Page 60 - Special Topic Session (STS) - Volume 2

STS459 Gan C.P. et al.
probability. Simons and Rolwes (2012) however found that GDP growth is
negatively related to probability of default while interest rate and exchange
rate are positively related to the probability of default in certain sectors.
Castrén, Dées, and Zaher, (2008), Wong, Choi and Fong (2006), Otani,
Shiratsuka, Tsurui and Yamada (2009), Avouyi-Dovi, Mireille, Jardet, Kendaoui
and Moquet (2009), Jakubik and Schmieder (2008), Küçüközmen and Yüksel
(2006), Atlintas (2012), and Figlewki, Frydman and Liang (2012) also
investigated the dependence of credit risk on some selected macroeconomic
variables. Rinaldi and Sanchis-Arellano (2006) reported that inflation rate, the
ratio of financial assets to disposable income, disposable income itself, the
ratio of household debt to household disposable income, and real lending
interest rates were important variables affecting non-performing loans.
Yurdakul (2014) reported that growth rate and ISE index reduce banks’ credit
risk, while money supply, foreign exchange rate, unemployment rate, inflation
rate, and interest rate increase banks’ credit risk.
The layout of this paper is as follows. In Section 2, the method based on
multivariate power-normal (MPN) distribution (Pooi, 2012) is used to fit the
data on credit ratings and macroeconomic variables. Section 3 presents the
results on the effects of macroeconomic variables on the credit rating
transition probabilities. Section 4 concludes this paper.

2. Method based on Multivariate Power Normal Distribution
The index of the -th company in a group of N companies may be
represented by the N – 1 binary codes 0 010 0 in which the value “1” takes
the -th position for 1 ≤ ≤ − 1 , while the N-th company may be
represented by N – 1 zeros. In the case when the number of possible credit
ratings is M, the rating j of a company may be represented by the M – 1 binary
codes 0 ⋯ 0 1 0 ⋯ 0 in which the value “1” takes the j-th position for 1 ≤ ≤
− 1, while the rating M may be represented by M – 1 zeros.
Consider a vector of N    3 M   components consisting of the value
 1

of a particular selected macroeconomic variable in the present quarter and the
codes for the index of a company together with those for the company’s
ratings in the previous, present and next quarters. We note that the use of
ratings in the previous quarter will enable us to form a non-Markovian model
(Gan et. al, 2017). From the credit rating data of N companies over quarters,

a table of × ( − 2) rows is formed with each row representing a value for

. The table can be partitioned into sub-tables with the -th (1 ≤ ≤ )

sub-table representing the values of derived from the -th company. Let

n be a positive integer which is less than ( − 2) . From the -th sub-table,

w
we form the -th sub-window using the first until the ( − 1 + )-th row

49 | I S I W S C 2 0 1 9