Lower Tail Dependence Covariance

PortfolioOptimisers.LowerTailDependenceCovariance — Type
struct LowerTailDependenceCovariance{__T_ve, __T_alpha, __T_ex, __T_cache} <: AbstractCovarianceEstimator

Measures co-movement in the lower tail, the share of the worst returns of one asset that fall on the same dates as the worst returns of another asset.

The statistic is the empirical lower tail dependence coefficient at the level alpha. It measures the dependence of joint drawdowns and of stress scenarios. A correlation over the full sample mixes that dependence with the dependence of the rest of the distribution.

The coefficient is the limit $\lambda_L = \lim_{v \to 0^{+}} \Pr[U_1 \leq v \mid U_2 \leq v]$ of the conditional probability that one series is in its lower tail when the other series is in its lower tail. Section 2 and Equation 1 of [38] state it, and they also state it with the copula $C$ as $\lim_{v \to 0^{+}} C(v,\, v) / v$. Definition 1 of [39] gives the dependence function $\Omega$, with $\Pr(X < x,\, Y < y) = \Omega(G, H)\,G\,H$, so $\lambda_L$ is the limit of $v\,\Omega(v,\, v)$. That work does not name the coefficient, and it does not state an estimator of it.

The library counts the joint exceedances and divides the count by $k = \lceil T \alpha \rceil$. This is the rank form of the empirical lower tail copula at $(1,\, 1)$ in Equations 12 and 16 of Section 4 of [40]. [38] estimates the coefficient from a fitted Archimedean copula, so it does not state this estimator. alpha sets the tail fraction $k / T$, so alpha is the threshold of that estimator.

Warning

The estimate is biased, not only noisy. At a fixed alpha, it converges to the ratio $C(\alpha,\, \alpha) / \alpha$ as $T$ grows, and not to the limit $\lambda_L$. A Gaussian copula has $\lambda_L = 0$ at each correlation below 1, but at correlation 0.6 and alpha = 0.05 its ratio is about 0.31. So a Gaussian sample shows a strong lower tail dependence that its model does not have. To compare a model with an estimate, compare the ratio of the model at the same alpha, not its limit. lower_tail_dependence gives the ratio at more levels.

Fields

  • ve: Variance estimator.
  • alpha: The quantile level of the lower tail.
  • ex: Parallel execution strategy.
  • cache: Optional partial-fit state. It is nothing until partial_fit! writes one. When the caller gives no data matrix, the batch verb of the estimator reads it. Each propagation channel does one thing with it. factory keeps it unchanged, because a factory call resolves the configuration, not the sample. port_opt_view copies the entries of the selected assets into a new state, so the viewed estimator holds the state of those assets alone. obs_weights_view removes it, because a state has no slice on the observation axis. Only a type with an observation-weights field has that channel. A type with no such field falls through to the identity method, so it keeps its state. A family whose state has no exact asset slice removes it on both axes and names the reason.

Constructors

LowerTailDependenceCovariance(;    ve::AbstractVarianceEstimator = SimpleVariance(),    alpha::Number = 0.05,    ex::FLoops.Transducers.Executor = ThreadedEx(),    cache::Option{<:AbstractPartialFitState} = nothing) -> LowerTailDependenceCovariance

Keywords correspond to the struct's fields.

Validation

  • 0 < alpha < 1.

Propagated parameters

factory passes its arguments on to these @fprop fields:

  • ve: Recursively updated through factory.
  • cache: Carried unchanged through factory.

View parameters

port_opt_view takes the view of these @vprop fields at the selected indices:

Examples

julia> LowerTailDependenceCovariance()LowerTailDependenceCovariance     ve ┼ SimpleVariance        │          me ┼ SimpleExpectedReturns        │             │   w ┴ nothing        │           w ┼ nothing        │   corrected ┴ Bool: true  alpha ┼ Float64: 0.05     ex ┴ Transducers.ThreadedEx{@NamedTuple{}}: Transducers.ThreadedEx()

Related

References

  • [39] M. Sibuya. Bivariate extreme statistics, I. Annals of the Institute of Statistical Mathematics 11, 195–210 (1960). Section 1, Definition 1, Equations 1.3 and 1.4.
  • [38] G. De Luca and P. Zuccolotto. A tail dependence-based dissimilarity measure for financial time series clustering. Advances in Data Analysis and Classification 5, 323–340 (2011). Section 2, Equation 1.
  • [40] R. Schmidt and U. Stadtmüller. Non-parametric estimation of tail dependence. Scandinavian Journal of Statistics 33, 307–335 (2006). Section 4, Equations 12 and 16.
source
Statistics.cor — Method
Statistics.cor(ce::LowerTailDependenceCovariance, X::MatNum; dims::Int = 1, kwargs...)

Compute the lower tail dependence correlation matrix with a LowerTailDependenceCovariance estimator.

The quantile level ce.alpha and the executor ce.ex configure the computation. Entry $(i, j)$ is the probability that assets $i$ and $j$ have returns in their lower tails at the same time. The generic covariance fallback reads the ve field of the estimator, and this method does not read it.

Algorithm

  1. Orient X with dims_oriented, which transposes it when dims is 2 and refuses any other value.
  2. Refuse a gapped sample with assert_finite_sample.
  3. Return lower_tail_dependence of the oriented matrix, with the alpha and the executor ex of the estimator.

Arguments

  • ce: The lower tail dependence covariance estimator.
  • X: The data matrix of asset returns, observations × assets.
  • dims: Dimension along which the computation runs.
  • kwargs...: More keyword arguments, which the method ignores.

Validation

  • dims is either 1 or 2.

Returns

  • rho::Matrix{<:Number}: The symmetric matrix of lower tail dependence coefficients.

Examples

julia> ce = LowerTailDependenceCovariance();julia> X = [0.01 0.02; 0.03 0.04; 0.02 0.03];julia> cor(ce, X)2×2 Matrix{Float64}: 1.0  1.0 1.0  1.0

Related

source

References

[38]
G. De Luca and P. Zuccolotto. A tail dependence-based dissimilarity measure for financial time series clustering. Advances in Data Analysis and Classification 5, 323–340 (2011).
[39]
M. Sibuya. Bivariate extreme statistics, I. Annals of the Institute of Statistical Mathematics 11, 195–210 (1960).
[40]
R. Schmidt and U. Stadtmüller. Non-parametric estimation of tail dependence. Scandinavian Journal of Statistics 33, 307–335 (2006).