In a long term bull market, those gaps are net positive

Bob Baerker
  • 76,304
  • 15
  • 99
  • 175
13

The following items could play a factor:

  1. Intraday margin is higher (e.g. 4x leverage for intraday but 2 x for overnight), causing people to buy at the open (push the price up) and sell at the close (push the price down). While people could also use intraday margin to create short positions, in general there are more long buyers than short sellers.

  2. Many investors are afraid of 'gaps'. For example that the market 'gaps up' or 'gaps down' between the close and the open, so may want to close out their positions at the end of the day to avoid those risks.

  3. Many of the great crashes have occurred intra-day.

  4. By employing this strategy over 26 years you are only gaining 20% greater return overall - that is less than 0.7% better per year on average. This kind of improvement could be eaten by bid-ask spreads and transaction costs at the close and the open. It is not clear what methodology was used to determine the prices. If the midpoint was used, that would not take into account transaction spreads.

xirt
  • 5,121
  • 10
  • 37
  • (1) More long buyers than short sellers during the day contradicts the idea that holding intraday under performs. (4) Even if B/A spreads ate up a chunk of the 20% overnight gain, there would be the same affect on the lagging intraday performance and the degree of outperformance of overnight versus intraday would remain. IOW, both would be worse due to what should be similar slippage. – Bob Baerker May 13 '20 at 11:34
  • @BobBaerker (1) not really. Due to margin rules buyers or sort sellers have to sell by the close, even if it means a loss. (4) true, but "full-day" could also be achieved with buy-and-hold over the entire period without any loss to spreads/slippage. – xirt May 13 '20 at 22:16
  • 1
    I think that you have wandered away from the premise of the question in several ways. Only Pattern Day Traders who are using the allowable 4x intraday margin "have to sell by the close, even if it means a loss". They must drop down to 2x margin, aka Reg T 50% so it's only a partial close. At 4x margin intraday, they get 4 times the gain or 4 times the loss. But given that the intraday gain severely lags the overnight gain, it's irrelevant because everyone in the intraday camp has to sell at the close regardless of margin or not. This is all not admissible as evidence. – Bob Baerker May 14 '20 at 03:43
  • cont. - Yes, full-day could also be achieved with buy-and-hold over the entire period without any loss to spreads/slippage". What does that have to do with the OP's premise of intraday hold versus overnight hold, with no hold beyond those time periods? Again, not admissible. – Bob Baerker May 14 '20 at 03:45
  • @BobBaerker [A] Agreed some people may hold 2x positions overnight and increase to 4x during the day. It is the additional 2x that they have to sell. The existing 2x could be from the day before. [B] The line shown for full day performance is identical to the line if you bought and held for the time period, clearly showing slippage is not taken into account. [C] Having to sell at the close puts acute downward pressure on the price, which might otherwise be supported by the market. – xirt May 14 '20 at 22:36
  • For #1, intraday longs are opening positions after 9:30am, so that doesn't impact the market's opening price. Their selling before 4pm may lower the market's closing price, but that's offset by their buying after the open that may have driven the market higher. – 7529 May 15 '20 at 01:04