Betting search salience and conditional SEIR scenarios in Ghana: a time-series public-proxy analysis

Understanding temporal changes in betting-related behavior is important for characterizing digital engagement with gambling in Ghana, yet direct monthly measures of population-level betting participation remain unavailable. This study therefore investigates betting-search salience using a preserved 72-month Google Trends composite spanning January 2020 to December 2025 and complements the empirical analysis with conditional SEIR-type behavioral scenarios. Six parsimonious autoregressive specifications were evaluated using an expanding validation period covering 2023–2024, with model selection fixed before an independent 2025 test period. The selected ARX2 specification achieved an in-sample ( R 2 = 0.690) and a validation RMSE of 0.1031 on the logarithmic scale. However, its 2025 RMSE of 0.1276 exceeded that of a no-change benchmark (0.1202), indicating limited out-of-sample predictive improvement and evidence of temporal instability. Model coefficients are therefore interpreted as conditional associations with search salience rather than estimates of betting prevalence or causal behavioral effects. The scenario analysis treats behavioral transition rates, initial state proportions, and the bounded mapping from fitted search salience to contact-rate trajectories as explicit assumptions rather than estimated Ghanaian quantities. Across most sensitivity configurations, enhanced-exit and combined scenarios reduced peak and cumulative active occupancy, although the magnitude of these effects varied substantially with assumed transition rates and initial conditions. The findings provide a cautious characterization of betting-search dynamics and their predictive limitations while avoiding unsupported inference about population-level betting prevalence or causal policy effects. More broadly, the study identifies the behavioral measurements required to move from digital search signals toward empirically calibrated latent-state models and quantitative evaluation of intervention effects in Ghana.