Accounting Information Reliability in Credit Evaluation by Financial Institutions

Abstract

This research investigates the critical yet underexplored nexus between the perceived reliability of accounting information and its subsequent utilization in credit
evaluation models employed by financial institutions. Departing from traditional studies that treat accounting data as a homogeneous, objective input, this paper posits
that reliability is a multidimensional, institutionally-constructed perception that significantly alters risk assessment outcomes. We introduce a novel methodological framework, the Reliability-Weighted Credit Evaluation (RWCE) model, which dynamically
adjusts financial ratios and metrics based on a composite reliability score. This score
is derived from a proprietary algorithm analyzing audit quality signals, reporting lag,
industry volatility benchmarks, and textual sentiment in management discussion and
analysis (MDA) sections. Our empirical analysis, conducted via a simulation engine
built on historical datasets from 1998-2004, demonstrates that integrating explicit reliability metrics reduces Type I (false positive) lending errors by an estimated 18.7%
and Type II (false negative) errors by 12.3% compared to conventional models, without a statistically significant increase in model complexity. The findings challenge the
implicit assumption of uniform reliability in financial statement analysis and propose a
paradigm shift towards adaptive, reliability-sensitive credit algorithms. This research
contributes to information economics, behavioral finance, and accounting theory by formalizing the processing of reliability as a distinct, quantifiable variable in automated
financial decision-making systems.