Revisiting Conventional Models of Electoral Prediction: Insights from Nigeria's 2023 Presidential Election

Abstract

Although electoral forecasting has received considerable scholarly attention, research systematically reassessing conventional prediction methods in relation to Nigeria’s electoral volatility, social media dynamics, and ethno-regional complexity remains limited. This study revisits opinion polls, historical voting patterns, and expert assessments in Nigeria’s 2023 presidential election and examines how social media sentiment can complement these approaches. A qualitative case study design was employed, drawing on eleven major pre-election polls, official results from the Independent National Electoral Commission, secondary sentiment-analysis datasets, and documentary sources. Comparative and thematic analyses were used to examine forecasting performance and identify methodological and contextual explanations for observed discrepancies. Conventional methods performed unevenly, with a reported failure rate of 63.6% among the eleven polling organisations reviewed. The analysis linked these shortcomings primarily to urban-centred sampling, inadequate turnout modelling, and insufficient attention to ethno-regional and religious cleavages. Approaches incorporating historical voting behaviour, turnout scenarios, and structural variables demonstrated greater predictive accuracy. Twitter sentiment provided timely indications of electoral competitiveness, although its contribution depended on recognising the limitations of social media data for representing the wider electorate. These findings support the development of hybrid forecasting frameworks that combine representative polling, historical and structural information, turnout modelling, and contextual interpretation of digital sentiment. The study highlights the importance of adapting forecasting methods to Nigeria’s social and institutional conditions. Its contributions include informing the integration of rational choice, socioeconomic, and institutional perspectives on voting behaviour and identifying methodological priorities for improving electoral information. The findings also have implications for public understanding of campaign forecasts and debates concerning electoral integrity. Further research should examine multi-platform social media analytics and compare forecasting performance across African electoral contexts.