Mentor:
Dr. S.K. Mathur
Assistant Professor of Economics
Department of Humanities and Social Sciences
Indian Institute of Technology, Kanpur
Submitted by:
Suyash Baderiya(Y9616)
Undergraduate Student
Department of Humanities and Social Sciences
Indian Institute of Technology, Kanpur
Abstract
This paper aims to develop a multiple regression model for country risk
assessment using its economic and political factor and analyze the effect of
changes in these factors on risk ratings. The ratings are as per Standard &
Poors sovereign foreign currency sovereign rating. The analysis is based
on data for 28 Asian countries provided by some international organizations
as well as data from various other trusted organizations.
Introduction
Currently, country risk is considered as the probability that a country will
fail to generate enough foreign exchange in order to pay its obligation
toward the foreign creditors (Kosmidou et al., 2008, p.1). In the present
scenario, where recession and debt has caused many governments to
default, the factors that influence the probability of default by a country and
the impact they have on it are issue of major concern.
Many institutions, like Standard & Poors, Moodys, etc. have developed
various methods to assess country risk. These estimates are presented in
a form of ratings, which are indicator of probability of default by the country.
For our analysis, we are going to use Standard & Poors rating. The ratings
by Standard and Poors are based on the information provided by the
debtors themselves and a few other sources. This analysis is mostly based
on the past financial history of the country.
This paper is an extension of the paper by Raluca, Zizi in 2011. In this
paper, we attempt to develop a multiple regression model for assessing
country risk and analyze the effect of change in economic and political
factors on a countrys rating. In our analysis, apart from economic
variables, we have also included political variables because in the current
scenario various political factors play an important role in economic
decision making and thus they have great influence on the economic
condition. The factors in our model are chosen on the basis of significance,
availability and uniformity of data.
Review of Literature
Most of the empirical research in this field suggests that there is a direct
effect of economic factors on risk rating. A recent research by Raluca,
Zizi(2011) also explained the effect of a political factor, i.e., corruption on a
countrys rating. This paper is an extension of the above paper where we
have included another political factor and removed a few factors, which we
found to be highly correlated or data was not available for the countries in
our analysis.
the
this
the
the
Methodology
In our analysis, we will use ordered logit, which is as follows:
Y* = X + u
where,
Y* is latent variable which cant be observed
X denotes matrix of explanatory variables
Y=0, if Y*< v1
1, if v1 < Y*< v2
2, if Y* > v2
where vi s are the cut off points which we get after regression and
v1 < v 2 < v 3
Before using regression, first we will eliminate those variables which can
cause multicollinearity. For this purpose, we will use the Variance Inflation
Factor method. In this method, we apply OLS procedure and eliminate
those variables whose VIF comes out to be greater than 5.
VIF i=
where
1
1R2i
2
Ri
is determined by regressing
Xi
Empirical Analysis
For all the analysis below, we are using Stata SE.
Before applying ordered logit regression, we need to remove those
variables which can cause multicollinearity for which we will be using VIF
method. Stata output after OLS and then finding VIF was as follows:
. vif
Variable
VIF
1/VIF
GDP
infl
cpi
er
tb
fb
col
fd
ir
eg
7.39
4.96
4.24
3.57
2.34
2.11
1.80
1.73
1.45
1.34
0.135288
0.201652
0.236097
0.279893
0.427647
0.473261
0.557093
0.577243
0.687579
0.743683
Mean VIF
3.09
Here, we can see that the VIF of GDP is greater than 5. Thus, we will
remove this variable from our analysis as it can cause multicollinearity. So,
our new model becomes:
Y i= 0 + 1 infl i + 2 eg i+ 3 fd i+ 4 cpi i+ 5 fb i + 6 col i+ 7 tb i+ 8 iri + 9 er i +ui
The model thus obtained is free from those variables which can cause
multicollinearity. Stata output after regressing using ordered logit model
was as follows:
. ologit Y infl eg fd cpi fb col tb ir er, level(90)
Iteration
Iteration
Iteration
Iteration
Iteration
Iteration
Iteration
Iteration
Iteration
0:
1:
2:
3:
4:
5:
6:
7:
8:
log
log
log
log
log
log
log
log
log
likelihood
likelihood
likelihood
likelihood
likelihood
likelihood
likelihood
likelihood
likelihood
=
=
=
=
=
=
=
=
=
-30.211782
-12.276868
-9.2919335
-7.7803876
-7.3095361
-7.2143445
-7.2122002
-7.2121958
-7.2121958
Number of obs
LR chi2(9)
Prob > chi2
Pseudo R2
Coef.
