Comparative analysis of data sources for back-testing cost estimation models

The article examines modern approaches to asset valuation with a focus on the selection, classification and comparability of data sources used for model testing and validation. It proposes a conceptual framework spanning both observable market indicators (quotes for liquid public securities, aggregated market measures) and expert-based inputs (independent appraisal reports, buy-side/sell-side assessments), as well as different value notions market, investment and liquidation value. The paper introduces the Asset Valuation Act (AVA) as a universal unit that records the outcome of a valuation action at a specific date, enabling heterogeneous sources to be captured within a single dataset for backtesting. A taxonomy of AVA sources is outlined along the dimensions of origin, observability, update frequency, methodological transparency, and bias risk, which facilitates cross-study comparability and practical implementation. The analysis shows that no single source is sufficient: market data provide timeliness and objectivity in liquid markets but may be distorted under illiquidity and stress, while expert assessments are indispensable for unique and non-traded assets yet call for stricter controls on independence, data quality, and reproducibility. The paper argues for standardization of backtesting procedures–including AVA metadata requirements, sampling and aggregation rules, and common criteria for accuracy and robustness–and for the development of formal reliability criteria tailored to different AVA types. Implementing the proposed framework improves transparency, comparability, and reproducibility of valuation research, supports model calibration, and reduces subjective assumptions in decision-making.