Abstract
This study develops an early warning signal (EWS) of government debt crisis using a panel data consisting of 43 developing countries over the period of 1960 to 2017. It employs two different methods: the noise to signal ratio to capture the signaling power of individual indicators; and the binomial logistic regression to construct a more general model. The binomial logistic regression offers a better predictive power relative to the noise to signal ratio. The binomial logistic regression can predict 61.5% of the government debt crisis 2 years in advance. An increase in inflation, government and private debt exposures, external debt to exports, ratio of short-term external debt to foreign exchange reserves, and the ratio of external interest payments to gross national income can signal an upcoming debt crisis. Similarly, a continuous decline in the gross domestic product (GDP) and government consumption also increase the likelihood of government debt crisis.