Using Israel-Darmois junction conditions to match a Vaidya exterior to a Friedmann-Robertson-Walker interior, we derive H = Ṁ/M: the Hubble parameter equals the fractional mass growth rate of the bounding horizon. For power-law growth Ṁ = κMα, the effective equation of state is weff = −1/3 − 2α/3. Constant fractional growth (α = 1) yields w = −1 exactly, reproducing the ΛCDM expansion history from boundary geometry without introducing a dark-energy fluid. Joint analysis of DESI DR2 BAO, Pantheon+ supernovae, and Planck CMB priors returns α = 0.968 ± 0.034, consistent with this prediction to within 3%. With the ΛCDM baseline locked (α ≡ 1), the data identify a localized growth-rate enhancement at zc = 0.65 +0.12/−0.06 with Planck-consistent background cosmology (H0 = 67.7 ± 0.7, Ωm = 0.307 ± 0.006), preferred over ΛCDM at Δχ² = 11.5 (p = 0.003, 2.9σ for 2 dof). The feature appears at consistent redshift and amplitude across three independent supernova compilations and in BAO+CMB data without supernovae. At equal parameter count with w0wa (5 vs 5), the Gaussian model fits Pantheon+ better (Δχ² = 11.2 vs 9.8 over ΛCDM) with 10:1 Bayesian preference. The framework predicts |Δw| > 0.2 divergence from w0wa at z > 1.5, testable with forthcoming surveys.