Empirical Bernstein Copula

• Empirical Bernstein Copula is a smooth nonparametric estimator that applies Bernstein polynomial smoothing to the empirical copula, enhancing differentiability and inference.
• It enables derivative-based inference and robust resampling by transforming a stepwise empirical copula into a smooth, polynomial function.
• It serves as a versatile building block for hypothesis testing, conditional estimation, and vine copula models, addressing bias-variance trade-offs in high-dimensional settings.

Searching arXiv for the cited papers and closely related work on empirical Bernstein copulas. The empirical Bernstein copula is a smooth nonparametric copula estimator obtained by applying multivariate Bernstein polynomials to the empirical copula on a regular grid of $[0,1]^d$. In the formulation used by Segers, Sibuya, and Tsukahara, if $C_n$ denotes the usual empirical copula and $m=(m_1,\dots,m_d)\in\mathbb N^d$ is a multi-degree, then

$$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$

Equivalently, it is the empirical copula averaged against a product-Binomial smoothing law, which makes the estimator polynomial, smooth, and amenable to derivative-based inference (Segers et al., 2016). In the bivariate notation used for symmetry testing,

$$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$

where $P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$ (Lyu et al., 2022). Across the literature, the estimator functions both as a direct smooth substitute for the stepwise empirical copula and as a building block for resampling, hypothesis testing, conditional distribution estimation, vine copulas, and semiparametric Bernstein-copula models (Kojadinovic et al., 2023).

1. Definition, rank construction, and copula structure

The starting point is the rank-based empirical copula. For i.i.d. observations $X_1,\dots,X_n\in\mathbb R^d$ with continuous margins, let $R_{i,j}^{(n)}$ be the rank of $X_{i,j}$ among $X_{1,j},\dots,X_{n,j}$, and define

$C_n$0

The empirical Bernstein copula replaces these hard indicators by Bernstein basis weights indexed on a finite grid (Segers et al., 2016). In bivariate form, with pseudo-observations

$C_n$1

the usual empirical copula is

$C_n$2

and the empirical Bernstein copula of order $C_n$3 is the Bernstein smoothing of $C_n$4 on the grid $C_n$5 (Lyu et al., 2022).

A central structural issue is whether Bernstein smoothing preserves the copula property. Segers, Sibuya, and Tsukahara give necessary and sufficient coefficient conditions for a Bernstein polynomial $C_n$6 to be a copula. If the coefficient array satisfies groundedness, uniform-margin constraints, and nonnegativity of the full forward difference $C_n$7, then the Bernstein polynomial is a copula; groundedness and marginal conditions are also necessary (Segers et al., 2016). This result explains why some Bernstein smoothers are genuine copulas and why others are not.

A related formulation replaces direct smoothing of $C_n$8 by smoothing of the empirical checkerboard copula $C_n$9. Lu and Ghosh define the multivariate empirical checkerboard Bernstein copula (ECBC)

$m=(m_1,\dots,m_d)\in\mathbb N^d$0

with $m=(m_1,\dots,m_d)\in\mathbb N^d$1. Because $m=(m_1,\dots,m_d)\in\mathbb N^d$2 is a copula and the Bernstein basis preserves the copula properties for any $m=(m_1,\dots,m_d)\in\mathbb N^d$3, $m=(m_1,\dots,m_d)\in\mathbb N^d$4 is itself a genuine copula (Lu et al., 2021).

2. Relation to the empirical beta copula and other smooth copula estimators

The most important special case is obtained by setting all Bernstein degrees equal to the sample size. Segers, Sibuya, and Tsukahara show that when $m=(m_1,\dots,m_d)\in\mathbb N^d$5, the empirical Bernstein copula becomes the empirical beta copula,

$m=(m_1,\dots,m_d)\in\mathbb N^d$6

which is therefore a particular case of the empirical Bernstein copula (Segers et al., 2016). In the equivalent representation,

$m=(m_1,\dots,m_d)\in\mathbb N^d$7

each coordinate is smoothed by the c.d.f. of a $m=(m_1,\dots,m_d)\in\mathbb N^d$8 kernel (Segers et al., 2016). The same relationship is reiterated in broader classes of smooth estimators: the empirical beta copula is the special case obtained when the smoothing law is product-Binomial with degree $m=(m_1,\dots,m_d)\in\mathbb N^d$9 (Kojadinovic et al., 2021).

