This blog covers the general topic of financial markets.


Total Return of Spot vs Rolling Future: the SP500 case

first posted: 2024-05-02 03:56:43.050501

The yahoo future time series ES=F does not reflect the total return of replicating the SP with rolling futures. It is missing the cost of rolling the future contracts which $d-r$, paid in 4 times during the year before the imm roll dates, where $d$ is the dividend yield and $r$ the interest earned on collateral. Compared to a spot investment or to a rolled future and US Treasury collateral investment strategy paying $r$, it is missing the dividend $d$.

There are many future time series in Yahoo Finance. We have the following time series for SP500:

  • SPY ETF quoting daily until 4pm NY. SPY quote is SPX/10
  • SPX index quoting daily until 5pm NY
  • ES=F future time series
  • ESM24.CME jun 24 future quoting continuously with daily settle snapshot at 3pm Chicago = 4pm NYT
  • ESU24.CME sep 24 future

The futures mature on 3rd friday of the H,M,U,Z month (Mar, Jun, Sep, Dec).

Yahoo Future Series Total Return: Omits the roll cost which is around dividend price

The SPY total return is higher than the price time series due to its approximately 2% dividend yield. The Yahoo ES=F time series does not seem to include future roll. Total return is wrong by $d=2\%$

img

Future vs Spot Quote: premium convergence

The future quote is usually higher than the spot, converging down to the ex-div price. As only a small margin is needed to take the future position, and the funds can be invested tax free in tbills, the premium should increase when treasury yields increase:

img

Dividend is paid on 15th of the IMM month, whereas settlement is done on the third friday. Contango measured that way depends on dividend payment date and settlement date. The apparent contango yields seems higher than the expected $r-d$.

Future vs Next Future Quote: a less noisy estimate of contango yield

Using the 2 next futures, we can measure the contango between two future maturities. This yield appears to be near $r-d$:

img

The data is relatively noisy and there is no evidence of a non 0 market implied dividend withholding tax rate.

Intraday Measured Contango Yield

The intraday measurement is noisy, but appears to show that contango yield is almost the same for the next future and its follower.

img