Stylized Facts of Intraday Financial Returns
Summary
The document gives a brief overview of empirical patterns reported in financial returns, particularly in response to a question about intraday equity and exchange-rate data. It notes that return distributions and volatility change over time, returns are roughly symmetric but become more heavy-tailed at finer sampling frequencies, and longer-horizon returns tend to look more Gaussian. It also mentions serial-correlation anomalies and empirical scaling behavior, while cautioning that these patterns do not make returns straightforward to forecast.
The discussion points readers toward research on asset-return properties and volatility clustering, but it does not present datasets, detailed methods, or evidence from a specific market or sampling interval. The claims are therefore a compact literature-oriented starting point rather than a complete account of intraday behavior. Results may vary with the asset, frequency, and statistical treatment.
Key ideas
- Return distributions and return volatility are not constant over time.
- Returns are approximately symmetric, with heavier tails reported at finer sampling frequencies.
- Returns appear more Gaussian when measured over longer horizons.
- Serial-correlation anomalies may exist, but the answer describes them as difficult to forecast.
- Empirical scaling patterns exist even though returns are not independent and identically distributed.
Tags
Full text
# Intraday Data - Stylized Facts? # Intraday Data - Stylized Facts? Can someone give an overview or literature on Intraday Data Stylized Facts? In particular for equity market returns or exchange rates. ## Answer by madilyn (score 2) https://quant.stackexchange.com/a/14898 - Stationarity. The distribution of returns is non-stationary. Moreover, standard deviation of returns is not constant over time. - Symmetry. The distribution of returns is approximately symmetric with increasing leptokurtosis as sampling frequency increases. However, large drawdowns are not matched with equally large upward movements. - Gaussian behavior. Returns become increasingly normal with decreasing frequency. Long horizon returns are approximately normal - Serial correlation. There exist anomalies in the serial correlation of returns, which are nevertheless impossibly difficult to forecast. - Scaling properties and asymmetry in time scale. Time scaling is nontrivial as returns are not iid, but there is documented evidence for fairly stable, empirical scaling properties. ## Answer by SiXUlm (score 1) https://quant.stackexchange.com/a/14936 This reminds me of a paper by Rama Cont: "Empirical properties of asset returns: stylized facts and statistical issues.". You can download here: http://www.cmap.polytechnique.fr/~rama/papers/empirical.pdf He also has a paper on volatility clustering: "Volatility clustering in financial markets: empirical facts and agent-based models.", which may be of your interest. http://www.cmap.polytechnique.fr/~rama/papers/clustering.pdf Hope this help.
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.