The literature distinguishes the empirical Bernstein copula from several neighboring constructions. The empirical copula is piecewise constant and not differentiable (Lyu et al., 2022). The checkerboard copula $$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$0 is a multilinear extension and is a genuine copula even for finite $$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$1 (Lu et al., 2021). ECBC smooths $$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$2 rather than $$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$3, and Lu and Ghosh use this to obtain a genuine copula estimator with data-dependent degrees (Lu et al., 2021).

A further distinction concerns the parametric or semiparametric “Bernstein copula” of Dou et al. Their Bernstein copula is a finite-mixture copula

$$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$4

where the weights satisfy marginal-uniformity constraints (Dou et al., 2013). This is not the same object as the empirical Bernstein copula, although the empirical estimator can be viewed as a specific rank-based smoothing device within the same Bernstein basis.

The main related estimators are summarized below.

Estimator	Construction	Relation to empirical Bernstein copula
Empirical copula $$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$5	Rank-based step function	Unsmoothed baseline (Segers et al., 2016)
Empirical Bernstein copula $$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$6	Bernstein smoothing of $$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$7	General smooth polynomial estimator (Segers et al., 2016)
Empirical beta copula $$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$8	Case $$\widehat C_{n,m}(u) :=B_m(C_n)(u) = \sum_{s_1=0}^{m_1}\cdots\sum_{s_d=0}^{m_d} C_n\!\Bigl(\tfrac{s_1}{m_1},\dots,\tfrac{s_d}{m_d}\Bigr) \prod_{j=1}^d \binom{m_j}{s_j}u_j^{\,s_j}(1-u_j)^{m_j-s_j}, \qquad u\in[0,1]^d.$$9	Special case of empirical Bernstein copula (Segers et al., 2016)
ECBC $$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$0	Bernstein smoothing of checkerboard copula $$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$1	Closely related genuine copula estimator (Lu et al., 2021)

This suggests a useful conceptual distinction: empirical Bernstein copulas are primarily defined by how the rank-based copula is smoothed, whereas the empirical beta copula and ECBC are particular structurally advantageous instances within the same Bernstein-polynomial ecosystem.

3. Large-sample properties, bias, variance, and smoothing rates

A basic result is that the empirical Bernstein copula is asymptotically equivalent to the empirical copula process under standard smoothness assumptions. Segers, Sibuya, and Tsukahara prove that if the minimum degree $$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$2 and $$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$3, then

$$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$4

where $$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$5 and $$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$6 is the same tight centered Gaussian limit as for the empirical copula (Segers et al., 2016). The proof decomposes the process into a stochastic equicontinuity term and a smoothing bias term (Segers et al., 2016).

For the bivariate setting studied in symmetry testing, the empirical Bernstein copula process

$$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$7

converges in $$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$8 to

$$C_{n,m}(u,v) = \sum_{k=0}^m\sum_{\ell=0}^m \widehat C_n(k/m,\ell/m)\, P_{m,k}(u)\,P_{m,\ell}(v),$$9

where $P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$0 is the $P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$1-Brownian bridge with covariance

$P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$2

provided $P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$3 has continuous first-order partials and $P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$4 with $P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$5 and $P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$6; the same source notes that $P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$7 suffices under stronger second-derivative control (Lyu et al., 2022).

Uniform consistency is established under mild smoothness. In the symmetry-testing paper, if $P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$8 with $P_{m,k}(u)=\binom mku^k(1-u)^{m-k}$9 and $X_1,\dots,X_n\in\mathbb R^d$0, then

$X_1,\dots,X_n\in\mathbb R^d$1

almost surely and in probability (Lyu et al., 2022). More precisely,

$X_1,\dots,X_n\in\mathbb R^d$2

while

$X_1,\dots,X_n\in\mathbb R^d$3

so that choosing $X_1,\dots,X_n\in\mathbb R^d$4 makes the overall error vanish (Lyu et al., 2022).

