Emotion-driven spending is a growing research frontier: it can lift consumption quality when constrained but, left unchecked, damages health, heightens household financial vulnerability and amplifies short-term market and credit risk; researchers should link long-term administrative and digital-trace data and apply causal machine-learning to measure algorithmic impacts and heterogeneity.

Main Finding

A systematic review of 197 studies (2000–May 2026) shows emotion-driven consumption is a growing, interdisciplinary research area. Its formation reflects interactions among psychological/behavioral traits, household attributes, and external stimuli. While emotion-driven consumption can support consumption upgrading when moderated, unchecked emotional spending harms health, increases household financial vulnerability, and raises short-term market volatility and consumer credit risk. Empirically, the literature is concentrated on micro-level individual data and econometric methods and needs richer long-term, multidimensional data and stronger methodological integration (e.g., natural experiments combined with machine learning).

Key Points

• Scope & growth: 197 high-quality studies reviewed; steady increase in publications, with rising international collaboration and interdisciplinary integration.
• Determinants: Emotion-driven consumption emerges from psychological/behavioral characteristics, individual and household attributes, and environmental stimuli (e.g., marketing, digital platforms).
• Effects:
  • Conditional benefit: Under moderate, rational constraints it can promote consumption upgrading.
  • Downsides: Without constraints, it can damage physical/mental health, worsen household financial vulnerability, and elevate short-term market volatility and consumer credit risk.
• Empirical practice: Predominant reliance on micro-level individual data and traditional econometric models.
• Gaps & recommendations: Need for interdisciplinary theoretical frameworks, long-term multidimensional datasets, and methodological advances combining natural experiments with machine learning.

Data & Methods

• Data used in reviewed studies: Mostly micro-level individual datasets (surveys, experiments, lab studies, panel microdata); fewer long-term or multidimensional databases.
• Empirical methods: Econometric models dominate (regressions, panel models, structural estimation). Experimental and behavioral methods are present but less widespread at scale.
• Methodological limitations identified: limited longitudinal coverage, sparse use of large-scale digital trace data, and underutilization of machine learning for prediction or heterogeneity exploration.
• Suggested methodological advances: build long-term, multidimensional databases (linking behavioral, financial, health, and digital platform data); combine natural/quasi-experiments for causal identification with machine learning methods for high-dimensional pattern discovery and prediction.

Implications for AI Economics

• Opportunity for AI methods:
  • Use ML to analyze large-scale digital trace and transaction data to detect emotion-driven consumption patterns and heterogeneity that traditional econometrics may miss.
  • Apply causal ML and double-robust methods to combine predictive power with causal inference (e.g., targeted policy evaluation, heterogeneous treatment effects).
• Data infrastructure:
  • Construct integrated, longitudinal datasets linking consumption, credit records, health outcomes, and platform interaction logs to study long-run welfare and financial stability implications.
  • Privacy-preserving ML techniques (federated learning, differential privacy) can enable cross-institutional research while protecting individuals.
• Policy and market design:
  • AI-driven platforms can both exacerbate and mitigate emotion-driven spending—recommendation systems and targeted ads may intensify impulses; conversely, AI-enabled nudges and real-time budgeting tools could restrain harmful spending.
  • Regulators and firms should monitor algorithmic impacts on consumer credit risk and short-term market volatility; stress-testing with agent-based simulations and ML-based scenario analysis can inform regulation.
• Research agenda for AI economists:
  • Integrate behavioral theory with computational methods to model emotion-influenced decision dynamics at scale.
  • Use natural experiments (policy changes, platform algorithm shifts) plus ML to estimate causal effects of algorithmic interventions on consumer welfare and financial stability.
  • Evaluate long-term welfare tradeoffs of personalization algorithms that exploit emotional states versus interventions designed to protect vulnerable households.

If you want, I can (a) expand any section with citations and examples from the reviewed studies, (b) propose a concrete data schema for a multidimensional longitudinal database, or (c) outline ML + causal-econometric workflows suited to this research agenda.