Page 393 - Contributed Paper Session (CPS) - Volume 4

CPS2444 Avijit Joarder et al.
in 2008 for Korea and then continued to fall but continued to increase for
India, Indonesia, Malaysia, Philippines and Thailand.
The above mentioned Asian countries except India experienced a rapid
expansion of total international claims on them from foreign banks. Ahead of
the AFC, the short-term international claims, in particular, on these countries
were booming. The most notable example is Thailand, whose share of short-
term international claims reached to an unprecedented level of 71% of total
international claim as of end-December 1993. Although it is not in the case of
Thailand only, the pre-AFC expansions in total international claims including
total short-term international claims on all remaining four countries were also
very sizable. In the case of India, both total international claims and total short-
term international claims followed similar upward trend in pre and post AFC.
Among selected Asian countries, the share of short-term international claims
on India was the lowest during the AFC. In terms of international claims as a
share of GDP, India remains still at the lowest at below 7% and short-term
international claims also are much lower (around 20%) compared to FX
reserves.

6
3.5: Probit regression for prediction of banking crisis
We finally aim at predicting banking crisis (warning signal) based on
short-term international claims as a share of country’s foreign reserves, long-
term international claims as a share of country’s GDP and growth in GDP. We
consider consolidated international claims on a country by non-resident
foreign banks, sourced from the BIS. In our analysis, for six countries (India,
Indonesia, Korea, Malaysia, Philippines and Thailand) of our interest, we use
the start of crisis years from the IMF Working Paper (WP/12/163) . The data
7
on foreign reserves and GDP at current price (USD equivalent) are sourced
from World Economic Outlook (April 2018), all on annual basis.
The results for the probit models are estimated by using lagged variables
to predict crisis over the period of 1983 to 2017. The results of the probit
regression, are in line with the hypothesis of an overheating economy that
grows too fast and therefore collapses during its growth path. Adding the GDP
growth as a control variable does not influence sign or significance of the core
variables of interest, and the probability of 0.0000 associated LR statistics =
1,065.21 rejects the null hypothesis that coefficients of all variables are
simultaneously equal to zero. The GDP growth with one-year lag appears to
be positive and significant showing that a crisis is most likely during an

6 We thank Philip Wooldridge (BIS) for suggesting us to use Probit model and also thank
Maximilian Jager, Centre for Doctoral Studies in Economics, University of Mannheim for his
inside on probit/logit models.
Systemic Banking Crises Database: An update (Table A1 on Banking Crises dates and Costs,
7
1970-2011), 1 June 2012 by Fabian Valencia and Luc Laeven.
382 | I S I W S C 2 0 1 9