Testing new property of elliptical model for stock returns distribution
Pith reviewed 2026-05-24 16:59 UTC · model grok-4.3
The pith
For elliptically contoured distributions Kendall's tau equals the probability two variables share the same sign, enabling tests that accept the model for US, UK and German stock returns but reject it for China in some years.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
The central claim is that for any pair of random variables with an elliptically contoured joint distribution the Kendall tau coefficient equals the probability of sign coincidence, and this identity yields distribution-free tests that can be applied to stock-return data. When the resulting Holm-corrected procedure is run on twelve years of daily returns, the elliptical model is accepted for the USA, Great Britain and Germany markets but rejected for China in some observation windows.
What carries the argument
The proved equality of Kendall's tau and sign-coincidence probability for any pair under an elliptically contoured distribution, which supplies the basis for the distribution-free tests.
If this is right
- The equality permits tests of the elliptical model without assuming any specific member of the elliptical class.
- Individual pair tests combined with the Holm procedure allow simultaneous assessment of the model for every pair of stocks in a market.
- The procedure accepts the elliptical model for the full 2003-2014 period in the US, British and German markets.
- The same procedure rejects the elliptical model for the Chinese market in some years of the same period.
Where Pith is reading between the lines
- Acceptance of the model would lend support to risk and portfolio calculations that assume elliptical symmetry in those three markets.
- Rejection for China could indicate stronger tail dependence or asymmetric dependence structures outside the elliptical class.
- Applying the tests to volatility-filtered residuals would check whether the i.i.d. assumption affects the results.
- The same sign-coincidence identity could be used to construct tests for other proposed dependence properties in financial data.
Load-bearing premise
Daily stock returns within each tested year are independent and identically distributed draws from one fixed elliptical distribution.
What would settle it
A large simulated sample from a known elliptical distribution such as the multivariate normal in which the sample Kendall tau differs significantly from the sample sign-coincidence probability would falsify the equality.
read the original abstract
Wide class of elliptically contoured distributions is a popular model of stock returns distribution. However the important question of adequacy of the model is open. There are some results which reject and approve such model. Such results are obtained by testing some properties of elliptical model for each pair of stocks from some markets. New property of equality of $\tau$ Kendall correlation coefficient and probability of sign coincidence for any pair of random variables with elliptically contoured distribution is proved in the paper. Distribution free statistical tests for testing this property for any pair of stocks are constructed. Holm multiple hypotheses testing procedure based on the individual tests is constructed and applied for stock markets data for the concrete year. New procedure of testing the elliptical model for stock returns distribution for all years of observation for some period is proposed. The procedure is applied for the stock markets data of China, USA, Great Britain and Germany for the period from 2003 to 2014. It is shown that for USA, Great Britain and Germany stock markets the hypothesis of elliptical model of stock returns distribution could be accepted but for Chinese stock market is rejected for some cases.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript proves a new property of elliptically contoured distributions: that Kendall's τ equals the probability of sign coincidence for any pair of random variables from this family. It constructs distribution-free tests for the property, combines them via the Holm procedure, and applies the resulting procedure to daily stock-return pairs from China, USA, Great Britain and Germany over 2003–2014, concluding that the elliptical model is acceptable for the last three markets but rejected for China in some years.
Significance. The mathematical identity and the associated distribution-free tests constitute a genuine addition to the toolkit for assessing elliptical models of asset returns. The decade-long, four-market application is a further strength. If the tests remain valid under realistic dependence, the work would supply a practical diagnostic; the current evidence for validity is, however, limited.
major comments (2)
- [Abstract and test-construction section] Abstract and the section constructing the tests: the distribution-free tests are derived under the explicit assumption that each pair of return series consists of i.i.d. draws from a fixed elliptical distribution within each yearly window. No pre-whitening, GARCH filtering, or regime-shift adjustment is described, yet daily equity returns exhibit serial dependence and volatility clustering; under even mild AR(1) or ARCH dependence the finite-sample null distribution of the test statistics deviates from the tabulated law, rendering the reported p-values and the subsequent market-level decisions unreliable.
- [Multi-year testing procedure] Section on the multi-year testing procedure: the claim that the elliptical model can be accepted for USA/UK/Germany rests on the per-year Holm-adjusted tests being correctly sized; without evidence that the procedure remains valid when the i.i.d. assumption is violated, the acceptance/rejection conclusions for the four markets cannot be taken at face value.
minor comments (2)
- [Abstract] Abstract: 'Wide class' should read 'The wide class'; 'is open' is better rendered 'remains open'.
- The probability of sign coincidence is used throughout but never given an explicit symbol or early definition; this notation should be introduced before the statement of the main theorem.
Simulated Author's Rebuttal
We are grateful to the referee for the thorough review and the recognition of the contribution of the new property and the tests. Below we provide point-by-point responses to the major comments.
read point-by-point responses
-
Referee: [Abstract and test-construction section] Abstract and the section constructing the tests: the distribution-free tests are derived under the explicit assumption that each pair of return series consists of i.i.d. draws from a fixed elliptical distribution within each yearly window. No pre-whitening, GARCH filtering, or regime-shift adjustment is described, yet daily equity returns exhibit serial dependence and volatility clustering; under even mild AR(1) or ARCH dependence the finite-sample null distribution of the test statistics deviates from the tabulated law, rendering the reported p-values and the subsequent market-level decisions unreliable.
Authors: The tests are derived under the i.i.d. assumption within each yearly window, as explicitly stated in the manuscript. This is a standard modeling choice for finite-period analyses of return properties. We agree that unmodeled serial dependence and volatility clustering may distort the finite-sample null distribution. In revision we will add an explicit discussion of this maintained assumption in the test-construction section, note the potential size distortion, and recommend that practitioners consider GARCH pre-filtering as a robustness check. This is a partial revision. revision: partial
-
Referee: [Multi-year testing procedure] Section on the multi-year testing procedure: the claim that the elliptical model can be accepted for USA/UK/Germany rests on the per-year Holm-adjusted tests being correctly sized; without evidence that the procedure remains valid when the i.i.d. assumption is violated, the acceptance/rejection conclusions for the four markets cannot be taken at face value.
Authors: The multi-year procedure aggregates per-year Holm-adjusted decisions. While the referee correctly identifies that dependence may affect size, the annual windows are chosen precisely to allow for possible non-stationarity across years. The rejections observed for China appear in several distinct years, lending some robustness. We will qualify the non-rejections for the other three markets by adding a caveat paragraph in the conclusions that reiterates the i.i.d. assumption and its possible violation. No change to the reported decisions themselves is required. revision: partial
Circularity Check
No significant circularity; mathematical identity and tests are independent of fitted data
full rationale
The paper proves a mathematical identity (equality of Kendall τ and sign-coincidence probability under elliptical distributions) and derives distribution-free tests from it. These tests are then applied to market data via Holm correction without any parameter fitting from the target returns that would force the outcome. No self-citations, ansatzes, or renamings appear in the provided text; the derivation chain consists of a standard necessary-condition test whose null distribution is derived independently of the observed series. The i.i.d. modeling choice affects validity but does not create definitional or fitted-input circularity.
Axiom & Free-Parameter Ledger
axioms (2)
- domain assumption Elliptically contoured distributions are closed under linear transformations and possess the stated sign-coincidence property.
- domain assumption Daily stock returns within each calendar year can be treated as i.i.d. draws from a fixed elliptical law.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.