Abstract

The widespread availability of free services offered by AI providers complicates the measurement of the value generated for users. Often, AI is substituting for market services (legal advice, tutoring, translation, basic medical triage or therapy) and economic activity moves from inside the production boundary to outside it, reducing measured GDP while raising welfare. AI is touching all of the major ways in which humans spend their time, including leisure, household management, education, personal care, and travel. This suggests that free digital products generate value for consumers that is at best partially captured in GDP.

The revised System of National Accounts 2025 recognizes the importance of unpaid household activities in providing a more complete picture of economic performance and growth. Two broad measurement approaches have emerged to gauge the value to households of their digitally enabled production activities. The first is estimating consumer surplus through willingness-to-accept or willingness-to-pay measures, typically via choice experiments. The second is capturing the time savings enabled by AI: by reducing the need for human intervention in routine or repetitive tasks, AI systems can significantly lower the time input required for both market and non-market activities.

Time Use Surveys are an important statistical tool for implementing either of these methodologies, providing an objective means of eliciting time use in household production and time savings associated with AI. We consider the scope for using Time Use Surveys to support the measurement of AI-enabled household production of services and the productivity and employment effects of AI. We suggest several areas for methodological development.