Std. Err.
infl
eg
fd
cpi
fb
col
tb
ir
er
.9163996
-.0311513
-.0368201
4.623579
.4007487
-6.448815
.1683288
.7909788
-54.07455
.7475124
.1114258
.0281814
2.758492
.5475104
3.671518
.0982859
.4277838
30.85314
/cut1
/cut2
-35.26547
-21.72721
22.95747
19.04497
z
1.23
-0.28
-1.31
1.68
0.73
-1.76
1.71
1.85
-1.75
P>|z|
0.220
0.780
0.191
0.094
0.464
0.079
0.087
0.064
0.080
=
=
=
=
28
46.00
0.0000
0.7613
2.145948
.1521278
.0095342
9.160894
1.301323
-.4097042
.3299947
1.49462
-3.325656
-73.02715
-53.05339
2.496215
9.59897
Here, I have done my analysis at 10% level of significance. The points cut1
and cut2 represent v1 and v2 respectively. To analyze the effect of change in
explanatory variables on probability of dependent variable attaining a
particular value, we need to calculate the marginal effects for being in the
best, moderate risk and worst region.
dy/dx
.0004774
-.0000162
-.0000192
.0024086
.0002088
-.0397507
.0000877
.0004121
-.0281697
Std. Err.
.0021
.00008
.00009
.01025
.00089
.12163
.00038
.00178
.12308
z
0.23
-0.19
-0.23
0.23
0.23
-0.33
0.23
0.23
-0.23
P>|z|
95% C.I.
0.820
0.847
0.822
0.814
0.815
0.744
0.819
0.817
0.819
-.003643
-.000182
-.000186
-.01768
-.001543
-.278149
-.000665
-.003069
-.269412
.004597
.000149
.000148
.022497
.001961
.198648
.00084
.003893
.213072
X
5.15
13.88
78.4704
4.09643
-2.63393
.678571
3.22286
12.3204
1.0075
dy/dx
.0018297
-.0000622
-.0000735
.0092314
.0008001
.02007
.0003361
.0015793
-.1079643
Std. Err.
.00486
.00026
.00021
.02367
.00206
.08865
.00092
.00414
.29153
z
0.38
-0.24
-0.36
0.39
0.39
0.23
0.37
0.38
-0.37
P>|z|
95% C.I.
0.706
0.811
0.720
0.697
0.698
0.821
0.714
0.703
0.711
-.007691
-.000571
-.000476
-.037158
-.00324
-.153676
-.001459
-.006541
-.679362
.01135
.000447
.000329
.05562
.00484
.193816
.002131
.0097
.463433
X
5.15
13.88
78.4704
4.09643
-2.63393
.678571
3.22286
12.3204
1.0075
dy/dx
-.0023071
.0000784
.0000927
-.01164
-.0010089
.0196807
-.0004238
-.0019913
.136134
Std. Err.
.00666
.00033
.00028
.03227
.00281
.04756
.00124
.00564
.39588
z
-0.35
0.23
0.33
-0.36
-0.36
0.41
-0.34
-0.35
0.34
P>|z|
95% C.I.
0.729
0.814
0.739
0.718
0.720
0.679
0.733
0.724
0.731
-.015354
-.000576
-.000452
-.074885
-.006526
-.073537
-.002857
-.013049
-.639775
.01074
.000733
.000637
.051605
.004508
.112898
.002009
.009067
.912043
X
5.15
13.88
78.4704
4.09643
-2.63393
.678571
3.22286
12.3204
1.0075
Interpretation of Results:
Using the co-efficients found in the regression, we can develop a
regression model for country risk assessment.
At 10% level of significance, corruption, coalition, international
reserves, trade balance and exchange rate are significant variables
as their p-value was less than the level of significance, i.e., 0.10.
With per unit increase in inflation rate, the probability of being in the
worst region decreases by 0.23% while the probability of being in the
best region is almost unaffected.
Export growth rate, trade balance and financial depth almost do not
affect of probability of being in either of the two regions.
Conclusions
Increase in inflation rate is likely to improve the credit rating or the
probability of making timely payment because sometimes
development may also cause inflation.
Less corrupt countries are more likely to have a good rating because
in these countries, the public funds are used in a more efficient
manner and thus their probability of defaulting is less.
Data Sources
http://www.standardandpoors.com
Transparency International
Central Bank of Bahrain
http://www.indexmundi.com
CIA World Factbook
IMF, World Economic Outlook, September 2011
United States Department of States
United Nations comtrade database
http://www.armbanks.am
Asian Development Bank Asian Development Outlook 2011
References
Danciu Aniela-Raluca, Goschin Zizi (2011)A Multiple Regression
Model for Country Risk Assessment for European countries
Academy of Economic Studies, Bucharest.
GUJARATI, D. (1995), Basic Econometrics, Third edition. New
York: McGraw-Hill.
Standard & Poor (2011), Ratings Performance. New York: Standard &
Poor.
Alexe S., Hammer P. L., Kogan A., Lejeune M. A. (2003b), A
Combinatorial Approach to Country Risk Rating. New Jersey:
Piscataway, RUTCOR, Rutgers University
Verbeek M(2000)., A guide to modern econometrics Third
Edition. Sussex: John Wiley and Sons Ltd
Greene, William H. (2008), Econometric Analysis Sixth Edition.
New Jersey: Pearson-Prentice Hall Press.
http://www.worldbank.org
http://www.standardandpoors.com
http://www.wikipedia.org