Scholarlink

<p>This study develops a two-step machine learning-based framework to forecast the likelihood and severity of recessions in India: The first step predicts the probability of a recession using binary classification, while the second step, conditional on a predicted recession, classifies its severity into three categories using multiclass classification. A diverse set of machine learning models has been deployed across multiple forecast horizons (3, 6, 9, and 12 months). Key macroeconomic indicators, including the 10–year–1–year government bond yield spread, 1-month weighted call money rate, WPI fuel inflation, and non-oil export, have emerged as consistent and significant predictors of the business cycle in India. Further, the Sub-sample analysis confirms that the inclusion of the Purchasing Managers’ Index (PMI) enhances forecasting performance, particularly in predicting the severity of recession in 6-, 9-, and 12- months ahead horizons. Robustness checks using alternative severity categorizations confirm the stability of the model. Machine learning-based classification indicated that India is currently experiencing a recessionary phase, and the recession is likely to persist in the near future but remain mild, with an expected severity range between + 2% and − 5%. The findings highlight the effectiveness of machine learning in providing early and reliable insights into economic downturns, thereby supporting timely and targeted policy interventions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting Likelihood and Severity of Recessions in India: A Machine Learning Approach

  • 期刊论文

  • 发表日期：

    2026年3月4日

查看全文
  • Nandhini Sridhar,
  • Saumitra Bhaduri

摘要

This study develops a two-step machine learning-based framework to forecast the likelihood and severity of recessions in India: The first step predicts the probability of a recession using binary classification, while the second step, conditional on a predicted recession, classifies its severity into three categories using multiclass classification. A diverse set of machine learning models has been deployed across multiple forecast horizons (3, 6, 9, and 12 months). Key macroeconomic indicators, including the 10–year–1–year government bond yield spread, 1-month weighted call money rate, WPI fuel inflation, and non-oil export, have emerged as consistent and significant predictors of the business cycle in India. Further, the Sub-sample analysis confirms that the inclusion of the Purchasing Managers’ Index (PMI) enhances forecasting performance, particularly in predicting the severity of recession in 6-, 9-, and 12- months ahead horizons. Robustness checks using alternative severity categorizations confirm the stability of the model. Machine learning-based classification indicated that India is currently experiencing a recessionary phase, and the recession is likely to persist in the near future but remain mild, with an expected severity range between + 2% and − 5%. The findings highlight the effectiveness of machine learning in providing early and reliable insights into economic downturns, thereby supporting timely and targeted policy interventions.