AI activity lifts demand for China’s education and training market primarily in bullish conditions, while policy uncertainty tends to depress it during calmer periods. Risk spillovers from AI to training are larger and more volatile than those from policy uncertainty, implying policymakers should balance AI-driven opportunity with market stability.

Main Finding

Using quantile-based, nonlinear time-series methods, the paper finds that AI activity and global economic policy uncertainty (EPU) jointly and asymmetrically affect China’s education and training market (ETM). There are unidirectional causal links (AI → EPU, AI → ETM, EPU → ETM). AI has a positive predictive effect on ETM concentrated in bullish market states, while EPU has a negative predictive effect concentrated in more stable policy periods. Risk spillovers are asymmetric: upside/extreme risk transmission (especially from AI to ETM) is much stronger and more volatile than downside transmission, whereas EPU’s effects are milder but more persistent.

Key Points

• Nonparametric quantile causality tests detect unidirectional causal relationships:
  • AI → EPU
  • AI → ETM
  • EPU → ETM
• Cross-quantilogram analysis reveals quantile-dependent predictive links:
  • AI positively predicts ETM mainly in upper-tail (bullish) states.
  • EPU negatively predicts ETM mainly in periods characterized by policy stability (non-crisis states).
  • Interaction: AI tends to raise EPU in bullish markets; EPU tends to stimulate AI activity during stable economic conditions.
• GARCH–conditional quantile regression exposes asymmetric risk spillovers:
  • Upside (extreme positive) spillovers are much larger than downside ones.
  • AI → ETM spillovers show high volatility and fat-tail/extreme behavior.
  • EPU’s impacts are lower in instantaneous intensity but more persistent over time.
• Practical conclusion: effective ETM policy must jointly consider AI dynamics and EPU, with attention to state-dependent (quantile) effects and asymmetric risk.

Data & Methods

• Data (described at a conceptual level): time-series indicators for China’s AI activity (e.g., AI investment/usage or an AI index), global or China-specific EPU index, and measures of the education & training market (ETM) — applied at a frequency suitable for time-series analysis.
• Main empirical tools:
  • Nonparametric quantile causality test: assesses directional predictive causality across different quantiles of the conditional distribution, allowing detection of state-dependent causal links that vary across market conditions.
  • Cross-quantilogram: measures lead–lag dependence across quantiles of two series to identify which parts of the distribution (e.g., tails vs center) drive predictability and interaction.
  • GARCH–conditional quantile regression: models time-varying volatility and estimates conditional quantiles to examine asymmetric tail risk spillovers and the extremeness of shocks (separating upside vs downside spillovers).
• What these methods enable: identification of nonlinear, state-dependent causality and asymmetric tail transmission that standard mean-based or linear models would miss. (Paper likely includes robustness checks across quantile ranges and lag structures.)

Implications for AI Economics

• Policy design:
  • Adopt state-contingent (quantile-aware) policies: interventions should differ by market state (e.g., support AI-driven ETM uptake in downturns; dampen overheating in bull states).
  • Monitor AI activity as a potential amplifier of EPU in bullish episodes and be prepared with macroprudential or information-stabilizing measures.
  • Because EPU effects are persistent, combine short-term stabilization with longer-term measures (e.g., regulatory clarity, predictable policy signaling) to reduce drag on ETM.
• For education & training providers:
  • Prepare for high-volatility, tail-driven changes from AI shocks (inventory, capacity, pricing strategies should account for extreme positive swings as well as persistent policy-driven demand shifts).
  • Use AI strategically to capture upside demand in bullish periods, while building resilience to policy-driven slowdowns.
• For researchers and forecasters:
  • Models of technology–market interaction should incorporate quantile-dependent links and asymmetric tail risk (GARCH-quantile or cross-quantile tools).
  • Microdata and regional/sectoral heterogeneity analysis would help unpack channels (e.g., which segments of ETM are most exposed to AI-driven extreme swings).
• For investors and risk managers:
  • Tail-aware risk management is crucial: leverage stress tests that incorporate asymmetric upside risk from AI and persistent downside pressure from elevated EPU.
  • Pay attention to leading quantile signals (e.g., upper-tail AI indicators) as early predictors of ETM booms.

Potential next steps: refine causal identification (instrumental or natural-experiment approaches), disaggregate ETM by subsector and region, and evaluate policy interventions tailored to quantile-specific dynamics.