A paper says price-volume correlations are always zero
A new paper argues the standard empirical price-volume correlation is zero by construction, and proposes measuring against squared volumes.
2 minAlgorithmic & AI Trading
A short paper on arXiv makes a provocative claim about one of the oldest measurements in market microstructure. Its author, Victor Olkhov, argues that calculating the price-volume correlation from the time series of values, volumes and prices of consecutive trades within an averaging interval, in a way that actually follows the definition of correlation, always gives zero.
The argument is about what the usual empirical formula really measures. According to the paper, the problem comes from the volume-weighted average price, which depends on the mean values and mean volumes of trades during the interval, not on the time series of prices alone. The common expression for the correlation therefore misleads researchers and does not fulfil the definition it is meant to estimate.
If that is right, many published findings would need rereading. The paper argues that the positive or negative price-volume correlations reported by numerous researchers in fact reflect the difference between the volume-weighted average price and the simple average price, a difference that vanishes only when all consecutive trade volumes are equal, multiplied by the mean trade volume.
Its proposed fix is to measure something else. To determine the statistical dependence between prices and trade volumes, the author suggests correlating prices with squared trade volumes, or squared prices with squared volumes, and derives expressions for calculating these empirically. The paper is 12 pages, filed under econometrics, and is a single-author argument rather than an empirical study; it has not yet faced replies. For anyone using volume as a trading signal, the practical point is narrower: check how a reported correlation was computed before relying on it.
Retold from arXiv. This is a summary in our own words; follow the link for the original reporting.