Herding dispersion tests
Test whether returns converge unusually during market moves
Return-dispersion approaches examine whether individual security returns converge unusually towards the market. Christie-Huang CSSD focuses on cross-sectional standard deviation in extreme market periods. Chang-Cheng-Khorana CSAD models absolute dispersion and nonlinear dependence on market returns. Reduced or nonlinear dispersion may be consistent with herding under the benchmark model, but does not directly observe investors copying one another.
Choose these for aggregate market herding hypotheses when synchronised returns from a sufficiently broad constituent set are available. Specify the benchmark and market-stress definition before inspecting significant coefficients.
Strengths
- Can be implemented using market and constituent returns
- Contrasts behaviour across market directions and stress conditions
Limitations
- Common information and changing exposures can mimic convergence
- Results depend on constituent selection, extreme cut-offs and nonlinear specification
Know the boundary
A negative nonlinear CSAD coefficient is evidence conditional on a dispersion model, not direct proof of intentional herding.