Other sources state slightly different rate prescriptions because they work with different loss criteria or asymptotic regimes. Ouimet and Susam report the conditions

$X_1,\dots,X_n\in\mathbb R^d$5

for uniform strong consistency and

$X_1,\dots,X_n\in\mathbb R^d$6

for the same limit distribution as the empirical copula. They also state that balancing $X_1,\dots,X_n\in\mathbb R^d$7 bias against variance reduction of order $X_1,\dots,X_n\in\mathbb R^d$8 leads to the rule-of-thumb $X_1,\dots,X_n\in\mathbb R^d$9, implemented in practice as $R_{i,j}^{(n)}$0 (Ouimet et al., 10 Jun 2025). In contrast, the two-sample testing paper states that, under mild regularity and continuous bounded first partial derivatives, the pointwise mean-squared error is

$R_{i,j}^{(n)}$1

which yields the familiar rule $R_{i,j}^{(n)}$2 for minimizing an integrated MSE (Lyu et al., 2023). The vine-copula paper, working with a contingency-table version of the estimator, reports bias $R_{i,j}^{(n)}$3 and stochastic error $R_{i,j}^{(n)}$4, motivating growth such as $R_{i,j}^{(n)}$5 (Weiß et al., 2012).

These differing prescriptions are not contradictory on their face; they arise from different estimands, approximations, and risk criteria. A plausible implication is that “the” optimal Bernstein degree is context-dependent: testing, tail functionals, generic copula estimation, and checkerboard-based constructions need not select the same rate.

4. Smoothness, derivatives, and resampling theory

A defining advantage of the empirical Bernstein copula is differentiability. Since $R_{i,j}^{(n)}$6 is a polynomial in the arguments, it is infinitely differentiable in the interior (Lyu et al., 2022). Under $R_{i,j}^{(n)}$7 with $R_{i,j}^{(n)}$8, one has, uniformly for $R_{i,j}^{(n)}$9, $X_{i,j}$0, and $X_{i,j}$1 slowly,

$X_{i,j}$2

(Lyu et al., 2022). This is crucial for multiplier bootstrap constructions, because the empirical copula itself is not differentiable (Lyu et al., 2022).

The symmetry-testing paper develops a full multiplier bootstrap Bernstein process. With i.i.d. weights $X_{i,j}$3 satisfying $X_{i,j}$4, one first constructs the Bernstein-smoothed bootstrap process

$X_{i,j}$5

and then corrects for estimated margins by subtracting the partial-derivative terms,

$X_{i,j}$6

Jointly,

$X_{i,j}$7

in $X_{i,j}$8 (Lyu et al., 2022).

Kojadinović and collaborators place empirical Bernstein copulas inside a broader class of smooth, possibly data-adaptive empirical copulas $X_{i,j}$9. Under mild smoothness of the true copula and weak mixing, the smooth sequential process $X_{1,j},\dots,X_{n,j}$0 differs from the classical sequential empirical copula process by $X_{1,j},\dots,X_{n,j}$1 uniformly, so both have the same weak limit (Kojadinovic et al., 2021). The same program yields valid smooth resampling procedures. In the i.i.d. case, a smooth bootstrap based on drawing from the fitted smooth estimator is asymptotically valid (Kojadinovic et al., 2023). In the time-series setting, a smooth extension of the sequential dependent multiplier bootstrap is also asymptotically valid (Kojadinovic et al., 2023).

The same line of work gives a Stute-type representation for empirical Bernstein copula processes. Under first-order smoothness, second-order control, and a variance condition on the smoothing law, one has

$X_{1,j},\dots,X_{n,j}$2

with an almost-sure remainder bound whose leading term is $X_{1,j},\dots,X_{n,j}$3 in the Bernstein case $X_{1,j},\dots,X_{n,j}$4 (Kojadinovic, 2022). This places the empirical Bernstein copula on the same asymptotic footing as the classical empirical copula while retaining finite-sample smoothness.

