THE DETERMINANTS OF SOVEREIGN CREDIT RATINGS: INDONESIA AND ITS NEIGHBORHOOD COUNTRIES 1998-2016

The aim of this research is to study the determinants of sovereign credit ratings of Indonesia and its neighborhood countries in the period of 1998-2016. Using secondary data and analyzed using ordered probit, it is found that every credit rating agency has its own variables influencing to its published credit ratings.In general, for Indonesia and its neighborhood countries, the variables with significant and positive relationship are fiscal balance and current account deficit to GDP, freedom index, and GDP per capita; while the variables with significant and negative relationship are external debt to GNI and real exchange rate. Gross domestic savings to GDP influences credit ratings in both ways. Interestingly, inflation does not affect the credit ratings. Indonesia and neighborhood governments could use this information to manage their macroeconomic indicators in order to get favorable ratings from credit rating agencies.


INTRODUCTION
Sovereign credit ratings are very important due to globalization of market and cross border investments. Sovereign credit ratings are not country ratings, but they address the credit risk of national government, do not address specific default risk of other issues (Beers & Cavanaugh, 1998). These ratings give insight into investing risk and political risk of a particular country.
Sovereign credit ratings affect economy of a country in terms of cost of debt and foreign direct investment. For example, OECD countries received high foreign direct investment (FDI) when their credit ratings were high (Cai, Gan, & Kim, 2018).Turkey observed two ways causality between sovereign credit ratings and FDI during 1995-2013 (Bayar & Kilic, 2014).
The downgrade of sovereign credit ratings would lead to reduction of investments and reliance of credit market due to rising cost of debt (Almeida, Cunha, Ferreira, & Restrepo, 2017). Furthermore, the bond yield of firms is found to increase significantly due to the downgrades. For short term government borrowing cost, a downgrade to subinvestment grade by one major rating agency increase Treasury bill yields by 138 basis points, on average (Hanusch, Hassan, Algu, Soobyah, & Kranz, 2016).
Research on determinants of sovereign credit ratings still attracts many studies up to now. One of the reasons might be credit rating agencies do not offer transparent criteria to determine ratings and their changes (Mora, 2006). Thus, many studies have been conducted in various countries in different periods. As shown in Table 1, the results of several previous studies on determinants of sovereign credit ratings are still ambiguous and inconsistent in different country studies.
The inconsistent results of previous research on different countries as exemplified in  Cantor and Packer (1996) and also Kabadayi and Çelik (2015) were used. The complete rating symbols and ordinal scale were shown in Table 2. Source: Cantor & Packer (1996); Kabadayi & Çelik (2015) There are seven independent variables used in this study, i.e. (1) (1) where SR stands for sovereign credit rating, j represents three different sovereign credit ratings (i.e. S&P, Moody's, and Fitch), i symbolizes country, and t denotes time. Thus, there are three empirical models, one for each credit rating agency to predict probability of getting certain sovereign credit ratings.
In addition, significance tests in terms of likelihood ratio statistic, Z-statistic, and pseudo-R 2 tests (Gujarati & Porter, 2009) were implemented to inquire the robustness of independent variables in explaining the dependent variable. Ordered probit model coefficient differ by a scale factor, therefore the magnitude of the coefficient cannot be interpreted directly.
The interpretation of coefficient is conducted through marginal effect test (Gujarati & Porter, 2009). Marginal effect is a measure of the instantaneous effect of a change in a particular explanatory variable on the predicted probability variable. The marginal effect tests were conducted for each rating level of each rating agency.

Regarding variable of external debt to GNI (EXDGNI), the variable is significant in
Moody's and Fitch models. The variable has negative coefficient as expected. This result is in accordance with previous studies (Cantor & Packer, 1996;Kabadayi & Çelik, 2015;Melki, Ftiti, & Ben Arab, 2017;Mellios & Paget-Blanc, 2006). Arefjevs and Braslins (2013) using slightly different measure, i.e. external debt to export ratio, also found that the variable has negative sign.Therefore, to increase credit ratings, especially from Moody's and Fitch, Indonesia and its neighborhood countries should reduce their external debt to GNI.
For freedom index, the variable is significant in all models and has positive sign as expected. The result is in line with Kabadayi and Çelik (2015). Calcagno and Benefield (2013)and Belasen, Hafer, and Jategaonkar (2015)also found positive relationship between economic freedom and bond ratings in 39 states and 50 states, respectively. Thus, countries with high economic freedom enjoy favorable bond ratings and pay lower borrowing costs. Similar situation may also be able to be inferred for sovereign credit ratings.

Regarding GDP per capita (GDPPC)
and sovereign credit rating, the variable is significant in all three models. It has positive sign as expected. The result confirms previous studies (Kabadayi & Çelik, 2015;Melki et al., 2017).
The sixth independent variable, i.e. real exchange rate (REXR) is significant in S&P and Moody's models. It shows negative signs, as expected, meaning that exchange rate depreciation leads to worse sovereign credit ratings. It is in line with Chodnicka (2015) and Kabadayi and Çelik (2015) but is contrary to Mellios and Paget-Blanc (2006). Therefore, a country in the sample should maintain its real exchange rate to get favorable sovereign credit ratings.
The last independent variable, i.e. gross domestic saving to GDP (SAVGDP) is significant in S&P and Fitch models. However, the coefficient has negative sign for S&P but positive sign for Fitch model. The positive sign is reported in Kabadayi and Çelik (2015), while positive and negative signs are also observed in literature (Chodnicka, 2015;Mellios & Paget-Blanc, 2006).

Possible explanation could be difference in
the level of economic development among sample. Chodnicka (2015) found that middle economic countriesin Europe have positive signs while low economic countries have negative signs.
As for marginal effect,   In Table 8 for Fitch Model, 1% increase in fiscal balance and current account will have higher probability to get rating "strong payment capacity" (SR=4) by 0.0120 point.
In addition, 1% increase in freedom index, saving to GDP, and GDP per capita will increase probability of getting SR=3; while 1% increase in external debt will decrease probability of getting SR=3 rating or increase probability of getting SR=2 rating. Focusing on column SR=3, the highest marginal effect is saving to GDP (SAVGDP).  (Kalloub et al., 2018). Researcher may also use other dependent variables, such as state bond ratings (Belasen et al., 2015;Calcagno & Benefield, 2013) or spread between state bonds and risk free bonds (Pačebutaitė, 2011).