Choosing Price Observations for Realized Variance Returns
Summary
The discussion explains how the choice of price observations affects return calculations for realized variance when intraday OHLC data are available. Its main recommendation is to calculate returns from one bar’s close to the next bar’s close, whether the bars are one minute or daily. Averaging each bar’s open and close is not presented as the standard approach.
The answers distinguish close-to-close returns from open-to-close returns. Measuring only from each session’s open to its close effectively assumes exposure is closed overnight and reopened in the morning, which may not match the strategy being evaluated. Close-to-close sampling includes the price change between successive closes and therefore carries a different exposure assumption. The guidance is conceptual rather than a comparison of estimators or a treatment of microstructure noise, missing observations, or market-specific session boundaries. The appropriate sampling convention should reflect the actual holding period and trading process being modeled.
Key ideas
- Returns are commonly computed from each interval’s close to the next interval’s close.
- Using open-to-close returns excludes overnight price changes from measured exposure.
- The price sampling convention should match the strategy’s actual holding period.
- The discussion does not compare statistical estimators or address market microstructure effects.
Tags
Full text
# Logarithmic returns for realized variance?
# Logarithmic returns for realized variance?
I am wondering which method makes more sense when computing log returns. I am trying to compute log returns for realized variance, and I have the opening and closing prices for every minute.
Since the log return is defined as
$$r_{t+1} = \ln \left(\frac{p_{t+1}}{p_t} \right)$$
should I take the average of the open and closing prices at every point as use that as $p_t$?
Or should I find $r_{t+1}$ at every $t$ by assuming $p_{t+1}$ = closing price and $p_t$ = opening price?
## Answer by SRKX (score 1, accepted)
https://quant.stackexchange.com/a/7399
It depends on your investment strategy. The most common approach is to use the close price of $p_t$ and $p_{t+1}$. The volatility you measure using this method implies the "assumption" that your are able to trade at close every day.
If you choose to compute the daily returns from open to close, then you assume that you are selling your position every night and buying it back every morning: you have no overnight exposure which is unlikely unless you are fairly sophisticated.
## Answer by Sidharath (score 0)
https://quant.stackexchange.com/a/7397
You should find r at every t by assuming p(t+1) = closing price and p(t) = Last period closing price.
## Answer by chrisaycock (score 0)
https://quant.stackexchange.com/a/7398
It took me a while to figure-out what you were asking. If I understand you correctly, you have minute-level OHLC market data and you want to compute the returns from this time series.
Returns are normally computed close-to-close. Whether you have one-minute bins or you have daily aggregates, you should ignore the opens and just compute returns from one close to the next.Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.