5. Statistical testing and inferential uses

A major inferential use of empirical Bernstein copulas is hypothesis testing based on smooth empirical processes. For bivariate symmetry, the process

$X_{1,j},\dots,X_{n,j}$5

is used to construct three statistics, and corresponding bootstrap versions

$X_{1,j},\dots,X_{n,j}$6

have the same weak limit (Lyu et al., 2022). The study reports that simulations under Gaussian, Clayton, Gumbel, and Frank copulas at various $X_{1,j},\dots,X_{n,j}$7, with $X_{1,j},\dots,X_{n,j}$8, show that Bernstein-based tests adhere more closely to nominal level, often under-rejection occurs with empirical-copula tests, and power is higher, especially for the sup-norm statistic (Lyu et al., 2022). For implementation, $X_{1,j},\dots,X_{n,j}$9–$C_n$00 multiplier replicates and a $C_n$01 integration grid are reported as sufficient for the double-integral statistics (Lyu et al., 2022).

The same methodology extends to two-sample equality testing. If $C_n$02 and $C_n$03 are empirical Bernstein copulas from two independent samples, then under $C_n$04, the process

$C_n$05

has a Gaussian limit, and three statistics are proposed: an $C_n$06 statistic $C_n$07, a weighted $C_n$08 statistic $C_n$09, and a sup-statistic $C_n$10 (Lyu et al., 2023). Under $C_n$11 these statistics diverge to infinity, guaranteeing consistency (Lyu et al., 2023). The same work studies both multiplier bootstrap and a subsampling Bernstein version, reporting that Bernstein tests outperform tests based on the empirical copula (Lyu et al., 2023).

Inference on functionals is another major theme. Since the smooth estimator shares the same first-order limit law as the empirical copula under standard conditions, Hadamard-differentiable functionals inherit asymptotic normality through the functional delta method (Kojadinovic, 2022). This suggests why the empirical Bernstein copula has been used for dependence measures, confidence intervals, and change-point procedures in the later smooth-copula literature (Kojadinovic et al., 2023).

A recurring practical message is that smoothness is not only cosmetic. It permits direct plug-in estimation of partial derivatives and copula functionals, and it stabilizes resampling procedures that are awkward or unavailable for the nonsmooth empirical copula (Lyu et al., 2022).

6. Applications, extensions, and methodological variants

The empirical Bernstein copula has been adapted well beyond unconditional copula estimation. In tail-focused concordance estimation, Ouimet and Susam define a Bernstein-smoothed lower-tail Spearman’s rho estimator using $C_n$12 and establish strong consistency and asymptotic normality under mild regularity conditions (Ouimet et al., 10 Jun 2025). Their pointwise expansions state

$C_n$13

$C_n$14

so smoothing lowers variance at the cost of an $C_n$15 bias (Ouimet et al., 10 Jun 2025). For lower-tail Spearman’s rho on $C_n$16, the reported Monte Carlo experiment with the Farlie–Gumbel–Morgenstern copula, $C_n$17, $C_n$18, $C_n$19, and $C_n$20 replications shows MSE reductions up to approximately $C_n$21 in deep-tail settings at $C_n$22 under weak to moderate dependence (Ouimet et al., 10 Jun 2025).

In high-dimensional dependence modeling, Weiss and Scheffer use nonparametric Bernstein copulas as pair-copulas in C-vine and D-vine decompositions (Weiß et al., 2012). Their construction starts from pseudo-observations, computes grid counts, and solves a quadratic program to enforce uniform margins and nonnegativity, using the Goldfarb–Idnani dual QP-solver (Weiß et al., 2012). The resulting pair-copulas are smooth and nonparametric, and the algorithms of Aas et al. for density evaluation and sequential simulation carry over once the Bernstein c.d.f.s, densities, and $C_n$23-functions are implemented (Weiß et al., 2012).

In conditional copula estimation, Lu and Ghosh employ the empirical checkerboard Bernstein copula estimator to build a fully nonparametric estimator of conditional copulas without selecting a parametric family (Lu et al., 2023). They derive closed-form estimators of conditional Kendall’s $C_n$24 and Spearman’s $C_n$25, and prove large-sample consistency under absolute continuity of the true three-dimensional copula and Lipschitz continuity of $C_n$26 (Lu et al., 2023). Their algorithm first adjusts the univariate conditional margins through lower-dimensional ECBC fits, then constructs a three-dimensional ECBC and differentiates in the covariate direction (Lu et al., 2023).

