A simple SUR-based approach to handling missing data

What the paper says

Purpose This study aims to explore practical strategies for addressing missing competitor data to support data-driven decision-making in the North American automotive industry. Design/methodology/approach The authors analyze 11 years of monthly sales data from GM, Ford, Honda, Toyota and Nissan to develop an approach for data-driven decisions in the presence of missing competitor information. The authors use autoregressive time series regressions to model four scenarios: full access to competitor information, substituting competitors’ missing data using historical marginal rates of substitution (MRS), a direct variable replacement without MRS and ignoring missing data entirely. Findings Results indicate that both the MRS-based approach (Method 2) and direct variable replacement (Method 3) offer predictive accuracy comparable to using full data, with minimal loss in explanatory power. While both methods are likely sufficient, due to the simplicity and time savings associated with direct variable replacement, the authors suggest using Method 3. In contrast, ignoring missing data significantly degrades model performance. These findings suggest that managers can maintain decision quality even with partial data by using simple substitution strategies. Practical implications While the proposed framework is validated using data from the North American automotive industry, the findings suggest its applicability to other competitive sectors. Originality/value By evaluating the above four distinct scenarios, we offer the novel application of marginal rates of substitution and the direct variable replacement strategy for replacing missing data, proving that decision-making could remain feasible even in the absence of complete information.