This research proposes a novel semiparametric regression framework, integrating truncated spline and wavelet estimators in the modeling of dynamic relationships among CG, IC, and institutional performance in Bali's Village Credit Institutions (LPDs) in Indonesia. The robustness of this estimator is initially tested by a Monte Carlo simulation under various conditions of nonstationarity. From these, it becomes quite clear that the proposed hybrid spline-wavelet model has produced the least MSE (MSE = 0.0076) and greatest coefficient of determination (R² = 0.968). However, the specific penalized estimation method used ensures an appropriate bias variance tradeoff, allowing the proper modeling of global smooth trends and local short term variations. Latent CG and IC constructs obtained through Partial Least Squares Structural Equation Modeling were applied as covariates to the hybrid regression model by using a longitudinal database from 86 LPDs over the period 2016-2023. From the empirical findings, it was manifested that CG and IC significantly influence institutional performance and account for as much as 78% of its variation. The time varying component depicted three phases: reform growth from 2016 to 2019, the pandemic contraction in 2020-2021, and post recovery stabilization in 2022-2023. In general, this hybrid spline-wavelet estimator showed superior precision, decreasing MSE by up to 31% compared to single basis models, and provided a novel methodological contribution to nonstationary financial and econometric modeling.