A related 2026 development uses empirical checkerboard/Bernstein approximations to estimate conditional distributions via Sklar’s theorem and then to obtain consistent nonparametric estimators of mean, quantile, and expectile regression functions (Schärer et al., 1 Feb 2026). In that setting, with $C_n$27, $C_n$28, uniform conditional convergence of the estimated Markov kernel is established almost surely under the assumption that the true copula allows a continuous Markov kernel (Schärer et al., 1 Feb 2026).

A distinct semiparametric direction is the finite-mixture Bernstein copula estimated by EM. Dou et al. propose EM algorithms for estimating the Bernstein copula weights $C_n$29, prove local linear convergence of the constrained M-step adjustment, and establish $C_n$30-consistency and asymptotic normality of the resulting semiparametric estimator $C_n$31 under regularity conditions (Dou et al., 2013). Although this model is different from the rank-smoothed empirical Bernstein copula, it demonstrates how the Bernstein basis also supports likelihood-based copula estimation.

7. Practical guidance, common misconceptions, and open issues

The choice of the Bernstein degree $C_n$32 is the main tuning problem. Several prescriptions appear in the literature. For symmetry testing, a rule of thumb $C_n$33 with $C_n$34 and $C_n$35 is reported to yield a good bias-variance trade-off, and power stabilizes rapidly in $C_n$36 for Cramér–von Mises-type statistics (Lyu et al., 2022). For lower-tail Spearman’s rho, the recommendation is $C_n$37 (Ouimet et al., 10 Jun 2025). For generic two-sample testing, $C_n$38 or $C_n$39 is stated to work well in practice (Lyu et al., 2023). In generalized smooth-copula frameworks, cross-validation or plug-in formulas are also contemplated (Kojadinovic et al., 2023).

A common misconception is that any Bernstein smoothing automatically yields a genuine copula. Segers, Sibuya, and Tsukahara show that this is false in general: when $C_n$40 does not divide $C_n$41, the empirical Bernstein copula is not a genuine copula and its bias inflates (Segers et al., 2016). By contrast, the empirical beta copula is always a genuine copula (Segers et al., 2016). This explains why later work often emphasizes empirical beta, checkerboard, or checkerboard-Bernstein constructions when exact copula constraints matter.

Another misconception is that smoothing merely introduces bias. The cited works repeatedly stress that smoothing also reduces variance and boundary artifacts. In the symmetry-testing study, Bernstein smoothing is described as asymptotically bias-free at the edges of $C_n$42, in contrast to the empirical copula, which has $C_n$43 jumps (Lyu et al., 2022). In the lower-tail Spearman setting, the variance term is explicitly reduced by $C_n$44 relative to the empirical copula (Ouimet et al., 10 Jun 2025). This suggests that the empirical Bernstein copula should be viewed as a bias-variance trade-off mechanism rather than as a one-sided smoothing correction.

There is also no universal dominance result among Bernstein-type estimators. Segers, Sibuya, and Tsukahara report that the empirical beta copula outperforms both the empirical copula and empirical Bernstein copulas with smaller smoothing degrees in terms of bias and variance, and remains significantly better in several cases, especially in terms of bias (Segers et al., 2016). By contrast, Lu and Ghosh report that two data-adaptive smooth estimators uniformly outperform the empirical beta copula in their Monte Carlo experiments (Kojadinovic et al., 2021). The comparison therefore depends on whether one restricts attention to classical empirical Bernstein smoothers, empirical beta, checkerboard-based smoothers, or broader data-adaptive classes.

Overall, the empirical Bernstein copula occupies a central position in nonparametric copula methodology. It provides a smooth polynomial enlargement of the empirical copula, admits rigorous weak-convergence and resampling theory, and has generated a broad family of descendants and variants, including the empirical beta copula, ECBC, data-adaptive smooth estimators, derivative estimators, and conditional copula procedures (Segers et al., 2016).