Behavioural early warning indicators of household sector vulnerability

Working paper 945

Abstract

Conventional macroprudential surveillance relies primarily on administrative data that are published with a lag, potentially delaying the identification of emerging financial vulnerabilities. This paper examines whether high-frequency online search behaviour provides additional early warning information for banking, credit and housing cycles in South Africa. Three behavioural early warning indicators from Google Trends data are constructed, capturing housing and mortgage demand, formal financial distress and informal credit substitution.  Their predictive content is evaluated using binary logistic regression models across multiple forecast horizons. The results show that behavioural indicators provide statistically significant incremental predictive information beyond macro-financial variables, although its performance differs across transmission channels. Informal credit substitution provides the strongest signals of banking fragility and household credit cycles, while housing and mortgage demand performs best in predicting house price booms. Formal financial distress performs best in predicting mortgage credit booms and longer-horizon house price busts.  Overall, the findings suggest that behavioural search data can support existing macroprudential surveillance by providing timely and low-cost complementary information on evolving household financial conditions, particularly in emerging market economies where information on informal credit activity is limited.