Детерминанты предлагаемой заработной платы в ИТ-секторе России: данные интернет-рекрутинга и ансамблевые модели

Вопросы статистики. 2026. № 5.

Okuskov I.

This article analyses the determinants of offered wages in Russia’s IT sector using internet recruitment data. The aim of the study is to assess the predictive power of the structural characteristics of job postings and to rank their contribution using non-linear ensemble machine learning methods. The empirical basis is a dataset of job postings for 2025 retrieved through the open HeadHunter API. This source compares favourably with traditional sources of vacancy data in its higher update frequency and more detailed occupational segmentation.

The influence of five groups of factors is examined: work experience, professional role, regional group, work schedule and type of employment. The study tests hypotheses on the dominant role of experience, the advantage of a non-linear ensemble over linear regression, and the weakening of regional differentiation under the influence of remote employment. A weighted ensemble of models (Random Forest, Extra Trees, Gradient Boosting) is proposed, in which the weights of the components are proportional to their predictive accuracy. Linear regression is rejected by diagnostic tests and is used as a reference.

The results show that non-linear models outperform linear regression in terms of the coefficient of determination. Work experience is found to be the key factor determining the level of pay, while the influence of work schedule and type of employment is minimal. Remote work only partially smooths regional wage differences.

The study confirms that internet recruitment data are an effective instrument for the timely monitoring of the labour market, complementing official statistics. A calibration adjustment is proposed for harmonising these sources.

Keywords: рынок труда, wages, заработная плата, региональная дифференциация, regional differentiation, ИТ-сектор, IT sector, Ансамблевые методы машинного обучения, labour market, internet recruitment, ensemble machine learning, интернет-рекрутинг