Перспективы медиа-мониторинга в исследованиях общественного мнения (на примере доверия президенту)

Социология: методология, методы, математическое моделирование. 2025. № 61. С. 165–203.

Ankudinov I.

The changing political mood of Russians is a constant subject of interest for sociological agencies. With the development of the Internet, conventional questionnaire research began to be supplemented by online surveys and, despite some skepticism, by social media mining. This article attempts to adjust an accidental web-sample so as to bring its estimates closer to representative omnibuses. We use media- and polling measures of trust in the president Putin to answer the questions: How far apart are these sources? Is it possible to bring their estimates closer using post- stratification weights? Has the relationship between them been disrupted by the onset of the Ukrainian conflict? Is this relationship – despite some absolute difference – stable over time, i.e. Is it possible to move from one indicator to another using linear transformations? The conclusions we reach based on the results of the statistical analysis are as follows: (1) Internet measurements based on average dictionary sentiment estimate trust in the president by 20 percentage points lower than surveys. (2) Adjusting the user sample for weights from the population census does not allow the estimates to be reconciled. (3) Clarification of estimates for separate networks and pollsters only confirms the detected difference. For some platforms this difference is more significant, which can be explained by the specifics of their audiences. (4) The “Ukrainian factor” had an ambiguous impact on the discrepancy between estimates. Tonality according to “Medialogia” records an increase in the gap, while Dostoevsky sees almost no difference between “before” and “after”. (5) The temporal dependence of offline and online estimates is no less ambiguous. Weighting the Internet sample gives some reason to consider the corresponding time series to be cointegrated, but this relationship does not appear in all tests.

Keywords: социальные сети, временные ряды, анализ текстов, text analysis, взвешивание, opinion mining, time series, политическое доверие, social media, public trust, pollsters, weighting, presidential rating, поллстеры, анализ настроений, президентский рейтинг